Browse the corpus

Cardiovascular disease (CVD) is now the leading cause of death globally (1). Marked ethnic differences in CVD exist, highlighted by a comparison of migrant and host populations. Mortality resulting from coronary heart disease (CHD) and stroke in South Asian migrants to the United Kingdom are 50% to 100% higher than the general United Kingdom population (2), mirroring risks in the Indian subcontinent (3). In contrast, people of black African and African Caribbean origin enjoy significant protection from CHD in the United Kingdom, although stroke mortality rates are even higher than those of South Asians (2). These observations reflect historical risks in black African migrants to the United States (4) and in Africa itself. These ethnic differentials in mortality have not been explained (5,6). However, previous analyses limited to deaths may be misleading. Both ethnic minority groups have more insulin resistance (IR) and diabetes than Europeans, but although South Asians display classical dyslipidemia and central obesity associated with IR, African Caribbeans have favorable lipoprotein profiles and less central obesity than Europeans. We hypothesized that diabetes and associated metabolic disturbances, measured in midlife, would account for ethnic differences in incident fatal and nonfatal CVD in a unique tri-ethnic community-based United Kingdom cohort followed for 20 years.

ve favorable lipoprotein profiles and less central obesity than Europeans. We hypothesized that diabetes and associated metabolic disturbances, measured in midlife, would account for ethnic differences in incident fatal and nonfatal CVD in a unique tri-ethnic community-based United Kingdom cohort followed for 20 years. Methods The SABRE (Southall and Brent Revisited) study examined a tri-ethnic community-based cohort from North and West London. Details of the cohort have been published (7). Briefly, participants 40 to 69 years of age at baseline (1988 through 1991) were selected randomly from 5-year age- and sex-stratified primary care physician lists (n = 4,063) and workplaces (n = 795) in the London districts of Southall and Brent (Fig. 1). The baseline studies initially were designed to study ethnic differences in metabolic risk factors in association with CVD in men; however, as the studies progressed, the importance of CVD in women was recognized and later recruitment included women. Because African Caribbeans were recruited a little later into the study, the gender rebalance was more complete in this than in the other ethnic groups (7). Ethnicity was agreed on with the interviewer at baseline based on self-report, parental place of origin, and appearance. All South Asians and black African and African Caribbeans were migrants. Most African Caribbeans (92.5%) were born in the Caribbean, and the remainder were born in West Africa. We previously reported similar cardiometabolic risks in these latter 2 groups (8). Most (82%) South Asians were born in the Indian subcontinent, and 14% were born in East Africa. Just more than half (52%) were of Punjabi Sikh origin.

bbeans (92.5%) were born in the Caribbean, and the remainder were born in West Africa. We previously reported similar cardiometabolic risks in these latter 2 groups (8). Most (82%) South Asians were born in the Indian subcontinent, and 14% were born in East Africa. Just more than half (52%) were of Punjabi Sikh origin. Participants attended a baseline clinic after an overnight fast. They underwent blood pressure measurements, electrocardiography, and anthropometry and completed a health and lifestyle questionnaire (7). Height was measured using a stadiometer. Body fat measurements included waist (halfway between costal margin and iliac crest), hip (over greater trochanter), and mid-thigh circumferences. Fasting bloods were drawn, and those not known to have diabetes underwent an oral glucose tolerance test (7). Bloods were analyzed for glucose, insulin, and lipids at the same hospital laboratory (7). Glycated hemoglobin was measured on stored whole blood samples (Southall center only) using an immunoassay on a clinically validated automated analyzer (c311, Roche, Burgess Hill, United Kingdom). Apolipoproteins were measured on available stored aliquots in a subsample of 2,349 participants (1,147 [56%] Europeans, 688 [45%] South Asians, and 514 [82%] African Caribbeans) using immunoturbidimetric methods.

ng an immunoassay on a clinically validated automated analyzer (c311, Roche, Burgess Hill, United Kingdom). Apolipoproteins were measured on available stored aliquots in a subsample of 2,349 participants (1,147 [56%] Europeans, 688 [45%] South Asians, and 514 [82%] African Caribbeans) using immunoturbidimetric methods. Baseline IR was estimated using the homeostasis model assessment 2 calculator (9). Baseline diabetes was defined according to World Health Organization criteria (10), self-report of doctor-diagnosed diabetes, or receipt of antidiabetes medications. Seated resting blood pressure was measured using a random zero sphygmomanometer (Hawksley, London, United Kingdom). Two measurements were obtained, and the mean of these was used in all analyses. Physical activity was summarized as the total weekly energy expended (in megajoules) in sports, walking, and cycling using questions and energy expenditure estimates based on the Allied Dunbar Fitness survey questionnaire (11) and earlier work by Durnin and Passmore (12,13). Frequency of fruit and green vegetable consumption was assessed by a simple dietary questionnaire. The Registrar General's Classification of Occupations (14) was used to assign midlife occupational status as manual or nonmanual.

d Dunbar Fitness survey questionnaire (11) and earlier work by Durnin and Passmore (12,13). Frequency of fruit and green vegetable consumption was assessed by a simple dietary questionnaire. The Registrar General's Classification of Occupations (14) was used to assign midlife occupational status as manual or nonmanual. Since baseline, participants have been flagged for death by the Office for National Statistics. From 2008 through 2011, survivors were invited to take part in a morbidity follow-up. This included a health and lifestyle questionnaire, primary care medical record review, attendance at our clinic at St. Mary's Hospital, London, or a combination thereof. Hospital Episode Statistics were obtained for traced participants since baseline. All participants gave written informed consent. Approval for the study at baseline was obtained from Ealing, Hounslow, and Spelthorne, and University College London research ethics committees, and at follow-up from St. Mary's Hospital Research Ethics Committee (ref. 07/H0712/109). Identification of first post-baseline CHD and stroke events For CHD, a composite endpoint comprised the first event after baseline identified from any of the following sources: 1 Cause of death included any of the following: angina, myocardial infarction or its sequelae, or atherosclerotic heart disease using International Classification of Disease-Ninth Edition (ICD-9) codes 410 through 415 or ICD-Tenth Edition (ICD-10) codes I200 through I259.

e identified from any of the following sources: 1 Cause of death included any of the following: angina, myocardial infarction or its sequelae, or atherosclerotic heart disease using International Classification of Disease-Ninth Edition (ICD-9) codes 410 through 415 or ICD-Tenth Edition (ICD-10) codes I200 through I259. 2 Primary care data were reviewed independently by 2 senior physicians blinded to participant ethnicity and identity. A CHD event was identified if both physicians agreed on definite or probable diagnosis of myocardial infarction or acute coronary syndrome, according to pre-determined criteria used in the ASCOT (Anglo-Scandinavian Cardiac Outcomes Trial) (15), based on symptoms, cardiac enzymes, electrocardiography findings, and hospital discharge diagnosis. Adjudication by a third physician was conducted if required. Coronary interventions (coronary artery bypass graft, angioplasty, stenting) were included as incident CHD events, as was angina confirmed on exercise testing. 3 Hospital Episode Statistics: for causes of death, diagnostic ICD-9 codes 410 through 415 or ICD-10 codes I200 through I259 were used, in addition to any of the following operation codes from the Office of Populations and Surveys classification of interventions and procedures: K401 through K469, K491 through K504, K751 through K759, or U541 (coronary revascularization interventions or rehabilitation for ischemic heart disease).

through I259 were used, in addition to any of the following operation codes from the Office of Populations and Surveys classification of interventions and procedures: K401 through K469, K491 through K504, K751 through K759, or U541 (coronary revascularization interventions or rehabilitation for ischemic heart disease). For stroke, the first event after baseline identified from any of the following sources: 1 Cause of death included ICD-9 codes 430 through 439 or ICD-10 codes I600 through I698. 2 Primary care data were reviewed in a similar manner to CHD, with definite or probable diagnosis of stroke made according to pre-determined criteria based on symptoms, duration of symptoms, and magnetic resonance image or computed tomography imaging. 3 Hospital Episode Statistics: diagnostic ICD-9 codes 430 through 439 or ICD-10 codes I600 through I698. 4 Participant report of physician-diagnosed stroke and duration of symptoms in excess of 24 h. Coronary heart disease or stroke that occurred before baseline was identified from participant report (at baseline) of physician-diagnosed disease (or presence of major Q waves on baseline electrocardiography for CHD). Statistical analyses Primary analyses related to participants with follow-up data from any source. We combined first post-baseline fatal and nonfatal CHD or stroke events as primary outcomes. Baseline characteristics were stratified by sex; ethnic group comparisons were made with Europeans as the reference group within each sex, using parametric (Student t) or nonparametric (Wilcoxon rank-sum or chi-square) statistical tests as appropriate.

seline fatal and nonfatal CHD or stroke events as primary outcomes. Baseline characteristics were stratified by sex; ethnic group comparisons were made with Europeans as the reference group within each sex, using parametric (Student t) or nonparametric (Wilcoxon rank-sum or chi-square) statistical tests as appropriate. Competing risk regression (competing risk = death from other causes) based on the proportional subhazards methods of Fine and Gray (16) was used to describe ethnic differences in incidence of primary outcomes. A priori testing revealed significant interactions between ethnicity and baseline diabetes for both South Asians and African Caribbeans compared with Europeans in stroke prediction. Hence, we show models predicting stroke events stratified by baseline diabetes status.

thnic differences in incidence of primary outcomes. A priori testing revealed significant interactions between ethnicity and baseline diabetes for both South Asians and African Caribbeans compared with Europeans in stroke prediction. Hence, we show models predicting stroke events stratified by baseline diabetes status. We assessed linearity of associations using tertiles of continuous covariates; no initial analyses suggested departure from linearity, and we included these as linear terms, with exceptions of blood pressure, where we used 4 categories (those receiving treatment in the top category and the remaining into thirds by ascending level of blood pressure). We grouped measures of glycemia and IR similarly, with diabetes in the top category. There were no significant sex and ethnicity interactions in association with stroke or CHD events, and we show results for men and women combined to maximize statistical power. We investigated potential mediators of observed ethnic differences in univariate analyses. Multivariable models included covariates that had the greatest positive or negative effects on ethnic differentials. We examined the composite Framingham (17) and INTERHEART (18) study risk factors for stroke and CHD outcomes. To assess the integrity of proportional hazards assumptions, we tested interactions between covariates and follow-up time. We plotted cumulative incidence curves for each ethnic-diabetes group and examined Schoenfeld-like residuals.

We assessed linearity of associations using tertiles of continuous covariates; no initial analyses suggested departure from linearity, and we included these as linear terms, with exceptions of blood pressure, where we used 4 categories (those receiving treatment in the top category and the remaining into thirds by ascending level of blood pressure). We grouped measures of glycemia and IR similarly, with diabetes in the top category. There were no significant sex and ethnicity interactions in association with stroke or CHD events, and we show results for men and women combined to maximize statistical power. We investigated potential mediators of observed ethnic differences in univariate analyses. Multivariable models included covariates that had the greatest positive or negative effects on ethnic differentials. We examined the composite Framingham (17) and INTERHEART (18) study risk factors for stroke and CHD outcomes. To assess the integrity of proportional hazards assumptions, we tested interactions between covariates and follow-up time. We plotted cumulative incidence curves for each ethnic-diabetes group and examined Schoenfeld-like residuals. Sensitivity analyses We compared baseline characteristics of 661 people lost to follow-up with those of 4,196 people followed-up. We repeated the above analyses as follows: (1) excluding 47 people who migrated directly from West Africa; (2) using data derived from direct follow-up only (i.e., without Hospital Episode Statistics data); (3) separately for fatal and nonfatal events; (4) excluding participants with baseline CHD or stroke; and (5) including diabetes identified during follow-up and before events in multivariable analyses and stratification. All analyses were conducted in Stata software version 12.0 (Stata Corporation, College Station, Texas). Statistical significance was accepted as p < 0.05.

ding participants with baseline CHD or stroke; and (5) including diabetes identified during follow-up and before events in multivariable analyses and stratification. All analyses were conducted in Stata software version 12.0 (Stata Corporation, College Station, Texas). Statistical significance was accepted as p < 0.05. Results We traced 4,534 participants (93%) to a United Kingdom address. Follow-up data were obtained for 4,196 participants (>92%) (Fig. 1). Participants were followed up for a median of 20.5 years. Baseline (1988 through 1991) Follow-up data were available for 2,049 Europeans (24% women), 1,517 South Asians (17% women), and 630 African Caribbeans (45% women). At baseline, the mean age of participants was 52.4 years (standard deviation: 6.9 years, range: 40 to 70 years). Baseline stroke was infrequent in all ethnic groups, and African Caribbean men had less CHD than European men. As expected, South Asians and African Caribbeans had more diabetes and hypertension than Europeans. Although they had lower body mass indices than European men, South Asian men were more centrally obese and dyslipidemic. African Caribbeans were less dyslipidemic than Europeans and African Caribbean men were less centrally obese than European men. South Asians and African Caribbeans smoked less and consumed less alcohol than Europeans (Table 1).

mass indices than European men, South Asian men were more centrally obese and dyslipidemic. African Caribbeans were less dyslipidemic than Europeans and African Caribbean men were less centrally obese than European men. South Asians and African Caribbeans smoked less and consumed less alcohol than Europeans (Table 1). There was more loss to follow-up in African Caribbeans (21%) than in Europeans (13%) or South Asians (11%). Baseline characteristics of participants lost to follow-up were similar within the ethnic group when compared with those who were followed up (Online Appendix, Online Table 1). CHD events during follow-up (to 2011) Coronary heart disease events occurred in 1,256 (30%) participants. Fatal CHD was the first recorded follow-up event in 159 participants. South Asians were most and African Caribbeans least at risk for CHD (Fig. 2). Incidence rates increased with age in all ethnic groups (Table 2). South Asians were on average 2.3 years younger at first post-baseline CHD event than Europeans (63.9 ± 8.6 years vs. 66.2 ± 8.6 years, p < 0.001), whereas African Caribbeans were 2.4 years older (68.6 ± 7.5 years, p = 0.008). Angina (exercise test confirmed) comprised 3.5% of first CHD events, whereas coronary revascularization procedures comprised the first event in 43 (7.8%) Europeans, 51 (8.5%) South Asians, and 2 (1.9%) African Caribbeans; the remainder of first events were identified as acute ischemic events or were related to atherosclerotic heart disease.

confirmed) comprised 3.5% of first CHD events, whereas coronary revascularization procedures comprised the first event in 43 (7.8%) Europeans, 51 (8.5%) South Asians, and 2 (1.9%) African Caribbeans; the remainder of first events were identified as acute ischemic events or were related to atherosclerotic heart disease. The age- and sex-adjusted subhazard ratio (SHR) for CHD between South Asians and Europeans was 1.70 (95% confidence interval [CI]: 1.52 to 1.91, p < 0.001) (Table 3). Of all measured individual risk factors, waist-to-hip ratio best attenuated the South Asian excess risk, although risk remained significantly elevated compared with that of Europeans (SHR: 1.45, 95% CI: 1.28 to 1.64, p < 0.001). Other attenuating factors included diabetes and measures of glycemia, IR, triglyceride, and HDL cholesterol. Adjustment for both protective (lower smoking prevalence) and adverse (waist-to-hip ratio and glycosylated hemoglobin) factors (adjusted SHR: 1.49, 95% CI: 1.25 to 1.77, p < 0.001) did not account for the South Asian CHD excess, nor did adjustment for Framingham or INTERHEART factors (Table 3). Markers of baseline glycemic status were (nonsignificantly) more predictive of CHD events in South Asians than in Europeans.

ylated hemoglobin) factors (adjusted SHR: 1.49, 95% CI: 1.25 to 1.77, p < 0.001) did not account for the South Asian CHD excess, nor did adjustment for Framingham or INTERHEART factors (Table 3). Markers of baseline glycemic status were (nonsignificantly) more predictive of CHD events in South Asians than in Europeans. CHD risk was lower in African Caribbeans than Europeans (age- and sex-adjusted SHR: 0.64, 95% CI: 0.52 to 0.79, p < 0.001). Our multivariate model, including favorable and unfavorable risk factors (HDL and LDL cholesterol, waist-to-thigh ratio, blood pressure, age, sex), did not account for the African Caribbean protection from CHD (adjusted SHR: 0.74, 95% CI: 0.59 to 0.93, p = 0.01) (Table 3). Baseline diabetes was associated most strongly with CHD events in South Asians and was associated least strongly in African Caribbeans. The age- and sex-adjusted SHR for diabetes were as follows: South Asians: 1.90 (95% CI: 1.59 to 2.27), Europeans: 1.61 (95% CI: 1.23 to 2.10), and African Caribbeans: 1.31 (95% CI: 0.85 to 2.02). Findings from sensitivity analyses were similar to those presented here. Stroke events during follow-up (to 2011) Stroke events occurred in 401 participants during follow-up. African Caribbeans had the highest and Europeans the lowest rates of stroke (Table 2).

CHD risk was lower in African Caribbeans than Europeans (age- and sex-adjusted SHR: 0.64, 95% CI: 0.52 to 0.79, p < 0.001). Our multivariate model, including favorable and unfavorable risk factors (HDL and LDL cholesterol, waist-to-thigh ratio, blood pressure, age, sex), did not account for the African Caribbean protection from CHD (adjusted SHR: 0.74, 95% CI: 0.59 to 0.93, p = 0.01) (Table 3). Baseline diabetes was associated most strongly with CHD events in South Asians and was associated least strongly in African Caribbeans. The age- and sex-adjusted SHR for diabetes were as follows: South Asians: 1.90 (95% CI: 1.59 to 2.27), Europeans: 1.61 (95% CI: 1.23 to 2.10), and African Caribbeans: 1.31 (95% CI: 0.85 to 2.02). Findings from sensitivity analyses were similar to those presented here. Stroke events during follow-up (to 2011) Stroke events occurred in 401 participants during follow-up. African Caribbeans had the highest and Europeans the lowest rates of stroke (Table 2). In South Asians and African Caribbeans, the age- and sex-adjusted SHR, 1.45 (95% CI: 1.17 to 1.80, p = 0.001) and 1.50 (95% CI: 1.13 to 2.00, p = 0.005), respectively, were strongly but incompletely attenuated on adjustment for baseline diabetes (SHR: 1.27 [95% CI: 1.02 to 1.58, p = 0.03] and 1.33 [95% CI: 1.01 to 1.76, p = 0.044]).

s and African Caribbeans, the age- and sex-adjusted SHR, 1.45 (95% CI: 1.17 to 1.80, p = 0.001) and 1.50 (95% CI: 1.13 to 2.00, p = 0.005), respectively, were strongly but incompletely attenuated on adjustment for baseline diabetes (SHR: 1.27 [95% CI: 1.02 to 1.58, p = 0.03] and 1.33 [95% CI: 1.01 to 1.76, p = 0.044]). Diabetes was predictive of stroke in all ethnic groups (Fig. 2), but most profoundly in African Caribbeans, in whom diabetes was associated with a 3.0-fold (95% CI: 1.8 to 4.8) age-adjusted incidence of stroke compared with a 1.3-fold (95% CI: 0.8 to 2.1) age-adjusted incidence of stroke in Europeans (p = 0.019 for ethnicity and diabetes interaction) (Online Appendix, Online Table 2). A similar interaction was observed for South Asians, in whom diabetes was associated with a 2.5-fold (95% CI: 1.8 to 3.4) incidence of stroke (p = 0.038 for interaction) (Online Table 2).

age-adjusted incidence of stroke in Europeans (p = 0.019 for ethnicity and diabetes interaction) (Online Appendix, Online Table 2). A similar interaction was observed for South Asians, in whom diabetes was associated with a 2.5-fold (95% CI: 1.8 to 3.4) incidence of stroke (p = 0.038 for interaction) (Online Table 2). In people with diabetes, the age-adjusted SHR in South Asians versus Europeans, at 1.96 (95% CI: 1.15 to 3.33, p = 0.013), was little affected on adjustment for any measured risk factors. Multivariable adjustment (smoking, years of education) had a small attenuating effect, whereas Framingham and INTERHEART adjustments did not explain the South Asian excess in those with diabetes. In African Caribbeans with diabetes, the age-adjusted SHR (2.30, 95% CI: 1.25 to 4.22, p = 0.007) was little changed on multivariable adjustment. The most marked attenuation was obtained by the INTERHEART adjustment (adjusted SHR: 1.62, 95% CI: 0.71 to 3.65, p = 0.25). In those without diabetes, there was only a modest and nonsignificant ethnic excess in risk of stroke (Table 4).

0, 95% CI: 1.25 to 4.22, p = 0.007) was little changed on multivariable adjustment. The most marked attenuation was obtained by the INTERHEART adjustment (adjusted SHR: 1.62, 95% CI: 0.71 to 3.65, p = 0.25). In those without diabetes, there was only a modest and nonsignificant ethnic excess in risk of stroke (Table 4). The diabetes and ethnicity interactions also were strong on analyses of fatal stroke events only. Date of diagnosis of pre-baseline diabetes was available for 440 (76%) participants with baseline diabetes; however, adjustment for age at diagnosis in these participants did not alter the ethnic differentials in stroke events. Sensitivity analyses demonstrated similar magnitude and direction of ethnicity-associated excess stroke risk and similar diabetes and ethnicity interactions. Inclusion of new diabetes (identified during follow-up and before stroke) in addition to baseline diabetes in multivariable analyses confirmed the additional risk of stroke in South Asians and African Caribbeans with diabetes compared with Europeans with diabetes, with age- and sex-adjusted SHR as follows: South Asians: 2.10 (95% CI: 1.34 to 3.30, p = 0.001), and African Caribbeans: 2.17 (95% CI: 1.28 to 3.68, p = 0.004).

s in multivariable analyses confirmed the additional risk of stroke in South Asians and African Caribbeans with diabetes compared with Europeans with diabetes, with age- and sex-adjusted SHR as follows: South Asians: 2.10 (95% CI: 1.34 to 3.30, p = 0.001), and African Caribbeans: 2.17 (95% CI: 1.28 to 3.68, p = 0.004). Discussion People of European, South Asian, and African Caribbean origin vary markedly in CVD risk, in parallel with differences in metabolic factors such as IR, dyslipidemia, and central adiposity. However, the between-ethnic group differences in CVD remained even after adjustment for conventional cardiometabolic risk factors measured in midlife. It is well recognized that explanations for disease risk within a group may not explain differences between groups (19). In 20 years of follow-up, CHD rates were significantly elevated in South Asian migrants compared with British Europeans. Multivariable adjustment for baseline risk factors explained little of the excess risk. In African Caribbeans (also migrants) rates of CHD continue to be 35% lower than in comparable Europeans, with adjustment for favorable midlife risk factors, such as HDL and LDL cholesterol, accounting for little of this protection. Rates of stroke remain elevated in both ethnic minority groups and more markedly so in people with diabetes in midlife, being 2.0 times greater in African Caribbeans and 1.7 times greater in South Asians compared with Europeans with diabetes.

s, such as HDL and LDL cholesterol, accounting for little of this protection. Rates of stroke remain elevated in both ethnic minority groups and more markedly so in people with diabetes in midlife, being 2.0 times greater in African Caribbeans and 1.7 times greater in South Asians compared with Europeans with diabetes. There are few population-based studies comparing cardiovascular risk in South Asians and Europeans (20). To our knowledge, SABRE is the only one to have published on longitudinal associations between risk factors and CHD events. We extend our earlier reports, confined to fatal events (5), showing that greater case fatality in South Asians is unlikely to be the explanation for their excess CHD. Metabolic risk factors accounted for approximately one third of the excess CHD risk in South Asians. Residual confounding, specifically imprecise measurement of key risk factors, could explain why more of the South Asian excess could not be explained by metabolic risk factors. Single measurements of risk factors in midlife would be poorly representative of lifetime exposure (21). Additionally, factors acting at specific critical periods of the life course, for example in utero and infancy, may play a strong and independent role in determining adult risk (22). The INTERHEART study showed that just 9 risk factors account for most of the population-attributable risk of CHD, even in South Asian populations (23), and of these, the main roles were played by lipids and smoking. Adjustment for these factors did not account for interethnic differences in our study. However, although we have fasting HDL and LDL cholesterol readings for our entire study population, we have data for apolipoproteins only on a subsample. Nevertheless, risk estimates on multivariable adjustment for the apolipoprotein subset were identical to those yielded by multivariable analysis on the full dataset using the HDL-to-LDL ratio as a proxy for lipoproteins, confirming that known clusters of risk factors identified from other studies do not account for the ethnic difference in CHD. We do not have comparable measures of psychosocial stress at baseline, but the likelihood that the association between stress and myocardial infarction in the INTERHEART study is the result of reverse causality (i.e., the infarct increasing stress levels rather than the reverse) cannot be discounted.

difference in CHD. We do not have comparable measures of psychosocial stress at baseline, but the likelihood that the association between stress and myocardial infarction in the INTERHEART study is the result of reverse causality (i.e., the infarct increasing stress levels rather than the reverse) cannot be discounted. It is tempting to suggest a role for genetic factors, and although these cannot be discounted, it is notable that, to date, no such factor has been identified (24). Further, CHD rates have escalated in India in the last 50 years, with marked urban and rural differences, suggesting that, as in the West, CHD is a consequence of industrialization and is associated with reduced physical activity and an adverse diet (3). Genetic factors alone are unlikely to have acted this rapidly. South Asian migrants to the United Kingdom will have grown up in circumstances where infant mortality was high, malnutrition rife, and infectious disease endemic (25). Migration to an obesogenic environment in the 1950s and 1960s, the like of which is only now emerging in India today, will have compounded those early insults. Epigenetic alterations in response to exposures can determine phenotype, perhaps dependent on genotype, offering an attractive explanation for the marked increase in cardiometabolic disease in South Asians (26,27). Similarly, we show that the African Caribbean protection from CHD is not wholly explained by conventional risk factors measured in middle age, although the greatest attenuation is observed on adjustment for the highly favorable lipid patterns of African Caribbeans, despite an excess of diabetes and IR. Again, residual confounding resulting from lack of longitudinal measures and early life influences on development and growth may play a role. Protection from CHD also was observed in African Americans in the early part of the last century (4), although this protection has been somewhat eroded (28), implicating the impact of changes in environmental risk factors.

ack of longitudinal measures and early life influences on development and growth may play a role. Protection from CHD also was observed in African Americans in the early part of the last century (4), although this protection has been somewhat eroded (28), implicating the impact of changes in environmental risk factors. Our previous analyses hinted at a potential excess stroke mortality associated with diabetes in African Caribbeans (6). We have now taken these earlier analyses forward with longer follow-up and the inclusion of nonfatal events that continue to suggest that diabetes and dysglycemia may have more potent roles as precursors of stroke in both ethnic minorities. This association has not been reported, or perhaps sought previously, and was not observed for CHD, so it must be treated with caution until it is confirmed or refuted in other datasets. If confirmed, this would identify potential diabetes-related mechanisms in understanding ethnic differences in stroke. A potential candidate is the impaired cerebral autoregulation in both African Caribbeans and South Asians that we have noted as a consequence of autonomic dysfunction resulting from hyperglycemia and IR (29,30). This could contribute to excess stroke in minority ethnic groups. Duration of exposure to hyperglycemia also may contribute, although we did not demonstrate a significant explanatory effect of age at diagnosis in this study population. Measurement of age at diagnosis of diabetes is imprecise, so we cannot discount duration of exposure to hyperglycemia as an explanation for the greater toxicity of diabetes for stroke risk in ethnic minority groups.

gh we did not demonstrate a significant explanatory effect of age at diagnosis in this study population. Measurement of age at diagnosis of diabetes is imprecise, so we cannot discount duration of exposure to hyperglycemia as an explanation for the greater toxicity of diabetes for stroke risk in ethnic minority groups. Study limitations To our knowledge, this is the largest tri-ethnic cohort in the United Kingdom with a 20-year follow-up between middle and older age, thus providing valuable and unique prospective ethnicity-specific information on CHD and stroke incidence. Our combined data sources for the primary analyses provide follow-up data on more than 90% of the original study population, and although loss to follow-up was more frequent in African Caribbeans (21%), baseline characteristics of those lost to follow-up were similar to those participants who were followed up. Although this is a relatively large cohort, the numbers of stroke events were small, in particular and with the exception of incident diabetes, our baseline measurements were limited to those made on only 1 occasion 20 years ago, meaning that we cannot account for changes in other risk factors during the follow-up period or in earlier life. We were unable to differentiate reliably between stroke subtypes, which may vary by ethnicity, although it is reported that the risks of all stroke types are elevated in South Asian and African Caribbean populations and that ischemic strokes are most frequent in all 3 groups (31–35). Classification of causes of death and hospital discharge codes may be imprecise, but sensitivity analyses of fatal events alone and nonfatal events restricted to those ascertained from primary care record review or participant report result in remarkably similar findings as those reported here. We also should point out that more than half of South Asians in our study population were of Punjabi Sikh origin, and although most South Asian populations are at increased risk of diabetes and CVD, our findings may not apply to all South Asians. Finally, South Asians and African Caribbeans who migrated to the United Kingdom in the second half of the 20th century largely did so for economic reasons, and they may not be wholly representative of their countries of origin or of migrants to other countries.

s and CVD, our findings may not apply to all South Asians. Finally, South Asians and African Caribbeans who migrated to the United Kingdom in the second half of the 20th century largely did so for economic reasons, and they may not be wholly representative of their countries of origin or of migrants to other countries. Conclusions Morbidity and mortality resulting from CHD are elevated in South Asians and are lower in African Caribbeans compared with European-origin populations. These differences were not explained by conventional risk factors measured in midlife. Factors across the life course, in particular the mismatch between early and later life environments in migrant cohorts, may be key (36). This is of critical importance in lower income countries, where CHD risks are increasing, and in African Caribbean populations, where there is evidence that protection from CHD erodes with time in industrialized environments. Diabetes may be associated more strongly with stroke risk in both minority ethnic groups, and although further confirmation is needed in larger studies, we suggest that early interventions to reduce cardiovascular risk may be of particular importance in these high-risk populations. Appendix Online Appendix, Online Tables 1 and 2 Acknowledgments The authors thank the SABRE study team, collaborators, and coinvestigators and Drs. Neil Chapman and Ajay Gupta for their assistance in identifying stroke events.

Conclusions Morbidity and mortality resulting from CHD are elevated in South Asians and are lower in African Caribbeans compared with European-origin populations. These differences were not explained by conventional risk factors measured in midlife. Factors across the life course, in particular the mismatch between early and later life environments in migrant cohorts, may be key (36). This is of critical importance in lower income countries, where CHD risks are increasing, and in African Caribbean populations, where there is evidence that protection from CHD erodes with time in industrialized environments. Diabetes may be associated more strongly with stroke risk in both minority ethnic groups, and although further confirmation is needed in larger studies, we suggest that early interventions to reduce cardiovascular risk may be of particular importance in these high-risk populations. Appendix Online Appendix, Online Tables 1 and 2 Acknowledgments The authors thank the SABRE study team, collaborators, and coinvestigators and Drs. Neil Chapman and Ajay Gupta for their assistance in identifying stroke events. The study was funded at baseline by the Medical Research Council, Diabetes UK, and British Heart Foundation and at follow-up by the Wellcome Trust and British Heart Foundation. The authors have reported that they have no relationships relevant to the contents of this paper to disclose. For supplemental tables, please see the online version of this article. Figure 1 Follow-Up of SABRE Study Cohort (2008 Through 2011)

The study was funded at baseline by the Medical Research Council, Diabetes UK, and British Heart Foundation and at follow-up by the Wellcome Trust and British Heart Foundation. The authors have reported that they have no relationships relevant to the contents of this paper to disclose. For supplemental tables, please see the online version of this article. Figure 1 Follow-Up of SABRE Study Cohort (2008 Through 2011) Flowchart showing status and responses at the 20-year follow-up. SABRE = Southall and Brent Revisited. Figure 2 Cumulative Incidence of Coronary Heart Disease and Stroke Events Cumulative incidence curves for combined fatal and nonfatal events, adjusted for age, sex, smoking, and baseline coronary heart disease (CHD) and stroke. Table 1 Baseline Characteristics (1988 Through 1991), Unadjusted

Flowchart showing status and responses at the 20-year follow-up. SABRE = Southall and Brent Revisited. Figure 2 Cumulative Incidence of Coronary Heart Disease and Stroke Events Cumulative incidence curves for combined fatal and nonfatal events, adjusted for age, sex, smoking, and baseline coronary heart disease (CHD) and stroke. Table 1 Baseline Characteristics (1988 Through 1991), Unadjusted Men Women European South Asian African Caribbean European South Asian African Caribbean n 1,564 1,259 347 485 258 283 Age (yrs) 53.0 ± 7.1 51.1 ± 7.0, p < 0.001 53.6 ± 5.8, p = 0.141 53.2 ± 6.8 50.3 ± 6.5, p < 0.001 52.8 ± 6.1, p = 0.48 Diabetes 112 (7) 278 (22), p < 0.001 64 (18), p < 0.001 21 (4) 44 (17), p < 0.001 60 (21), p < 0.001 Treated hypertension 139 (9) 169 (13), p < 0.001 69 (20), p < 0.001 55 (11) 34 (13), p = 0.46 79 (28), p < 0.001 Known CHD 115 (8) 86 (7), p = 0.59 9 (3), p = 0.001 17 (4) 3 (1), p = 0.060 10 (4), p = 0.98 Known stroke 21 (1) 22 (2), p = 0.38 11 (3), p = 0.018 8 (2) 5 (2), p = 0.76 7 (2), p = 0.43 SBP (mm Hg) 123 ± 17 125 ± 17, p < 0.001 128 ± 17, p < 0.001 120 ± 17 125 ± 21, p < 0.001 131 ± 17, p < 0.001 DBP (mm Hg) 77 ± 11 81 ± 10, p < 0.001 81 ± 12, p < 0.001 75 ± 10 77 ± 10, p = 0.010 82 ± 12, p < 0.001 Waist circumference (cm) 92.1 ± 11.1 93.3 ± 9.6, p = 0.002 89.7 ± 10.2, p < 0.001 80.2 ± 12.2 86.2 ± 10.7, p < 0.001 88.9 ± 11.6, p < 0.001 Waist-to-hip ratio 0.94 (0.90–0.99) 0.98 (0.94–1.02), p < 0.001 0.94 (0.90–0.99), p = 0.89 0.79 (0.75–0.85) 0.87 (0.81–0.93), p < 0.001 0.86 (0.81–0.92), p < 0.001 Body mass index (kg/m2) 26.2 ± 3.9 25.9 ± 3.3, p = 0.008 26.4 ± 3.3, p = 0.44 26.1 ± 4.6 27.5 ± 4.6, p < 0.001 29.4 ± 5.0, p < 0.001 Weight (kg) 78.5 (71.0, 87.0) 73.8 (67.1–81.4), p < 0.001 77.5 (69.2–85.1), p = 0.067 65.5 (58.5–73.9) 64.6 (57.4–72.4), p = 0.100 74.6 (67.0–84.0), p < 0.001 Height (cm) 174.3 ± 6.9 169.8 ± 6.6, p < 0.001 171.9 ± 6.9, p < 0.001 161.2 ± 6.4 154.7 ± 5.6, p < 0.001 161.2 ± 5.5, p = 0.86 Total cholesterol (mmol/l) 6.0 (5.3–6.8) 5.9 (5.2–6.6), p = 0.002 5.4 (4.8–6.3), p < 0.001 6.0 (5.2–6.9) 5.7 (5.0–6.5), p < 0.001 5.6 (4.8–6.5), p < 0.001 Triglycerides (mmol/l) 1.5 (1.0–2.1) 1.8 (1.2–2.6), p < 0.001 1.1 (0.8–1.5), p < 0.001 1.3 (0.9–1.8) 1.4 (1.1–1.9), p < 0.001 1.1 (0.8–1.4), p < 0.001 HDL cholesterol (mmol/l) 1.2 (1.1–1.5) 1.1 (1.0–1.3), p < 0.001 1.4 (1.2–1.7), p < 0.001 1.6 (1.3–1.9) 1.4 (1.2–1.6), p < 0.001 1.6 (1.4–1.9), p = 0.61 Apolipoprotein B-to-A1 ratio 0.69 (0.56–0.83), n = 6,700 0.73 (0.61–0.85), n = 451 0.56 (0.44–0.71), n = 240 0.55 (0.45–0.67), n = 477 0.60 (0.49–0.70), n = 237 0.51 (0

HDL cholesterol (mmol/l) 1.2 (1.1–1.5) 1.1 (1.0–1.3), p < 0.001 1.4 (1.2–1.7), p < 0.001 1.6 (1.3–1.9) 1.4 (1.2–1.6), p < 0.001 1.6 (1.4–1.9), p = 0.61 Apolipoprotein B-to-A1 ratio 0.69 (0.56–0.83), n = 6,700 0.73 (0.61–0.85), n = 451 0.56 (0.44–0.71), n = 240 0.55 (0.45–0.67), n = 477 0.60 (0.49–0.70), n = 237 0.51 (0 .39–0.64), n = 274 Fasting glucose (mmol/l) 5.4 (5.1–5.9) 5.6 (5.2–6.4), p < 0.001 5.6 (5.2–6.4), p < 0.001 5.3 (4.9–5.7) 5.1 (4.7–5.5), p = 0.003 5.6 (5.1–6.3), p < 0.001 Fasting insulin (μIU/ml) 7.5 (5.0–10.9) 10.8 (7.3–15.7), p < 0.001 8.7 (5.7–12.4), p < 0.001 5.3 (3.8–8.1) 7.7 (5.3–11.5), p < 0.001 9.3 (6.2–13.0), p < 0.001 HbA1c (%) 5.6 (5.4–5.8), n = 1,233 5.9 (5.6–6.3), p < 0.001, n = 939 5.5 (5.3–5.8), n = 189 5.8 (5.5–6.1), p < 0.001, n = 212 — HOMA2 IR 0.9 (0.6–1.3) 1.2 (0.8–1.9), p < 0.001 1.0 (0.7–1.5), p < 0.001 0.6 (0.4–1.0) 0.9 (0.6–1.3), p < 0.001 1.1 (0.7–1.5), p < 0.001 Physical activity (leisure time; MJ/week) 4.0 (1.5–6.1) 3.5 (1.0–4.0), p < 0.001 3.7 (1.2–4.5), p < 0.001 3.7 (1.2–4.4) 1 (1–3.5), p = 0.001 3.7 (1.2–4.1), p = 0.46 Smoking categories, current/former/never (%) 34/40/26 16/11/73, p < 0.001 27/19/54, p < 0.001 30/24/46 2/0.5/98, p < 0.001 9/9/83, p < 0.001 Alcohol (units/week) 10.9 (2.4–24.5) 3 (0–14.0), p < 0.001 9.1 (2.3–23.0), p = 0.26 1.6 (0.2–6.4) 0 (0–0), p < 0.001 0.8 (0.1–3.3), p = 0.002 Green vegetables or fruit, daily/most days 1,215 (68) 953 (68), p = 0.81 272 (61), p = 0.004 409 (73.3) 223 (77.4), p = 0.19 273 (79.1), p = 0.048 Manual occupation 981 (63) 959 (77), p < 0.001 289 (85), p < 0.001 251 (53) 159 (72), p < 0.001 167 (62), p = 0.030 Years of education 10 (9–11) 12 (10–14), p < 0.001 11 (9–12), p = 0.015 10 (9–11) 11 (8–12), p = 0.97 11 (9–12), p = 0.37 Values are mean± SD, n (%), or median (25th–75th centiles).

273 (79.1), p = 0.048 Manual occupation 981 (63) 959 (77), p < 0.001 289 (85), p < 0.001 251 (53) 159 (72), p < 0.001 167 (62), p = 0.030 Years of education 10 (9–11) 12 (10–14), p < 0.001 11 (9–12), p = 0.015 10 (9–11) 11 (8–12), p = 0.97 11 (9–12), p = 0.37 Values are mean± SD, n (%), or median (25th–75th centiles). p Values for comparisons with Europeans of same sex. HbA1c indicates that baseline HbA1c is not available for African Caribbeans. CHD = coronary heart disease; DBP = diastolic blood pressure; HbA1c = glycosylated hemoglobin; HDL = high-density lipoprotein; HOMA2 = homeostasis model assessment 2; IR = insulin resistance; SBP = systolic blood pressure. Table 2 Incidence Rates for First Post-Baseline CHD and Stroke Events (Nonfatal and Fatal) by Ethnicity, Sex, and Age Group at End of Follow-Up

CHD = coronary heart disease; DBP = diastolic blood pressure; HbA1c = glycosylated hemoglobin; HDL = high-density lipoprotein; HOMA2 = homeostasis model assessment 2; IR = insulin resistance; SBP = systolic blood pressure. Table 2 Incidence Rates for First Post-Baseline CHD and Stroke Events (Nonfatal and Fatal) by Ethnicity, Sex, and Age Group at End of Follow-Up Person-Years of Follow-up European South Asian African Caribbean Men Women Men Women Men Women CHD No. of first events 473 78 526 73 67 38 Age < 55 yrs 16,631 4.3 (2.9–6.4) 1.7 (0.6–5.3) 12.6 (10.0–15.8) 3.1 (1.2–8.2) 2.1 (0.5–8.4) 1.0 (0.1–7.3) Age 56–70 yrs 41,644 17.5 (15.6–19.8) 8.7 (6.5–11.7) 26.7 (23.9–29.8) 19.1 (14.5–25.2) 8.6 (6.1–12.1) 5.6 (3.5–9.0) Age 70+ yrs 12,448 34.7 (29.9–40.2) 16.6 (11.6–23.8) 50.2 (42.6–59.2) 33.5 (21.1–53.2) 23.8 (16.8–33.6) 22.3 (14.4–34.6) Stroke No. of first events 139 34 130 27 44 27 Age < 55 yrs 16,877 0.3 (0.1–1.4) 0 1.2 (0.6–2.4) 0.8 (0.1–5.4) 3.2 (1.0–9.9) 0 Age 56–70 yrs 44,910 3.4 (2.6–4.4) 3.1 (1.9–5.0) 5.4 (4.3–6.8) 5.1 (3.0–8.5) 3.8 (2.3–6.3) 4.9 (2.9–8.1) Age 70+ yrs 15,006 12.6 (10.0–16.0) 9.2 (5.8–15.0) 12.9 (10.0–17.0) 16.9 (9.3–30.0) 17.5 (12.0–26.0) 12.5 (7.1–22.0) Data are rates/1,000 person-years (95% confidence interval). Abbreviation as in Table 1. Table 3 Coronary Heart Disease Events: Subhazard Ratios for First Post-Baseline Events (Nonfatal and Fatal) During 20.5 Years of Follow-Up in South Asians and African Caribbeans Compared With Europeans

Person-Years of Follow-up European South Asian African Caribbean Men Women Men Women Men Women CHD No. of first events 473 78 526 73 67 38 Age < 55 yrs 16,631 4.3 (2.9–6.4) 1.7 (0.6–5.3) 12.6 (10.0–15.8) 3.1 (1.2–8.2) 2.1 (0.5–8.4) 1.0 (0.1–7.3) Age 56–70 yrs 41,644 17.5 (15.6–19.8) 8.7 (6.5–11.7) 26.7 (23.9–29.8) 19.1 (14.5–25.2) 8.6 (6.1–12.1) 5.6 (3.5–9.0) Age 70+ yrs 12,448 34.7 (29.9–40.2) 16.6 (11.6–23.8) 50.2 (42.6–59.2) 33.5 (21.1–53.2) 23.8 (16.8–33.6) 22.3 (14.4–34.6) Stroke No. of first events 139 34 130 27 44 27 Age < 55 yrs 16,877 0.3 (0.1–1.4) 0 1.2 (0.6–2.4) 0.8 (0.1–5.4) 3.2 (1.0–9.9) 0 Age 56–70 yrs 44,910 3.4 (2.6–4.4) 3.1 (1.9–5.0) 5.4 (4.3–6.8) 5.1 (3.0–8.5) 3.8 (2.3–6.3) 4.9 (2.9–8.1) Age 70+ yrs 15,006 12.6 (10.0–16.0) 9.2 (5.8–15.0) 12.9 (10.0–17.0) 16.9 (9.3–30.0) 17.5 (12.0–26.0) 12.5 (7.1–22.0) Data are rates/1,000 person-years (95% confidence interval). Abbreviation as in Table 1. Table 3 Coronary Heart Disease Events: Subhazard Ratios for First Post-Baseline Events (Nonfatal and Fatal) During 20.5 Years of Follow-Up in South Asians and African Caribbeans Compared With Europeans Europeans South Asians African Caribbeans Unadjusted Reference group 1.61 (1.43–1.80), p < 0.001 0.58 (0.47–0.71), p < 0.001 Adjusted for Age 1.77 (1.58–1.99), p < 0.001 0.59 (0.48–0.72), p < 0.001 Sex and age (model 1) 1.70 (1.52–1.91), p < 0.001 0.64 (0.52–0.79), p < 0.001 Model 1+ Smoking (never, former, current) 1.98 (1.72–2.27), p < 0.001 0.68 (0.55–0.84), p = 0.001 Baseline CHD 1.73 (1.54–1.94), p < 0.001 0.68 (0.55–0.84), p < 0.001 Diabetes 1.55 (1.38–1.75), p < 0.001 0.60 (0.49–0.75), p < 0.001 SBP/treated hypertension⁎ 1.62 (1.44–1.82), p < 0.001 0.58 (0.47–0.71), p < 0.001 Waist-to-hip ratio 1.45 (1.28–1.64), p < 0.001 0.59 (0.48–0.73), p < 0.001 Waist-to-thigh ratio 1.49 (1.32–1.69), p < 0.001 0.72 (0.58–0.88), p = 0.002 Body mass index 1.71 (1.52–1.92), p < 0.001 0.61 (0.50–0.75), p < 0.001 Total cholesterol 1.78 (1.58–2.00), p < 0.001 0.71 (0.57–0.87), p = 0.002 Triglycerides 1.60 (1.42–1.80), p < 0.001 0.71 (0.58–0.88), p = 0.002 HDL and LDL cholesterol 1.68 (1.48–1.90), p < 0.001 0.74 (0.60–0.92), p = 0.008 Fasting glucose† 1.56 (1.38–1.75), p < 0.001 0.59 (0.48–0.73), p < 0.001 HOMA2 IR† 1.51 (1.34–1.72), p < 0.001 0.58 (0.47–0.72), p < 0.001 HbA1c‡ 1.48 (1.28–1.70), p < 0.001 Not available Alcohol consumption 1.67 (1.49–1.88), p < 0.001 0.64 (0.52–0.79), p < 0.001 Green vegetable/fruit consumption 1.70 (1.51–1.91), p < 0.001 0.65 (0.53–0.80), p < 0.001 Physical activity 1.70 (1.51–1.91), p < 0.001 0.66 (0.53–0.81), p < 0.001 Years of education 1.76 (1.56–1.99), p < 0.001 0.66 (0.54–0.82), p < 0.001 Manual occupation 1.68 (1.49–1.90), p < 0.001 0.65 (0.53–0.80), p < 0.001 Multivariable Include covariates with largest effects on ethnic differences Model 1+ Waist-to-hip ratio and HbA1c,‡ smoking 1.49 (1.25–1.77), p < 0.001 — Waist-to-thigh ratio, LDL + HDL cholesterol, SBP/treated hypertension†, HOMA IR — 0.74 (0.59–0.93), p = 0.011 Composite models Framingham variables§ 1.77 (1.53–2.05), p < 0.001 0.66 (0.52–0.82), p < 0.001 INTERHEART variables∥ 1.52 (1.32–1.74), p < 0.001 0.65 (0.52–0.82), p < 0.001 INTERHEART variables (subset)¶ 1.52 (1.26–1.85), p < 0.001 0.68 (0.52–0.89), p = 0.004 Data are rates/1,000 person-years (95% confidence interval).

Composite models Framingham variables§ 1.77 (1.53–2.05), p < 0.001 0.66 (0.52–0.82), p < 0.001 INTERHEART variables∥ 1.52 (1.32–1.74), p < 0.001 0.65 (0.52–0.82), p < 0.001 INTERHEART variables (subset)¶ 1.52 (1.26–1.85), p < 0.001 0.68 (0.52–0.89), p = 0.004 Data are rates/1,000 person-years (95% confidence interval). Abbreviations as in Table 1. ⁎ Tertiles 1–3 (= 4 for those with treated hypertension). † Tertiles 1–3 (= 4 for those with diabetes). ‡ HbA1c available only for 1,422 Europeans and 1,151 South Asians (age- and sex-adjusted subhazard ratio in this group, 1.68 [1.46–1.92]). § Framingham variables: age, sex, diabetes, current smoking, total and HDL cholesterol, SBP and treated hypertension (complete data available for multivariable adjustment 97% Europeans, 96% South Asians, 95% African Caribbeans). ∥ Age, sex, current smoking, diabetes, treated hypertension, HDL and LDL cholesterol, waist-to-hip ratio (sex-specific tertiles), alcohol consumption, daily green vegetable/fruit consumption, tertiles of physical activity (complete data available for multivariable adjustment for 93% Europeans, 90% South Asians, 89% African Caribbeans).

nt smoking, diabetes, treated hypertension, HDL and LDL cholesterol, waist-to-hip ratio (sex-specific tertiles), alcohol consumption, daily green vegetable/fruit consumption, tertiles of physical activity (complete data available for multivariable adjustment for 93% Europeans, 90% South Asians, 89% African Caribbeans). ¶ Age, sex, current smoking, diabetes, treated hypertension, apolipoprotein B-to-A1 ratio, waist-to-hip ratio (sex-specific tertiles), alcohol consumption, daily green vegetable/fruit consumption, tertiles of physical activity. Apolipoprotein data were available only for 1,147 Europeans, 688 South Asians, and 514 African Caribbeans (age- and sex-adjusted subhazard ratios in this subsample: South Asian vs. European: 1.77 [1.49–2.10], p < 0.001, African Caribbean vs. European: 0.65 [0.51–0.84]). Table 4 Stroke: Subhazard Ratios for First Post-Baseline Events (Nonfatal and Fatal) During 20.5 Years of Follow-Up in South Asians and African Caribbeans Compared With Europeans

¶ Age, sex, current smoking, diabetes, treated hypertension, apolipoprotein B-to-A1 ratio, waist-to-hip ratio (sex-specific tertiles), alcohol consumption, daily green vegetable/fruit consumption, tertiles of physical activity. Apolipoprotein data were available only for 1,147 Europeans, 688 South Asians, and 514 African Caribbeans (age- and sex-adjusted subhazard ratios in this subsample: South Asian vs. European: 1.77 [1.49–2.10], p < 0.001, African Caribbean vs. European: 0.65 [0.51–0.84]). Table 4 Stroke: Subhazard Ratios for First Post-Baseline Events (Nonfatal and Fatal) During 20.5 Years of Follow-Up in South Asians and African Caribbeans Compared With Europeans Europeans South Asians African Caribbeans Without diabetes at baseline Unadjusted Reference group 0.93 (0.71–1.20), p = 0.56 1.07 (0.76–1.52), p = 0.69 Adjusted for Age 1.13 (0.87–1.46), p = 0.36 1.08 (0.77–1.53), p = 0.65 Sex and age 1.12 (0.86–1.40), p = 0.38 1.12 (0.79–1.60), p = 0.52 With diabetes at baseline Unadjusted 1.72 (1.01–2.92), p = 0.047 2.03 (1.12–3.67), p = 0.020 Adjusted for Age 1.96 (1.15–3.33), p = 0.013 2.30 (1.25–4.22), p = 0.007 Age plus Sex 1.97 (1.16–3.35), p = 0.012 2.21 (1.14–4.30), p = 0.019 Smoking (never, former, current) 1.68 (0.94–2.98), p = 0.078 2.41 (1.22–4.78), p = 0.011 Baseline stroke 1.96 (1.15–3.33), p = 0.014 2.01 (1.09–3.70), p = 0.025 SBP/treated hypertension⁎ 1.96 (1.14–3.35), p = 0.014 2.01 (1.07–3.81), p = 0.031 Waist-to-hip ratio 2.04 (1.19–3.50), p = 0.009 2.32 (1.27–4.26), p = 0.006 Waist-to-thigh ratio 1.96 (1.15–3.35), p = 0.013 2.51 (1.23–5.12), p = 0.012 Body mass index 1.82 (1.07–3.08), p = 0.027 2.26 (1.23–4.15), p = 0.009 Total cholesterol 2.11 (1.22–3.65), p = 0.008 2.60 (1.37–4.96), p = 0.004 Triglycerides 1.94 (1.13–3.31), p = 0.016 2.53 (1.37–4.67), p = 0.003 HDL and LDL cholesterol 1.92 (1.06–3.45), p = 0.030 2.13 (1.08–4.19), p = 0.029 Fasting glucose† 1.94 (1.14–3.31), p = 0.093 2.29 (1.25–4.21), p = 0.009 HOMA2 IR† 1.98 (1.11–3.56), p = 0.022 2.45 (1.30–4.61), p = 0.006 HbA1c§ 1.71 (0.82–3.57), p = 0.150 Not available Alcohol consumption 1.94 (1.12–3.34), p = 0.018 2.19 (1.13–4.24), p = 0.021 Green vegetable/fruit consumption 1.91 (1.12–3.27), p = 0.017 2.21 (1.13–4.29), p = 0.020 Physical activity 2.01 (1.18–3.43), p = 0.010 2.34 (1.26–4.34), p = 0.007 Years of education 2.12 (1.24–3.65), p = 0.006 2.25 (1.22–4.13), p = 0.009 Manual occupation 1.86 (1.09–3.19), p = 0.023 2.24 (1.19–4.19), p = 0.012 Multivariable models (include covariates with largest effects on ethnic differences) Smoking, years of education 1.81 (1.01–32.4), p = 0.045 — Total cholesterol, baseline stroke, SBP/treated hypertension† — 2.32 (1.16–4.65), p = 0.017 Composite models Framingham variables‡ 2.05 (1.14–3.71), p = 0.017 2.04 (0.93–4.47), p = 0.074 INTERHEART variables∥ 1.95 (1.05–3.63), p = 0.034 1.62 (0.71–3.65), p = 0.25 D

king, years of education 1.81 (1.01–32.4), p = 0.045 — Total cholesterol, baseline stroke, SBP/treated hypertension† — 2.32 (1.16–4.65), p = 0.017 Composite models Framingham variables‡ 2.05 (1.14–3.71), p = 0.017 2.04 (0.93–4.47), p = 0.074 INTERHEART variables∥ 1.95 (1.05–3.63), p = 0.034 1.62 (0.71–3.65), p = 0.25 D ata are rates/1,000 person-years (95% confidence interval). Abbreviations as in Table 1. ⁎ Tertiles 1–3 (= 4 for those with treated hypertension). † Tertiles 1–3 (= 4 for those with diabetes). § HbA1c available only for 72 Europeans and 227 South Asians; age-adjusted subhazard ratio in this group with diabetes: 1.70 (0.82–3.51). ‡ Age, sex, (diabetes), current smoking, total and HDL cholesterol, SBP and treated hypertension (complete data available for multivariable adjustment for 97% Europeans, 96% South Asians, and 95% African Caribbeans). ∥ Age, sex, current smoking, (diabetes), treated hypertension, HDL and LDL cholesterol, waist-to-hip ratio (sex-specific tertiles), alcohol consumption, daily green vegetables/fruit, tertiles of physical activity (complete data available for multivariable adjustment for 93% Europeans, 90% South Asians, and 89% African Caribbeans).

brosis, as well as reduced gap junction signaling protein Cx43 in the RVOTs of those with BrS compared with tissue from victims of noncardiac death. Myocardial biopsies before epicardial ablation also display fibrosis at sites of delayed activation in patients with BrS. These data support the depolarization hypothesis. TRANSLATIONAL OUTLOOK: Future studies should address the roles of quantification of fibrosis and gap junction proteins in the diagnosis of and risk stratification for SCD among patients with known or suspected BrS and identify the predictors and determinants of these structural abnormalities. Appendix Online Methods Acknowledgment The authors thank W. Banya, Imperial College, London, for his statistical input.

There is growing evidence that obese adults without metabolic risk factor clustering (the so-called “healthy obese”) progress to unhealthy obesity over time (1). However, the pathophysiological changes underlying the long-term transition into an unhealthy obese state have not been well characterized. To inform clinical management of healthy obesity, we aimed to identify the metabolic risk factors responsible for this transition, as well as the timing of their onset. Repeat clinical data were drawn from the Whitehall II cohort study of British adults. We grouped participants as normal-weight (body mass index [BMI] 18.5 to 24.9 kg/m2), overweight (BMI 25 to 29 kg/m2), or obese (BMI ≥30 kg/m2), and as healthy (2) if they were free of any the following characteristics: high-density lipoprotein cholesterol <1.03 mmol/l (men) and <1.29 mmol/l (women); blood pressure ≥130/85 mm Hg or antihypertension medication use; fasting plasma glucose ≥5.6 mmol/l or diabetic medication use; triglycerides ≥1.7 mmol/l; and homeostatic model–assessed insulin resistance >2.83 (baseline 90th percentile value). Participants provided written informed consent. The University College London research ethics committee provided ethical approval.

tion use; fasting plasma glucose ≥5.6 mmol/l or diabetic medication use; triglycerides ≥1.7 mmol/l; and homeostatic model–assessed insulin resistance >2.83 (baseline 90th percentile value). Participants provided written informed consent. The University College London research ethics committee provided ethical approval. Cross-tabulations were used to describe incidence of each of these 5 metabolic risk factors at 5- (1997 to 1999), 10- (2002 to 2004), 15- (2007 to 2009), and 20-year (2012 to 2014) follow-ups for healthy normal-weight, overweight, or obese participants at baseline (1992 to 1994). Poisson regression models with robust error variances were used to estimate age-, sex-, and ethnicity-adjusted incidence ratios and 95% confidence intervals for having each metabolic risk factor at follow-up for healthy obese compared with healthy normal-weight adults at baseline. Among 2,878 adults with anthropometric and metabolic risk factor data at each time point, 1,120 adults (39 to 61 years of age; 68% male) were free of all metabolic risk factors at baseline. This initially healthy status was progressively rarer among those in higher BMI groups, representing 51.5%, 25.8%, and 13.4% of normal-weight, overweight, and obese adults, respectively. Of the healthy obese participants, 57.1% had at least 1 metabolic risk factor at the 5-year follow-up, with corresponding proportions being 64.3% at 10 years and 78.6% at 20 years. These proportions were smaller (32.8%, 46.7%, and 60.3%) among initially healthy normal-weight participants.

nd obese adults, respectively. Of the healthy obese participants, 57.1% had at least 1 metabolic risk factor at the 5-year follow-up, with corresponding proportions being 64.3% at 10 years and 78.6% at 20 years. These proportions were smaller (32.8%, 46.7%, and 60.3%) among initially healthy normal-weight participants. After 5 years (Figure 1), relative to initially healthy normal-weight adults, initially healthy obese adults were 4.40 times more likely to be insulin resistant, 3.35 times more likely to have high blood glucose, and 1.92 times more likely to be hypertensive (all p < 0.05). Incident insulin resistance remained higher over all subsequent follow-ups among baseline healthy obese compared with healthy normal-weight adults. Case numbers were small for 20-year incidence of low high-density lipoprotein cholesterol (2 cases) and high triglycerides (3 cases) among healthy obese adults, with little difference in risk compared with healthy normal-weight adults over time. Additional data are available on request. The risk of developing insulin resistance, high blood glucose, and hypertension was 2 to 5 times higher among initially healthy obese adults compared with their healthy normal-weight counterparts, and these changes were evident after only 5 years of follow-up. There was little difference in progression to dyslipidemia. However, the key factor explaining the long-term decline of healthy obesity was insulin resistance, which was consistently most common among healthy obese adults over time.

al-weight counterparts, and these changes were evident after only 5 years of follow-up. There was little difference in progression to dyslipidemia. However, the key factor explaining the long-term decline of healthy obesity was insulin resistance, which was consistently most common among healthy obese adults over time. Healthy obese adults are known to experience an elevated future risk of type 2 diabetes (3) and cardiovascular disease (4) compared with healthy normal-weight counterparts. That insulin resistance is an established indicator of future impaired glucose metabolism (5) may explain their much higher incidence of type 2 diabetes (relative risk near 4.0) (3) and slightly higher incidence of cardiovascular disease (relative risk near 1.2) (4), given that earlier onset of risk factors leads to a greater cumulated exposure and higher disease risk. Overall, our findings suggest that healthy obesity is strongly linked with future insulin resistance that subsequently induces cardiometabolic pathology.

idence of cardiovascular disease (relative risk near 1.2) (4), given that earlier onset of risk factors leads to a greater cumulated exposure and higher disease risk. Overall, our findings suggest that healthy obesity is strongly linked with future insulin resistance that subsequently induces cardiometabolic pathology. Please note: Mr. Bell is supported by an Economic and Social Research Council (ESRC) studentship. Dr. Hamer is supported by the British Heart Foundation (RE/10/005/28296). Dr. Singh-Manoux has received research support from the U.S. National Institutes of Health (NIH) National Institute of Aging (NIA) (R01AG013196; R01AG034454). Dr. Sabia has received research support from the NIH NIA (R01AG034454) and ESRC (ES/J023299/1). Dr. Kivimäki has received research support from the Medical Research Council (MR/K013351/1); the National Heart, Lung, and Blood Institute (R01HL36310); the NIA (R01AG034454); NordForsk (75021); and an ESRC professorial fellowship (ES/J023299/1). The funders had no role in the study design; in the collection, analysis and interpretation of data; in writing of the report; or in the decision to submit the paper for publication. The developers and funders of Whitehall II do not bear any responsibility for the analyses or interpretations presented here. The authors declare that there is no duality of interest associated with this manuscript. Dr. Batty has reported that he has no relationships relevant to the contents of this paper to disclose. The authors thank all of the participating civil service departments and their welfare, personnel, and establishment officers; the British Occupational Health and Safety Agency; the British Council of Civil Service Unions; all participating civil servants in the Whitehall II study; and all members of the Whitehall II study team. The Whitehall II Study team comprises research scientists, statisticians, study coordinators, nurses, data managers, administrative assistants and data entry staff, who make the study possible. Whitehall II data, protocols, and other metadata are available to bona fide researchers for research purposes. Please refer to the Whitehall II data sharing policy.

research scientists, statisticians, study coordinators, nurses, data managers, administrative assistants and data entry staff, who make the study possible. Whitehall II data, protocols, and other metadata are available to bona fide researchers for research purposes. Please refer to the Whitehall II data sharing policy. Figure 1 Incidence of Metabolic Risk Factors Among Initially Healthy Obese Compared With Initially Healthy Normal-Weight Adults Over 20 Years (n = 1,120) Results are incidence ratios and 95% confidence intervals (CI) for having each metabolic risk factor at follow-up, on the basis of Poisson regression models with robust error variances. Models are adjusted for age, sex, and ethnicity. Little difference in high-density lipoprotein cholesterol or triglycerides was observed between groups. Baseline healthy status is defined as having none of 5 metabolic risk factors (hypertension, low high-density lipoprotein cholesterol, high triglycerides, insulin resistance, and high blood glucose).

Brugada syndrome (BrS) is an inherited arrhythmia syndrome diagnosed by the presence of the type 1 Brugada electrocardiogram (ECG) (1). It was initially described in survivors of cardiac arrest without structural disease (2), and it is partly responsible for sudden arrhythmic death syndrome (SADS) 1, 3, 4. Potential causal variants in the cardiac sodium channel gene SCN5A are identified in 20% of cases (5). It was initially proposed that the basis for BrS was an abnormal transmural repolarization in the right ventricular outflow tract (RVOT) due to heterogeneous loss of the cardiomyocyte action potential dome in the epicardium (6). However, electrophysiological, imaging, and histopathological studies have identified subtle structural abnormalities in patients with BrS 7, 8, 9. Myocardial fibrosis has been suggested by abnormal, low-voltage, fractionated electrograms localized to the RVOT at the epicardium 9, 10. Ablation at these sites has eliminated the type 1 Brugada ECG pattern and successfully reduced arrhythmic events (10), as was seen in a previous experimental model (11). A study of sudden cardiac death (SCD) cases associated the type 1 ECG with arrhythmogenic right ventricular cardiomyopathy (8). Furthermore, SCD cases with a familial diagnosis of BrS showed structural abnormalities that were insufficient to fulfill the diagnostic criteria for cardiomyopathy or myocarditis (12). Other myocardial anomalies have been reported in selected cases 13, 14. Therefore, there is significant debate about the underlying substrate in BrS (15).

es with a familial diagnosis of BrS showed structural abnormalities that were insufficient to fulfill the diagnostic criteria for cardiomyopathy or myocarditis (12). Other myocardial anomalies have been reported in selected cases 13, 14. Therefore, there is significant debate about the underlying substrate in BrS (15). To resolve this controversy, we tested the hypothesis that BrS is associated with fibrosis in the RVOT and altered expression of the gap junction protein connexin-43 (Cx43), which may be critical for correct cellular migration and maintenance of RVOT zonation 16, 17. We expected this to manifest as abnormal late and fractionated potentials at the RVOT epicardium. Methods Study setting and cohorts Post-mortem BrS cohort From 2005 to 2010, 1,304 unexpected SCD cases were referred for specialist cardiac autopsy. We studied 6 male cases (B1 to B6; mean age 23.2 years) (Table 1), which fulfilled the following criteria for SADS (1): 1) age 1 to 64 years; 2) unexpected sudden death; 3) whole heart available; 4) heart morphologically normal at coronial/medical examiner and specialist cardiac autopsies; 5) no antemortem cardiac conditions; and 6) negative toxicological analysis. In addition, 1 or more first-degree blood relatives had to be diagnosed with BrS (Online Methods) following familial evaluation 1, 18, 19. All 6 cases were asymptomatic before death, according to primary care records and family interview, with no family history of premature death. Five died at rest (4 during sleep) and 1 during exertion. None had undergone previous cardiac investigation.

Methods Study setting and cohorts Post-mortem BrS cohort From 2005 to 2010, 1,304 unexpected SCD cases were referred for specialist cardiac autopsy. We studied 6 male cases (B1 to B6; mean age 23.2 years) (Table 1), which fulfilled the following criteria for SADS (1): 1) age 1 to 64 years; 2) unexpected sudden death; 3) whole heart available; 4) heart morphologically normal at coronial/medical examiner and specialist cardiac autopsies; 5) no antemortem cardiac conditions; and 6) negative toxicological analysis. In addition, 1 or more first-degree blood relatives had to be diagnosed with BrS (Online Methods) following familial evaluation 1, 18, 19. All 6 cases were asymptomatic before death, according to primary care records and family interview, with no family history of premature death. Five died at rest (4 during sleep) and 1 during exertion. None had undergone previous cardiac investigation. Post-mortem control cohort Six control cases (C1 to C6) (Table 1) of premature noncardiac death were identified from 407 consecutive homograft valve donors from Harefield Hospital, London (2010 to 2012). These were matched to the post-mortem BrS cases by random risk set sampling selection for age (within 3 years) and sex in a 1:1 ratio. Inclusion criteria for control cases were: 1) age 1 to 64 years; 2) absence of antemortem cardiac symptoms (syncope or seizures); 3) normal specialist cardiac autopsy; and 4) intact RVOT.

ese were matched to the post-mortem BrS cases by random risk set sampling selection for age (within 3 years) and sex in a 1:1 ratio. Inclusion criteria for control cases were: 1) age 1 to 64 years; 2) absence of antemortem cardiac symptoms (syncope or seizures); 3) normal specialist cardiac autopsy; and 4) intact RVOT. In vivo BrS ablation cohort Six symptomatic male BrS patients (mean age 39.8 years) (Table 1) undergoing mapping and RVOT ablation during open thoracotomy were studied at Bhumibol Adulyadej Air Force Hospital (cases V1 to V5, Bangkok) and Yokohama Rosai Hospital (case V6, Japan). All had an implantable cardioverter defibrillator (ICD) before recruitment, with a clinical diagnosis (Online Methods) of BrS 1, 19, and normal echocardiography, computed tomography/magnetic resonance imaging, and coronary angiography. Thoracotomy was indicated for ICD lead extraction (V1, V2, V5, and V6) or to permit epicardial access for ablation after a failed percutaneous attempt (V3 and V4). Mutation analysis In vivo BrS subjects and clinically affected blood relatives of post-mortem cases were counseled and offered SCN5A mutation analysis. Mutation analysis was not undertaken in the autopsy cases due to lack of suitable unfixed material.

In vivo BrS ablation cohort Six symptomatic male BrS patients (mean age 39.8 years) (Table 1) undergoing mapping and RVOT ablation during open thoracotomy were studied at Bhumibol Adulyadej Air Force Hospital (cases V1 to V5, Bangkok) and Yokohama Rosai Hospital (case V6, Japan). All had an implantable cardioverter defibrillator (ICD) before recruitment, with a clinical diagnosis (Online Methods) of BrS 1, 19, and normal echocardiography, computed tomography/magnetic resonance imaging, and coronary angiography. Thoracotomy was indicated for ICD lead extraction (V1, V2, V5, and V6) or to permit epicardial access for ablation after a failed percutaneous attempt (V3 and V4). Mutation analysis In vivo BrS subjects and clinically affected blood relatives of post-mortem cases were counseled and offered SCN5A mutation analysis. Mutation analysis was not undertaken in the autopsy cases due to lack of suitable unfixed material. Specialist cardiac post-mortem examination A systematic specialist post-mortem of the whole heart was undertaken, with macroscopic and microscopic evaluation in all referred SCD cases and control hearts, blinded to the results of familial evaluation (20). At least 20 tissue sections were sampled from each case, including the following: coronary arteries; ascending aorta; 4 sequential sections from the atrioventricular node to the branches of the His-Purkinje system; 4 sinoatrial node sections; and 2 RVOT sections. Sectioning of the anterior, lateral, and posterior left ventricle (LV), anterior and posterior interventricular septum, and right ventricle (RV) was performed at the midventricular level. Histological examination (Online Methods) was performed with hematoxylin and eosin and elastic Van Gieson stains.

d 2 RVOT sections. Sectioning of the anterior, lateral, and posterior left ventricle (LV), anterior and posterior interventricular septum, and right ventricle (RV) was performed at the midventricular level. Histological examination (Online Methods) was performed with hematoxylin and eosin and elastic Van Gieson stains. Detailed post-mortem RVOT examination Up to 14 parallel longitudinal sections of 3-mm thickness were taken from the RVOT in each post-mortem subject to ensure complete examination of this region. Morphometric analysis for post-mortem myocardial collagen/fibrosis All post-mortem RVOT sections were stained with the picrosirius red (PSR) technique, with RV free wall and LV tissue for comparison. These sections (n = 267, total area quantified 6,505 mm2) were digitized (Scanscope CS, Aperio, California) at 20× magnification in 24-bit color. Computational semiautomated morphometric analysis was performed on 5× magnification images of transmural tissue sections on the basis of green color depth thresholds (ImageJ, National Institutes of Health, Bethesda, Maryland), with blinding to the diagnosis and cardiac wall. Epicardial, mid-myocardial, and endocardial zones and fat cells were defined by consensus (Figure 1A). Regions of collagen and fat were defined by color threshold, with proportions calculated by cardiac wall and tissue zone relative to tissue area.

, Bethesda, Maryland), with blinding to the diagnosis and cardiac wall. Epicardial, mid-myocardial, and endocardial zones and fat cells were defined by consensus (Figure 1A). Regions of collagen and fat were defined by color threshold, with proportions calculated by cardiac wall and tissue zone relative to tissue area. Confocal microscopy analysis of post-mortem Cx43 distribution An RVOT section from each post-mortem case underwent Cx43 immunofluorescent staining (Online Methods) to evaluate gap junction remodeling. Three transmural tissue strips of 450 μm width with intact myocardium per case were identified using 4′,6-diamidino-2-phenylindole immunofluorescence, blinded to the Cx43 signal. A Zeiss LSM-780 (Carl Zeiss Ltd., Cambridge, United Kingdom) inverted confocal microscope (20×, 0.8 numerical aperture objective lens) with sequential channel scanning (Alexa Fluor 488, 4′,6-diamidino-2-phenylindole, and cyanine Cy3 fluorescence) in a single optical plane was used. Cx43 was defined by color threshold (ImageJ). Perinuclear lipofuscin was excluded. Morphometric analysis of Cx43 was performed as for collagen. Serial sections immediately adjacent to the Cx43-stained strip were imaged with PSR to permit correction for collagen content (Figure 1B) by dividing by the proportion representing the noncollagenous component. Adjusted and unadjusted Cx43 proportions were aggregated per subject.

of Cx43 was performed as for collagen. Serial sections immediately adjacent to the Cx43-stained strip were imaged with PSR to permit correction for collagen content (Figure 1B) by dividing by the proportion representing the noncollagenous component. Adjusted and unadjusted Cx43 proportions were aggregated per subject. In vivo open thoracotomy mapping and ablation of RVOT Cases V1 to V4 underwent mini-lateral thoracotomy to expose the anterior RVOT, whereas cases V5 and V6 had a midline thoracotomy. For cases V1 to V5, epicardial mapping was performed with a 3.5-mm-tip ThermoCool catheter (Biosense Webster, Diamond Bar, California) limited to the anterior RVOT (Figure 2). Radiofrequency ablations with 20- to 45-W energy were performed off pump at substrate sites identified by abnormal late and fractionated electrograms. For case V6, electroanatomical mapping was performed with the CARTO 3 System (Biosense Webster) intraoperatively, with manual confirmation of abnormal electrogram amplitudes. Cryoablation was then performed at sites of abnormal late potentials following total cardiopulmonary bypass with aorta-bicaval cannulation. The ablation endpoint for all cases was elimination of abnormal late and fractionated electrograms in the RVOT epicardium.

h manual confirmation of abnormal electrogram amplitudes. Cryoablation was then performed at sites of abnormal late potentials following total cardiopulmonary bypass with aorta-bicaval cannulation. The ablation endpoint for all cases was elimination of abnormal late and fractionated electrograms in the RVOT epicardium. Biopsy of in vivo substrate sites in the RVOT All sites identified with abnormal electrograms were biopsied under direct vision: off-pump sampling (cases V1 to V5) was limited to small samples of epicardial surface and myocardial tissue to minimize complications; transmural biopsies were taken during heart–lung bypass in case V6. Biopsy tissue was stained with PSR. Clinical endpoints In vivo BrS subjects were reviewed 1 month post-ablation and every 3 months thereafter with ICD interrogation and ECG. Ajmaline provocation was performed at 6 months for patients recruited from Bangkok. Research governance The following institutional review boards approved the study: London Stanmore Research Ethics Committee; Bhumibol Adulyadej Air Force Hospital; and Yokohama Rosai Hospital. Informed consent was obtained from subjects and/or next of kin. Statistical analysis Analysis was undertaken using Stata v12.1 (StataCorp LP, College Station, Texas). Natural log transformation corrected skew in measured tissue proportions of fibrosis and fat before analysis by simple and multiple regression (using independent factors for disease status, myocardial wall, and myocardial region) with robust variances; analyses are reported as odds ratios (OR). A p value ≤0.05 was considered significant.

rmation corrected skew in measured tissue proportions of fibrosis and fat before analysis by simple and multiple regression (using independent factors for disease status, myocardial wall, and myocardial region) with robust variances; analyses are reported as odds ratios (OR). A p value ≤0.05 was considered significant. Results Post-mortem diagnosis of Brugada syndrome on familial cardiac evaluation A mean of 3.7 first-degree blood relatives per post-mortem BrS case underwent familial evaluation, with 1.7 diagnosed with BrS on average. One relative of B4 was diagnosed with BrS on the basis of a spontaneous type 1 Brugada ECG pattern, with other relatives identified following ajmaline provocation (Figure 3). No relatives had evidence of structural or functional myocardial disease on cardiac imaging. Genetic mutation analysis Five of the 6 families of post-mortem BrS cases consented to genetic analysis; 2 affected relatives of B4 were found to carry the p.Leu1462Gln mutation in SCN5A. Poor quality of extracted DNA prevented confirmation in B4. All in vivo cases underwent genetic testing and 2 SCN5A mutation carriers were identified (case V4 p.Ser528Cys and case V6 p.Leu846Arg). Collagen staining and myocardial architecture of the RVOT Myocardial collagen in the control group was seen in the epicardial surface and around blood vessels. Linear collagen was distributed parallel to myocytes, but did not surround the individual myocytes (Figure 4A2 and 4A3). This collagen distribution pattern is normal in the RV.

g and myocardial architecture of the RVOT Myocardial collagen in the control group was seen in the epicardial surface and around blood vessels. Linear collagen was distributed parallel to myocytes, but did not surround the individual myocytes (Figure 4A2 and 4A3). This collagen distribution pattern is normal in the RV. In the post-mortem BrS group, there was an appearance of increased epicardial surface collagen that was thicker than that in control hearts, indicating epicardial fibrosis (Figure 4B1). There was infiltration of the epicardial surface fibrosis into the underlying epicardial myocardium, with individual myocytes surrounded by collagen, which was considered interstitial myocardial fibrosis (Figure 4B2). There was also evidence of replacement of myocytes by collagen, focal replacement fibrosis, admixed with fat in the epicardial myocardium (Figure 4B3). The in vivo tissue samples taken in the regions of late potentials showed similar epicardial and myocardial fibrosis patterns (Figure 4C1 to 4C3). The epicardial fibrosis appeared to be separated from the underlying myocardium by fat in some sections, whereas in others, it infiltrated directly into the underlying myocardium. Morphometric analysis of post-mortem collagen by PSR The BrS cohort had greater collagen content than control hearts, with maximal differences seen in the RVOT epicardium (13.9% vs. 10.5%; p = 0.024) (Figure 5A). Multivariable analysis (Table 2) identified that the diagnosis of BrS was associated with an OR of 1.42 (p = 0.026) for collagen proportion, regardless of the cardiac chamber.

had greater collagen content than control hearts, with maximal differences seen in the RVOT epicardium (13.9% vs. 10.5%; p = 0.024) (Figure 5A). Multivariable analysis (Table 2) identified that the diagnosis of BrS was associated with an OR of 1.42 (p = 0.026) for collagen proportion, regardless of the cardiac chamber. Control hearts and cases also showed similar patterns of collagen distribution, but this was greater in cases. The RVOT (OR: 1.98; p = 0.003) and RV (OR: 1.66; p = 0.020) walls had higher collagen content in comparison with the LV, irrespective of diagnosis. Similarly, a gradient of decreasing collagen content was seen from the epicardial to endocardial zones (OR: 2.00; p = 0.001) in all chambers. Morphometric analysis of post-mortem fat cells Regression analysis for the proportion of fat content in the myocardium showed no significant difference between BrS and control hearts (p = 0.133). Post-mortem Cx43 signal distribution and quantification In control myocardial tissue, Cx43 localized to the intercalated disc (Figure 4A4 and 4A5). BrS cases showed a reduced Cx43 signal and a decreased punctate pattern in the intercalated disc (Figure 4B4 and 4B5). BrS cases had reduced Cx43 signal in the RVOT compared with control hearts (OR: 0.59; p = 0.001) (Figure 5B, Table 3), even following correction for collagen content (OR: 0.58; p = 0.036). No significant difference was observed between myocardial zones of the RVOT (p = 0.476).

Post-mortem Cx43 signal distribution and quantification In control myocardial tissue, Cx43 localized to the intercalated disc (Figure 4A4 and 4A5). BrS cases showed a reduced Cx43 signal and a decreased punctate pattern in the intercalated disc (Figure 4B4 and 4B5). BrS cases had reduced Cx43 signal in the RVOT compared with control hearts (OR: 0.59; p = 0.001) (Figure 5B, Table 3), even following correction for collagen content (OR: 0.58; p = 0.036). No significant difference was observed between myocardial zones of the RVOT (p = 0.476). Clinical outcomes The mean radiofrequency ablation time was 14 ± 6 min per in vivo BrS case; no surgical complications occurred. In the 5 patients who underwent radiofrequency ablation, fractionated electrograms disappeared immediately, with a drastic reduction of ventricular electrograms after radiofrequency was turned off. The ECG pattern normalized (i.e., reversion from type 1 Brugada ECG pattern) within a week in all cases, and a negative ajmaline test was seen in those who underwent subsequent provocation 3 months later (n = 5 of 6). No further ventricular tachycardia (VT) or ventricular fibrillation (VF) episodes were seen during the follow-up period (mean 24.6 ± 9.7 months, median 25 months), and quinidine therapy was not required.

es, and a negative ajmaline test was seen in those who underwent subsequent provocation 3 months later (n = 5 of 6). No further ventricular tachycardia (VT) or ventricular fibrillation (VF) episodes were seen during the follow-up period (mean 24.6 ± 9.7 months, median 25 months), and quinidine therapy was not required. Discussion This study systematically describes increased collagen content in the RVOT that shows epicardial surface and intramyocardial fibrosis, as well as diminished gap junction protein expression. In vivo human evidence of conduction delay in the RVOT was associated with similar patterns of fibrosis, corroborating the post-mortem findings (Central Illustration). Ablation at these sites eliminated the type 1 ECG pattern with successful suppression of VT/VF recurrence, giving support to the hypothesis that conduction delay is responsible for the BrS phenotype.

he RVOT was associated with similar patterns of fibrosis, corroborating the post-mortem findings (Central Illustration). Ablation at these sites eliminated the type 1 ECG pattern with successful suppression of VT/VF recurrence, giving support to the hypothesis that conduction delay is responsible for the BrS phenotype. Myocardial fibrosis Despite the a priori exclusion at expert autopsy of overt structural abnormalities in SADS cases, the diagnosis of BrS was associated with increased collagen content in all ventricular walls. This was over and above the normal collagen seen in age- and sex-matched control hearts. In addition, the in vivo cases all had normal cardiac imaging, including computed tomography/magnetic resonance imaging, as well as macroscopically normal hearts on direct visualization during thoracotomy. These cases, therefore, represent minimally structurally perturbed candidates for the diagnosis of BrS, yet they showed distinctive patterns of fibrosis. This reveals the limitations of current imaging technology for detecting subtle changes in the myocardium that can still give rise to physiologically detectable changes.

se cases, therefore, represent minimally structurally perturbed candidates for the diagnosis of BrS, yet they showed distinctive patterns of fibrosis. This reveals the limitations of current imaging technology for detecting subtle changes in the myocardium that can still give rise to physiologically detectable changes. We have identified previously that one-third of unexplained SCDs with idiopathic fibrosis and/or hypertrophy had familial diagnoses of BrS (12). LV and RV free-wall, age-related fibrosis has also been seen in mouse models of BrS 21, 22. In addition, we identified epicardial and intramyocardial fibrosis at the site of epicardial late potentials in the RVOT of BrS patients. A detailed study of a single patient with BrS who underwent transplantation has previously colocalized interstitial fibrosis with conduction delay (14). Moreover, murine models of BrS, including epicardial electrophysiological study of Langendorff perfused hearts, have shown RVOT pathology: increased collagen; delayed conduction; and a propensity for ventricular arrhythmia with programmed stimulation in the RVOT (23). It is therefore plausible that BrS may reflect a generalized disease of myocardial architecture, with baseline properties of the RVOT predisposing it to fibrosis, which is likely to underlie the condition and arrhythmic risk (24). Interestingly, although fibrosis and conduction delay have been identified in carriers of SCN5A mutations (25), all cases demonstrated some evidence of fibrosis, whether they harbored an SCN5A mutation or not. The reported increase in profibrotic markers secondary to sodium channel inactivation, independent of messenger ribonucleic acid expression, suggests that fibrosis may be a feature irrespective of mutation status (26).

5), all cases demonstrated some evidence of fibrosis, whether they harbored an SCN5A mutation or not. The reported increase in profibrotic markers secondary to sodium channel inactivation, independent of messenger ribonucleic acid expression, suggests that fibrosis may be a feature irrespective of mutation status (26). Fat infiltration of myocardium No significant difference in fat content was observed between BrS cases and control hearts. In contrast, transmural fat infiltration in the absence of fibrosis predominated in Italian post-mortem cases with the Brugada ECG pattern (8). This difference may reflect the inclusion of patients with overt antemortem and post-mortem features of arrhythmogenic right ventricular cardiomyopathy in the Italian study without suitable age- and sex-matched controls. Significance of Cx43 The Cx43 signal was diminished in BrS compared with the control myocardium. This raises the possibility that changes at the intercalated disc that affect Cx43 expression may cause cardiomyocyte electrical uncoupling, and therefore, may be important in the pathogenesis of BrS. Royer et al. (21) describe diminished Cx43 expression in the scn5a-knockout mouse model’s myocardium, which is a clear correlation with the human phenotype.

ges at the intercalated disc that affect Cx43 expression may cause cardiomyocyte electrical uncoupling, and therefore, may be important in the pathogenesis of BrS. Royer et al. (21) describe diminished Cx43 expression in the scn5a-knockout mouse model’s myocardium, which is a clear correlation with the human phenotype. Open thoracotomy catheter ablation As previously reported (10), abolition of the type 1 ECG and suppression of VT/VF episodes in a high-risk BrS patient cohort were seen following epicardial ablation at sites of late potentials in the RVOT. To our knowledge, this study reports, for the first time, a surgical approach with either midline or mini-lateral thoracotomy to access the epicardial surface of the RVOT for ablation. Depolarization versus repolarization Our findings reinforce other human studies that have identified conduction delay in the RVOT in BrS in vivo 10, 27, 28, 29, 30. Two of these studies used noncontact intracardiac mapping or noninvasive ECG imaging and proposed additional repolarization abnormalities 27, 30. We have correlated directly acquired delayed, prolonged, and fragmented epicardial electrograms and histopathological evidence for fibrosis that support depolarization delay as the primary substrate.

noncontact intracardiac mapping or noninvasive ECG imaging and proposed additional repolarization abnormalities 27, 30. We have correlated directly acquired delayed, prolonged, and fragmented epicardial electrograms and histopathological evidence for fibrosis that support depolarization delay as the primary substrate. Study limitations Subject recruitment was limited by the rarity of thoracotomy in BrS patients and the availability of whole hearts post-mortem in which families were diagnosed with BrS. Thus, our cohorts represent a unique collection. Both control and case hearts went through similar processing after death, with an approximate 24- to 48-h delay before fixation and an intervening period of refrigeration. We were unable to establish more accurate timing.

mortem in which families were diagnosed with BrS. Thus, our cohorts represent a unique collection. Both control and case hearts went through similar processing after death, with an approximate 24- to 48-h delay before fixation and an intervening period of refrigeration. We were unable to establish more accurate timing. The etiology of death in the 6 BrS post-mortem cases was established by identifying BrS in blood relatives in the absence of alternative explanations. This methodology forms the basis of internationally accepted guidelines for the diagnosis of genetic disorders in unexplained SCD and BrS 1, 3. However, we recognize that without previous ECG evidence, we cannot be absolutely certain of the diagnosis. Nonetheless, it is a reasonable assumption, as the deceased young person does, at a minimum, have a 50% chance of having the same diagnosis. The chance of any other diagnosis is much smaller. In addition, the finding of 1 SCN5A mutation in the 5 families tested is consistent with the established prevalence of 20% in BrS (5). Retrospective investigation by molecular autopsy was not possible in our cases, although the absence of a mutation would not exclude BrS due to the low molecular genetic yield (5). Our study only included symptomatic BrS cases. Thus, our observations may reflect a biased population of high-risk subjects. However, myocardial fibrosis has also been identified in low-risk living patients on magnetic resonance imaging 31, 32 and histopathology (33).

The etiology of death in the 6 BrS post-mortem cases was established by identifying BrS in blood relatives in the absence of alternative explanations. This methodology forms the basis of internationally accepted guidelines for the diagnosis of genetic disorders in unexplained SCD and BrS 1, 3. However, we recognize that without previous ECG evidence, we cannot be absolutely certain of the diagnosis. Nonetheless, it is a reasonable assumption, as the deceased young person does, at a minimum, have a 50% chance of having the same diagnosis. The chance of any other diagnosis is much smaller. In addition, the finding of 1 SCN5A mutation in the 5 families tested is consistent with the established prevalence of 20% in BrS (5). Retrospective investigation by molecular autopsy was not possible in our cases, although the absence of a mutation would not exclude BrS due to the low molecular genetic yield (5). Our study only included symptomatic BrS cases. Thus, our observations may reflect a biased population of high-risk subjects. However, myocardial fibrosis has also been identified in low-risk living patients on magnetic resonance imaging 31, 32 and histopathology (33). Conclusions BrS, in the absence of overt structural or functional abnormalities, is unequivocally associated with increased collagen, fibrosis, and reduced gap junction expression in the RVOT. Myocardial late potentials indicative of the arrhythmic substrate anatomically collocate with fibrosis in the RVOT of BrS subjects. Therefore, it is plausible that BrS represents a disease of myocardial architecture and cardiomyocyte electrical coupling in the RVOT. The reduction in arrhythmic burden and reversal of electrocardiographic signature of BrS following ablation at these sites supports our hypothesis that these myocardial changes result in discontinuity of cardiac conduction responsible for arrhythmogenesis. These data are the strongest yet to support the depolarization theory of the pathogenesis of BrS 29, 34.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: Two main theories have been proposed to explain the pathophysiology of BrS: 1) that it is due to either dispersed repolarization; or 2) to abnormal depolarization due to conduction delay. Tissue from cases of SCD due to BrS without evident structural disease exhibits increased collagen throughout the heart and fibrosis, as well as reduced gap junction signaling protein Cx43 in the RVOTs of those with BrS compared with tissue from victims of noncardiac death. Myocardial biopsies before epicardial ablation also display fibrosis at sites of delayed activation in patients with BrS. These data support the depolarization hypothesis.

TRANSLATIONAL OUTLOOK: Future studies should address the roles of quantification of fibrosis and gap junction proteins in the diagnosis of and risk stratification for SCD among patients with known or suspected BrS and identify the predictors and determinants of these structural abnormalities. Appendix Online Methods Acknowledgment The authors thank W. Banya, Imperial College, London, for his statistical input. This project was funded in part by Cardiac Risk in the Young and by an unrestricted grant from Biotronik. Dr. Raju was supported by the British Heart Foundation. Fellowship FS/11/71/28918. Dr. Behr was supported by the Higher Education Funding Council for England. Drs. Wilde and Tan were supported by the Netherlands CardioVascular Research Initiative (Dutch Heart Foundation, Dutch Federation of University Medical Centers, Netherlands Organisation for Health Research and Development, and the Royal Netherlands Academy of Sciences). Dr. Tan was supported by the grant ZonMW VICI 918·86·616. Dr. Nademanee was supported by the Adventist Health Care at White Memorial Medical Center, and the Vejdusit and Duangtawan Foundation of Thailand, Bangkok Medical Center and Bumrungrad Hospital. Dr. Nademanee is a consultant for and has received research grants and royalties from Biosense Webster. Dr. Wilde is a consultant for and a member of the scientific advisory board for Sorin. Dr. Nogami has consulting agreements and has received research grants and royalties from Biosense Webster; has received speaker honoraria from St. Jude Medical and Boston Scientific; and has received research grants from Medtronic and Johnson & Johnson. Dr. Behr has received unrestricted research funds from Biotronik and St. Jude Medical. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Nademanee, Raju, and de Noronha contributed equally to this work. Drs. Nogami, Sheppard, Veerakul, and Behr are joint senior authors.

unrestricted research funds from Biotronik and St. Jude Medical. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Nademanee, Raju, and de Noronha contributed equally to this work. Drs. Nogami, Sheppard, Veerakul, and Behr are joint senior authors. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section, please see the online version of this article. Central Illustration Pathophysiology of Brugada Syndrome: Conduction Delay Due to Fibrosis and Connexin-43 Abnormalities Conduction delay in the right ventricular outflow tract (RVOT) is caused by myocyte electrical uncoupling due to a reduction in connexin-43 at endplates and subtle interstitial and replacement fibrosis. As a result, epicardial electrograms are abnormal, slowed, and fragmented. This provides the substrate for the Brugada type 1 electrocardiographic (ECG) pattern, re-entry, and the generation of polymorphic ventricular tachycardia (VT) and ventricular fibrillation. Figure 1 Morphometric Analysis of Histological Stained Sections

Conduction delay in the right ventricular outflow tract (RVOT) is caused by myocyte electrical uncoupling due to a reduction in connexin-43 at endplates and subtle interstitial and replacement fibrosis. As a result, epicardial electrograms are abnormal, slowed, and fragmented. This provides the substrate for the Brugada type 1 electrocardiographic (ECG) pattern, re-entry, and the generation of polymorphic ventricular tachycardia (VT) and ventricular fibrillation. Figure 1 Morphometric Analysis of Histological Stained Sections (A) Morphometric analysis of a single tissue section. (A1) Visually defined tissue zones (yellow polygons) defining the epicardium, mid-myocardium, and endocardium. (A2) Collagen (black). (A3) Fat cells (black). (B) Representative serial myocardial strips from the post-mortem control group for collagen correction for Cx43 morphometric analysis. (B1) Myocardial strip of Cx43 expression. (B2) Serial section aligned to B1, stained with PSR. (B3) A threshold drawing generated from the PSR-stained myocardium image (as created by ImageJ). Scale bar = 200 μm; Cx43 = connexin-43; PSR = picrosirius red. Figure 2 Computed Tomography Scan, Epicardial Electrograms, and Histology of RVOT of In Vivo BrS Patient

(A) Morphometric analysis of a single tissue section. (A1) Visually defined tissue zones (yellow polygons) defining the epicardium, mid-myocardium, and endocardium. (A2) Collagen (black). (A3) Fat cells (black). (B) Representative serial myocardial strips from the post-mortem control group for collagen correction for Cx43 morphometric analysis. (B1) Myocardial strip of Cx43 expression. (B2) Serial section aligned to B1, stained with PSR. (B3) A threshold drawing generated from the PSR-stained myocardium image (as created by ImageJ). Scale bar = 200 μm; Cx43 = connexin-43; PSR = picrosirius red. Figure 2 Computed Tomography Scan, Epicardial Electrograms, and Histology of RVOT of In Vivo BrS Patient Computed tomography scan of the heart (center) of in vivo BrS patient V2 showing an anatomical grid over the anterior RVOT. ECG lead II and a distal bipolar (0.4 mV/cm voltage scale at 30- to 300-Hz filter settings) and unipolar (5 mV/cm voltage scale at 0.05- to 300-Hz filter settings) electrogram at labeled sites are given in surrounding panels, with pacing stimuli indicated by red arrowheads. Abnormal fractionated electrograms are on the (A to C) left and normal electrograms on the (D to E) right. (F) Epicardial biopsy and histology (PSR) at the site of the abnormal electrogram shows epicardial fibrosis with focal finger-like projections of collagen into myocardium. ABL d = distal bipolar ablation catheter electrogram; ABL uni = unipolar ablation catheter electrogram; BrS = Brugada syndrome; RVOT = right ventricular outflow tract; other abbreviations as in Figure 1.

f the abnormal electrogram shows epicardial fibrosis with focal finger-like projections of collagen into myocardium. ABL d = distal bipolar ablation catheter electrogram; ABL uni = unipolar ablation catheter electrogram; BrS = Brugada syndrome; RVOT = right ventricular outflow tract; other abbreviations as in Figure 1. Figure 3 Right Precordial ECG Traces From Blood Relatives of Post-Mortem BrS Cases During Ajmaline Provocation ECG traces acquired following cranial displacement of electrode positions V1 and V2 into 2ics. 2ics = second intercostal space; other abbreviations as in Figures 1 and 2. Figure 4 RVOT Histological Sections Stained for Collagen and Immunoconfocal Images of Cx43 Expression

Figure 3 Right Precordial ECG Traces From Blood Relatives of Post-Mortem BrS Cases During Ajmaline Provocation ECG traces acquired following cranial displacement of electrode positions V1 and V2 into 2ics. 2ics = second intercostal space; other abbreviations as in Figures 1 and 2. Figure 4 RVOT Histological Sections Stained for Collagen and Immunoconfocal Images of Cx43 Expression RVOT histological sections stained for collagen (purple-red) with PSR and immunoconfocal images of gap junction protein Cx43 expression (green fluorescence). Sections from (A) post-mortem control, (B) post-mortem BrS cases, and (C) in vivo BrS patients. (A) Post-mortem control. PSR: (A1) normal epicardial collagen thickness, with (A2) linear collagen between myocytes and (A3) around blood vessels, but no evidence of complete circumscription of myocytes by collagen. Cx43: (A4) normal appearance of gap junction signal concentrated to form transverse stripes, with an organized parallel orientation. (A5) Clusters of gap junctions in a typical ring-like formation at the intercalated disc, with large gap junctions circumscribing the periphery of the disc and smaller junctions in the inner region. (B) Post-mortem BrS. PSR: (B1) thickened epicardial collagen layer, with (B2) evidence of interstitial fibrosis, identified by collagen circumscribing myocytes, and (B3) replacement fibrosis, identified by replacement of myocytes by collagen in a region of infiltration by fat. Cx43: (B4) notable dispersion of the signal along the axis of the cell and (B5) sparse junctional plaque with an ill-defined border. (C) In vivo BrS. PSR: (C1) thickened epicardial collagen layer with (C2) evidence of interstitial fibrosis, identified by collagen circumscribing myocytes, and (C3) replacement fibrosis, identified by replacement of myocytes by collagen. Abbreviations as in Figures 1 and 2.

junctional plaque with an ill-defined border. (C) In vivo BrS. PSR: (C1) thickened epicardial collagen layer with (C2) evidence of interstitial fibrosis, identified by collagen circumscribing myocytes, and (C3) replacement fibrosis, identified by replacement of myocytes by collagen. Abbreviations as in Figures 1 and 2. Figure 5 Scatterplot of Collagen and Cx43 Quantification in the Epicardial Myocardium of the Right Ventricular Outflow Tract of BrS and Control Post-Mortem Cases (A) PSR quantification of collagen content and (B) immunofluorescence quantification of Cx43. Orange data points represent distribution means. Blue data points represent individual cases and controls. Abbreviations as in Figures 2 and 3. Table 1 Demographic Data, Familial Evaluation Results, and Index Presentation for the Included Post-Mortem BrS, Post-Mortem Control, and In Vivo BrS Cases

(A) PSR quantification of collagen content and (B) immunofluorescence quantification of Cx43. Orange data points represent distribution means. Blue data points represent individual cases and controls. Abbreviations as in Figures 2 and 3. Table 1 Demographic Data, Familial Evaluation Results, and Index Presentation for the Included Post-Mortem BrS, Post-Mortem Control, and In Vivo BrS Cases Case Sex Age (yrs) Index Presentation Clinical Abnormality Cardiac Morphology Relatives Evaluated Relatives Affected Post-mortem BrS cohort B1 M 15 SCD in sleep Diagnosis in relative Normal 2 2 B2 M 18 SCD in sleep Diagnosis in relative Normal 4 1 B3 M 19 SCD in sleep Diagnosis in relative Normal 5 1 B4 M 23 SCD with exercise Diagnosis in relative Tunneled RCA 3 2 B5 M 24 SCD in sleep Diagnosis in relative Atrial septal defect 3 1 B6 M 40 SCD with minimal activity Diagnosis in relative Normal 5 3 Post-mortem control cohort C1 M 17 RTA None Normal — — C2 M 18 RTA None Normal — — C3 M 22 Suicide None Normal — — C4 M 22 RTA None Normal — — C5 M 22 RTA None Normal — — C6 M 37 Homicide None Normal — — In vivo BrS cohort V1 M 48 Multiple syncope Spontaneous type 1 ECG Normal — — V2 M 28 Multiple syncope Ajmaline-provoked type 1 ECG Normal — — V3 M 59 VF arrest Spontaneous type 1 ECG Normal — — V4 M 29 VF arrest with fever Spontaneous type 1 ECG Normal — — V5 M 47 Syncope Spontaneous Type 1 ECG Normal — — V6 M 27 Multiple syncope Spontaneous type 1 ECG Normal — — BrS = Brugada syndrome; ECG = electrocardiogram; M = male; RCA = right coronary artery; RTA = road traffic accident; SCD = sudden cardiac death; VF = ventricular fibrillation.

r Spontaneous type 1 ECG Normal — — V5 M 47 Syncope Spontaneous Type 1 ECG Normal — — V6 M 27 Multiple syncope Spontaneous type 1 ECG Normal — — BrS = Brugada syndrome; ECG = electrocardiogram; M = male; RCA = right coronary artery; RTA = road traffic accident; SCD = sudden cardiac death; VF = ventricular fibrillation. Table 2 Univariable and Multivariate Regression Analysis of Proportional Collagen Content, as Evaluated by Morphometric Analysis of PSR Staining in BrS Cases Versus Control Hearts Variable BrS vs. Control Hearts Univariable Analysis Multivariate Analysis OR (95% CI) p Value OR (95% CI) p Value Disease 1.42 (1.06–1.90) 0.024 1.42 (1.05–191) 0.026 LV 1.00 N/A 1.00 N/A RV 1.66 (1.11–2.50) 0.019 1.66 (1.10–2.51) 0.020 RVOT 1.98 (1.34–2.91) 0.003 1.98 (1.33–2.93) 0.003 Endo 1.00 N/A 1.00 N/A Mid 1.27 (1.02–1.58) 0.033 1.27 (1.02–1.58) 0.035 Epi 2.00 (1.46–2.73) <0.001 2.00 (1.45–2.74) 0.001 BrS = Brugada syndrome; CI = confidence interval; Endo = endocardium; Epi = epicardium; LV = left ventricle; Mid = mid-myocardium; OR = odds ratio; PSR = picrosirius red; RV = right ventricle; RVOT = right ventricular outflow tract. Table 3 Multivariable Regression Analysis of Proportional Connexin43 Content in BrS Post-Mortem Cases Versus Control Hearts Variable BrS vs. Control Hearts OR (95% CI) p Value Disease 0.59 (0.44–0.79) 0.001 Endocardium 1.00 N/A Mid-myocardium 0.97 (0.64–1.49) 0.897 Epicardium 1.16 (0.76–1.78) 0.476 Disease (corrected for collagen) 0.58 (0.36–0.96) 0.036 Expression according to zone and after correction for collagen content is also shown.

Table 3 Multivariable Regression Analysis of Proportional Connexin43 Content in BrS Post-Mortem Cases Versus Control Hearts Variable BrS vs. Control Hearts OR (95% CI) p Value Disease 0.59 (0.44–0.79) 0.001 Endocardium 1.00 N/A Mid-myocardium 0.97 (0.64–1.49) 0.897 Epicardium 1.16 (0.76–1.78) 0.476 Disease (corrected for collagen) 0.58 (0.36–0.96) 0.036 Expression according to zone and after correction for collagen content is also shown. Abbreviations as in Table 2.

The renin-angiotensin-aldosterone system (RAAS) maintains cardiovascular homeostasis through angiotensin II (Ang II). Clinically, angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers are mainstay treatments for hypertension and heart failure (HF). Following myocardial infarction (MI), RAAS inhibition stabilizes adverse cardiac remodeling and function and limits progression to HF. A natural counter-regulatory axis of the RAAS exists, centered on ACE2, an ACE homologue that metabolizes Ang II to angiotensin-(1-7) [Ang-(1-7)] 1, 2. Currently being explored therapeutically in cardiovascular diseases including HF and pulmonary hypertension, ACE2 shows promising therapeutic effects (3). Ang-(1-7) acts via the receptor Mas to block detrimental effects of Ang II and mediates direct therapeutic effects in cardiovascular disease 4, 5. Ang-(1-7) is in clinical trials to treat diabetic foot ulcers and cancer 6, 7, emphasizing translational approaches targeting the counter-regulatory RAAS axis. Less studied than Ang-(1-7), the alternative counter-regulatory RAAS peptide angiotensin-(1-9) [Ang-(1-9)] reduces adverse cardiovascular remodeling in rat models of hypertension and MI following peptide administration via osmotic mini-pump 8, 9, 10. Ang-(1-9) attenuates cardiomyocyte hypertrophy and cardiac fibrosis in hypertensive models; these effects are blocked by coadministration of the angiotensin type 2 receptor (AT2R) antagonist PD123,319, further supporting independent effects of Ang-(1-9) as a new counter-regulatory RAAS axis peptide 8, 11.

ump 8, 9, 10. Ang-(1-9) attenuates cardiomyocyte hypertrophy and cardiac fibrosis in hypertensive models; these effects are blocked by coadministration of the angiotensin type 2 receptor (AT2R) antagonist PD123,319, further supporting independent effects of Ang-(1-9) as a new counter-regulatory RAAS axis peptide 8, 11. Assessment of RAAS peptides as therapeutics is limited by short circulatory half-life, requiring osmotic mini-pumps for sustained release in vivo in experimental models. Accordingly, alternative delivery strategies are required for clinical translation. Viral gene therapy is being pursued for HF, including clinical trials using adeno-associated virus (AAV) vector-mediated delivery of sarcoplasmic endoreticulum calcium adenosine triphosphatase 2a (SERCA2a), emphasizing safety and clinical utility (12). Angiotensin peptides are not produced from genes, but are generated extracellularly in the circulation. Synthetic expression cassettes for Ang II, Ang-(1-7), and Ang-(1-9) have been utilized in transgenic models and in gene transfer approaches 13, 14, 15. Here, for the first time, in vivo AAV-mediated gene transfer of Ang-(1-9) via a synthetic expression cassette has been utilized to study cardiac effects in a murine model of MI.

expression cassettes for Ang II, Ang-(1-7), and Ang-(1-9) have been utilized in transgenic models and in gene transfer approaches 13, 14, 15. Here, for the first time, in vivo AAV-mediated gene transfer of Ang-(1-9) via a synthetic expression cassette has been utilized to study cardiac effects in a murine model of MI. Methods Detailed methods are presented in the Online Appendix. Briefly, an Ang-(1-9) expression cassette (13) was sub-cloned into plasmid adeno-associated virus-multiple cloning site (pAAV-MCS) and AAV9 vectors produced via standard protocols (16). Surgical procedures were performed in accordance with the Animals Scientific Procedures Act (1986) and approved by the University of Glasgow Animal Welfare and Ethical Review Panel and UK Home Office. For MI, the left anterior descending artery (LAD) was ligated. Sham animals had identical procedures without ligation. AAVAng-(1-9) or AAV green fluorescent protein (GFP) were delivered intravenously via tail vein following MI as described (17). Echocardiography was performed weekly (Figure 1A) and pressure volume (PV) loop measurements made. Fibrosis was assessed by Picrosirius red staining as described (8). Hypertrophy was measured by wheat germ agglutinin staining. Quantitative reverse transcription polymerase chain reaction was assessed with inventoried gene expression assays. Ventricular cardiomyocytes were isolated from adult C57BL/6 mice, loaded with Fura–4FAM, and the Fura–4FAM fluorescence ratio (340/380 nm excitation) was measured using a spinning wheel monochromator and converted to [Ca2+]i (18). Cardiomyocytes were incubated for 15 min with 1 μmol/l Ang-(1-9), field-stimulated (1.0 Hz), and perfused with 1.8 mmol/l [Ca2+]o HEPES superfusate containing 1 μmol/l Ang-(1-9). Calcium transients and contractility in human-induced pluripotent stem cell-derived cardiomyocytes (hiPS-CM; iCell2 cardiomyocytes, Cellular Dynamics International [Madison, Wisconsin, USA]) were measured in the optical platform CellOPTIQ (Clyde Biosciences Ltd, Glasgow, United Kingdom) in cells loaded with 3 μmol/l Fura-4F-AM. Calcium transients were obtained from the 360/380 ratio and contraction was assessed using a high-resolution camera coupled to CellOPTIQ. Male adult Wistar rats were sacrificed, hearts excised, and Langendorff perfused at 37°C and constant flow (10 ml/min) (19). A fluid-filled balloon was inserted into the left ventricle and connected to a solid-state pressure transducer.

o and contraction was assessed using a high-resolution camera coupled to CellOPTIQ. Male adult Wistar rats were sacrificed, hearts excised, and Langendorff perfused at 37°C and constant flow (10 ml/min) (19). A fluid-filled balloon was inserted into the left ventricle and connected to a solid-state pressure transducer. Hearts were paced and perfused with 1 μmol/l Ang-(1-9). Statistical analysis Data are represented as mean ± SE of the mean (SEM). Paired Student t test for direct comparisons and 1-way analysis of variance with Tukey’s post-test for multiple comparison were performed. Echocardiography was analyzed using repeated measures analysis of variance with Tukey’s post-test. Statistical significance was demonstrated with a p < 0.05.

± SE of the mean (SEM). Paired Student t test for direct comparisons and 1-way analysis of variance with Tukey’s post-test for multiple comparison were performed. Echocardiography was analyzed using repeated measures analysis of variance with Tukey’s post-test. Statistical significance was demonstrated with a p < 0.05. Results Previously, tail vein delivery of 1 × 1011 viral genomes (vg) AAV9 demonstrated robust cardiac transduction (17). To confirm this, AAVGFP-mediated transduction was assessed at 1, 2, and 8 weeks following LAD ligation (Figure 1A). High enhanced GFP expression was observed throughout the myocardium at all time points (Figure 1B). Quantification of enhanced GFP expression in cardiac lysates revealed enhanced GFP expression was detectable at 1 week and increased at 4 and 8 weeks (Figure 1C). Next, animals were subjected to sham procedure, MI, MI/AAVGFP, or MI/AAVAng-(1-9) to assess effects on cardiac function and remodeling. MI in presence or absence of AAVGFP produced higher mortality than sham in the acute recovery phase due to cardiac rupture (sham: 100% survival; MI: 73% survival; MI/AAVGFP: 67% survival) (Figures 1D and 1E). Delivery of AAVAng-(1-9) increased survival to 91% in MI-induced animals.

s on cardiac function and remodeling. MI in presence or absence of AAVGFP produced higher mortality than sham in the acute recovery phase due to cardiac rupture (sham: 100% survival; MI: 73% survival; MI/AAVGFP: 67% survival) (Figures 1D and 1E). Delivery of AAVAng-(1-9) increased survival to 91% in MI-induced animals. Assessment of cardiac function Serial echocardiography was performed (Figure 2A) and a significant reduction in fractional shortening (FS) observed 1 week post-MI in MI and MI/AAVGFP, which progressively decreased at 4 and 8 weeks (Figure 2B). Decreased FS was associated with increased left ventricular end-systolic and end-diastolic dimension (LVESD and LVEDD) (Figures 2C and 2D). AAVAng-(1-9) infusion significantly attenuated reduced FS at all time points. At 8 weeks, FS in MI/AAVAng-(1-9) was significantly reduced compared to sham (38.5 ± 1.9% vs. 49.1 ± 1.6%; p < 0.05), although it was significantly increased compared to MI and MI/AAVGFP (MI = 25.8 ± 2.2%; MI/AAVGFP = 26.6 ± 0.7%; p < 0.05). Importantly, in MI/AAVAng-(1-9), FS remained stable from 1 week, in contrast to the progressive decline in other groups (Figure 2B). At 1 week, LVESD in MI/AAVAng-(1-9) was significantly reduced compared to MI/AAVGFP (Figure 2C). No significant changes in posterior left ventricular (LV) wall thickness were detected at any time point (Figure 2E). Additionally, ejection fraction (EF) was significantly reduced 1 week post-MI in MI and MI/AAVGFP and further decreased at 4 and 8 weeks (Figure 2F). AAVAng-(1-9) delivery significantly attenuated reduced EF at all time points. E/A wave ratio was not different between groups (Figure 2G).

at any time point (Figure 2E). Additionally, ejection fraction (EF) was significantly reduced 1 week post-MI in MI and MI/AAVGFP and further decreased at 4 and 8 weeks (Figure 2F). AAVAng-(1-9) delivery significantly attenuated reduced EF at all time points. E/A wave ratio was not different between groups (Figure 2G). Eight-week PV loop measurements in MI/AAVAng-(1-9) revealed significant attenuation of the decreased systolic indexes observed in MI and MI/AAVGFP (Figure 3A). AAVAng-(1-9) significantly increased end-systolic pressure (Figure 3B) (p < 0.001), EF (Figure 3C) (p < 0.001), and cardiac output (CO) (Figure 3D) (p < 0.05). Importantly, EF was normalized to sham level, whereas CO was significantly increased compared to sham (p < 0.05). However, maximum derivative of change in systolic pressure over time (dP/dtmax) remained significantly reduced to 78.5% of sham (Figure 3E) (p < 0.001). There were no significant differences in end-diastolic pressure, dP/dtmin, and the rate constant of LV pressure decline (Tau) following AAVAng-(1-9) delivery (Figures 3F to 3H). The end-diastolic pressure volume relationship (EDPVR) in MI and MI/AAVGFP was significantly increased (p < 0.05) to 363.3% and 400% of sham, respectively (Figure 3I). Following AAVAng-(1-9), EDPVR was normalized to sham levels (Figure 3I), while there was no detectable change in end-diastolic volume (Figure 3J). End-systolic volume was significantly increased in MI and MI/AAVGFP (p < 0.01); however, it was not different between sham and MI/AAVAng-(1-9) (Figure 3K). Stroke volume was significantly increased (p < 0.05) in MI/AAVAng-(1-9) compared to sham and MI/AAVGFP (Figure 3L). Additionally, the end-systolic pressure volume relationship (ESPVR) was significantly decreased in MI and MI/AAVGFP but normalized by AAVAng-(1-9) (Figure 3M).

etween sham and MI/AAVAng-(1-9) (Figure 3K). Stroke volume was significantly increased (p < 0.05) in MI/AAVAng-(1-9) compared to sham and MI/AAVGFP (Figure 3L). Additionally, the end-systolic pressure volume relationship (ESPVR) was significantly decreased in MI and MI/AAVGFP but normalized by AAVAng-(1-9) (Figure 3M). Effects on hypertrophy and fibrosis Heart weight/tibia length (HW:TL) ratios were significantly increased in all MI groups to 121%, 118%, and 125% of sham for MI, MI/AAVGFP (p < 0.05), and MI/AAVAng-(1-9) (p < 0.01), respectively (Figures 4A and B). Cardiomyocyte diameter was significantly increased compared to sham in all MI groups (sham: 15.1 ± 0.3 μm; MI: 20.9 ± 0.5 μm; MI/AAVGFP: 19.4 ± 0.4 μm; MI/AAVAng-(1-9) = 20.2 ± 0.4 μm; p <0.001) (Figures 4C and D). No significant differences in cell length were observed (Figures 4E and F). LV and right ventricular fibrosis was significantly increased in all MI groups (p < 0.01) (Figures 5A and B). Septal fibrosis in MI and MI/AAVGFP was significantly increased compared to sham, but significantly reduced in MI/AAVAng-(1-9) (MI: 10 ± 2.4; MI/AAVGFP: 6.3 ± 0.4; MI/AAVAng-(1-9): 3.4 ± 0.6%; p < 0.01). Perivascular fibrosis was significantly elevated in MI and MI/AAVGFP; however, delivery of AAVAng-(1-9) normalized this (Online Figure 1). Scar size was consistent among all MI groups (MI: 35.9 ± 2.8%; MI/AAVGFP: 35.2 ± 2.1%; MI/AAVAng-(1-9): 36.9 ± 2.5%) (data not shown). However, in MI and MI/AAVGFP, scar thickness was 329 ± 25 μm and 276 ± 3.9 μm, respectively, whereas in MI/AAVAng-(1-9), scar thickness was significantly increased versus MI/AAVGFP to 383 ± 14 μm (p < 0.05) (Figure 5C).

MI groups (MI: 35.9 ± 2.8%; MI/AAVGFP: 35.2 ± 2.1%; MI/AAVAng-(1-9): 36.9 ± 2.5%) (data not shown). However, in MI and MI/AAVGFP, scar thickness was 329 ± 25 μm and 276 ± 3.9 μm, respectively, whereas in MI/AAVAng-(1-9), scar thickness was significantly increased versus MI/AAVGFP to 383 ± 14 μm (p < 0.05) (Figure 5C). Quantitative polymerase chain reaction of levels of RAAS genes in cardiac complementary DNA revealed significantly increased ACE in all MI groups compared to sham, whereas ACE2 expression remained unchanged (Online Figures 2A and 2B). Furthermore, significantly increased AT2R expression in MI/AAVAng-(1-9) was observed, while the angiotensin type 1 receptors were significantly decreased in all MI groups (Online Figures 2C and 2D). Mas expression was significantly downregulated in MI/AAVAng-(1-9) (Online Figure 2E). There were no significant changes in gene expression of the inflammatory markers tumor necrosis factor alpha; interleukin (IL) 1β, IL6, or IL12a; or interferon γ (Online Figure 3). Additionally, gene expression of matrix metalloproteinase (MMP)-2 and -12 and tissue inhibitor of metalloproteinase-1 were significantly increased in MI groups compared to sham, whereas MMP-9 and -14 were not changed (Online Figure 4). MMP-2 was significantly reduced in MI/AAVGFP and MI/AAVAng-(1-9) and MMP-12 was significantly reduced in MI/AAVAng-(1-9). SERCA2a was significantly reduced in all MI groups (Online Figure 5).

oteinase-1 were significantly increased in MI groups compared to sham, whereas MMP-9 and -14 were not changed (Online Figure 4). MMP-2 was significantly reduced in MI/AAVGFP and MI/AAVAng-(1-9) and MMP-12 was significantly reduced in MI/AAVAng-(1-9). SERCA2a was significantly reduced in all MI groups (Online Figure 5). Effects in cardiomyocytes and whole hearts Calcium (Ca2+) handling, in particular sarcoplasmic reticulum (SR)–mediated Ca2+ release, is the major determinant of cardiomyocyte contractility. Therefore, characteristics of SR-mediated Ca2+ release and uptake (Ca2+ transients) were determined in murine cardiomyocytes acutely exposed to soluble Ang-(1-9) peptide. Ang-(1-9) significantly increased Ca2+ transient amplitude (control: 561.0 ± 86.5 nmol/l; Ang-(1-9): 933.7 ± 107.0 nmol/l; p <0.05) (Figures 6A and 6B); an observation also observed in cardiomyocytes isolated from MI hearts (Online Figure 6). In parallel, Ang-(1-9) significantly increased cell shortening compared to control cardiomyocytes [control: 6.8 ± 0.9%; Ang-(1-9): 10.2 ± 1.1%; p < 0.05] (Figures 6C and 6D). The rate of decline of the Ca2+ transient was not significantly altered by Ang-(1-9) (data not shown), suggesting no change in rate of Ca2+ removal from the cytosol through SR uptake via SERCA or the sodium calcium exchanger. To determine SR Ca2+ content, a major determinant of Ca2+ transient amplitude (20), a rapid bolus of 10 mmol/l caffeine was applied at the end of the protocol to release all SR Ca2+ into the cytosol. The caffeine-induced Ca2+-transient amplitude in Ang-(1-9)–incubated cardiomyocytes was significantly increased compared to control, indicating an increased SR Ca2+ content (control: 987.5 ± 101.4 nmol/l; Ang-(1-9): 1,535.2 ± 188.8 nmol/l; p < 0.05) (Figure 6E). SERCA-mediated Ca2+ uptake is bypassed during application of 10 mmol/l caffeine and cytosolic Ca2+ removal occurs predominately via the sodium calcium exchanger. The rate constant of decline of the caffeine-induced Ca2+-transient (Tau) was unaltered by Ang-(1-9), supporting the conclusion that Ang-(1-9) does not alter cardiomyocyte Ca2+ extrusion (Online Figure 7). One possible route through which the SR Ca2+ content and transient could be elevated is through increased influx of Ca2+ (e.g., via L-type Ca2+ channels).

hen n = 60 and all patients had completed 3 months of treatment, respectively, with the intention of halting the study if adverse measures were identified or for futility but not for interim positive efficacy. The interim analyses were performed by independent personnel not directly associated with the study’s conduct. The study sponsor and funder (Novartis) participated in discussions about the design and conduct of this study; they also provided the drugs used in the trial and logistical support for its execution. The trial design, endpoints, and statistical analyses were largely derived from the academic investigators’ previously published studies 19, 23, 24. Following the final database lock, all patient data were analyzed independently by the Centre for Statistics in Medicine, Oxford (J.B.). The manuscript was drafted by the academic investigators (R.P.C., J.S.B., J.-C.T., and Z.A.F.), in accordance with the written agreement between Novartis and the academic institutions, and reviewed and revised by the writing committee. All authors had full access to all the data in the study and assume responsibility for publication. All statistical analyses were performed by using Stata 14 (StataCorp LP, College Station, Texas).

eine-induced Ca2+-transient (Tau) was unaltered by Ang-(1-9), supporting the conclusion that Ang-(1-9) does not alter cardiomyocyte Ca2+ extrusion (Online Figure 7). One possible route through which the SR Ca2+ content and transient could be elevated is through increased influx of Ca2+ (e.g., via L-type Ca2+ channels). To assess this, cardiomyocytes were continuously perfused with Ang-(1-9) and a 10 mmol/l bolus of caffeine applied for 10 s after 15 min followed by 2 min of steady state measurements while cells were stimulated. The amplitude of the first Ca2+ transient after caffeine was taken as an index of Ca2+ influx via the L-type Ca2+ channel 21, 22, 23. Ang-(1-9) significantly increased the L-type Ca2+-transient amplitude versus controls (191.8 ± 28.4 nmol/l vs. 74.6 ± 17.3 nmol/l; p < 0.05) (Figures 6F and 6G). To assess whether the positive inotropy observed in isolated cardiomyocytes translated to whole heart contractile function, hearts isolated according to the Langendorff model were perfused with Ang-(1-9). After 4 min of perfusion, Ang-(1-9) induced a significant increase in developed pressure with a concomitant elevation in dP/dtmax, confirming a positive inotropic response to Ang-(1-9) (Figure 7). Since protein kinase A (PKA) has been previously reported to modulate calcium flux via the L type Ca2+ channel following application of Ang-(1-7) (24), we used the inhibitor H-89, which did indeed abolish the response to Ang-(1-9), thus supporting a role for PKA in the positive inotropic effect of Ang-(1-9).

Figure 7). Since protein kinase A (PKA) has been previously reported to modulate calcium flux via the L type Ca2+ channel following application of Ang-(1-7) (24), we used the inhibitor H-89, which did indeed abolish the response to Ang-(1-9), thus supporting a role for PKA in the positive inotropic effect of Ang-(1-9). To extrapolate the findings in murine cardiomyocytes and rat hearts to a human model, hiPSC-CMs were used (25), and intracellular Ca2+ and contraction were measured before (baseline) and after 15 min incubation with different Ang-(1-9) concentrations. A dose-dependent increase in Ca2+ transient and contraction amplitudes was observed within concentrations 0.5 μmol/l to 2 μmol/l (data not shown) with no effect on parameters such as calcium transient upstroke, rate of decline, or contraction/relaxation times. Ca2+ transient amplitude and contraction following incubation with 1 μmol/l Ang-(1-9) compared to control cells was measured and a 210 ± 10% change from baseline in Ca2+ transient amplitude for cells incubated with 1 μmol/l Ang-(1-9) was observed, an effect significantly different from control cells (98 ± 13% change from baseline) (Figures 8A and B). A parallel effect was observed for contraction in terms of increased amplitude (160 ± 13% vs. 93 ± 12%) (Figures 8C and D).

eline in Ca2+ transient amplitude for cells incubated with 1 μmol/l Ang-(1-9) was observed, an effect significantly different from control cells (98 ± 13% change from baseline) (Figures 8A and B). A parallel effect was observed for contraction in terms of increased amplitude (160 ± 13% vs. 93 ± 12%) (Figures 8C and D). Discussion Our study focused on an innovative gene therapy approach to deliver Ang-(1-9) directly to the heart to assess therapeutic effects and mechanisms of action in a murine MI model. AAV9-mediated delivery of Ang-(1-9) reduced acute rupture and mildly affected cardiac hypertrophy and fibrosis, but preserved LV systolic function, even at 8 weeks post-MI (Central Illustration). The effects of Ang-(1-9) were mediated via a direct positive inotropic effect. In isolated cardiomyocytes, Ang-(1-9) enhanced Ca2+ handling by increasing SR Ca2+ content and Ca2+ transient amplitude (Central Illustration).

and fibrosis, but preserved LV systolic function, even at 8 weeks post-MI (Central Illustration). The effects of Ang-(1-9) were mediated via a direct positive inotropic effect. In isolated cardiomyocytes, Ang-(1-9) enhanced Ca2+ handling by increasing SR Ca2+ content and Ca2+ transient amplitude (Central Illustration). While rupture rates in MI and MI/AAVGFP groups were consistent with previous studies 26, 27, AAVAng-(1-9) reduced acute rupture. Although the reasons for this are not entirely clear, because Ang-(1-9) delivery increased scar thickness, the mechanism underlying this effect might entail stabilization of cardiac architecture in the acute phase post-MI. This is a potentially beneficial finding because overall incidence of cardiac rupture in acute ST-elevation MI patients is 6.4% (28). At 8 weeks, there were no detectable differences in gene expression of tumor necrosis factor alpha, IL1β, IL6, IL12α, and interferon-γ associated with inflammation; this is not unexpected because these cytokines are upregulated acutely following MI (29). We also measured expression of genes involved in tissue remodeling in MI, including MMP-2, -9, -12, -14 and tissue inhibitor of metalloproteinase-1 (30). Differences in MMP-2 and -12 could be detected at 8 weeks following AAVAng-(1-9) delivery, suggesting one possible mechanism by which Ang-(1-9) can modulate remodeling during scar evolution. Understanding how AAVAng-(1-9) delivery contributes to healing post-MI and scar thickening will be important to investigate by assessing a range of acute time points within the first few days post-delivery, when inflammation is high and the scar is rapidly remodeling and evolving. Since AAV-mediated delivery has been detectable as early as 2 days post-delivery and is accelerated in damaged tissue 31, 32, 33, future studies may reveal other mechanisms of AAVAng-(1-9) action.

points within the first few days post-delivery, when inflammation is high and the scar is rapidly remodeling and evolving. Since AAV-mediated delivery has been detectable as early as 2 days post-delivery and is accelerated in damaged tissue 31, 32, 33, future studies may reveal other mechanisms of AAVAng-(1-9) action. AAVAng-(1-9) delivery significantly reduced fibrosis, although not specifically in the LV, suggesting this did not significantly contribute to Ang-(1-9)’s inotropic effect. The reduced fibrosis (albeit regional) aligned with previous studies where osmotic mini-pump delivery attenuated cardiac fibrosis (8). This supports a general antifibrotic effect for the counter-regulatory RAAS axis, given the ACE2/Ang-(1-7)/Mas system’s well-established antifibrotic effects on the myocardium 34, 35. Ang-(1-9) did not mediate any antihypertrophic effect, contrary to previous reports 9, 11, possibly because the previous work only assessed hypertrophy at 2 weeks compared to at 8 weeks here. Therefore, early acute effects of Ang-(1-9) on limiting hypertrophy might not be maintained once significant structural remodeling has taken place at 8 weeks.

antihypertrophic effect, contrary to previous reports 9, 11, possibly because the previous work only assessed hypertrophy at 2 weeks compared to at 8 weeks here. Therefore, early acute effects of Ang-(1-9) on limiting hypertrophy might not be maintained once significant structural remodeling has taken place at 8 weeks. AAVAng-(1-9)–transduced hearts consistently had greater contraction and blood ejection, evidenced by dramatically increased CO and stroke volume and normalized EF, showing that regardless of MI-induced dilation, function was maintained. This contrasted with a previous study assessing osmotic mini-pump-mediated Ang-(1-9) delivery on cardiac function in rats post-MI that showed significantly reduced LV dimensions and volumes and reported reduced wall thickness, but no change in LV systolic function (9). Therefore, while certain parameters were consistent (e.g., alterations in LVESD), this current study clearly demonstrated markedly improved systolic function with AAVAng-(1-9), corroborated via echocardiography and PV loop measurements. A major reason for the difference may be method of peptide delivery: Direct gene transfer in the heart via AAV9 utilized here (vs. osmotic mini-pump) achieved high local cardiac concentrations 17, 36. Local tissue-specific effects of the RAAS might differ from systemic effects; for instance, local Ang II production in the heart does not produce acute cardiac remodeling, whereas systemic infusion does (37). Tissue-specific effects were also reported for Ang-(1-7) in MI in transgenic mice (38), and lentiviral delivery of Ang-(1-7) in rat MI improved cardiac function (39), supporting the concept that local cardiac Ang-(1-7) and Ang-(1-9) produce beneficial effects.

te cardiac remodeling, whereas systemic infusion does (37). Tissue-specific effects were also reported for Ang-(1-7) in MI in transgenic mice (38), and lentiviral delivery of Ang-(1-7) in rat MI improved cardiac function (39), supporting the concept that local cardiac Ang-(1-7) and Ang-(1-9) produce beneficial effects. Additionally, AAVAng-(1-9) delivery significantly increased myocardial AT2R gene expression. AT2R expression is reported to increase acutely following MI (40). Given that AT2R is associated with cardioprotective effects (41), including reduced remodeling and improved function post-MI, this might underlie some therapeutic effects of AAVAng-(1-9). A small but significant change in Mas expression was also observed in the MI/AAVAng-(1-9) group. The reason for this is not clear because Mas is upregulated in dysfunctional hearts 4 weeks post-MI in rats (42); however, since in our studies, cardiac function was preserved in AAVAng-(1-9)-infused mice, Mas downregulation might be compensatory. This requires confirmation.

also observed in the MI/AAVAng-(1-9) group. The reason for this is not clear because Mas is upregulated in dysfunctional hearts 4 weeks post-MI in rats (42); however, since in our studies, cardiac function was preserved in AAVAng-(1-9)-infused mice, Mas downregulation might be compensatory. This requires confirmation. To gain further insight into potential mechanisms underlying the positive inotropic effects of Ang-(1-9), excitation contraction coupling was studied in isolated murine cardiomyocytes (normal and after MI) and the whole rat heart and hiPSC-CMs. We demonstrated a direct inotropic effect of Ang-(1-9), mediated through increasing Ca2+ transient amplitude leading to increased contraction, and possibly explained via increased L-type Ca2+ influx paralleled by increased SR Ca2+ content. Although a direct inotropic effect of Ang-(1-9) has not been reported previously, when Ang-(1-7) is applied intracellularly to cardiomyocytes, PKA is activated, leading to increased L-type Ca2+ channel activity (24). Ang II is reported to increase Ca2+ transient amplitude and intracellular Ang II is reported to increase Ca2+ transient amplitude via modulating L-type Ca2+ current and releasing SR Ca2+ 43, 44. Several cardiomyocyte Ca2+ handling proteins, including the L-type Ca2+ channel, are regulated by PKA-mediated pathways (45). The increased contractility in the isolated hearts perfused with Ang-(1-9) in this study and the effect of PKA inhibition suggest that Ang-(1-7) and Ang-(1-9) may act by similar mechanisms leading to PKA activation.

Ca2+ handling proteins, including the L-type Ca2+ channel, are regulated by PKA-mediated pathways (45). The increased contractility in the isolated hearts perfused with Ang-(1-9) in this study and the effect of PKA inhibition suggest that Ang-(1-7) and Ang-(1-9) may act by similar mechanisms leading to PKA activation. Study limitations Our studies were performed in a murine model of permanent LAD ligation and future studies in larger animal models following ischemic reperfusion would be helpful to inform translation of the gene therapy. Furthermore, the inotropic effects studied in isolated cardiomyocytes were performed via peptide perfusion and further work to isolate cardiomyocytes from hearts infused in vivo with the gene therapy combined with use of patch clamping would enable full dissection of the inotropic effects of Ang-(1-9). Nonetheless, the current studies strongly support a beneficial effect of cardiac Ang-(1-9) gene therapy in the setting of MI. Conclusions This study suggested that gene therapy to augment Ang-(1-9) levels in the heart produces clear benefit in a murine MI model. Our data supported the notion that administration of the counter-regulatory RAAS peptide Ang-(1-9) via translational gene therapy is a novel and promising approach in heart disease that preserves cardiac systolic function post-MI and is maintained in a sustained manner.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: In a preclinical model of MI, gene therapy with Ang-(1-9) preserved systolic function by mediating a direct positive inotropic effect on cardiomyocytes.

mising approach in heart disease that preserves cardiac systolic function post-MI and is maintained in a sustained manner.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: In a preclinical model of MI, gene therapy with Ang-(1-9) preserved systolic function by mediating a direct positive inotropic effect on cardiomyocytes. TRANSLATIONAL OUTLOOK: Further work is needed to assess whether Ang-(1-9) gene delivery in other large animal models of myocardial infarction preserves systolic function and prevents heart failure. Appendix Online Data Acknowledgments The authors thank Nicola Britton, Gregor Aitchison, and Catherine Hawksby for technical support and Dr. John McClure for advice on statistical analysis. All data is included in the manuscript. This work was supported by a British Heart Foundation PhD Studentship (FS/09/052/28032), BHF project grant (PG/11/43/28901), and an MRC Confidence in Concept Award (MC_PC_13063), and MRC Research Grant (G0901161). Dr. Zamora is recipient of a postdoctoral fellowship from Fundacion Alfonso Martin Escudero, Spain. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For supplemental text, references, and figures, please see the online version of this article. Figure 1 AAV Delivery

This work was supported by a British Heart Foundation PhD Studentship (FS/09/052/28032), BHF project grant (PG/11/43/28901), and an MRC Confidence in Concept Award (MC_PC_13063), and MRC Research Grant (G0901161). Dr. Zamora is recipient of a postdoctoral fellowship from Fundacion Alfonso Martin Escudero, Spain. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For supplemental text, references, and figures, please see the online version of this article. Figure 1 AAV Delivery (A) Study design. (B) Immunohistochemistry for enhanced GFP at 1, 2, and 8 weeks following intravenous delivery of AAVGFP. (Original magnification ×4 for upper panel and ×40 for lower panel; scale = 100 μm.) (C) Quantification of GFP in transduced heart lysates using GFP assay. Fluorescence normalized to negative control heart tissue basal fluorescence and total protein concentration. (D) Mortality for each animal group. Group sizes are n = 10, n = 15, n = 15, and n = 11 for sham, MI, MI/AAVGFP, and MI/AAV Ang-(1-9), respectively. (E) Percent survival and cause of mortality. AAV = adeno-associated virus; Ang-(1-9) = angiotensin-(1-9); GFP = green fluorescence protein; MI = myocardial infarction. Figure 2 Cardiac Function

(A) Study design. (B) Immunohistochemistry for enhanced GFP at 1, 2, and 8 weeks following intravenous delivery of AAVGFP. (Original magnification ×4 for upper panel and ×40 for lower panel; scale = 100 μm.) (C) Quantification of GFP in transduced heart lysates using GFP assay. Fluorescence normalized to negative control heart tissue basal fluorescence and total protein concentration. (D) Mortality for each animal group. Group sizes are n = 10, n = 15, n = 15, and n = 11 for sham, MI, MI/AAVGFP, and MI/AAV Ang-(1-9), respectively. (E) Percent survival and cause of mortality. AAV = adeno-associated virus; Ang-(1-9) = angiotensin-(1-9); GFP = green fluorescence protein; MI = myocardial infarction. Figure 2 Cardiac Function (A) Eight-week M-mode images (scale = 5 mm and 1 s). Effect of AAVAng-(1-9) varied by parameter: (B) Serial FS; (C) LVESD; (D) LVEDD; (E) posterior LV thickness; and (F) EF. *p <0.05 versus sham; #p <0.05 versus MI and MI/AAVGFP; ∼p <0.05 MI/AAVGFP versus MI/AAVAng-(1-9). (G) Average E/A ratio measurements (n = 6 per group). Data presented as mean ± SEM. A = after wave; E = early wave; EF = ejection fraction; FS = fractional shortening; LV = left ventricular; LVEDD = left ventricular end diastolic dimension; LVESD = left ventricular end systolic dimension; other abbreviations as in Figure 1. Figure 3 Hemodynamic Indexes

(A) Eight-week M-mode images (scale = 5 mm and 1 s). Effect of AAVAng-(1-9) varied by parameter: (B) Serial FS; (C) LVESD; (D) LVEDD; (E) posterior LV thickness; and (F) EF. *p <0.05 versus sham; #p <0.05 versus MI and MI/AAVGFP; ∼p <0.05 MI/AAVGFP versus MI/AAVAng-(1-9). (G) Average E/A ratio measurements (n = 6 per group). Data presented as mean ± SEM. A = after wave; E = early wave; EF = ejection fraction; FS = fractional shortening; LV = left ventricular; LVEDD = left ventricular end diastolic dimension; LVESD = left ventricular end systolic dimension; other abbreviations as in Figure 1. Figure 3 Hemodynamic Indexes LV hemodynamic measurements at 8 weeks were determined using a PV-loop system with true blood volume calculated using Wei’s equation. Shown are (A) PV-loop relationship example; the systolic functional indexes of (B) ESP, (C) EF, (D) CO; and (E) dP/dtmax, the diastolic functional indexes of (F) end-diastolic pressure (EDP), (G) dP/dtmin, (H) Tau, and (I) EDPVR; and the volume indexes of (J) EDV, (K) ESV, (L) SV, and (M) ESPVR. *p <0.05, **p <0.01, ***p <0.001 versus sham; #p <0.05, ##p <0.01, ###p <0.001 versus MI and MI/AAVGFP; ∼p <0.05 versus MI/AAVGFP only. n = 9, 9, 9, and 8 for, sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM. CO = cardiac output; EDP = end diastolic pressure; EDPVR = EDP volume relationship; EDV = end diastolic volume; ESP = end systolic pressure; ESPVR = ESP volume relationship; ESV = end systolic volume; PV = pressure volume; SV = stroke volume; other abbreviations as in Figures 1 and 2.

ely. Data presented as mean ± SEM. CO = cardiac output; EDP = end diastolic pressure; EDPVR = EDP volume relationship; EDV = end diastolic volume; ESP = end systolic pressure; ESPVR = ESP volume relationship; ESV = end systolic volume; PV = pressure volume; SV = stroke volume; other abbreviations as in Figures 1 and 2. Figure 4 Cardiomyocyte Hypertrophy (A) Heart images at 8 weeks (scale bar = 5 mm). (B) Ratio of HW to TL. *p <0.05, **p <0.01 versus sham. n = 10, 10, 9, and 8 for, sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM. (C) Cardiac cross sections in transverse axis (original main image magnification ×25; scale = 50 μm; inset zoom image scale = 12.5 μm). (D) LV cardiomyocyte diameter in hearts. ***p <0.001 versus sham. n = 10, 10, 9, and 8 for sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. (E) Cardiac cross sections in longitudinal axis (original main image magnification = ×25; scale = 50 μm; inset zoom image scale = 50 μm). (F) LV cardiomyocyte length. n = 10, 10, 9, and 8 for sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM with average cell size taken as average of a group of cells evenly distributed across LV. HW = heart weight; TL = tibia length; other abbreviations as in Figures 1 and 2. Figure 5 Cardiac Fibrosis

(A) Heart images at 8 weeks (scale bar = 5 mm). (B) Ratio of HW to TL. *p <0.05, **p <0.01 versus sham. n = 10, 10, 9, and 8 for, sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM. (C) Cardiac cross sections in transverse axis (original main image magnification ×25; scale = 50 μm; inset zoom image scale = 12.5 μm). (D) LV cardiomyocyte diameter in hearts. ***p <0.001 versus sham. n = 10, 10, 9, and 8 for sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. (E) Cardiac cross sections in longitudinal axis (original main image magnification = ×25; scale = 50 μm; inset zoom image scale = 50 μm). (F) LV cardiomyocyte length. n = 10, 10, 9, and 8 for sham, MI, MI/AAVGFP, and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM with average cell size taken as average of a group of cells evenly distributed across LV. HW = heart weight; TL = tibia length; other abbreviations as in Figures 1 and 2. Figure 5 Cardiac Fibrosis (A) Picrosirius red staining of heart sections (original magnification ×1.25; scale = 1 mm; zoom insert image scale = 0.5 mm). (B) Quantification of total cardiac fibrosis of the scar, LV, right ventricular, and septum regions. (C) Scar thickness for each MI group. n = 10, 10, 9, and 8 for sham, MI, MI/AAVGFP and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM. *p <0.05, **p <0.01, ***p <0.001 versus sham region; #p <0.05, ##p <0.01 versus MI and MI/AAVGFP region. Abbreviations as in Figures 1 and 2. Figure 6 Excitation Contraction Coupling in Isolated Cardiomyocytes

(A) Picrosirius red staining of heart sections (original magnification ×1.25; scale = 1 mm; zoom insert image scale = 0.5 mm). (B) Quantification of total cardiac fibrosis of the scar, LV, right ventricular, and septum regions. (C) Scar thickness for each MI group. n = 10, 10, 9, and 8 for sham, MI, MI/AAVGFP and MI/AAVAng-(1-9), respectively. Data presented as mean ± SEM. *p <0.05, **p <0.01, ***p <0.001 versus sham region; #p <0.05, ##p <0.01 versus MI and MI/AAVGFP region. Abbreviations as in Figures 1 and 2. Figure 6 Excitation Contraction Coupling in Isolated Cardiomyocytes (A) Representative Ca2+-transient traces from Ang-(1-9)–treated cells. (B) Average Ca2+-transient amplitude for cardiomyocytes untreated (control; n = 21) or pre-treated with 1 μmol/l Ang-(1-9) (n = 21). (C) Cell shortening traces. (D) Cell shortening. (E) Average sarcoplasmic reticulum Ca2+ content. *p < 0.05 versus control. (F) Average first Ca2+ transient traces after 10 mmol/l caffeine. (G) Average L-type Ca2+-transient amplitude [control, n = 10; Ang-(1-9), n = 11]. *p < 0.05 versus control. Data presented as mean ± SEM. Red trace = 1 Hz stimulation. Abbreviations as in Figure 1. Figure 7 Inotropic Effects of Ang-(1-9)

(A) Representative Ca2+-transient traces from Ang-(1-9)–treated cells. (B) Average Ca2+-transient amplitude for cardiomyocytes untreated (control; n = 21) or pre-treated with 1 μmol/l Ang-(1-9) (n = 21). (C) Cell shortening traces. (D) Cell shortening. (E) Average sarcoplasmic reticulum Ca2+ content. *p < 0.05 versus control. (F) Average first Ca2+ transient traces after 10 mmol/l caffeine. (G) Average L-type Ca2+-transient amplitude [control, n = 10; Ang-(1-9), n = 11]. *p < 0.05 versus control. Data presented as mean ± SEM. Red trace = 1 Hz stimulation. Abbreviations as in Figure 1. Figure 7 Inotropic Effects of Ang-(1-9) Upon achieving maximum developed pressure, hearts were paced at 320 beats/min and allowed to reach steady state for 10 min before addition of 1 μmol/l Ang-(1-9). The protein kinase A inhibitor H89 (1 μmol/l) was perfused 10 min prior to adding Ang-(1-9) and was present throughout perfusion. (A) There were significant differences in (A) LV developed pressure compared to control hearts, but not in (B) the first derivative of LV developed pressured (dP/dtmax). Control n = 5, Ang-(1-9) n = 6, and H89 + Ang-(1-9) n = 5. *p <0.05 versus control hearts; †p <0.05 versus Ang-(1-9). Abbreviations as in Figure 1. Figure 8 Excitation Contraction Coupling in hiPSC-CMs

Upon achieving maximum developed pressure, hearts were paced at 320 beats/min and allowed to reach steady state for 10 min before addition of 1 μmol/l Ang-(1-9). The protein kinase A inhibitor H89 (1 μmol/l) was perfused 10 min prior to adding Ang-(1-9) and was present throughout perfusion. (A) There were significant differences in (A) LV developed pressure compared to control hearts, but not in (B) the first derivative of LV developed pressured (dP/dtmax). Control n = 5, Ang-(1-9) n = 6, and H89 + Ang-(1-9) n = 5. *p <0.05 versus control hearts; †p <0.05 versus Ang-(1-9). Abbreviations as in Figure 1. Figure 8 Excitation Contraction Coupling in hiPSC-CMs (A) Representative Ca2+-transient traces from Ang-(1-9)-treated cells. (B) Average Ca2+-transient amplitude for iCell2 hiPSC-CM (Cellular Dynamics International, Madison, Wisconsin). (C) Contractility traces. (D) Average contraction amplitude for iCell2 hiPSC-CMs. *p <0.05 versus control. Data presented as mean ± SEM (n = 5). Red trace = 1 Hz stimulation. hiPSC-CM = human-induced pluripotent stem cell-derived cardiomyocytes; other abbreviation as in Figure 1. Central Illustration Post-MI Gene Therapy With Ang-(1-9)

(A) Representative Ca2+-transient traces from Ang-(1-9)-treated cells. (B) Average Ca2+-transient amplitude for iCell2 hiPSC-CM (Cellular Dynamics International, Madison, Wisconsin). (C) Contractility traces. (D) Average contraction amplitude for iCell2 hiPSC-CMs. *p <0.05 versus control. Data presented as mean ± SEM (n = 5). Red trace = 1 Hz stimulation. hiPSC-CM = human-induced pluripotent stem cell-derived cardiomyocytes; other abbreviation as in Figure 1. Central Illustration Post-MI Gene Therapy With Ang-(1-9) Adeno-associated virus serotype 9–mediated delivery of Ang-(1-9) via tail vein in a murine model of MI following coronary artery ligation produced significantly improved CO, SV, EF, and FS. Incubating freshly isolated adult murine cardiomyocytes or human induced pluripotent stem cell-derived cardiomyocytes with Ang-(1-9) leads to elevated SR Ca2+ content through stimulation of the L type calcium channel and enhanced contraction. Ang-(1-9) = angiotensin-(1-9); CO = cardiac output; EF = ejection fraction; FS = fractional shortening; MI = myocardial infarction; SR = sarcoplasmic reticulum; SV = stroke volume.

Left ventricular noncompaction (LVNC) is characterized as a primary genetic cardiomyopathy by the American Heart Association, but is characterized by the European Society of Cardiologists as an “unclassified cardiomyopathy,” aptly demonstrating some of the controversy that surrounds this condition 1, 2, 3. Previously considered a rare cardiomyopathy, there has been a rapid proliferation in publications regarding this entity, raising the question of whether this is a result of better identification of those with the disease or whether it is being over-diagnosed due to the rapid expansion in the utilization of cardiac imaging and the ever-improving visualization of cardiac structures 4, 5. More than 8% of athletes meet 1 of the 3 current echocardiographic criteria for LVNC, whereas 43% of a healthy population cohort meet the most commonly used cardiac magnetic resonance imaging (CMR) threshold for diagnosis measured on long axis cine sequences as proposed by Peterson et al. 6, 7, 8. In addition, a high prevalence of LVNC has been observed in both dilated and hypertrophic cardiomyopathies 9, 10.

s 43% of a healthy population cohort meet the most commonly used cardiac magnetic resonance imaging (CMR) threshold for diagnosis measured on long axis cine sequences as proposed by Peterson et al. 6, 7, 8. In addition, a high prevalence of LVNC has been observed in both dilated and hypertrophic cardiomyopathies 9, 10. Since the original CMR criteria was proposed by Petersen et al. (7), several other groups have developed alternate diagnostic criteria with improved sensitivity and specificity, utilizing measurements on both short axis systolic and diastolic views of the left ventricle as well as quantifying the compacted-to-noncompacted myocardial mass ratio 11, 12, 13. However, given the earlier findings from multiple studies utilizing multiple imaging modalities of significant noncompaction in asymptomatic cohorts free from known cardiovascular disease (CVD), it is not clear whether these new criteria help identify those with genuine disease, or whether, when applied to the general population, they will serve to further strengthen the notion of LVNC as an anatomical phenotype rather than a pathological entity. This is of significant clinical importance due to the long-term implications that currently receiving a diagnosis of LVNC entails—impacting insurance costs and necessitating long-term monitoring and follow-up. The aim of this study was to determine the prevalence of the population exhibiting LVNC, the predictors for the presence LVNC, and the physiological implications of noncompaction on cardiac function.

ly receiving a diagnosis of LVNC entails—impacting insurance costs and necessitating long-term monitoring and follow-up. The aim of this study was to determine the prevalence of the population exhibiting LVNC, the predictors for the presence LVNC, and the physiological implications of noncompaction on cardiac function. Methods Study population Following local ethical committee approval, a cohort of 2,047 volunteers was invited to the imaging arm of the TASCFORCE (Tayside Screening for Cardiovascular Events) study. Volunteers were enrolled into the study if they: 1) were more than 40 years of age; 2) were free from CVD or other indication for statin therapy as recommended by the Scottish Intercollegiate Guidelines Network (SIGN) report 97 for “Risk Estimation and the Prevention of Cardiovascular Disease” published in February 2007; 3) had a serum B-type natriuretic peptide (BNP) level greater than their gender specific median; and 4) had a 10-year risk of coronary heart disease <20% as predicted by the Adult Treatment Panel III algorithm (14). Exclusion criteria included the following: 1) pregnancy; 2) known primary muscle disease; 3) known atherosclerotic disease—including angina, previous myocardial infarction, peripheral arterial disease, amputation, previous revascularization surgery, hypertension, heart failure, or cerebrovascular event; 4) known diabetes; 5) active liver disease; 6) other known illness or contraindication to magnetic resonance imaging (MRI); 7) participation in another clinical trial; 8) inability to give informed consent; 9) known alcohol abuse; and 10) a blood pressure >145/95 mm Hg. Details of the TASCFORCE study arms and design are encapsulated within Figure 1.

ver disease; 6) other known illness or contraindication to magnetic resonance imaging (MRI); 7) participation in another clinical trial; 8) inability to give informed consent; 9) known alcohol abuse; and 10) a blood pressure >145/95 mm Hg. Details of the TASCFORCE study arms and design are encapsulated within Figure 1. Image acquisition The MRI protocol has been described in detail elsewhere (15). In brief, imaging was performed using a 3-T Magnetom Trio Scanner (Siemens, Erlangen, Germany). Whole-body magnetic resonance angiography was performed using a dual-bolus injection technique with the CMR cines performed before the first contrast injection, and the late gadolinium enhancement sequences performed between the first and second contrast bolus injections. For CMR, a body matrix radiofrequency coil (6 elements) was used in combination with a spine array (up to 24 elements). Electrocardiograph (EKG)–gated segmented breath-hold cinematic (CINE) TrueFISP (Siemens, Erlangen, Germany) images were acquired in the horizontal and vertical long axes, and in the short axis from the atrioventricular ring to the left ventricular (LV) apex using a 2-dimensional ECG-gated breath-hold segmented (CINE) TrueFISP sequence. Retrospective ECG gating was used, with 25 cardiac phases reconstructed (25 lines per segment) and 2 image slices acquired per breath-hold. Parallel imaging was also implemented (integrated parallel acquisition technique [iPAT x2]).

V) apex using a 2-dimensional ECG-gated breath-hold segmented (CINE) TrueFISP sequence. Retrospective ECG gating was used, with 25 cardiac phases reconstructed (25 lines per segment) and 2 image slices acquired per breath-hold. Parallel imaging was also implemented (integrated parallel acquisition technique [iPAT x2]). Image analysis LV mass and volume quantification was performed as previously described (15). Values were normalized to height1.7. For noncompaction assessment, each of the 4 diagnostic criteria was measured as follows (Central Illustration):1. Long axis noncompaction (LAX) was measured on the horizontal and vertical LAX cine sequences, which were analyzed at end-diastole. The thickness of the compacted and noncompacted myocardium was measured at the location of maximum noncompaction as described by Petersen et al. (7). Where uncertainty existed, multiple sites were measured and the maximum noncompaction ratio recorded. An LAX noncompacted: compacted myocardial ratio ≥2.3 was considered to meet the LAX diagnosis of LVNC (7). The location of maximum noncompaction was recorded using the American Heart Association (AHA) 17-segment model of the left ventricle.

ted, multiple sites were measured and the maximum noncompaction ratio recorded. An LAX noncompacted: compacted myocardial ratio ≥2.3 was considered to meet the LAX diagnosis of LVNC (7). The location of maximum noncompaction was recorded using the American Heart Association (AHA) 17-segment model of the left ventricle. 2. Short axis noncompaction (SAX) was performed using the short axis cine images, with the region of highest noncompacted myocardium to compacted myocardium ratio measured both in diastole and systole, as described by Stacey et al. (12) and Grothoff et al. (13). The apical LV (segment 17) was excluded from analysis. A diastolic SAX (SAXDIAS) noncompacted: compacted myocardial ratio ≥3 was considered to meet the SAXDIAS diagnosis of LVNC, whereas a systolic SAX (SAXSYST) noncompacted:compacted myocardial ratio ≥2 was considered to meet the SAXSYST diagnosis of LVNC. The location of maximum noncompaction in both systole and diastole was recorded using the AHA 17-segment model of the left ventricle. 3. Noncompacted myocardial mass (NCMASS) was measured as described by Jacquier et al. (11). Endocardial and epicardial contours were defined on the SAX stack at end-diastole with the papillary muscles included in the compacted mass. A new endocardial contour was then defined to incorporate the noncompacted trabeculae to calculate the global LV mass. The NCMASS was calculated as the difference between the global LV mass and the compacted LV mass. A noncompacted mass >20% of the global LV mass was considered to meet the NCMASS diagnosis of LVNC.

ss. A new endocardial contour was then defined to incorporate the noncompacted trabeculae to calculate the global LV mass. The NCMASS was calculated as the difference between the global LV mass and the compacted LV mass. A noncompacted mass >20% of the global LV mass was considered to meet the NCMASS diagnosis of LVNC. All individuals had the LAX noncompacted: compacted myocardial ratio measured, however only those with a maximum LAX noncompacted: compacted myocardial ratio ≥2 underwent SAXSYST, SAXDIAST and NCMASS measurements. A lower ratio threshold of ≥2 was chosen to widen the population sampled to ensure adequate capture of all individuals likely to meet any of the diagnostic criteria. Those who met all 4 diagnostic criteria for LVNC were taken as demonstrating the LVNC phenotype. The Central Illustration demonstrates the measurements performed using the 4 techniques. All analysis was performed using commercial software (Argus, Siemens Multi-Modality Work Platform, version VB 15, Seimens) by 1 of 2 observers. Fifteen study datasets were read by both observers, with 1 observer reading them twice to calculate intraobserver and interobserver variability for each of the 4 measures.

ues. All analysis was performed using commercial software (Argus, Siemens Multi-Modality Work Platform, version VB 15, Seimens) by 1 of 2 observers. Fifteen study datasets were read by both observers, with 1 observer reading them twice to calculate intraobserver and interobserver variability for each of the 4 measures. Statistical analysis Data are expressed as mean ± SD for continuous variables, median (range) for ordinal variables and number of patients (%) for nominal variables. Normality tests were performed; if the test failed, where possible standard transformations such as square root, reciprocal, or logarithmic transforms were used to generate a Gaussian distribution. An independent sample Student t test was used to test the null hypothesis that samples originate from the same source. Chi-square or Fisher’s exact tests were used as appropriate to compare nominal data. Two-way mixed, absolute agreement, average measure intraclass correlation coefficients (ICC) for each of the 4 measures of noncompaction were determined with ICC >0.75 = excellent, 0.4 to 0.75 = good, <0.40 = poor, and <0.20 = slight. All data were analyzed using SPSS statistical package (version 21.0, IBM SPSS, Chicago, Illinois). Significance was adjusted for multiple comparisons using a Bonferroni correction. Results Of the 1,528 volunteers who completed the imaging protocol, 48 were excluded due to inadequate image quality. This resulted in 1,480 (age 54.1 ± 8.3 years, 38% male) undergoing complete imaging with diagnostic quality images allowing measurement of all 4 measures of noncompaction.

Statistical analysis Data are expressed as mean ± SD for continuous variables, median (range) for ordinal variables and number of patients (%) for nominal variables. Normality tests were performed; if the test failed, where possible standard transformations such as square root, reciprocal, or logarithmic transforms were used to generate a Gaussian distribution. An independent sample Student t test was used to test the null hypothesis that samples originate from the same source. Chi-square or Fisher’s exact tests were used as appropriate to compare nominal data. Two-way mixed, absolute agreement, average measure intraclass correlation coefficients (ICC) for each of the 4 measures of noncompaction were determined with ICC >0.75 = excellent, 0.4 to 0.75 = good, <0.40 = poor, and <0.20 = slight. All data were analyzed using SPSS statistical package (version 21.0, IBM SPSS, Chicago, Illinois). Significance was adjusted for multiple comparisons using a Bonferroni correction. Results Of the 1,528 volunteers who completed the imaging protocol, 48 were excluded due to inadequate image quality. This resulted in 1,480 (age 54.1 ± 8.3 years, 38% male) undergoing complete imaging with diagnostic quality images allowing measurement of all 4 measures of noncompaction. The average maximum LAX noncompacted ratio within the entire cohort was 1.78 ± 0.63. A total of 296 (20.0%) of 1,482 analyzed datasets demonstrated an LAX ratio of ≥2 and were therefore included in subsequent analysis. Of the 296 who underwent all 4 diagnostic tests for LVNC, 219 (74%) met at least 1 diagnostic criterion for LVNC, 117 (39.5%) met 2 criteria, 63 (21.3%) met 3 criteria, and 19 (6.4%) met all 4 diagnostic criteria (Table 1). A total of 186 (62.8%) met the LAX criteria with the most common location for the maximum LAX noncompaction ratio found in the apical lateral wall. A total of 106 met the SAXDIAST (35.8%) criterion, with the most common site of maximum noncompaction found in the apical lateral wall (segment 16). A total of 65 (22%) met the SAXSYST criterion, with the most common site of maximum noncompaction found in the apical lateral wall (segment 16). A total of 61 (20.6%) met the NCMASS criteria. Thus, 14.8% of the normal population met at least 1 of the current CMR criteria for LVNC, whereas 1.3% met all 4 of the proposed diagnostic criteria for LVNC.

XSYST criterion, with the most common site of maximum noncompaction found in the apical lateral wall (segment 16). A total of 61 (20.6%) met the NCMASS criteria. Thus, 14.8% of the normal population met at least 1 of the current CMR criteria for LVNC, whereas 1.3% met all 4 of the proposed diagnostic criteria for LVNC. Those who met all 4 of the diagnostic criteria (and were therefore considered in our study to exhibit the LVNC phenotype) demonstrated no significant differences in demographics, allometric measures, or cardiovascular risk factors (Table 2). They did however show significantly lower LV mass index (LVMI) (36.1 ± 9.2 g/m1.7 vs. 42.5 ± 9.5 g/m1.7, p = 0.004), higher LV end systolic volumes (LVESV) (20.6 ± 6.1 g/m1.7 vs. 17.1 ± 5.5 ml/m1.7; p = 0.006), lower ejection fraction (EF) (64.7 ± 9.2% vs. 69.0 ± 6.5%; p = 0.005), and lower LV mass volume ratio (LVMVR) (0.62 ± 0.10 g/ml vs. 0.79 ± 0.15 g/ml; p < 0.001) (Table 3). A significant but weak inverse correlation was seen between systolic blood pressure and the LAX noncompaction ratio (B = −0.004; p = 0.001) with an inverse correlation observed between the degree of myocardial noncompaction and LV mass (B = −0.006; p < 0.001) (Table 4).

g/ml vs. 0.79 ± 0.15 g/ml; p < 0.001) (Table 3). A significant but weak inverse correlation was seen between systolic blood pressure and the LAX noncompaction ratio (B = −0.004; p = 0.001) with an inverse correlation observed between the degree of myocardial noncompaction and LV mass (B = −0.006; p < 0.001) (Table 4). Repeatability for the LAX measures was good for intraobserver repeated measures (ICC: 0.59; 95% confidence interval [CI]: -0.28 to 0.87), and poor for interobserver measures (ICC: 0.28; 95% CI: −1.3 to 0.76). Repeatability for the SAXDIAS measures was good for intraobserver repeated measures (ICC: 0.65; 95% CI: −7.42 to 0.57), and good for interobserver measures (ICC: 0.73; 95% CI: -0.18 to 0.93). Repeatability for the SAXSYST measures was only slight for intraobserver repeated measures (ICC: 0.19; 95% CI: −1.92 to 0.77), and good for interobserver measures (ICC: 0.50; 95% CI: −0.49 to 0.87). Repeatability for the NCMASS measures was good for intraobserver repeated measures (ICC: 0.70; 95% CI: −0.08 to 0.92), and excellent for interobserver measures (ICC: 0.88; 95% CI: −0.51 to 0.97). Discussion In our study, almost 15% of the population meet at least 1 of the current CMR diagnostic criteria for LVNC, and 1.3% of an asymptomatic population free from known CVD meet all 4 current CMR criteria. In addition, we demonstrated that the presence of LVNC is not associated with demographics, body shape, or biochemical markers of CVD.

almost 15% of the population meet at least 1 of the current CMR diagnostic criteria for LVNC, and 1.3% of an asymptomatic population free from known CVD meet all 4 current CMR criteria. In addition, we demonstrated that the presence of LVNC is not associated with demographics, body shape, or biochemical markers of CVD. Our findings are comparable with previous work in the MESA (Multi Ethnic Study of Atherosclerosis) population, in which an LAX noncompaction ratio >2.3 was observed in 43% of 323 participants free from cardiac disease and hypertension (8). The lower incidence in our cohort is likely due to 2 factors. The first is the additional use of the LV outflow tract LAX view of the LV in the MESA cohort, thereby increasing the likelihood of detecting a region of greater noncompaction using 3 LAX views compared with 2 views alone. The second is the multiethnic cohort examined in this previous study, since there is a greater noncompacted mass in healthy blacks compared with healthy whites 6, 16. We have thus validated the previous observations made in the MESA cohort within a second, much larger population study, and further developed and strengthened the original observations demonstrating that even when alternate or more stringent combined criteria are used, a significant proportion of the general population would still be considered to have LVNC. A significant but weak correlation was seen between systolic blood pressure and LAX noncompaction ratios, consistent with prior observations by others (17). No correlation was seen between allometric measures and noncompaction ratios, suggesting that the presence and thickness of trabeculations are not determined by body size or composition. In our study we observed that those with LVNC had a higher ESV with a lower LV mass (LVM) and EF. Previous work in the Framingham study has shown that inclusion of the trabeculae within the myocardial mass contours results in a significant increase in LVM and a decrease in LV volumes consistent with our observations (18). Thus, these findings are most likely due to the technique used for the measurements of mass, volume, and function in the current study in which trabeculations were included in the blood pool rather than within the LVM. However, follow-up of this cohort is required to confirm that this is the case and that these findings are not indicative of early pathological changes.

hnique used for the measurements of mass, volume, and function in the current study in which trabeculations were included in the blood pool rather than within the LVM. However, follow-up of this cohort is required to confirm that this is the case and that these findings are not indicative of early pathological changes. Interestingly, although a difference in LV metrics was observed between groups when only those meeting all 4 criteria were looked at, only a very weak correlation was seen between the noncompaction ratio and LV measures when only the LAX measure was used. This suggests that LAX noncompaction is not only the least specific criterion (resulting in the most over diagnosis) in our study cohort, but also has limited implications for LV remodeling.

ked at, only a very weak correlation was seen between the noncompaction ratio and LV measures when only the LAX measure was used. This suggests that LAX noncompaction is not only the least specific criterion (resulting in the most over diagnosis) in our study cohort, but also has limited implications for LV remodeling. The observation of high incidence of LVNC in 2 separate population studies indicates 1 of 2 possibilities. One is that the current diagnostic criteria lack specificity for the accurate identification and diagnosis of LVNC, with resultant extensive over diagnosis in normal individuals. Indeed, the poor correlation between the measures, with 15% of the study cohort meeting at least 1 criterion but <2% meeting all 4, suggest that the feature they are measuring is poorly captured by any 1 of the techniques. This is in keeping with previous observations using echocardiographic diagnostic criteria in which, in a population with known heart failure and diagnosis of LVNC, only 29.8% met all 3 criteria, whereas 36.3% fulfilled only 1 criterion (19). In the current study we have not assessed the use of fractal analysis, which has been previously described to better differentiate healthy volunteers from those with pathological LVNC (20). While this holds some potential, it must be noted that the cohort in whom they demonstrated a lack of over-diagnosis was free from hypertension and nonobese. Because both of these increase trabecular complexity, its ability to differentiate a typical patient presenting with shortness of breath and both of these comorbidities from true LVNC remains to be proven. Finally, trabecular complexity using fractal analysis in both gene-negative and gene-positive hypertrophic cardiomyopathy is both within the same diagnostic range seen in LVNC. Thus, its specificity in the diagnosis of LVNC is questionable (10). One potential solution has been proposed that moves from a purely imaging-based diagnosis to a diagnosis that is more holistic and closer to that of arrhythmogenic right ventricular cardiomyopathy — requiring, in addition to meeting imaging criteria either a family member with LVNC, a regional wall motion abnormality, LVNC-related complications (arrhythmia, heart failure, or thromboembolism), or carrier status of a genetic mutation known to be associated with LVNC (21).

thmogenic right ventricular cardiomyopathy — requiring, in addition to meeting imaging criteria either a family member with LVNC, a regional wall motion abnormality, LVNC-related complications (arrhythmia, heart failure, or thromboembolism), or carrier status of a genetic mutation known to be associated with LVNC (21). The second possibility is that noncompaction is an anatomical phenotype rather than a pathological cardiomyopathy. At 10-year follow-up of the aforementioned MESA study, those who met the Peterson et al. (7) criterion for LVNC demonstrated no significant difference in LVEF over the follow-up period, nor any difference in cardiovascular events compared to those who did not meet the criterion (22). Planned 5- and 10-year follow-up within our TASCFORCE study group will provide further useful information on the clinical impact of noncompaction within this population. It may simply be that those currently diagnosed with LVNC are those with the anatomical LVNC phenotype who subsequently develop dilated or hypertrophic cardiomyopathy. The argument in favor of this is strengthened by a recent study in patients with heart failure that demonstrated a lack of significant association between noncompaction ratios and subsequent major adverse cardiovascular events (9).

The study sponsor and funder (Novartis) participated in discussions about the design and conduct of this study; they also provided the drugs used in the trial and logistical support for its execution. The trial design, endpoints, and statistical analyses were largely derived from the academic investigators’ previously published studies 19, 23, 24. Following the final database lock, all patient data were analyzed independently by the Centre for Statistics in Medicine, Oxford (J.B.). The manuscript was drafted by the academic investigators (R.P.C., J.S.B., J.-C.T., and Z.A.F.), in accordance with the written agreement between Novartis and the academic institutions, and reviewed and revised by the writing committee. All authors had full access to all the data in the study and assume responsibility for publication. All statistical analyses were performed by using Stata 14 (StataCorp LP, College Station, Texas). Results Of 450 patients screened, 189 were randomized to receive either placebo (n = 94) or canakinumab 150 mg (n = 95). The proportions of patients with diabetes, duration of diabetes, and glycemic control (estimated from HbA1c) were similar in the groups, and there was a high prevalence for each of hypertension, dyslipidemia, and background coronary artery disease (Table 1); the latter was slightly higher proportionately in the placebo group. There was no significant difference between the canakinumab group compared with the placebo group for the overall rate of completion of the study, which was 70.5% with canakinumab versus 77.7% for placebo (risk ratio: 1.32; 95% confidence interval [CI]: 0.81 to 2.15; p = 0.26) or the rate of discontinuation due to adverse events (14.7% vs. 11.7%; risk ratio: 1.26; 95% CI: 0.60 to 2.63; p = 0.54) (Table 2). The most common cause for discontinuation was for adverse events. A full list of adverse events is provided in Online Table 1. Seven patients withdrew consent, and 7 patients were excluded from analysis due to significant protocol deviation.

atomical LVNC phenotype who subsequently develop dilated or hypertrophic cardiomyopathy. The argument in favor of this is strengthened by a recent study in patients with heart failure that demonstrated a lack of significant association between noncompaction ratios and subsequent major adverse cardiovascular events (9). Study limitations First, we only conducted a full analysis of all diagnostic criteria in those with an LAX ratio of ≥2, and thus may have underestimated the total number who may have met 1 or more of the other 3 diagnostic criteria. However, this is only likely to further strengthen our observation of overdiagnosis if more participants without this criterion happened to meet 1 of the other 3 criteria. Second, a selection criterion for recruitment into the imaging arm of the TASCFORCE study was BNP above the gender specific median. Given the known association between BNP and heart failure, this could bias the results towards detecting a higher prevalence of a phenotype that is traditionally associated with heart failure and a poor clinical outcome. It could also increase the prevalence of those with diastolic dysfunction, which may in turn affect trabeculation quantification if there is a significant remodeling epiphenomenon component to these measures. No previous reports have described an association between these 2, nor is there any evidence of absence of association. However, of some reassurance, we did not observe any correlation between BNP and noncompaction ratios, nor were BNP levels significantly higher in those who met all 4 criteria, suggesting the impact of any potential bias of this is likely to be small.

etween these 2, nor is there any evidence of absence of association. However, of some reassurance, we did not observe any correlation between BNP and noncompaction ratios, nor were BNP levels significantly higher in those who met all 4 criteria, suggesting the impact of any potential bias of this is likely to be small. Conclusions A significant proportion of an asymptomatic population free from CVD satisfy all currently used CMR diagnostic criteria for LVNC, suggesting that either these all have poor specificity for LVNC, or that LVNC is an anatomical phenotype rather than a distinct cardiomyopathy.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: The current cardiac MRI criteria for diagnosis of LVNC lead to over-representation of its frequency in asymptomatic patients without other manifestations of heart disease. This suggests that LVNC is an anatomical phenotype rather than a distinct cardiomyopathy. TRANSLATIONAL OUTLOOK: Further studies are needed to validate more specific, comprehensive criteria beyond simple anatomical measures on cardiac imaging that identify patients with LVNC at risk of developing adverse clinical events such as arrhythmia, heart failure, or thromboembolism.

Conclusions A significant proportion of an asymptomatic population free from CVD satisfy all currently used CMR diagnostic criteria for LVNC, suggesting that either these all have poor specificity for LVNC, or that LVNC is an anatomical phenotype rather than a distinct cardiomyopathy.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: The current cardiac MRI criteria for diagnosis of LVNC lead to over-representation of its frequency in asymptomatic patients without other manifestations of heart disease. This suggests that LVNC is an anatomical phenotype rather than a distinct cardiomyopathy. TRANSLATIONAL OUTLOOK: Further studies are needed to validate more specific, comprehensive criteria beyond simple anatomical measures on cardiac imaging that identify patients with LVNC at risk of developing adverse clinical events such as arrhythmia, heart failure, or thromboembolism. The present study was funded by the Souter Charitable Foundation and the Chest, Heart and Stroke Scotland Charity. Dr. Weir-McCall is supported by the Wellcome Trust through the Scottish Translational Medicine and Therapeutics Initiative (Grant no. WT 085664) in the form of a Clinical Research Fellowship. Neither group had any role in: study design, the collection, analysis, and interpretation of data; in the writing of the manuscript; nor in the decision to submit the manuscript for publication. Dr. Houston has received grants from Guerbet; and is nonexecutive director for Tayside Flow Technologies. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

a; in the writing of the manuscript; nor in the decision to submit the manuscript for publication. Dr. Houston has received grants from Guerbet; and is nonexecutive director for Tayside Flow Technologies. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. Central Illustration Measurement of Myocardial Noncompaction Using Each of the 4 Techniques (A, B) Images demonstrate long axis noncompaction ratio measurement (orange line = compacted myocardium, blue line = noncompacted myocardium) with a maximum long axis noncompaction ratio of 3.4 obtained in the anterior apical wall. (C, D) Images show short axis noncompaction measurements are demonstrated at diastole (C) where the maximum noncompaction ratio = 3.6 and systole (D) where the maximum noncompaction ratio = 2.2. (E, F) Images delineate compacted and total myocardial mass contours giving a noncompacted mass of 24% of the total mass. Figure 1 Consolidated Standards of Reporting Trials Flow Diagram of the Tayside Screening for Cardiac Events Study Diagram describes the recruitment, exclusions, final study numbers, and planned follow-up. BNP = B-type natriuretic peptide; CV = cardiovascular; LVNC = left ventricular noncompaction; MRI = magnetic resonance imaging. Table 1 Breakdown of the Currently Used Cardiac Magnetic Resonance Imaging Diagnostic Criteria for Left Ventricular Noncompaction in the Whole Population and by Sex

Diagram describes the recruitment, exclusions, final study numbers, and planned follow-up. BNP = B-type natriuretic peptide; CV = cardiovascular; LVNC = left ventricular noncompaction; MRI = magnetic resonance imaging. Table 1 Breakdown of the Currently Used Cardiac Magnetic Resonance Imaging Diagnostic Criteria for Left Ventricular Noncompaction in the Whole Population and by Sex Long-Axis Short-Axis Diastole Short-Axis Systole Noncompacted Mass All 4 Criteria Total (n = 1,480) 186 (12.6) 106 (7.2) 65 (4.4) 61 (4.1) 19 (1.3) Male (n = 565) 71 (12.6) 42 (7.4) 20 (3.5) 23 (4.1) 6 (1.1) Female (n = 915) 115 (12.6) 64 (7.0) 45 (4.9) 38 (4.2) 13 (1.4) Values are n (%). Table 2 Comparison of Cohort Characteristics Between Those Meeting 1, 2, 3, or 4 Left Ventricular Noncompaction Criteria

Long-Axis Short-Axis Diastole Short-Axis Systole Noncompacted Mass All 4 Criteria Total (n = 1,480) 186 (12.6) 106 (7.2) 65 (4.4) 61 (4.1) 19 (1.3) Male (n = 565) 71 (12.6) 42 (7.4) 20 (3.5) 23 (4.1) 6 (1.1) Female (n = 915) 115 (12.6) 64 (7.0) 45 (4.9) 38 (4.2) 13 (1.4) Values are n (%). Table 2 Comparison of Cohort Characteristics Between Those Meeting 1, 2, 3, or 4 Left Ventricular Noncompaction Criteria Criteria Met 0 1 2 3 4 p Value∗ N 1,262 102 54 44 19 Male 480 (38) 47 (46) 19 (35) 16 (36) 6 (32) 0.64 Age, yrs 54.2 ± 8.2 53.9 ± 7.9 53.4 ± 9.0 53.5 ± 9.7 54.1 ± 8.6 0.97 Pulse, beats/min 63.4 ± 9.3 64.3 ± 16.1 62.3 ± 7.9 63.4 ± 10.7 60.3 ± 6.0 0.038 Systolic BP, mm Hg 123 ± 12 123 ± 12 120 ± 11 121 ± 11 118 ± 13 0.10 Diastolic BP, mm Hg 73 ± 9 73 ± 9 71 ± 9 70 ± 8 71 ± 9 0.44 ASSIGN 10-yr risk score 9.4 ± 6.7 8.9 ± 5.6 8.1 ± 5.5 9.4 ± 7.9 8.1 ± 5.2 0.42 Height, m 1.68 ± 0.09 1.68 ± 0.09 1.68 ± 0.08 1.70 ± 0.10 1.67 ± 0.09 0.69 Weight, kg 75.5 ± 14.5 76.1 ± 13.4 75.1 ± 14.1 75.3 ± 13.1 69.9 ± 12.1 0.09 BMI, kg/m2 26.8 ± 4.3 26.9 ± 3.5 26.7 ± 4.6 26.1 ± 3.7 25.0 ± 3.0 0.019 Current smoker 152 (12) 9 (9) 5 (9) 3 (7) 0 (0) 0.15 Ex smoker 328 (26) 30 (29) 15 (28) 18 (39) 5 (26) 1.00 Nonsmoker 482 (62) 61 (61) 34 (63) 23 (52) 14 (68) 0.64 Smoking pack-yrs 6.0 ± 11.7 8.2 ± 16.5 4.7 ± 9.1 4.6 ± 8.2 6.3 ± 12.9 0.90 FH of CVD 328 (26) 26 (25) 13 (24) 12 (27) 6 (32) 0.60 Total cholesterol, mmol/l 5.49 ± 0.98 5.28 ± 0.78 5.38 ± 0.85 5.60 ± 1.21 5.33 ± 0.96 0.48 LDL cholesterol, mmol/l 3.40 ± 0.88 3.27 ± 0.76 3.27 ± 0.72 3.51 ± 1.11 3.43 ± 0.81 0.88 HDL cholesterol, mmol/l 1.44 ± 0.43 1.39 ± 0.42 1.46 ± 0.38 1.43 ± 0.42 1.44 ± 0.37 0.95 Triglycerides, mmol/l 1.48 ± 0.86 1.39 ± 0.79 1.51 ± 1.01 1.51 ± 0.91 1.19 ± 0.59 0.15 Random glucose, mmol/l 5.18 ± 0.92 5.18 ± 0.69 4.99 ± 0.81 5.25 ± 0.97 5.38 ± 0.42 0.56 BNP, pg/ml 27.5 ± 15.6 24.1 ± 13.8 29.4 ± 20.5 32.8 ± 27.3 31.0 ± 23.4 0.39 Values are N, n (%) or mean ± SD. N for diagnostic criteria met is mutually exclusive, and based on the maximum number of criteria met by each study participant.

m glucose, mmol/l 5.18 ± 0.92 5.18 ± 0.69 4.99 ± 0.81 5.25 ± 0.97 5.38 ± 0.42 0.56 BNP, pg/ml 27.5 ± 15.6 24.1 ± 13.8 29.4 ± 20.5 32.8 ± 27.3 31.0 ± 23.4 0.39 Values are N, n (%) or mean ± SD. N for diagnostic criteria met is mutually exclusive, and based on the maximum number of criteria met by each study participant. ASSIGN = assessing cardiovascular risk using Scottish Intercollegiate Guidelines Network; BMI = body mass index; BP = blood pressure; BNP = B-type natriuretic peptide; CVD = cardiovascular disease; FH = family history; HDL = high-density lipoprotein; LDL = low-density lipoprotein. ∗ p values are derived from comparing those meeting 0 criteria and those meeting all 4 criteria, with binary outcome variables compared using the Fisher exact test. Significance level set at p = 0.0025 after Bonferroni correction for multiple comparisons. Table 3 Comparison of Left Ventricular Measures Between Those Meeting 1, 2, 3, or 4 Left Ventricular Noncompaction Criteria Criteria Met 0 1 2 3 4 p Value LVM, g/m1.7 42.5 ± 9.5 42.9 ± 10.2 39.8 ± 10.2 39.9 ± 8.3 36.1 ± 9.2 0.004 LVEDV, ml/m1.7 54.5 ± 9.7 55.8 ± 9.2 56.0 ± 12.2 57.5 ± 10.5 58.7 ± 2.7 0.069 LVESV, ml/m1.7 17.1 ± 5.5 17.7 ± 5.3 18.1 ± 5.4 18.6 ± 5.7 20.6 ± 6.1 0.006 LVSV, ml/m1.7 37.4 ± 6.5 38.1 ± 6.0 37.9 ± 8.4 38.9 ± 6.4 38.0 ± 9.7 0.78 LVEF, % 69.0 ± 6.5 68.7 ± 6.1 67.8 ± 5.5 68.1 ± 5.9 64.7 ± 9.2 0.005 LVMVR, g/ml 0.79 ± 0.15 0.77 ± 0.14 0.72 ± 0.11 0.70 ± 0.13 0.62 ± 0.10 <0.001 Values are mean ± SD.

LVESV, ml/m1.7 17.1 ± 5.5 17.7 ± 5.3 18.1 ± 5.4 18.6 ± 5.7 20.6 ± 6.1 0.006 LVSV, ml/m1.7 37.4 ± 6.5 38.1 ± 6.0 37.9 ± 8.4 38.9 ± 6.4 38.0 ± 9.7 0.78 LVEF, % 69.0 ± 6.5 68.7 ± 6.1 67.8 ± 5.5 68.1 ± 5.9 64.7 ± 9.2 0.005 LVMVR, g/ml 0.79 ± 0.15 0.77 ± 0.14 0.72 ± 0.11 0.70 ± 0.13 0.62 ± 0.10 <0.001 Values are mean ± SD. LVEDV = left ventricular end-diastolic volume; LVEF = left ventricular ejection fraction; LVESV = left ventricular end-systolic volume; LVM = left ventricular mass; LVMVR = left ventricular mass volume ratio; LVSV = left ventricular stroke volume. Table 4 Linear Regression Coefficients Change in the “Maximum Long Axis Noncompaction Ratio” Per Unit Increase in Demographic, Biochemical, and Cardiac Magnetic Resonance Imaging Measures

LVEDV = left ventricular end-diastolic volume; LVEF = left ventricular ejection fraction; LVESV = left ventricular end-systolic volume; LVM = left ventricular mass; LVMVR = left ventricular mass volume ratio; LVSV = left ventricular stroke volume. Table 4 Linear Regression Coefficients Change in the “Maximum Long Axis Noncompaction Ratio” Per Unit Increase in Demographic, Biochemical, and Cardiac Magnetic Resonance Imaging Measures B SE Intercept p Value N 1,480 Sex 0.011 0.034 1.75 0.75 Age, yrs −0.001 0.002 1.78 0.79 Pulse, beats/min 0.002 0.002 1.87 0.26 Systolic BP, mm Hg −0.004 0.001 2.29 0.001 Diastolic BP, mm Hg −0.003 0.002 1.99 0.073 ASSIGN risk score, % −0.003 0.003 1.78 0.28 Height, m 0.12 0.18 1.55 0.50 Weight, kg 0.00 0.001 1.76 0.91 BMI, kg/m2 −0.002 0.004 1.81 0.60 Smoking pack-yrs 0.00 0.001 1.76 0.93 Total cholesterol, mmol/l −0.02 0.017 1.89 0.16 LDL-cholesterol, mmol/l −0.006 0.039 1.85 0.75 HDL-cholesterol, mmol/l −0.06 0.02 1.78 0.11 Triglycerides, mmol/l −0.02 0.019 1.78 0.36 Random glucose, mmol/l 0.008 0.027 1.72 0.77 BNP, pg/ml 0.000 0.001 1.75 0.95 LVM, g/m2 −0.006 0.002 2.02 <0.001 LVEDV, ml/m2 0.005 0.002 1.48 0.003 LVESV, ml/m2 0.008 0.003 1.63 0.012 LVSV, ml/m2 0.006 0.003 1.54 0.02 LVEF, % −0.004 0.003 2.00 0.17 LVMVR, g/ml −0.78 0.11 2.37 <0.001 B = gradient; SE = standard error; other abbreviations as in Tables 2 and 3.

Atherosclerosis is well-established as a disease with an important inflammatory component 1, 2, 3. Systemic markers of inflammation such as C-reactive protein and serum amyloid A are strongly related to cardiovascular prognosis in various populations and clinical settings 4, 5. Furthermore, therapeutic interventions that reduce cardiovascular risk have also been associated with a reduction in systemic inflammatory markers 6, 7. However, whether specifically targeting inflammation reduces cardiovascular risk remains unknown. Interleukins are important mediators of inflammation, both locally and systemically. Macrophages are key cellular components of atherosclerotic plaque and produce interleukin (IL)-1β (8), which is also promoted by cellular cholesterol activation of inflammasomes (9). IL-1β and interleukin-1α exert proinflammatory effects that are inhibited by the endogenous antagonist interleukin-1 receptor antagonist (IL-1RA). Atherosclerosis-prone mice that are deficient in IL-1β develop smaller lesions (10), and administration of IL-1RA reduces early atherogenesis in mice (11). IL-1RA–deficient mice have shown increased atherosclerosis (12) and vascular inflammation, associated with destruction of elastic tissues (13).

antagonist (IL-1RA). Atherosclerosis-prone mice that are deficient in IL-1β develop smaller lesions (10), and administration of IL-1RA reduces early atherogenesis in mice (11). IL-1RA–deficient mice have shown increased atherosclerosis (12) and vascular inflammation, associated with destruction of elastic tissues (13). Therefore, given the key role for IL-1β as a mediator of innate immunity and the effects of interleukin inhibition in experimental atherosclerosis, interventions to reduce inflammation through IL-1β have been proposed to treat atherosclerosis. Although evidence of benefit to vascular disease in humans remain sparse, administration of the IL-1RA anakinra to patients with rheumatoid arthritis improved several measures of vascular function, including aortic distensibility, flow-mediated vasodilation, and coronary flow reserve (14).

sed to treat atherosclerosis. Although evidence of benefit to vascular disease in humans remain sparse, administration of the IL-1RA anakinra to patients with rheumatoid arthritis improved several measures of vascular function, including aortic distensibility, flow-mediated vasodilation, and coronary flow reserve (14). In addition to its key role in vascular disease, IL-1β has been implicated in the pathogenesis of type 2 diabetes mellitus (T2DM). IL-1RA expression is reduced in pancreatic islets of patients with T2DM, and high glucose concentrations induce the production of IL-1β in human pancreatic beta cells, leading to impaired insulin secretion, decreased cell proliferation, and apoptosis (15). Blockade of the interleukin-1 receptor with anakinra improved glycemia and beta-cell secretory function and reduced markers of systemic inflammation in patients with T2DM (16). Patients with T2DM are at high risk for cardiovascular disease (17) and have evidence of both increased plaque inflammation (18) and reduced arterial distensibility (19). This group might, therefore, derive “metabolic” and “vascular” benefits from targeting IL-1β, including reduced risk of atherothrombotic complications.

tients with T2DM are at high risk for cardiovascular disease (17) and have evidence of both increased plaque inflammation (18) and reduced arterial distensibility (19). This group might, therefore, derive “metabolic” and “vascular” benefits from targeting IL-1β, including reduced risk of atherothrombotic complications. A human monoclonal anti-human IL-1β antibody of the immunoglobulin G1/k isotype canakinumab functionally neutralizes IL-1β through steric hindrance of its receptor interaction. It is effective in reducing systemic markers of inflammation, including C-reactive protein and IL-6 (20). Its effects on cardiovascular outcomes are under investigation in the CANTOS (Canakinumab Anti-inflammatory Thrombosis Outcomes Study) trial (21). Vascular magnetic resonance imaging (MRI) has emerged as a precise, highly reproducible, and versatile tool to assess both vascular structure and function at multiple arterial loci 22, 23, 24, 25. Accordingly, we designed a randomized, placebo-controlled Phase II clinical trial to test the effects of IL-1β inhibition, using canakinumab, on: 1) MRI-derived measures of vascular structure and function; 2) measures of diabetes control; and 3) indicators of systemic inflammation in patients with atherosclerotic vascular disease and impaired glucose tolerance (IGT) or T2DM.

Phase II clinical trial to test the effects of IL-1β inhibition, using canakinumab, on: 1) MRI-derived measures of vascular structure and function; 2) measures of diabetes control; and 3) indicators of systemic inflammation in patients with atherosclerotic vascular disease and impaired glucose tolerance (IGT) or T2DM. Methods Novartis Pharmaceuticals (Cambridge, Massachusetts) initiated this Phase II, double-blind, randomized, placebo-controlled trial, with the final study protocol designed in collaboration with the investigators (R.P.C., J.-C.T., and Z.A.F.), based on their previously published methods 19, 23, 24. The study was undertaken at 9 centers in Canada, the United Kingdom, the United States, Germany, and Israel, in compliance with the principles of the Declaration of Helsinki and according to Good Clinical Practice guidelines. The protocol was reviewed and approved by the institutional review board, or equivalent, for each center. All participants provided written informed consent before undertaking any study procedures.

rael, in compliance with the principles of the Declaration of Helsinki and according to Good Clinical Practice guidelines. The protocol was reviewed and approved by the institutional review board, or equivalent, for each center. All participants provided written informed consent before undertaking any study procedures. Patients Patients (ages 18 to 74 years) were eligible for inclusion if they had clinically evident atherosclerotic vascular disease: previous myocardial infarction; history of angina; carotid stenosis (>30%); peripheral vascular disease (ankle–brachial index <0.9); endarterectomy >3 months previously; or transient ischemic attack or stroke. In addition, patients must also have had either T2DM (for ≤14 years and glycosylated hemoglobin [HbA1c] levels between 6% and 10%) or IGT (defined as a peak 2-h glucose reading ≥140 mg/dl but <200 mg/dl after an oral glucose tolerance test during screening). Patients were to have been on stable statin therapy for a period of ≥6 weeks before screening (or have physician-documented statin intolerance).

d hemoglobin [HbA1c] levels between 6% and 10%) or IGT (defined as a peak 2-h glucose reading ≥140 mg/dl but <200 mg/dl after an oral glucose tolerance test during screening). Patients were to have been on stable statin therapy for a period of ≥6 weeks before screening (or have physician-documented statin intolerance). After consent, eligible patients were randomized (1:1) to receive canakinumab 150 mg or placebo, subcutaneously, monthly for 12 months. Exclusion criteria included: pregnancy; systemic steroid use; baseline high-sensitivity C-reactive protein (hs-CRP) levels >30 mg/l; history of significant multiple drug allergies; history or evidence of chronic infection, including tuberculosis and liver disease; or a standard contraindication to MRI. Randomization occurred between December 2009 and November 2012. Patient groups were assigned centrally according to a validated computer-generated randomization code, stratified according to glycemic status (T2DM or IGT).

hronic infection, including tuberculosis and liver disease; or a standard contraindication to MRI. Randomization occurred between December 2009 and November 2012. Patient groups were assigned centrally according to a validated computer-generated randomization code, stratified according to glycemic status (T2DM or IGT). Imaging procedures An integrated vascular MRI was performed at baseline and after 3 and 12 months of treatment. If the imaging data did not meet an evaluable standard at any time point, the patient was rescanned. Randomization required an evaluable baseline scan. The imaging procedure comprised measures of aortic wall area and distensibility, as well as carotid wall area bilaterally. The imaging protocol was adapted from Lee et al. (19), and staff at each imaging site underwent individualized training to ensure consistency of method and data acquisition. All trial sites used a 3.0-T whole-body MRI scanner, including Trio, TIM Trio, or Verio (Siemens Medical Solutions USA, Inc., Malvern, Pennsylvania) or Achieva (Philips, Amsterdam, the Netherlands) platforms. For carotid imaging, a bilateral 4-channel carotid array (Machnet B.V., Roden, the Netherlands) was used on the Siemens scanners and an equivalent multi-channel (4 to 8) phased array carotid coil (Shanghai Medical, Shanghai, China) was used on the Philips scanners. Detailed information regarding imaging protocols and analysis is available in the Online Appendix.

Imaging procedures An integrated vascular MRI was performed at baseline and after 3 and 12 months of treatment. If the imaging data did not meet an evaluable standard at any time point, the patient was rescanned. Randomization required an evaluable baseline scan. The imaging procedure comprised measures of aortic wall area and distensibility, as well as carotid wall area bilaterally. The imaging protocol was adapted from Lee et al. (19), and staff at each imaging site underwent individualized training to ensure consistency of method and data acquisition. All trial sites used a 3.0-T whole-body MRI scanner, including Trio, TIM Trio, or Verio (Siemens Medical Solutions USA, Inc., Malvern, Pennsylvania) or Achieva (Philips, Amsterdam, the Netherlands) platforms. For carotid imaging, a bilateral 4-channel carotid array (Machnet B.V., Roden, the Netherlands) was used on the Siemens scanners and an equivalent multi-channel (4 to 8) phased array carotid coil (Shanghai Medical, Shanghai, China) was used on the Philips scanners. Detailed information regarding imaging protocols and analysis is available in the Online Appendix. The velocity at which the arterial pulse propagates is termed pulsed wave velocity (PWV). A measure of arterial stiffness, it is an independent predictor of mortality in both T2DM and IGT (26). The Sphygmocor platform (AtCor Medical Pty. Ltd., West Ryde, Australia) was applied immediately before MRI scanning to acquire the aortic central pulse pressure from radial artery applanation tonometry and the PWV from the carotid-femoral pulse waves.

is an independent predictor of mortality in both T2DM and IGT (26). The Sphygmocor platform (AtCor Medical Pty. Ltd., West Ryde, Australia) was applied immediately before MRI scanning to acquire the aortic central pulse pressure from radial artery applanation tonometry and the PWV from the carotid-femoral pulse waves. Safety assessments Monitoring of vital signs, electrocardiogram, standard hematology, and biochemistry (including measurement of lipoproteins, liver function, and creatine kinase) were conducted throughout the study. A data monitoring committee oversaw subject safety on an ongoing basis. In addition, 3 adjudication committees made blinded assessments of adverse events in relation to cardiac, malignant, and infection-related events. Endpoints The primary efficacy objectives were the effects of the drug on aortic distensibility and total plaque burden in the aorta and carotid arteries; the primary safety objective was the safety and tolerability of canakinumab in this population. Secondary objectives included the effects of canakinumab on aortic PWV; hs-CRP; HbA1c; homeostasis model assessment (HOMA)–insulin resistance; and peak blood glucose level 2 h after an oral glucose challenge. Exploratory analyses of peripheral biomarkers of inflammation (including IL-6, serum amyloid A, and plasma lipoproteins) were also performed.

the effects of canakinumab on aortic PWV; hs-CRP; HbA1c; homeostasis model assessment (HOMA)–insulin resistance; and peak blood glucose level 2 h after an oral glucose challenge. Exploratory analyses of peripheral biomarkers of inflammation (including IL-6, serum amyloid A, and plasma lipoproteins) were also performed. Statistical analyses Sample size was calculated from the study of Lee et al. (19). To detect a 35% change in aortic distensibility or an 11% change in plaque burden at 12 months to achieve a power of 0.8 and nominal p < 0.05 (2-sided), 60 patients per group were required. To ensure 120 datasets with 12-month follow-up data, it was planned to randomize 190 patients. Given the exploratory nature of the study, corrections were not made for the multiplicity of statistical tests. As defined by the protocol, patients were included in the 3-month data analysis if they had no missing doses at 3 months; for the 12-month analysis, participants were required to have no missing doses by 3 months and 1 or no missing doses between 3 and 12 months.

were not made for the multiplicity of statistical tests. As defined by the protocol, patients were included in the 3-month data analysis if they had no missing doses at 3 months; for the 12-month analysis, participants were required to have no missing doses by 3 months and 1 or no missing doses between 3 and 12 months. To compare treatment with placebo, we conducted an analysis of covariance on change from baseline, including the glycemic status as a factor and the baseline as a covariate at 3 and 12 months. All data were checked for normality and log-transformed, if appropriate. When a log transformation of the change from baseline was not possible because of negative values, an analysis of covariance was conducted on the log-transformed 3- and 12-month data, including glycemic index status as a factor and the log-transformed baseline as a covariate. We considered a p value <0.05 as significant. Results are reported as means with 95% confidence limits. Two interim analyses were pre-specified, when n = 60 and all patients had completed 3 months of treatment, respectively, with the intention of halting the study if adverse measures were identified or for futility but not for interim positive efficacy. The interim analyses were performed by independent personnel not directly associated with the study’s conduct.

(14.7% vs. 11.7%; risk ratio: 1.26; 95% CI: 0.60 to 2.63; p = 0.54) (Table 2). The most common cause for discontinuation was for adverse events. A full list of adverse events is provided in Online Table 1. Seven patients withdrew consent, and 7 patients were excluded from analysis due to significant protocol deviation. As a biomarker of atherosclerotic plaque burden, vessel wall area was quantified in the aorta and carotid arteries (Table 3). There was no statistically significant difference in mean carotid wall area between these 2 groups at either time point. Baseline mean carotid wall areas were 27.7 ± 9.79 mm2 and 27.1 ± 9.6 mm2 (p = NS) for the canakinumab and placebo groups, respectively. Change in mean carotid artery wall area was –3.37 mm2 (p = 0.06) after 12 months for canakinumab versus placebo. There was an increase (12 months vs. baseline) in wall area for each of the left and right carotid arteries (and mean of the left and right) with placebo but, on average, neither progression nor regression in the canakinumab group (Figure 1A). Change from baseline at 12 months was compared for individual patients between the left and right carotid arteries for both treatment groups. In each, there were strong correlations between changes in the left and right carotid arteries for canakinumab and placebo, respectively (p < 0.0001 for both) (Figure 1B). The vessel lumen area was not changed by canakinumab treatment. There was no statistically significant difference in wall area between canakinumab treatment and placebo at any of the 3 aortic sites at either 3 or 12 months.

ht carotid arteries for canakinumab and placebo, respectively (p < 0.0001 for both) (Figure 1B). The vessel lumen area was not changed by canakinumab treatment. There was no statistically significant difference in wall area between canakinumab treatment and placebo at any of the 3 aortic sites at either 3 or 12 months. Aortic distensibility was calculated from measurements made at 3 sites in the aorta. There were no statistically significant differences between canakinumab treatment and placebo for change in aortic distensibility, and no significant changes occurred in systolic or diastolic blood pressure at either 3 or 12 months of treatment versus baseline. There were also no significant differences in measures of PWV between these 2 groups at either time point (Table 3). Compared with placebo, canakinumab reduced hs-CRP at 3 months (geometric mean ratio [GMR]: 0.568; 95% CI: 0.436 to 0.740; p < 0.0001) and 12 months (GMR: 0.56; 95% CI: 0.414 to 0.758; p = 0.0002) (Figure 2). Similarly, IL-6 was reduced by canakinumab at the 3-month time point (GMR: 0.580; 95% CI: 0.483 to 0.697; p < 0.0001). Neither serum amyloid A nor adiponectin changed significantly in response to canakinumab treatment compared with placebo (Table 4).

12 months (GMR: 0.56; 95% CI: 0.414 to 0.758; p = 0.0002) (Figure 2). Similarly, IL-6 was reduced by canakinumab at the 3-month time point (GMR: 0.580; 95% CI: 0.483 to 0.697; p < 0.0001). Neither serum amyloid A nor adiponectin changed significantly in response to canakinumab treatment compared with placebo (Table 4). In this study population of patients with T2DM or IGT and near-universal statin use, canakinumab had no effect on either plasma low-density lipoprotein or high-density lipoprotein cholesterol levels compared with placebo at either of the 3- or 12-month time points. Levels of total plasma cholesterol were mildly elevated from baseline in the canakinumab-treated patients at 3 months compared with the placebo group (GMR: 1.120; 95% CI: 1.050 to 1.195; p = 0.0008) but not at 12 months (GMR: 1.084; 95% CI: 0.991 to 1.185; p = 0.08). The elevation in total cholesterol level most likely reflected an elevation in triglyceride-rich lipoproteins, given the coinciding elevation in triglyceride level that also accompanied canakinumab treatment compared with placebo at 3 months (GMR: 1.21; 95% CI: 1.082 to 1.358; p = 0.001) and 12 months (GMR: 1.20; 95% CI: 1.046 to 1.380; p = 0.01). Lipoprotein(a) levels were reduced by canakinumab compared with placebo. Changes from baseline values were as follows: at 3 months, mean change was –3.719 mg/dl (95% CI: –6.809 to –0.628; p = 0.02); at 12 months, mean change was –4.300 mg/dl (95% CI: –8.052 to –0.548; p = 0.025).

In this study population of patients with T2DM or IGT and near-universal statin use, canakinumab had no effect on either plasma low-density lipoprotein or high-density lipoprotein cholesterol levels compared with placebo at either of the 3- or 12-month time points. Levels of total plasma cholesterol were mildly elevated from baseline in the canakinumab-treated patients at 3 months compared with the placebo group (GMR: 1.120; 95% CI: 1.050 to 1.195; p = 0.0008) but not at 12 months (GMR: 1.084; 95% CI: 0.991 to 1.185; p = 0.08). The elevation in total cholesterol level most likely reflected an elevation in triglyceride-rich lipoproteins, given the coinciding elevation in triglyceride level that also accompanied canakinumab treatment compared with placebo at 3 months (GMR: 1.21; 95% CI: 1.082 to 1.358; p = 0.001) and 12 months (GMR: 1.20; 95% CI: 1.046 to 1.380; p = 0.01). Lipoprotein(a) levels were reduced by canakinumab compared with placebo. Changes from baseline values were as follows: at 3 months, mean change was –3.719 mg/dl (95% CI: –6.809 to –0.628; p = 0.02); at 12 months, mean change was –4.300 mg/dl (95% CI: –8.052 to –0.548; p = 0.025). In this population of patients with T2DM (86%) or IGT (14%) and with median baseline HbA1c levels <7%, canakinumab also had no significant effect compared with placebo on fasting blood glucose, HbA1c, HOMA–insulin resistance or HOMA-β, or 2-h glucose, obtained as part of an oral glucose tolerance test.

Lipoprotein(a) levels were reduced by canakinumab compared with placebo. Changes from baseline values were as follows: at 3 months, mean change was –3.719 mg/dl (95% CI: –6.809 to –0.628; p = 0.02); at 12 months, mean change was –4.300 mg/dl (95% CI: –8.052 to –0.548; p = 0.025). In this population of patients with T2DM (86%) or IGT (14%) and with median baseline HbA1c levels <7%, canakinumab also had no significant effect compared with placebo on fasting blood glucose, HbA1c, HOMA–insulin resistance or HOMA-β, or 2-h glucose, obtained as part of an oral glucose tolerance test. Major adverse cardiovascular events occurred in 9.0% of patients, with no significant difference between the active treatment group (11%) compared with the placebo group (7%) (risk ratio: 1.41; 95% CI: 0.56 to 3.56; p = 0.46) (Table 5). A full list of adverse events is given in Online Table 1. Discussion Inflammation contributes to the pathogenesis of vascular dysfunction and atherogenesis, as well as to the complications of atherosclerosis (1). Data from preclinical studies have suggested that inhibiting IL-1β may directly affect atherosclerosis and vascular inflammation (10).

Major adverse cardiovascular events occurred in 9.0% of patients, with no significant difference between the active treatment group (11%) compared with the placebo group (7%) (risk ratio: 1.41; 95% CI: 0.56 to 3.56; p = 0.46) (Table 5). A full list of adverse events is given in Online Table 1. Discussion Inflammation contributes to the pathogenesis of vascular dysfunction and atherogenesis, as well as to the complications of atherosclerosis (1). Data from preclinical studies have suggested that inhibiting IL-1β may directly affect atherosclerosis and vascular inflammation (10). In the present randomized clinical trial of IL-1β inhibition in patients with T2DM or IGT, canakinumab 150 mg monthly reduced blood levels of IL-6 and hs-CRP (Central Illustration). Similar reductions in inflammatory indexes have been reported previously in patients with T2DM (20). Significantly, given the target population, the present study showed that this effect persists even with near-universal statin use. Canakinumab had no effect on fasting glucose, HbA1c, or measures of insulin sensitivity. In common with earlier studies 19, 23, 24, atherosclerosis burden was quantified in the common carotid arteries and the aorta. We found no statistically significant effect of 12 months’ treatment with canakinumab on magnetic resonance–derived measures of vascular structure or function. However, in each of the common carotid arteries individually, and in the combined vessel average, there was a suggestion of possible retarded progression of atherosclerotic burden.

cally significant effect of 12 months’ treatment with canakinumab on magnetic resonance–derived measures of vascular structure or function. However, in each of the common carotid arteries individually, and in the combined vessel average, there was a suggestion of possible retarded progression of atherosclerotic burden. Because atherosclerosis is a systemic disease, we performed a further analysis, comparing changes in the right and left carotid arteries within individual patients. There were strong relationships between changes in the left- versus right-sided vessels evident in both the canakinumab and the placebo groups but with a tendency toward progression bilaterally in the placebo group. As detailed earlier, analyses of wall area were conducted by operators who were blinded to the timing and treatment allocation of the images. Moreover, measurements were made separately on individual slices, but the data presented are for whole arteries. Given this high level of analytical stringency, these consistent correlations supported the technical robustness of the measurements.

rs who were blinded to the timing and treatment allocation of the images. Moreover, measurements were made separately on individual slices, but the data presented are for whole arteries. Given this high level of analytical stringency, these consistent correlations supported the technical robustness of the measurements. In common with previous studies, we chose to quantify vessel wall area in the tubular carotid arteries because they are less susceptible to error in serial measurements due to “volume averaging” effects. Previously described techniques allow plaque lipid quantification with the use of T2 mapping, but these were not current at the time of protocol design (27). Lipid elements may be the most readily mobilized components of atherosclerotic plaque, although the mechanisms by which IL-1β inhibition might affect plaque lipid are not clear. Vascular 18fludeoxyglucose positron emission tomography has been used effectively to evaluate carotid and aortic plaque macrophage activity in clinical trials, and this test may have been an alternative imaging modality 23, 28, 29. However, changes in macrophage function are associated with changes in substrate utilization and mode of energy generation (30). Therefore, using 18fludeoxyglucose positron emission tomography to identify macrophages on the basis of their glycolytic activity may provide only partial insight into the relevant biology with respect to IL-1β inhibition, and could even be misleading (31).

s in substrate utilization and mode of energy generation (30). Therefore, using 18fludeoxyglucose positron emission tomography to identify macrophages on the basis of their glycolytic activity may provide only partial insight into the relevant biology with respect to IL-1β inhibition, and could even be misleading (31). This study focused on changes in vessel wall structure and function. In doing so, we intended to obtain insights into the possible effects of canakinumab on different manifestations, or stages, of vessel wall pathology. Despite this expansive approach, we observed no significant effects on the vessel wall. Given the discordance with animal studies, one should also consider the possibility that previously observed beneficial effects of IL-1β inhibition need not have been realized directly at the arterial level. Accumulating evidence suggests roles for peripheral monocytes in accelerating atherosclerosis in response to inflammatory stimuli (32). Indeed, IL-1β enhances hematopoietic stem cell proliferation and leukocyte production after acute myocardial infarction in mice and is reduced by administration of anti–IL-1β antibodies (33).

ating evidence suggests roles for peripheral monocytes in accelerating atherosclerosis in response to inflammatory stimuli (32). Indeed, IL-1β enhances hematopoietic stem cell proliferation and leukocyte production after acute myocardial infarction in mice and is reduced by administration of anti–IL-1β antibodies (33). In mice, most studies of IL-1β inhibition have shown reduced atherosclerosis and/or plaque inflammation; however, 1 study found that inactivation of IL-1 signaling through loss of the IL-1 receptor type 1 in apolipoprotein E–/– mice promoted multiple indexes of atherosclerotic plaque instability, including reduced plaque smooth muscle cell content, reduced plaque collagen content, and impaired outward vessel remodeling, leading to reduced lumen size (34). Our study in humans found no evidence of changes in aortic characteristics according to distensibility measures at multiple sites. In theory, it would be possible for changes in lumen area/outward remodeling to be present despite no change in wall area, because wall area is derived by subtracting the lumen area from the total vessel area. Examining these measures individually provides information on tendencies for luminal constriction or outward remodeling. In human atherosclerosis, we found no change in lumen area or adverse effect on outward vessel remodeling.

area, because wall area is derived by subtracting the lumen area from the total vessel area. Examining these measures individually provides information on tendencies for luminal constriction or outward remodeling. In human atherosclerosis, we found no change in lumen area or adverse effect on outward vessel remodeling. Canakinumab had no effect on fasting glucose, HbA1c, or measures of insulin sensitivity in the present study. This outcome does not agree with the findings of Larsen et al. (16), who used the IL-1RA anakinra. The apparent discrepancy may reflect better glycemic control at baseline in our study, in which >50% of patients had HbA1c levels <7% compared with a mean HbA1c level >8.5% in the treatment arm of the anakinra trial. However, canakinumab administration did result in an increase in plasma triglyceride levels, as reported previously for this agent (20). The underlying mechanism is not clear. Other treatments targeting inflammation have also been associated with elevated levels of very-low-density lipoprotein triglycerides (35). There was also a marginal but statistically significant elevation in total cholesterol levels, with an increase by 18.5 mg/dl after 3 months’ treatment. Conversely, lipoprotein(a) was significantly reduced by canakinumab at both time points.

iated with elevated levels of very-low-density lipoprotein triglycerides (35). There was also a marginal but statistically significant elevation in total cholesterol levels, with an increase by 18.5 mg/dl after 3 months’ treatment. Conversely, lipoprotein(a) was significantly reduced by canakinumab at both time points. Study limitations This study did not evaluate atherosclerosis at the carotid bifurcation, where the burden of disease is often greatest, nor did it quantify plaque lipid content. Although MRI scans provide highly reproducible measures of plaque structure and vessel function, they offer no direct measure of plaque inflammation. It is possible for drugs to markedly affect plaque biology without altering plaque size, particularly on short-term follow-up (36). In this study, no attempt was made to select subjects on the basis of inflammatory status. More refined patient selection may enhance the effectiveness of drugs that are directed toward specific processes and pathways. Finally, we cannot exclude the possibility that the dose of canakinumab was not high enough to generate a maximal effect on atherosclerotic burden, although the dose was sufficient to lower both hs-CRP and IL-6 levels.

tion may enhance the effectiveness of drugs that are directed toward specific processes and pathways. Finally, we cannot exclude the possibility that the dose of canakinumab was not high enough to generate a maximal effect on atherosclerotic burden, although the dose was sufficient to lower both hs-CRP and IL-6 levels. There were no significant differences in adverse events between the 2 groups, nor were any unexpected adverse outcomes identified. However, this small study cannot thoroughly or systematically evaluate the safety of canakinumab. The question of clinical efficacy clearly remains open. This question will be answered by the Phase III trial, CANTOS, which randomized >10,000 patients with hs-CRP ≥2 mg/l to treatment in a secondary prevention population; the goal is to evaluate the composite primary endpoint of nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death 20, 21.

remains open. This question will be answered by the Phase III trial, CANTOS, which randomized >10,000 patients with hs-CRP ≥2 mg/l to treatment in a secondary prevention population; the goal is to evaluate the composite primary endpoint of nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death 20, 21. Conclusions In patients with T2DM and established cardiovascular disease, canakinumab reduced markers of inflammation (hs-CRP and IL-6) compared with placebo. Treatment with canakinumab also increased levels of triglycerides and total cholesterol but reduced lipoprotein(a) levels. Despite measurable effects on systemic markers of inflammation, there was no statistically significant effect on measures of vascular structure or function. Effects of canakinumab on plaque inflammation or on leukocyte function elsewhere may be undetectable with current imaging technologies. The results of this Phase II trial leave open the important question of clinical efficacy, which will be addressed by the multinational Phase III trial CANTOS.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: Markers of inflammation such as C-reactive protein and serum amyloid A are related to cardiovascular prognosis in patients with atherosclerosis. Canakinumab, a monoclonal antibody that inhibits IL-1β, reduces these markers without measurable effects on arterial structure or function. TRANSLATIONAL OUTLOOK: The therapeutic efficacy of canakinumab in patients with vascular disease and T2DM or glucose intolerance will be assessed in a multicenter clinical trial. Appendix Online Data Online Data

Conclusions In patients with T2DM and established cardiovascular disease, canakinumab reduced markers of inflammation (hs-CRP and IL-6) compared with placebo. Treatment with canakinumab also increased levels of triglycerides and total cholesterol but reduced lipoprotein(a) levels. Despite measurable effects on systemic markers of inflammation, there was no statistically significant effect on measures of vascular structure or function. Effects of canakinumab on plaque inflammation or on leukocyte function elsewhere may be undetectable with current imaging technologies. The results of this Phase II trial leave open the important question of clinical efficacy, which will be addressed by the multinational Phase III trial CANTOS.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: Markers of inflammation such as C-reactive protein and serum amyloid A are related to cardiovascular prognosis in patients with atherosclerosis. Canakinumab, a monoclonal antibody that inhibits IL-1β, reduces these markers without measurable effects on arterial structure or function. TRANSLATIONAL OUTLOOK: The therapeutic efficacy of canakinumab in patients with vascular disease and T2DM or glucose intolerance will be assessed in a multicenter clinical trial. Appendix Online Data Online Data Acknowledgments The authors acknowledge with gratitude the contributions of Dr. Kimberley Bailey, Dr. Peter Bernhardt, Dr. Thomas Forst, Ms. Megan Gaston, Ms. Jamie Hutchinson, Dr. Tomas Jax, Dr. Alistair Lindsay, Dr. Nirmala Nanguneri, Dr. Eli Rothj, and Dr. Dain Wahl.

TRANSLATIONAL OUTLOOK: The therapeutic efficacy of canakinumab in patients with vascular disease and T2DM or glucose intolerance will be assessed in a multicenter clinical trial. Appendix Online Data Online Data Acknowledgments The authors acknowledge with gratitude the contributions of Dr. Kimberley Bailey, Dr. Peter Bernhardt, Dr. Thomas Forst, Ms. Megan Gaston, Ms. Jamie Hutchinson, Dr. Tomas Jax, Dr. Alistair Lindsay, Dr. Nirmala Nanguneri, Dr. Eli Rothj, and Dr. Dain Wahl. This work was funded by Novartis. Dr. Choudhury has received honoraria/consultancy fees from Amgen, AstraZeneca, Boehringer Ingelheim, Isis Pharmaceuticals, GlaxoSmithKline, Merck, Roche, and Sanofi. Dr. Mani has received grant support from Novartis, Daiichi-Sankyo, Amgen, and Aegerion; has received honoraria from Aegerion; and he has served as a consultant for MedLion Inc. Dr. McLaughlin has received research support from AbbVie. Drs. Basson, Svensson, Zhang, and Yates are employees of and hold equity shares in Novartis. Dr. Tardif has received research support from Novartis, AstraZeneca, Merck, Eli Lilly, Sanofi, DalCor, and Pharmascience; has received honoraria from Servier, Thrasos and DalCor; and holds equity interest in DalCor. Dr. Schecter is an employee of Baxalta and Shire; during the study; and was an employee of Novartis. Dr. Fayad has received research support from Novartis. Prof. Choudhury was a Wellcome Trust Senior Research Fellow in Clinical Science. Dr. Tardif holds the Canada Research Chair in translational and personalized medicine and the University of Montreal endowed research chair in atherosclerosis. This work was also supported by the National Institute for Health Research Oxford Biomedical Research Centre (Drs. Choudhury, Biasiolli, and Robson and Ms. Birks). All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Chun Yuan, PhD, served as Guest Editor for this paper.

also supported by the National Institute for Health Research Oxford Biomedical Research Centre (Drs. Choudhury, Biasiolli, and Robson and Ms. Birks). All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Chun Yuan, PhD, served as Guest Editor for this paper. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For detailed information regarding imaging protocols and analysis as well as a supplemental table, please see the online version of this article. Figure 1 Changes in Carotid Wall Area We evaluated 12-month change from baseline in wall area as an indicator of atherosclerosis progression. (A) In both left and right carotid arteries, canakinumab retarded progression of wall area compared with placebo. Consistent in magnitude and direction, the changes did not reach statistical significance in the mean carotid wall area (pre-stated endpoint) or in either carotid artery analyzed separately. Bars = means with 95% confidence intervals. (B) Within the same patient, there was a striking concordance of change in wall area between left- and right-sided arteries, which was maintained in both treatment groups (p < 0.0001 for each). Upper right quadrant = patients with progression in both carotid arteries. Figure 1Figure 2 Changes in Lipids and C-Reactive Protein Absolute level or change from baseline at 3 and 12 months are shown for (A) C-reactive protein; (B) lipoprotein(a) (Lp[a]); (C) total cholesterol; (D) high-density lipoprotein (HDL) cholesterol; and (E) triglycerides.

We evaluated 12-month change from baseline in wall area as an indicator of atherosclerosis progression. (A) In both left and right carotid arteries, canakinumab retarded progression of wall area compared with placebo. Consistent in magnitude and direction, the changes did not reach statistical significance in the mean carotid wall area (pre-stated endpoint) or in either carotid artery analyzed separately. Bars = means with 95% confidence intervals. (B) Within the same patient, there was a striking concordance of change in wall area between left- and right-sided arteries, which was maintained in both treatment groups (p < 0.0001 for each). Upper right quadrant = patients with progression in both carotid arteries. Figure 1Figure 2 Changes in Lipids and C-Reactive Protein Absolute level or change from baseline at 3 and 12 months are shown for (A) C-reactive protein; (B) lipoprotein(a) (Lp[a]); (C) total cholesterol; (D) high-density lipoprotein (HDL) cholesterol; and (E) triglycerides. Figure 2Central Illustration Effects of IL-1β Inhibition Interleukin (IL)-1β seems important in the pathogenesis of atherosclerosis. In this placebo-controlled trial in patients with evidence of clinical atherosclerosis and either type 2 diabetes mellitus or impaired glucose tolerance, the IL-1β inhibitor canakinumab reduced measures of inflammation but did not significantly affect measures of vascular structure or function. LDL = low-density lipoprotein. Central IllustrationTable 1 Baseline Characteristics

Interleukin (IL)-1β seems important in the pathogenesis of atherosclerosis. In this placebo-controlled trial in patients with evidence of clinical atherosclerosis and either type 2 diabetes mellitus or impaired glucose tolerance, the IL-1β inhibitor canakinumab reduced measures of inflammation but did not significantly affect measures of vascular structure or function. LDL = low-density lipoprotein. Central IllustrationTable 1 Baseline Characteristics Table 1 Canakinumab (n = 95) Placebo (n = 94) Male 82 (86) 80 (85) Age, yrs 61.7 ± 7.8 61.9 ± 6.9 Diabetes 81 (85) 81 (86) Hypertension 81 (85) 83 (88) Systolic blood pressure, mm Hg 128.7 ± 10.3 128.4 ± 10.0 Diastolic blood pressure, mm Hg 76.6 ± 8.1 76.9 ± 9.1 Current smoker 7 (7) 7 (7) Peripheral vascular disease 8 (8) 8 (8) Previous stroke 21 (22) 22 (23) CAD 79 (83) 90 (96) Medications ACE inhibitor 66 (69) 72 (77) Beta-blocker 62 (65) 75 (80) Statin 92 (97) 94 (100) Insulin 29 (31) 26 (28) Antiplatelet agent∗ 90 (95) 89 (95) BMI, kg/m2 30.3 ± 4.1 30.3 ± 4.0 hs-CRP, mg/l 1.77 (0.84–3.74) 1.85 (0.83–3.88) IL-6, ng/l 2.20 (1.61–3.63) 2.24 (1.60–3.30) Serum amyloid A, mg/l 3.8 (1.9–6.3) 3.0 (1.9–6.8) Adiponectin, ng/ml 3,685 (2,860–5,740) 3,880 (2,620–4,950) Lp(a), mg/dl 18.8 (4.1–37.7) 15.1 (2.0–40.7) LDL cholesterol, mg/dl 90.5 ± 33.3 88.6 ± 36.3 HDL cholesterol, mg/dl 42.5 ± 10.4 41.8 ± 10.4 Triglycerides, mg/dl 147 ± 106 138 ± 72 Duration of diabetes, yrs ≤1 5 7 >1 and <5 22 16 ≥5 54 58 Fasting blood glucose, mg/dl 144 ± 48 146 ± 48 Insulin, pmol/l 77 (53–133) 72 (51–120) HOMA-IR 3.7 (2.4–6.6) 3.6 (2.4–6.1) HbA1c, % 7.03 ± 1.02 6.85 ± 0.93 HbA1c ≤7% 50 52 >7% and <7.5% 14 20 ≥7.5% 29 21 Values are n (%), mean ± SD, or median (interquartile range).

ation of diabetes, yrs ≤1 5 7 >1 and <5 22 16 ≥5 54 58 Fasting blood glucose, mg/dl 144 ± 48 146 ± 48 Insulin, pmol/l 77 (53–133) 72 (51–120) HOMA-IR 3.7 (2.4–6.6) 3.6 (2.4–6.1) HbA1c, % 7.03 ± 1.02 6.85 ± 0.93 HbA1c ≤7% 50 52 >7% and <7.5% 14 20 ≥7.5% 29 21 Values are n (%), mean ± SD, or median (interquartile range). ACE = angiotensin-converting enzyme; BMI = body mass index; CAD = coronary artery disease; HbA1c = glycosylated hemoglobin; HDL = high-density lipoprotein; HOMA-IR = homeostatic model assessment–insulin resistance; hs-CRP = high-sensitivity C-reactive protein; IL = interleukin; LDL = low-density lipoprotein; Lp(a) = lipoprotein(a). ∗ Aspirin, clopidogrel, prasugrel, or ticagrelor. For HbA1c, there are 2 missing values for canakinumab and 1 for the control group. Table 2 Subject Disposition Table 2 Placebo (n = 94) Canakinumab (n = 95) Total (N = 189) Patients Completed 73 (77.7) 67 (70.5) 140 (74.1) Discontinued∗ 21 (22.3) 28 (29.5) 49 (25.9) Main cause of discontinuation Adverse event(s)∗ 11 (11.7) 14 (14.7) 25 (13.2) Consent withdrawal 3 (3.2) 4 (4.2) 7 (3.7) Lost to follow-up 1 (1.1) 2 (2.1) 3 (3.2) Administrative 4 (4.3) 2 (2.1) 6 (3.2) Death 0 (0) 1 (1.1) 1 (0.5) Protocol deviation 2 (2.1) 5 (5.3) 7 (3.7) Values are n (%). ∗ There was no significant difference between canakinumab compared with placebo for the overall rate of discontinuation from the study (risk ratio: 1.32; 95% confidence interval: 0.81 to 2.15; p = 0.26) or the rate of discontinuation due to adverse events (risk ratio: 1.26; 95% confidence interval: 0.60 to 2.63; p = 0.54).

ere was no significant difference between canakinumab compared with placebo for the overall rate of discontinuation from the study (risk ratio: 1.32; 95% confidence interval: 0.81 to 2.15; p = 0.26) or the rate of discontinuation due to adverse events (risk ratio: 1.26; 95% confidence interval: 0.60 to 2.63; p = 0.54). Table 3 Change From Baseline in MRI Measures

ere was no significant difference between canakinumab compared with placebo for the overall rate of discontinuation from the study (risk ratio: 1.32; 95% confidence interval: 0.81 to 2.15; p = 0.26) or the rate of discontinuation due to adverse events (risk ratio: 1.26; 95% confidence interval: 0.60 to 2.63; p = 0.54). Table 3 Change From Baseline in MRI Measures Table 3 3 Months 12 Months No. of Patients: LSM (95% CI), p Value No. of Patients: LSM (95% CI), p Value Mean (R & L) carotid wall area, mm2 Canakinumab n = 63; 0.97 (–1.55 to 3.48) n = 48; 0.59 (–2.40 to 3.59) Placebo n = 67; 2.13 (–0.37 to 4.64) n = 55; 3.96 (0.94 to 6.98) Canakinumab vs. placebo –1.17 (–4.17 to 1.84), p = 0.44 –3.37 (–6.90 to 0.16), p = 0.06 Proximal ascending aorta wall area, mm2 Canakinumab n = 64; 2.12 (–10.61 to 14.85) n = 50; 11.20 (–9.77 to 32.16) Placebo n = 69; 16.12 (2.90 to 29.34) n = 59; 30.26 (9.57 to 50.96) Canakinumab vs. placebo –14.00 (–29.82 to 1.82), p = 0.08 –19.07 (–44.00 to 5.87), p = 0.13 Proximal descending aorta wall area, mm2 Canakinumab n = 75; 9.01 (–2.17 to 20.19) n = 63; 20.92 (5.06 to 36.78) Placebo n = 80; 6.26 (–5.18 to 17.69) n = 67; 25.28 (8.54 to 42.02) Canakinumab vs. placebo 2.75 (–10.57 to 16.08), p = 0.68 –4.36 (–23.63 to 14.92), p = 0.66 Distal descending aorta wall area, mm2 Canakinumab n = 67; 3.08 (–7.08 to 13.25) n = 56; 14.65 (0.86 to 28.45) Placebo n = 73; –4.41 (–14.76 to 5.95) n = 64; 21.02 (6.67 to 35.36) Canakinumab vs. placebo 7.49 (–4.74 to 19.72), p = 0.23 –6.36 (–23.49 to 10.77), p = 0.46 Ascending aorta distensibility, × 103 mm Hg–1 Canakinumab n = 66; 0.00 (–0.26 to 0.27) n = 51; –0.11 (–0.38 to 0.17) Placebo n = 61; –0.13 (–0.41 to 0.16) n = 53; –0.13 (–0.43 to 0.17) Canakinumab vs. placebo 0.13 (–0.20 to 0.46), p = 0.44 0.03 (–0.32 to 0.370), p = 0.87 Proximal descending aorta distensibility, × 103 mm Hg–1 Canakinumab n = 70; 0.08 (–0.23 to 0.39) n = 56; –0.19 (–0.53 to 0.16) Placebo n = 71; 0.10 (–0.22 to 0.41) n = 62; –0.25 (–0.60 to 0.11) Canakinumab vs. placebo –0.02 (–0.39 to 0.35), p = 0.93 0.06 (–0.36 to 0.48), p = 0.78 Distal descending aorta distensibility, × 103 mm Hg–1 Canakinumab n = 72; 0.04 (–0.32 to 0.40) n = 54; –0.14 (–0.59 to 0.31) Placebo n = 74; 0.12 (–0.24 to 0.49) n = 64; –0.20 (–0.66 to 0.25) Canakinumab vs.

62; –0.25 (–0.60 to 0.11) Canakinumab vs. placebo –0.02 (–0.39 to 0.35), p = 0.93 0.06 (–0.36 to 0.48), p = 0.78 Distal descending aorta distensibility, × 103 mm Hg–1 Canakinumab n = 72; 0.04 (–0.32 to 0.40) n = 54; –0.14 (–0.59 to 0.31) Placebo n = 74; 0.12 (–0.24 to 0.49) n = 64; –0.20 (–0.66 to 0.25) Canakinumab vs. placebo –0.09 (–0.51 to 0.34), p = 0.69 0.06 (–0.48 to 0.61), p = 0.82 Least squared means (LSM) of outcomes (95% confidence intervals [CIs]) and number of patients in each group reported from the analysis of covariance of the change from baseline at 3 and 12 months, adjusted for baseline of outcome and including the 2-level factor type 2 diabetes mellitus or impaired glucose tolerance as a covariate. Inclusion of patients is according to the protocol. Data from an end-of-study scan of 8 patients who left the trial early were included if the time of the scan was within the allowed time limits for the 3-month scan (82 to 130 days). L = left; MRI = magnetic resonance imaging; R = right; other abbreviations as in Table 2. Table 4 Blood Measures for Lipids, Diabetes Control, and Markers of Inflammation

placebo –0.09 (–0.51 to 0.34), p = 0.69 0.06 (–0.48 to 0.61), p = 0.82 Least squared means (LSM) of outcomes (95% confidence intervals [CIs]) and number of patients in each group reported from the analysis of covariance of the change from baseline at 3 and 12 months, adjusted for baseline of outcome and including the 2-level factor type 2 diabetes mellitus or impaired glucose tolerance as a covariate. Inclusion of patients is according to the protocol. Data from an end-of-study scan of 8 patients who left the trial early were included if the time of the scan was within the allowed time limits for the 3-month scan (82 to 130 days). L = left; MRI = magnetic resonance imaging; R = right; other abbreviations as in Table 2. Table 4 Blood Measures for Lipids, Diabetes Control, and Markers of Inflammation Table 4 3 Months 12 Months No. of Patients; LSM (95% CI), p Value LDL cholesterol, mg/dl Log-transformed results Canakinumab n = 64; 4.425 (4.343 to 4.508) n = 55; 4.515 (4.414 to 4.616) Placebo n = 75; 4.361 (4.278 to 4.444) n = 65; 4.460 (4.357 to 4.563) Canakinumab vs. placebo 0.065 (–0.029 to 0.158), p = 0.18 0.055 (–0.064 to 0.174), p = 0.36 Back-transformed results Canakinumab 83.52 (76.94 to 90.75) 91.38 (82.60 to 101.10) Placebo 78.34 (72.10 to 85.12) 86.49 (78.03 to 95.88) Canakinumab/placebo 1.067 (0.971 to 1.171) 1.057 (0.938 to 1.190) Total cholesterol, mg/dl Log-transformed results Canakinumab n = 49; 5.153 (5.098 to 5.208) n = 40; 5.193 (5.119 to 5.267) Placebo n = 58; 5.040 (4.983 to 5.096) n = 52; 5.113 (5.037 to 5.188) Canakinumab vs. placebo 0.113 (0.048 to 0.178), p = 0.0008 0.081 (–0.009 to 0.170), p = 0.08 Back-transformed results Canakinumab 173.0 (163.7 to 182.7) 180.0 (167.2 to 193.8) Placebo 154.5 (145.9 to 163.4) 166.2 (154.0 to 179.1) Canakinumab/placebo 1.120 (1.050 to 1.195) 1.084 (0.991 to 1.185) Triglycerides, mg/dl Log-transformed results Canakinumab n = 73; 4.938 (4.843 to 5.033) n = 62; 4.979 (4.865 to 5.093) Placebo n = 79; 4.746 (4.646 to 4.845) n = 70; 4.795 (4.676 to 4.915) Canakinumab vs. placebo 0.192 (0.079 to 0.306), p = 0.001 0.184 (0.045 to 0.322), p = 0.01 Back-transformed results Canakinumab 139.5 (126.8 to 153.4) 145.3 (129.7 to 162.9) Placebo 115.1 (104.2 to 127.1) 120.9 (107.3 to 136.3) Canakinumab/placebo 1.212 (1.082 to 1.358) 1.202 (1.046 to 1.380) HDL cholesterol, mg/dl∗ Canakinumab n = 73; 1.779 (0.309 to 3.252) n = 62; 0.731 (–1.017 to 2.479) Placebo n = 79; –0.282 (–1.817 to 1.257) n = 70; –0.182 (–1.910 to 1.640) Canakinumab vs. placebo 2.061 (0.302 to 3.821), p = 0.02 0.913 (–1.214 to 3.039), p = 0.40 Lp(a), mg/dl∗ Canakinumab n = 56; –3.325 (–5.835 to –0.816) n = 47; –3.083 (–6.066 to –0.100) Placebo n = 59; 0.394 (–2.333 to 3.120) n = 51; 1.217 (–2.059 to 4.493) Canakinumab vs. placebo –3.719 (–6.809 to –0.628), p = 0.02 –4.300 (–8.052 to –0.548), p = 0.025 No.

to 3.821), p = 0.02 0.913 (–1.214 to 3.039), p = 0.40 Lp(a), mg/dl∗ Canakinumab n = 56; –3.325 (–5.835 to –0.816) n = 47; –3.083 (–6.066 to –0.100) Placebo n = 59; 0.394 (–2.333 to 3.120) n = 51; 1.217 (–2.059 to 4.493) Canakinumab vs. placebo –3.719 (–6.809 to –0.628), p = 0.02 –4.300 (–8.052 to –0.548), p = 0.025 No. of Patients; Mean (95% CI on Log Scale) HbA1c, % Log-transformed results Canakinumab n = 73; 1.919 (1.900 to 1.942) n = 60; 1.930 (1.897 to 1.963) Placebo n = 80; 1.918 (1.894 to 1.943) n = 69; 1.922 (1.886 to 1.958) Canakinumab vs. placebo 0.001 (–0.025 to 0.027), p = 0.95 0.008 (–0.031 to 0.047), p = 0.68 Back-transformed results Canakinumab 6.814 (6.686 to 6.973) 6.890 (6.666 to 7.121) Placebo 6.807 (6.646 to 6.980) 6.835 (6.593 to 7.085) Canakinumab/placebo 1.001 (0.975 to 1.027) 1.008 (0.969 to 1.048) Fasting blood glucose, mg/dl Log-transformed results Canakinumab n = 71; 4.928 (4.868 to 4.989) n = 58; 4.922 (4.845 to 4.999) Placebo n = 78; 4.849 (4.786 to 4.912) n = 68; 4.853 (4.773 to 4.934) Canakinumab vs. placebo 0.079 (0.013 to 0.145), p = 0.02 0.068 (–0.017 to 0.154), p = 0.12 Back-transformed results Canakinumab 138 (130 to 147) 137 (127 to 148) Placebo 128 (120 to 136) 128 (118 to 139) Canakinumab/placebo 1.082 (1.013 to 1.156) 1.070 (0.983 to 1.166) HOMA-IR Log-transformed results Canakinumab n = 67; 1.443 (1.282 to 1.603) n = 56; 1.324 (1.133 to 1.514) Placebo n = 76; 1.358 (1.190 to 1.527) n = 68; 1.278 (1.082 to 1.473) Canakinumab vs. placebo 0.084 (–0.094 to 0.262), p = 0.35 0.046 (–0.170 to 0.262), p = 0.67 Back-transformed results Canakinumab 4.233 (3.604 to 4.968) 3.758 (3.105 to 4.545) Placebo 3.888 (3.287 to 4.604) 3.589 (2.951 to 4.362) Canakinumab/placebo 1.088 (0.910 to 1.300) 1.047 (0.844 to 1.300) HOMA-B Log-transformed results Canakinumab n = 66; 4.066 (3.886 to 4.247) n = 56; 3.969 (3.773 to 4.165) Placebo n = 76; 4.224 (4.033 to 4.415) n = 68; 4.133 (3.929 to 4.337) Canakinumab vs. placebo –0.158 (–0.361 to 0.046), p = 0.13 –0.164 (–0.389 to 0.062), p = 0.15 Back-transformed results Canakinumab 58.32 (48.72 to 69.90) 52.93 (43.51 to 64.39) Placebo 68.31 (56.43 to 82.68) 62.36 (50.86 to 76.48) Canakinumab/placebo 0.854 (0.697 to 1.047) 0.849 (0.678 to 1.064) C-reactive protein, mg/l Log-transformed results Canakinumab n = 74; –0.278 (–0.500 to –0.056) n = 61; –0.170 (–0.419 to 0.078) Placebo n = 79; 0.287 (0.051 to 0.523) n = 69; 0.409 (0.145 to 0.673) Canakinumab vs.

(56.43 to 82.68) 62.36 (50.86 to 76.48) Canakinumab/placebo 0.854 (0.697 to 1.047) 0.849 (0.678 to 1.064) C-reactive protein, mg/l Log-transformed results Canakinumab n = 74; –0.278 (–0.500 to –0.056) n = 61; –0.170 (–0.419 to 0.078) Placebo n = 79; 0.287 (0.051 to 0.523) n = 69; 0.409 (0.145 to 0.673) Canakinumab vs. placebo –0.565 (–0.829 to –0.301), p < 0.0001 –0.579 (–0.881 to –0.277), p = 0.0002 Back-transformed results Canakinumab 0.757 (0.607 to 0.946) 0.848 (0.658 to 1.081) Placebo 1.332 (1.052 to 1.687) 1.505 (1.156 to 1.9690) Canakinumab/placebo 0.568 (0.436 to 0.740) 0.561 (0.414 to 0.758) IL-6, pg/ml Log-transformed results Canakinumab n = 48; 0.367 (0.205 to 0.529) n = 13; 0.744 (0.313 to 1.745) Placebo n = 48; 0.913 (0.742 to 1.081) n = 14; 1.147 (0.729 to 1.565) Canakinumab vs. placebo –0.544 (–0.728 to –0.361), p < 0.0001 –0.403 (–0.888 to 0.081), p = 0.10 Back-transformed results Canakinumab 1.443 (1.228 to 1.697) 2.104 (1.368 to 5.726) Placebo 2.492 (2.100 to 2.948) 3.149 (2.073 to 4.783) Canakinumab/placebo 0.580 (0.483 to 0.697) 0.668 (0.411 to 1.084) Serum amyloid A, mg/l Log-transformed results Canakinumab n = 67; 0.963 (0.760 to 1.166) n = 56; 1.000 (0.795 to 1.205) Placebo n = 76; 1.129 (0.922 to 1.336) n = 69; 1.243 (1.032 to 1.453) Canakinumab vs. placebo –0.166 (–0.408 to 0.077), p = 0.18 –0.243 (–0.493 to 0.008), p = 0.06 Back-transformed results Canakinumab 2.620 (2.138 to 3.209) 2.718 (2.214 to 3.337) Placebo 3.093 (2.514 to 3.804) 3.466 (2.807 to 4.276) Canakinumab/placebo 0.847 (0.665 to 1.080) 0.784 (0.611 to 1.008) Adiponectin, ng/ml Log-transformed results Canakinumab n = 36; 8.265 (8.208 to 8.323) Placebo n = 36; 8.284 (8.227 to 8.342) Canakinumab vs. placebo –0.019 (–0.084 to 0.046), p = 0.57 Back-transformed results Canakinumab 3,885 (3,670 to 4,117) Placebo 3,960 (3,741 to 4,196) Canakinumab/placebo 0.981 (0.919 to 1.047) Abbreviations as in Tables 1 and 3.

d results Canakinumab n = 36; 8.265 (8.208 to 8.323) Placebo n = 36; 8.284 (8.227 to 8.342) Canakinumab vs. placebo –0.019 (–0.084 to 0.046), p = 0.57 Back-transformed results Canakinumab 3,885 (3,670 to 4,117) Placebo 3,960 (3,741 to 4,196) Canakinumab/placebo 0.981 (0.919 to 1.047) Abbreviations as in Tables 1 and 3. ∗ Change from baseline, not log-transformed. LSMs of outcomes (95% CIs) with number of patients reported from the analysis of covariance of the log-transformed outcome at 3 and 12 months, adjusted for log-transformed baseline of outcome and including the 2-level factor type 2 diabetes mellitus or impaired glucose tolerance as a covariate. Inclusion of patients is according to the protocol. The treatment effect is reported as the difference between the canakinumab arm and the placebo arm and after back-transforming the treatment effect as the ratio of the levels in the canakinumab arm to the placebo arm. The LSMs are back-transformed (geometric means). Table 5 MACE Table 5 MACE Placebo Canakinumab p Value Group All N 94 95 Yes 7 (7) 10 (11) 0.612 No 87 (93) 85 (89) Type 2 diabetes mellitus N 81 81 Yes 6 (7) 9 (11) 0.589 No 75 (93) 72 (89) Impaired glucose tolerance N 13 14 Yes 1 (8) 1 (7) 1.000 No 12 (92) 13 (93) Values are n (%) unless otherwise indicated. MACE = major adverse cardiac events.

Atrial fibrillation (AF) causes cardiovascular death, frequent hospitalization, and cognitive decline even in patients treated according to guidelines 1, 2, 3. Antiarrhythmic drug (AAD) therapy remains the most commonly used treatment to maintain sinus rhythm in AF patients, but AAD effectiveness remains limited (3). Unfortunately, we lack a basic understanding of why AADs prevent AF over long periods in some patients but not in others 4, 5. Identifying factors that modify the effects of AADs would allow the selection of responsive patients and could help guide development of novel AADs (6). Paired like homeodomain-2 transcription factor (PITX2) is a transcription factor that regulates the development of the left atrium (LA) and thoracic organs. Its c isoform is expressed in the adult LA and regulates the expression of LA ion channels 7, 8, 9. Low atrial Pitx2 expression renders mice susceptible to AF and shortens the LA action potential 8, 10, 11. In this study, we investigated how atrial PITX2 modifies the effects of AADs. We detected variable LA PITX2 messenger ribonucleic acid (mRNA) expression in AF patients requiring rhythm control therapy. After finding that low Pitx2c enhanced the effect of flecainide, mediated by a more positive resting membrane potential (RMP), we identified reduced TWIK-related acid-sensitive K+ channel 2 (TASK-2) expression as a possible driver of this effect and replicated these effects in cells expressing human sodium (Na) channels and in a human atrial action potential model.

ct of flecainide, mediated by a more positive resting membrane potential (RMP), we identified reduced TWIK-related acid-sensitive K+ channel 2 (TASK-2) expression as a possible driver of this effect and replicated these effects in cells expressing human sodium (Na) channels and in a human atrial action potential model. Methods All experiments were conducted under the Animals (Scientific Procedures) Act 1986, and approved by the home office (PPL number 30/2967) and the institutional review board at the University of Birmingham. Analyses of human atrial tissue were approved by the institutional review board of Academic Medical Center, Amsterdam, the Netherlands. All patients provided written informed consent. Left atrial appendages (LAAs) were excised from 95 patients undergoing bilateral thoracoscopic AF ablation either in the AFACT (Atrial Fibrillation Ablation and Autonomic Modulation via Thoracoscopic Surgery) trial (12) or undergoing similar procedures in the same centers using an endoscopic stapling device, snap frozen in liquid nitrogen and stored at –80°C (13). Deoxyribonucleic acid and ribonucleic acid were extracted using DNeasy and RNeasy kits (Qiagen Ltd., Manchester, United Kingdom), respectively. PITX2 mRNA content was quantified by quantitative polymerase chain reaction. Single nucleotide polymorphisms (SNPs) rs2200733, rs6838973, and rs1448818 (14) were identified using TaqMan assays (Thermo Fisher Scientific Inc., Waltham, Massachusetts). Adult mice (age 12 to 16 weeks) on an MF1 background with normal or reduced (Pitx2c+/−) atrial Pitx2c expression were studied (8).

Left atrial appendages (LAAs) were excised from 95 patients undergoing bilateral thoracoscopic AF ablation either in the AFACT (Atrial Fibrillation Ablation and Autonomic Modulation via Thoracoscopic Surgery) trial (12) or undergoing similar procedures in the same centers using an endoscopic stapling device, snap frozen in liquid nitrogen and stored at –80°C (13). Deoxyribonucleic acid and ribonucleic acid were extracted using DNeasy and RNeasy kits (Qiagen Ltd., Manchester, United Kingdom), respectively. PITX2 mRNA content was quantified by quantitative polymerase chain reaction. Single nucleotide polymorphisms (SNPs) rs2200733, rs6838973, and rs1448818 (14) were identified using TaqMan assays (Thermo Fisher Scientific Inc., Waltham, Massachusetts). Adult mice (age 12 to 16 weeks) on an MF1 background with normal or reduced (Pitx2c+/−) atrial Pitx2c expression were studied (8). LA epicardial monophasic action potentials were recorded from Langendorff-perfused murine hearts 8, 15. Programmed stimulation was performed at baseline and with flecainide 1 μmol/l or d,l-sotalol 10 μmol/l. Arrhythmia inducibility and effective refractory period (ERP) were measured by using single right atrial extrastimuli after steady-state pacing in 1-ms decrements 15, 16, 17, 18. Transmembrane action potentials were recorded using borosilicate glass microelectrodes from superfused murine LAs (17), RMP, action potential duration (APD), upstroke velocity, and activation times were analyzed 15, 17, 18.

ng single right atrial extrastimuli after steady-state pacing in 1-ms decrements 15, 16, 17, 18. Transmembrane action potentials were recorded using borosilicate glass microelectrodes from superfused murine LAs (17), RMP, action potential duration (APD), upstroke velocity, and activation times were analyzed 15, 17, 18. The human atrial cell model of Courtemanche et al. (19) was used. Pitx2c+/– deficiency was modeled by reducing IK1 conductance by 25% and doubling IKr conductance. Simulations were run in strands of 100 atrial cells (cell length 100 μm). The 5 leftmost cells of the strand were paced (S1) for 2 min at 1,000- and 500-ms basic cycle lengths. Premature stimulation (S2) was applied to determine the ERP and conduction velocity as measured from cells 25 to 75. Values for all other parameters were measured from the 50th cell. For the modeling, post-repolarization refractoriness (PRR) was calculated as the difference between APD at –60 mV repolarization and ERP. LA cell isolation was performed as previously reported (20). Standard INa and IK1 currents were recorded as previously published 18, 19, 20. Background K+ (TASK-like) currents sensitive to high Ba2+ (10 mM) were measured 21, 22, 23. Human embryonic kidney (HEK) 293 cells stably expressing the human Nav1.5 channel were obtained (SB Ion Channels, Glasgow, UK).

ously reported (20). Standard INa and IK1 currents were recorded as previously published 18, 19, 20. Background K+ (TASK-like) currents sensitive to high Ba2+ (10 mM) were measured 21, 22, 23. Human embryonic kidney (HEK) 293 cells stably expressing the human Nav1.5 channel were obtained (SB Ion Channels, Glasgow, UK). Ribonucleic acid and complementary deoxyribonucleic acid were synthesized from murine LA, (SuperScript VILO, Thermo Fisher Scientific Inc.) to quantify expression of 20 atrial ion channels and genes with suspected PITX2-dependent regulation (9) using custom-designed Taqman low density array plates (Thermo Fisher Scientific Inc.). Western immunoblotting was performed on murine LA tissue lysates with antibodies detecting TASK-2, Kv1.6, Na/K ATPase alpha-1, Na/K ATPase alpha-2, Na/Ca exchanger 1, Serca2a, Nav1.5, or calnexin, using standard methods. Optical action potentials and calcium ion (Ca2+) transients were recorded in murine LA and analyzed using custom-made MATLAB algorithms (MathWorks, Natick, Massachusetts) as previously described (17).

Ribonucleic acid and complementary deoxyribonucleic acid were synthesized from murine LA, (SuperScript VILO, Thermo Fisher Scientific Inc.) to quantify expression of 20 atrial ion channels and genes with suspected PITX2-dependent regulation (9) using custom-designed Taqman low density array plates (Thermo Fisher Scientific Inc.). Western immunoblotting was performed on murine LA tissue lysates with antibodies detecting TASK-2, Kv1.6, Na/K ATPase alpha-1, Na/K ATPase alpha-2, Na/Ca exchanger 1, Serca2a, Nav1.5, or calnexin, using standard methods. Optical action potentials and calcium ion (Ca2+) transients were recorded in murine LA and analyzed using custom-made MATLAB algorithms (MathWorks, Natick, Massachusetts) as previously described (17). Statistical analysis All experiments were performed and analyzed in a blinded fashion. Murine studies were performed and analyzed blinded to genotype in littermate pairs. Categorical data were compared using the Fisher exact test. Numerical data were compared by 2-sided paired parametric Student t tests (e.g., measurements before and after perfusion of flecainide or sotalol) and Wilcoxon signed rank tests. Multiple measurements were assessed by repeated measures of analysis of variance followed by correction for multiple comparison (Bonferroni test) if the overall test was significant. Two-sided p < 0.05 were considered significant. Box plots depict individual measurements (points), mean, and SEM. Statistics and figures were created using Prism 5 (GraphPad Software, San Diego, California).

f variance followed by correction for multiple comparison (Bonferroni test) if the overall test was significant. Two-sided p < 0.05 were considered significant. Box plots depict individual measurements (points), mean, and SEM. Statistics and figures were created using Prism 5 (GraphPad Software, San Diego, California). Results PITX2 mRNA varied markedly in human LAA (Central Illustration) harvested from AF patients (Table 1) (13), suggesting that a 50% lowered PITX2 expression defines a large, potentially clinically relevant group of AF patients. This did not directly correlate with SNP haplotype (Table 2), although we found numerically lower PITX2c levels in patients with 5 risk alleles. Flecainide suppressed atrial arrhythmias in murine Pitx2c+/– hearts. Flecainide abolished induced atrial arrhythmias in hearts with reduced Pitx2c expression (0 of 17 hearts with atrial arrhythmias) but not in hearts with normal Pitx2c expression (atrial arrhythmias remained in 3 of 12 hearts) (Figures 1A to 1C). Flecainide prolonged ERPs and refractoriness beyond the end of repolarization (PRR) calculated as the difference between ERP and APD90 (ms). Flecainide prolonged PRR more in hearts with reduced Pitx2c expression (Figures 1D and 1E, Table 3). PITX2c+/– hearts had shorter atrial action potentials (8). Flecainide abolished APD differences between Pitx2c+/– and wild-type LA by prolonging early repolarization (APD30, APD50, and APD70) (Table 3). Murine atrial PITX2 expression did not modulate the effects of sotalol on atrial APD or ERP (Table 4).

1E, Table 3). PITX2c+/– hearts had shorter atrial action potentials (8). Flecainide abolished APD differences between Pitx2c+/– and wild-type LA by prolonging early repolarization (APD30, APD50, and APD70) (Table 3). Murine atrial PITX2 expression did not modulate the effects of sotalol on atrial APD or ERP (Table 4). RMP was slightly depolarized in LA murine cells with reduced Pitx2c expression (range of mean depolarization 1.2 to 2.4 mV over 5 cycle lengths; all p < 0.05) (Figures 2A and 2B). Atrial Pitx2c levels did not significantly affect dV/dtmax (100-ms paced cycle length: wild-type: 104.4 ± 4.3 V/s; Pitx2c+/–: 93.7 ± 4.5 V/s) (Figure 2C). Flecainide did not modify atrial RMP (Figure 2B) but reduced action potential amplitude consistent with its Na-channel blocking effect, specifically at 100-ms cycle length: wild-type baseline: 77.5 ± 1.2 mV (n = 30); wild-type flecainide: 71.3 ± 1.2 mV (n = 31); Pitx2c+/– baseline: 73.4 ± 1.3 mV (n = 22); and Pitx2c+/– flecainide: 65.1 ± 1.45 mV (n = 24).

(Figure 2B) but reduced action potential amplitude consistent with its Na-channel blocking effect, specifically at 100-ms cycle length: wild-type baseline: 77.5 ± 1.2 mV (n = 30); wild-type flecainide: 71.3 ± 1.2 mV (n = 31); Pitx2c+/– baseline: 73.4 ± 1.3 mV (n = 22); and Pitx2c+/– flecainide: 65.1 ± 1.45 mV (n = 24). Because the Courtemanche–Ramirez–Nattel model does not incorporate background K+ currents (19), we simulated a depolarized RMP in this model by a 25% reduction in IK1. This reduced the RMP at 500-ms paced cycle length by 2 mV from 79.9 mV (“normal PITX2”) to –77.9 mV (“low PITX2”). Na channels recovered from inactivation more slowly upon partial INa block (50% or 60%) (Figure 3A). Furthermore, PRR was enhanced in the PITX2 deficiency model (Figure 3B and Table 5). Inhibition of INa reduced upstroke velocity (dV/dtmax) and conduction velocity in both models, and reproduced the prolongation of PRR (Figure 3B). Kcna6 and Kcnk5 mRNA expression were reduced in Pitx2c+/– murine LA (Figure 4A, Online Table 1), whereas mRNA concentrations of 20 other ion channels or related genes were not altered. Kv1.6 protein concentration was unaltered, whereas TASK-2 protein concentration was reduced in murine atria with reduced Pitx2c expression (Figure 4B). Nav1.5 mRNA and protein expression were not changed (Figures 4A and 4B).

ble 1), whereas mRNA concentrations of 20 other ion channels or related genes were not altered. Kv1.6 protein concentration was unaltered, whereas TASK-2 protein concentration was reduced in murine atria with reduced Pitx2c expression (Figure 4B). Nav1.5 mRNA and protein expression were not changed (Figures 4A and 4B). Atrial Pitx2c expression did not modify peak Na+ currents (INa) recorded from isolated murine cardiomyocytes at holding potentials ranging from –100 to –65 mV (Figures 5A to 5C). Peak INa was reduced at more depolarized holding potentials (Figure 5). Flecainide inhibited INa better at more positive holding potentials (inhibition at –70 mV: 68 ± 5%; inhibition at –65 mV: 75 ± 5%; n = 86 cells from n = 17 atria) in cells from murine atria with normal or reduced Pitx2c expression, suggesting that the greater efficiency of flecainide in atria with reduced Pitx2c expression is secondary to RMP depolarization (Figure 5C). Consistent with this, flecainide inhibited human Nav1.5 channels expressed in HEK cells more potently at more depolarized test potentials (–65 to –75 mV) (Figures 5D and 5E). Background K+ currents, which include TASK currents, were reduced in Pitx2c+/– murine atria, whereas IK1 did not differ between genotypes (Figure 6).

Atrial Pitx2c expression did not modify peak Na+ currents (INa) recorded from isolated murine cardiomyocytes at holding potentials ranging from –100 to –65 mV (Figures 5A to 5C). Peak INa was reduced at more depolarized holding potentials (Figure 5). Flecainide inhibited INa better at more positive holding potentials (inhibition at –70 mV: 68 ± 5%; inhibition at –65 mV: 75 ± 5%; n = 86 cells from n = 17 atria) in cells from murine atria with normal or reduced Pitx2c expression, suggesting that the greater efficiency of flecainide in atria with reduced Pitx2c expression is secondary to RMP depolarization (Figure 5C). Consistent with this, flecainide inhibited human Nav1.5 channels expressed in HEK cells more potently at more depolarized test potentials (–65 to –75 mV) (Figures 5D and 5E). Background K+ currents, which include TASK currents, were reduced in Pitx2c+/– murine atria, whereas IK1 did not differ between genotypes (Figure 6). Reduced Pitx2c expression did not alter atrial conduction velocities or activation patterns (Online Figures 1A to 1C, Table 6), consistent with published data (8). We found that 1 μmol/l flecainide decreased atrial conduction velocities without differences between wild-type and Pitx2c+/– mice (Online Figures 1B and 1C). Calcium transient relaxation times at 50% relaxation were not different between wild-type and Pitx2c+/– (Online Figures 1D and 1E). Flecainide 1 μmol/l shortened 50% Ca2+ relaxation times by approximately 10% and decreased Ca2+ transient amplitude by approximately 50% in murine atria with normal and reduced Pitx2c expression (Online Figures 1E and 1F). Additionally, expression of the Na/Ca exchanger Serca2a and Na/K ATPase alpha-1 and alpha-2 subunit protein did not differ between wild-type and Pitx2c+/– atria (Online Figure 2).

ecreased Ca2+ transient amplitude by approximately 50% in murine atria with normal and reduced Pitx2c expression (Online Figures 1E and 1F). Additionally, expression of the Na/Ca exchanger Serca2a and Na/K ATPase alpha-1 and alpha-2 subunit protein did not differ between wild-type and Pitx2c+/– atria (Online Figure 2). Discussion This study demonstrated that LA PITX2 mRNA concentrations vary in patients with AF requiring rhythm control therapy (Central Illustration). Furthermore, flecainide increases PRR and suppresses arrhythmias more effectively in atria with halved Pitx2c expression, mediated by a more depolarized RMP (Central Illustration). Drug-induced PRR is thought to prevent arrhythmias, as reactivation can then occur only after full recovery of excitability, avoiding slow propagation during the vulnerable period 16, 24, 25. We found similar effects in cells expressing human Na channels and in the Courtemanche–Ramirez–Nattel model of human atrial action potentials. Thus, this study highlighted modulation of the atrial RMP by PITX2, possibly mediated by background currents such as TASK-2, as a target for AAD therapy, including atrial-selective therapy. Furthermore, the results suggested that markers for atrial PITX2 expression may identify AF patients who benefit from Na-channel blocker therapy (Central Illustration).

n of the atrial RMP by PITX2, possibly mediated by background currents such as TASK-2, as a target for AAD therapy, including atrial-selective therapy. Furthermore, the results suggested that markers for atrial PITX2 expression may identify AF patients who benefit from Na-channel blocker therapy (Central Illustration). Low atrial PITX2 expression was identified as an important determinant of the antiarrhythmic effects of Na channel blockers. Low LA Pitx2c mRNA depolarized atrial RMP (Figure 2), consistent with a previous report (11). A depolarized RMP increased flecainide-induced PRR (Figure 1) 26, 27, 28, 29, 30. The conduction-slowing effect of flecainide was not modulated by reduced atrial Pitx2c (Online Figure 1), an important surrogate for drug safety. Both the modeling experiments (Figure 3) and the experiments in HEK cells expressing human Na channels (Figure 5) confirmed that small changes in RMP can markedly modulate Na-channel inhibition.

ng effect of flecainide was not modulated by reduced atrial Pitx2c (Online Figure 1), an important surrogate for drug safety. Both the modeling experiments (Figure 3) and the experiments in HEK cells expressing human Na channels (Figure 5) confirmed that small changes in RMP can markedly modulate Na-channel inhibition. Resting membrane potential Open-state Na-channel blockers such as flecainide and propafenone bind preferentially to Na channels integrated in membranes with slightly depolarized resting potentials, where more channels are in the open or inactivated state 31, 32. Our data can be interpreted as suggesting that AAD combinations that include a Na-channel blocker with a membrane potential modifying substance, such as amiodarone 16, 33 or the combination of dronedarone and ranolazine 29, 34, 35, may have synergistic antiarrhythmic effects because they modulate atrial RMP and thereby enhance the effect of Na-channel blockade. Further studies of such drug combinations and the relationship between their effectiveness and the patient’s atrial PITX2 mRNA levels are warranted. Our data also suggested that such combined effects may be of special relevance in patients who have a depolarized RMP, such as secondary to low LA PITX2. Because PITX2 expression is confined to the LA in the heart, AAD therapy that leverages modifications in RMP may achieve “atrial-specific” AAD therapy.

are warranted. Our data also suggested that such combined effects may be of special relevance in patients who have a depolarized RMP, such as secondary to low LA PITX2. Because PITX2 expression is confined to the LA in the heart, AAD therapy that leverages modifications in RMP may achieve “atrial-specific” AAD therapy. RMP is maintained by an intricate balance of different transmembrane currents and is closely related to the potassium equilibrium potential. We identified that PITX2 modifies expression of the genes encoding Kv1.6 and TASK-2 (Figure 4). Complete deletion of PITX2 regulates other potassium and Na channels such as Kcnj2 8, 36, which alter the RMP, but these were not responsible for the depolarized RMP observed in our study. Two-pore domain potassium channels, such as TASK-2, contribute to RMP in various cells, including skeletal and cardiac muscle 37, 38. To date, an altered function of the TASK-1 channel and of IK1 has been implicated in atrial remodeling and AF 39, 40. This study demonstrated that TASK-2 is expressed in atrial myocardium (Figure 4B), suggesting that a reduced function of TASK-2 could depolarize RMP (Figures 1 and 5) 8, 11, analogous to the effect of TASK-2 in neuronal and cartilage tissue 41, 42.

nnel and of IK1 has been implicated in atrial remodeling and AF 39, 40. This study demonstrated that TASK-2 is expressed in atrial myocardium (Figure 4B), suggesting that a reduced function of TASK-2 could depolarize RMP (Figures 1 and 5) 8, 11, analogous to the effect of TASK-2 in neuronal and cartilage tissue 41, 42. Developing clinical markers for patients with depolarized RMP It will be challenging to directly assess LA RMP in AF patients, but our data suggested that differences in atrial RMP could explain the effectiveness of Na-channel blockers in carriers of common gene variants on chromosome 4q25 (43), although LA PITX2 levels are modulated by factors other than SNP status (Table 2) (44). It seems desirable to develop and validate drivers that modify RMP and clinical markers for patients prone to a depolarized atrial RMP to select appropriate AADs for individual patients in the future, thus enabling personalized AAD selection 6, 45.

2 levels are modulated by factors other than SNP status (Table 2) (44). It seems desirable to develop and validate drivers that modify RMP and clinical markers for patients prone to a depolarized atrial RMP to select appropriate AADs for individual patients in the future, thus enabling personalized AAD selection 6, 45. Study limitations This study provided robust evidence that LA PITX2 expression varies in AF patients and that reduced PITX2c expression enhances the antiarrhythmic effects of Na-channel blockers by modulating atrial RMP. The study was partly motivated by the assumption that gene variants on chromosome 4q25 modify PITX2 expression, an assumption that has not been definitively proven 9, 11, 44, 46. Our analysis (Table 2) and that of others indicate that SNP status does not always correlate with PITX2 levels 47, 48. Our findings are relevant to AAD therapy even if the presumed link between PITX2 expression and genetic variants on chromosome 4q25 proves elusive. The mechanisms by which reduced PITX2 mRNA concentrations shorten the LA action potential at high heart rates remain to be fully elucidated 8, 20. Validating our findings in patients is desirable but will be challenging because access to fresh LA cardiomyocytes and LA tissue is limited.

chromosome 4q25 proves elusive. The mechanisms by which reduced PITX2 mRNA concentrations shorten the LA action potential at high heart rates remain to be fully elucidated 8, 20. Validating our findings in patients is desirable but will be challenging because access to fresh LA cardiomyocytes and LA tissue is limited. Due to the novelty of our findings, we could not perform a priori power calculations for our mechanistic experiments, and we analyzed several functional parameters to identify potential mechanisms conveying the antiarrhythmic effects of flecainide in atria with low Pitx2c concentrations. Our findings thus require independent validation. Conclusions This study shows that low LA PITX2 mRNA levels increase atrial RMP and thereby increase the effectiveness of flecainide (Central Illustration). This finding calls for appropriately designed clinical studies to assess whether AF patients with low atrial PITX2 levels respond favorably to Na-channel blockade. Further studies exploring the relevance of TASK channels to atrial RMP also are warranted.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: PITX2, a transcription factor linked to left–right asymmetry in the chest during development, modulates the expression of LA ion channels maintaining the RMP and modulates the antiarrhythmic effects of Na-channel blockers. TRANSLATIONAL OUTLOOK: Clinical studies are needed to assess whether reduced PITX2 expression identifies patients with AF who respond favorably to Na-channel blocking drugs. Appendix Online Data Online Data

Conclusions This study shows that low LA PITX2 mRNA levels increase atrial RMP and thereby increase the effectiveness of flecainide (Central Illustration). This finding calls for appropriately designed clinical studies to assess whether AF patients with low atrial PITX2 levels respond favorably to Na-channel blockade. Further studies exploring the relevance of TASK channels to atrial RMP also are warranted.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: PITX2, a transcription factor linked to left–right asymmetry in the chest during development, modulates the expression of LA ion channels maintaining the RMP and modulates the antiarrhythmic effects of Na-channel blockers. TRANSLATIONAL OUTLOOK: Clinical studies are needed to assess whether reduced PITX2 expression identifies patients with AF who respond favorably to Na-channel blocking drugs. Appendix Online Data Online Data Acknowledgments The authors thank Sian Marie O’Brien, Sarah Hopkins, Syeeda Nashitha Kabir, Pushpa Patel, and Charles Carey for technical support; Marta Coric for help with HEK cells; and Ilaria Piccini for advice on TLDA.

TRANSLATIONAL OUTLOOK: Clinical studies are needed to assess whether reduced PITX2 expression identifies patients with AF who respond favorably to Na-channel blocking drugs. Appendix Online Data Online Data Acknowledgments The authors thank Sian Marie O’Brien, Sarah Hopkins, Syeeda Nashitha Kabir, Pushpa Patel, and Charles Carey for technical support; Marta Coric for help with HEK cells; and Ilaria Piccini for advice on TLDA. This work was supported by the European Union (EUTRAF 25105 to Drs. Kirchhof and Rohr; and Grant Agreement No. 633196 [CATCH ME] to Drs. Kirchhof and Fabritz); British Heart Foundation (FS/13/43/30324 to Drs. Kirchhof and Fabritz); Leducq Foundation to Dr. Kirchhof; Physical Science of Imaging in Biomedical Sciences (PSIBS) University of Birmingham for TY to Dr. Fabritz (EP/F50053X/1); DFG (FA 413 3/1) to Dr. Fabritz; Swiss National Science Foundation (138297) to Dr. Rohr; and Boehringer Ingelheim Foundation to Mr. Kuhlmann. Dr. de Groot is supported by NWO/ZonMW VIDI Grant 016.146.310. Dr. Riley is currently employed by Bio-Techne (R&D Products). Dr. Fabritz has received further institutional research grant support from DFG, MRC, and Gilead Inc. Dr. Kirchhof has received further research support from the German Centre for Heart Research and from several drug and device companies active in atrial fibrillation; and has received honoraria from several such companies. Drs. Syeda, Fabritz, and Kirchhof are listed as inventors on a patent (WO2015/140571) held by the University of Birmingham on genotype-specific antiarrhythmic drug therapy of atrial fibrillation. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Syeda and Holmes contributed equally to this work.

inventors on a patent (WO2015/140571) held by the University of Birmingham on genotype-specific antiarrhythmic drug therapy of atrial fibrillation. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Syeda and Holmes contributed equally to this work. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section as well as supplemental figures and a table, please see the online version of this article. Figure 1 Atrial Arrhythmia Inducibility in Pitx2c+/– Murine Whole Hearts

inventors on a patent (WO2015/140571) held by the University of Birmingham on genotype-specific antiarrhythmic drug therapy of atrial fibrillation. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Syeda and Holmes contributed equally to this work. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section as well as supplemental figures and a table, please see the online version of this article. Figure 1 Atrial Arrhythmia Inducibility in Pitx2c+/– Murine Whole Hearts (A) Image and schematic representation of the Langendorff-perfused heart. (B) Atrial arrhythmia inducibility in isolated, beating hearts from wild-type (WT) and reduced paired like homeodomain 2 messenger ribonucleic acid (Pitx2c+/–) mice. Flecainide abolished atrial arrhythmia inducibility in Pitx2c+/– hearts only. *p < 0.05 flecainide versus baseline. (C) Representative trace of atrial fibrillation (AF) induced during programmed stimulation at baseline, showing reduced severity of arrhythmias with 1 μmol/l flecainide in Pitx2c+/– atria. (D) Effects of flecainide on atrial effective refractory period (ERP) in wild-type and Pitx2c+/– isolated, beating hearts. Shown is the difference in atrial ERP between baseline and 1 μmol/l flecainide at 80- to 120-ms paced cycle length following a single extrastimulus (S2) in WT and Pitx2c+/– isolated, beating hearts. *p < 0.05 between genotypes across all cycle lengths. (E) Whereas flecainide prolonged ERP in both genotypes, this effect was more pronounced in Pitx2c+/– atria. Flecainide caused post-repolarization refractoriness (PRR), the difference between ERP (orange and grey lines) and APD90(blue lines), in WT and Pitx2c+/– atria. Flecainide-induced PRR in Pitx2c+/– is almost 3 times that of WT atria. *p < 0.05 WT versus Pitx2c+/–. #p < 0.05 baseline versus 1 μmol/l flecainide. APD = action potential duration; EG = intracardiac electrogram; EP = electrophysiology; LA = left atrium; LV = left ventricle; MAP = monophasic action potential; RA = right atrium; RV = right ventricle.

+/– is almost 3 times that of WT atria. *p < 0.05 WT versus Pitx2c+/–. #p < 0.05 baseline versus 1 μmol/l flecainide. APD = action potential duration; EG = intracardiac electrogram; EP = electrophysiology; LA = left atrium; LV = left ventricle; MAP = monophasic action potential; RA = right atrium; RV = right ventricle. Figure 1Figure 2 Resting Membrane Potential and Maximum Upstroke Velocity (A) Schematic representation of sharp microelectrode electrode recordings from the superfused whole LA. (B) Resting membrane potential (RMP) in WT and Pitx2c+/– LA at baseline and with 1 μmol/l flecainide. Pitx2c+/– LA have depolarized RMP. The difference between WT and Pitx2c+/– is exaggerated with flecainide. *p < 0.05; ***p < 0.001 across all cycle lengths. (C) Flecainide unmasked a lower maximum upstroke velocity (dV/dtmax) in Pitx2c+/– LA compared to WT. **p < 0.01 versus WT. Abbreviations as in Figure 1. Figure 2Figure 3 Modeling of the Electrophysiological Consequences of Pitx2c Deficiency

(A) Schematic representation of sharp microelectrode electrode recordings from the superfused whole LA. (B) Resting membrane potential (RMP) in WT and Pitx2c+/– LA at baseline and with 1 μmol/l flecainide. Pitx2c+/– LA have depolarized RMP. The difference between WT and Pitx2c+/– is exaggerated with flecainide. *p < 0.05; ***p < 0.001 across all cycle lengths. (C) Flecainide unmasked a lower maximum upstroke velocity (dV/dtmax) in Pitx2c+/– LA compared to WT. **p < 0.01 versus WT. Abbreviations as in Figure 1. Figure 2Figure 3 Modeling of the Electrophysiological Consequences of Pitx2c Deficiency (A) Propagated action potentials (top) simulated with the Courtemanche–Ramirez–Nattel model modified to reflect the more positive RMP of murine Pitx2c+/– atria and the effect of 60% sodium current (INa) block at pacing cycle lengths of 500 and 1,000 ms, with 2-min pre-pacing, and corresponding time courses of the product of the 2 inactivation gates h and j of INa(bottom), reflecting INa availability. (B) Reduced sodium conductance (gNa) increased post-repolarization refractoriness (PRR) in the reference model and the Pitx2c deficiency model. After reducing gNa, PRR was greater in the Pitx2c deficiency model than in the reference model (grey lines). Lines denote APD60(blue) and ERP (orange: with 100% gNa; grey: with 40% gNa). Abbreviations as in Figures 1 and 2. Figure 3Figure 4 mRNA and Protein Expression in Murine LA

(A) Propagated action potentials (top) simulated with the Courtemanche–Ramirez–Nattel model modified to reflect the more positive RMP of murine Pitx2c+/– atria and the effect of 60% sodium current (INa) block at pacing cycle lengths of 500 and 1,000 ms, with 2-min pre-pacing, and corresponding time courses of the product of the 2 inactivation gates h and j of INa(bottom), reflecting INa availability. (B) Reduced sodium conductance (gNa) increased post-repolarization refractoriness (PRR) in the reference model and the Pitx2c deficiency model. After reducing gNa, PRR was greater in the Pitx2c deficiency model than in the reference model (grey lines). Lines denote APD60(blue) and ERP (orange: with 100% gNa; grey: with 40% gNa). Abbreviations as in Figures 1 and 2. Figure 3Figure 4 mRNA and Protein Expression in Murine LA (A) Relative quantity (RQ) of messenger ribonucleic acid (mRNA) of selected ion channels in LA from Pitx2c+/– and WT mice. Expression levels were measured relative to WT sample 1. Kcnk5 encodes TWIK-related acid-sensitive K+ channel 2 (TASK-2), Kcna6 encodes Kv1.6, and Scn5a encodes Nav1.5. *p < 0.05. (B) Concentration of TASK-2, KV1.6, and Nav1.5 proteins relative to calnexin (arbitrary units). Representative immunoblots are displayed below the corresponding dot plot. *p < 0.05. Abbreviations as in Figures 1 and 2. Figure 4Figure 5 Effect of Membrane Potential on Sodium-Channel Inhibition

(A) Relative quantity (RQ) of messenger ribonucleic acid (mRNA) of selected ion channels in LA from Pitx2c+/– and WT mice. Expression levels were measured relative to WT sample 1. Kcnk5 encodes TWIK-related acid-sensitive K+ channel 2 (TASK-2), Kcna6 encodes Kv1.6, and Scn5a encodes Nav1.5. *p < 0.05. (B) Concentration of TASK-2, KV1.6, and Nav1.5 proteins relative to calnexin (arbitrary units). Representative immunoblots are displayed below the corresponding dot plot. *p < 0.05. Abbreviations as in Figures 1 and 2. Figure 4Figure 5 Effect of Membrane Potential on Sodium-Channel Inhibition (A) Schematic representation of patch clamp experiments carried out in isolated atrial cardiomyocytes. Human embryonic kidney (HEK) cells were patch clamped using the same setup. (B) Representative traces showing INa measured by patch clamping in isolated atrial cardiomyocytes, at baseline and with flecainide, at increasing holding potentials (–100 to –65 mV). (C) Reduction in peak INa was enhanced with increased holding potentials, irrespective of cardiomyocyte origin. (D) Representative traces showing INa in Nav1.5-transfected HEK cells at baseline and with flecainide at increasing holding potentials (–100 to –65 mV). (E) Reduction in peak INa with flecainide in Nav1.5-transfected HEK cells was enhanced with increased holding potentials. The greatest difference in percent reduction was between –70 and –65 mV. Abbreviations as in Figures 1, 2, and 3. Figure 5Figure 6 Currents Responsible for RMP

(A) Schematic representation of patch clamp experiments carried out in isolated atrial cardiomyocytes. Human embryonic kidney (HEK) cells were patch clamped using the same setup. (B) Representative traces showing INa measured by patch clamping in isolated atrial cardiomyocytes, at baseline and with flecainide, at increasing holding potentials (–100 to –65 mV). (C) Reduction in peak INa was enhanced with increased holding potentials, irrespective of cardiomyocyte origin. (D) Representative traces showing INa in Nav1.5-transfected HEK cells at baseline and with flecainide at increasing holding potentials (–100 to –65 mV). (E) Reduction in peak INa with flecainide in Nav1.5-transfected HEK cells was enhanced with increased holding potentials. The greatest difference in percent reduction was between –70 and –65 mV. Abbreviations as in Figures 1, 2, and 3. Figure 5Figure 6 Currents Responsible for RMP (A) Representative raw traces showing IK1 currents (50 μmol/l Ba2+ sensitive) in isolated WT and Pitx2c+/– LA cardiomyocytes at test potentials ranging from –120 to 50mV. (B) No difference is seen in the Ik1 current/voltage relationship for LA cardiomyocytes between WT and Pitx2c+/– LA at test potentials ranging from –120 to 50 mV. (C) Representative raw traces show background current response to 10 mmol/l Ba2+ in isolated WT and Pitx2c+/– LA cardiomyocytes. (D)Pitx2c+/– cardiomyocytes had significantly reduced background K+ (10 mmol/l Ba2+ sensitive) currents than WT cardiomyocytes. *p < 0.05. Abbreviations as in Figures 1, 2, and 3.

0 to 50 mV. (C) Representative raw traces show background current response to 10 mmol/l Ba2+ in isolated WT and Pitx2c+/– LA cardiomyocytes. (D)Pitx2c+/– cardiomyocytes had significantly reduced background K+ (10 mmol/l Ba2+ sensitive) currents than WT cardiomyocytes. *p < 0.05. Abbreviations as in Figures 1, 2, and 3. Figure 6Central Illustration PITX2 Deficiency Potentiates Flecainide's Antiarrhythmic Effects Paired like homeodomain 2 (PITX2) might play an important role in regulating gene expression and electrical function of the left atrium. (Left)Pitx2 messenger ribonucleic acid (mRNA) levels are variable in left atrial appendages of patients undergoing thoracoscopic atrial fibrillation (AF) ablation therapy. The differentiation between “low” (orange) and “high” (blue)PITX2 levels is somewhat arbitrary. (Right) A low left atrial PITX2c mRNA expression slightly depolarizes left atrial resting membrane potential (RMP). A depolarized RMP, in turn, enhances the antiarrhythmic effect of sodium-channel blockers such as flecainide. PRR = post-repolarization refractoriness. Central IllustrationTable 1 Baseline Characteristics (N= 101)∗

Paired like homeodomain 2 (PITX2) might play an important role in regulating gene expression and electrical function of the left atrium. (Left)Pitx2 messenger ribonucleic acid (mRNA) levels are variable in left atrial appendages of patients undergoing thoracoscopic atrial fibrillation (AF) ablation therapy. The differentiation between “low” (orange) and “high” (blue)PITX2 levels is somewhat arbitrary. (Right) A low left atrial PITX2c mRNA expression slightly depolarizes left atrial resting membrane potential (RMP). A depolarized RMP, in turn, enhances the antiarrhythmic effect of sodium-channel blockers such as flecainide. PRR = post-repolarization refractoriness. Central IllustrationTable 1 Baseline Characteristics (N= 101)∗ Table 1Age, yrs 59.7 ± 8.4 (40–76) Male 79 Congestive heart failure 6 Hypertension 34 Age ≥75 yrs 1 Diabetes 9 Stroke/transient ischemic attack/embolus 10 Vascular disease 10 Female 22 Age ≥65 yrs 31 CHA2DS2-VASc score 0 60 1 24 ≥2 17 Previous catheter ablation for AF 20 Type of AF Paroxysmal 44 Persistent 56 Longstanding persistent 1 AF duration, yrs 6.0 (1–35) Antiarrhythmic drugs and rate control agents Quinidine or disopyramide 4 Flecainide or propafenone 33 Amiodarone, dronedarone, or sotalol 41 Beta blockers 53 Verapamil or diltiazem 17 Digoxin 15 Anticoagulant agents (before PVI procedure) Vitamin K antagonists 89 Antiplatelets 6 Values are mean ± SD (range), n, or mean (range). PVI = pulmonary vein isolation. ∗ Left atrial appendages were collected from these patients with atrial fibrillation (AF).

Table 1Age, yrs 59.7 ± 8.4 (40–76) Male 79 Congestive heart failure 6 Hypertension 34 Age ≥75 yrs 1 Diabetes 9 Stroke/transient ischemic attack/embolus 10 Vascular disease 10 Female 22 Age ≥65 yrs 31 CHA2DS2-VASc score 0 60 1 24 ≥2 17 Previous catheter ablation for AF 20 Type of AF Paroxysmal 44 Persistent 56 Longstanding persistent 1 AF duration, yrs 6.0 (1–35) Antiarrhythmic drugs and rate control agents Quinidine or disopyramide 4 Flecainide or propafenone 33 Amiodarone, dronedarone, or sotalol 41 Beta blockers 53 Verapamil or diltiazem 17 Digoxin 15 Anticoagulant agents (before PVI procedure) Vitamin K antagonists 89 Antiplatelets 6 Values are mean ± SD (range), n, or mean (range). PVI = pulmonary vein isolation. ∗ Left atrial appendages were collected from these patients with atrial fibrillation (AF). Table 2 PITX2 mRNA Expression in Left Atrial Appendages From AF Ablation Patients∗ Table 2Risk Alleles 25% IQR Median 75% IQR Mean SEM No. of Patients 0 3.22 3.69 5.22 4.04 0.6 3 1 2.96 4.25 6.25 4.54 0.5 13 2 2.65 3.78 4.75 3.94 0.3 22 3 2.74 3.72 4.92 3.83 0.4 17 4 3.00 4.29 5.41 4.39 0.5 10 5 1.96 2.66 4.66 3.10 0.7 4 6 4.95 4.95 4.95 4.95 0.0 1 IQR = interquartile range; LA = left atrium; other abbreviations as in Table 1. ∗ This dataset was grouped according to the number of risk single nucleotide polymorphism (SNP) alleles for AF on chromosome 4q25 (rs2200733, SNP2 rs6838973, rs1448818 [13]). Although PITX2 mRNA is numerically lower in patients with 5 or 6 risk alleles, we did not find a PITX2 mRNA gradient according to AF risk.

Table 2Risk Alleles 25% IQR Median 75% IQR Mean SEM No. of Patients 0 3.22 3.69 5.22 4.04 0.6 3 1 2.96 4.25 6.25 4.54 0.5 13 2 2.65 3.78 4.75 3.94 0.3 22 3 2.74 3.72 4.92 3.83 0.4 17 4 3.00 4.29 5.41 4.39 0.5 10 5 1.96 2.66 4.66 3.10 0.7 4 6 4.95 4.95 4.95 4.95 0.0 1 IQR = interquartile range; LA = left atrium; other abbreviations as in Table 1. ∗ This dataset was grouped according to the number of risk single nucleotide polymorphism (SNP) alleles for AF on chromosome 4q25 (rs2200733, SNP2 rs6838973, rs1448818 [13]). Although PITX2 mRNA is numerically lower in patients with 5 or 6 risk alleles, we did not find a PITX2 mRNA gradient according to AF risk. Table 3 Effect of Flecainide on Refractoriness and Repolarization in Mouse Hearts

∗ This dataset was grouped according to the number of risk single nucleotide polymorphism (SNP) alleles for AF on chromosome 4q25 (rs2200733, SNP2 rs6838973, rs1448818 [13]). Although PITX2 mRNA is numerically lower in patients with 5 or 6 risk alleles, we did not find a PITX2 mRNA gradient according to AF risk. Table 3 Effect of Flecainide on Refractoriness and Repolarization in Mouse Hearts Table 3Paced CL, ms Wild-Type Pitx2c+/– 120 100 80 120 100 80 Baseline Flecainide Baseline Flecainide Baseline Flecainide Baseline Flecainide Baseline Flecainide Baseline Flecainide LA ERP, ms 23.5 ± 2.3 (11) 29.8 ± 3.0 (11) 22.2 ± 2.1 (11) 29.6 ± 3.3 (11) 21.9 ± 2.4 (10) 28.7 ± 3.5∗ (10) 30.5 ± 2.4 (11) 38.5 ± 3.3∗ (11) 28.0 ± 2.3 (13) 40.2 ± 2.8∗ (13) 27.5 ± 2.5 (13) 41.2 ± 3.0∗† (13) LA monophasic APD, ms APD50 10.2 ± 1.3 (8) 14.5 ± 1.7 (8) 10.8 ± 1 (8) 11.9 ± 1.6 (8) 10.4 ± 0.7 (7) 12.0 ± 1.1 (7) 12.4 ± 1.1 (15) 14.4 ± 1.3 (15) 11.5 ± 1.0 (15) 12.4 ± 1.1 (15) 10.6 ± 0.9 (11) 10.3 ± 1.0 (11) APD70 17.8 ± 2.2 (9) 23 ± 2.1 (9) 18.4 ± 1.6 (9) 18.7 ± 2.2 (9) 18.1 ± 1.2 (8) 18.1 ± 1.9 (8) 18.0 ± 1.6 (15) 19.2 ± 1.8 (15) 16.0 ± 1.4 (13) 16.2 ± 1.4 (13) 14.9 ± 1.0† (10) 13.1 ± 0.7 (10) APD90 31.3 ± 3.0 (8) 37.4 ± 2.8 (8) 31.5 ± 2.5 (9) 29.9 ± 2.9 (9) 31.0 ± 1.4 (8) 28.4 ± 2.7 (8) 28.3 ± 2.2 (13) 29.9 ± 2.2 (13) 27.4 ± 2.2 (13) 26.6 ± 1.5 (13) 26.8 ± 1.7 (10) 23.1 ± 1.4 (10) LA transmembrane APD, ms APD30 4.5 ± 0.1 (30) 5.5 ± 0.3 (22) 4.5 ± 0.1 (30) 5.4 ± 0.3 (22) 4.4 ± 0.1 (30) 5.2 ± 0.3 (22) 4.0 ± 0.1 (31) 4.7 ± 0.2 (24) 3.9 ± 0.1 (31) 4.6 ± 0.2 (24) 3.8 ± 0.1† (31) 4.4 ± 0.2 (24) APD50 6.7 ± 0.2 (30) 8.2 ± 0.5 (22) 6.6 ± 0.2 (30) 8.0 ± 0.4 (22) 6.4 ± 0.2 (30) 7.8 ± 0.5 (22) 5.9 ± 0.2 (31) 7.1 ± 0.4 (24) 5.7 ± 0.2 (31) 7.0 ± 0.4 (24) 5.6 ± 0.2† (31) 6.7 ± 0.3 (24) APD70 10.5 ± 0.4 (30) 12.7 ± 0.8 (22) 10.1 ± 0.4 (30) 12.1 ± 0.7 (22) 9.6 ± 0.4 (30) 11.8 ± 0.7 (22) 8.9 ± 0.4 (31) 10.7 ± 0.6 (24) 8.6 ± 0.4 (31) 10.3 ± 0.6 (24) 8.3 ± 0.3† (31) 9.8 ± 0.5 (24) APD90 20.9 ± 1.0 (30) 23.4 ± 1.5 (22) 19.9 ± 0.9 (30) 22.2 ± 1.4 (22) 18.4 ± 0.8 (30) 21.6 ± 1.3 (22) 17.6 ± 0.9 (31) 20.3 ± 1.1 (24) 16.5 ± 0.8 (31) 19.2 ± 1.0 (24) 15.7 ± 0.8† (31) 17.9 ± 0.9 (24) LA optical APD, ms APD30 6.1 ± 0.3 (10) 7.3 ± 0.6 (6) 6.4 ± 0.8 (10) 5.9 ± 1.0 (6) 6.1 ± 0.4 (10) 6.9 ± 1.3 (6) 4.9 ± 0.4 (10) 7.7 ± 0.9 (8) 4.6 ± 0.3 (10) 5.4 ± 0.7 (8) 4.3 ± 0.4† (10) 5.7 ± 0.7 (8) APD50 8.5 ± 0.6 (10) 10.7 ± 1.2 (6) 8.9 ± 1.1 (10) 8.5 ± 1.2 (6) 8.3 ± 0.7 (10) 10.3 ± 1.8 (6) 6.9 ± 0.4 (10) 10.0 ± 1.0 (8) 6.6 ± 0.4 (10) 8.1 ± 0.9 (8) 6.1 ± 0.4† (10) 8.0 ± 0.9 (8) APD70 11.7 ± 1.2 (10) 15.0 ± 2.1 (6) 12

1.3 (6) 4.9 ± 0.4 (10) 7.7 ± 0.9 (8) 4.6 ± 0.3 (10) 5.4 ± 0.7 (8) 4.3 ± 0.4† (10) 5.7 ± 0.7 (8) APD50 8.5 ± 0.6 (10) 10.7 ± 1.2 (6) 8.9 ± 1.1 (10) 8.5 ± 1.2 (6) 8.3 ± 0.7 (10) 10.3 ± 1.8 (6) 6.9 ± 0.4 (10) 10.0 ± 1.0 (8) 6.6 ± 0.4 (10) 8.1 ± 0.9 (8) 6.1 ± 0.4† (10) 8.0 ± 0.9 (8) APD70 11.7 ± 1.2 (10) 15.0 ± 2.1 (6) 12 .5 ± 1.5 (10) 12.8 ± 1.7 (6) 11.5 ± 1.1 (10) 14.4 ± 2.5 (6) 9.4 ± 0.0 (10) 13.3 ± 1.2 (8) 9.4 ± 0.6 (10) 11.6 ± 1.5 (8) 9.1 ± 0.5† (10) 11.2 ± 1.2 (8) Values are mean ± SEM (number of atria). APD = action potential duration; CL = cycle length; ERP = effective refractory period; Pitx2c+/– = PITX2 deficient; other abbreviations as in Tables 1 and 2. ∗ p < 0.05 vs. baseline. † p < 0.05 vs. wild-type. Table 4 Electrophysiological Effects of Sotalol Table 4Paced CL, ms Wild-Type Pitx2c+/– 120 100 120 100 Baseline Sotalol Baseline Sotalol Baseline Sotalol Baseline Sotalol LA ERP, ms 38.7 ± 7.8 (7) 33.9 ± 6.3 (7) 32.2 ± 6.1 (6) 29.2 ± 5.3 (6) 39.3 ± 4.0 (4) 26.8 ± 3.5 (4) 37.0 ± 5.7 (4) 24.0 ± 3.7 (4) LA APD, ms APD50 11.5 ± 1.2 (9) 13.4 ± 1.2 (9) 10.9 ± 2.0 (7) 12.2 ± 1.3 (7) 10.8 ± 1.1 (7) 11.2 ± 1.0 (7) 8.3 ± 0.9 (4) 11.1 ± 1.7 (4) APD70 17.6 ± 2.2 (9) 20.0 ± 1.9 (9) 16.0 ± 1.3 (7) 18.2 ± 2.3 (7) 16.5 ± 1.6 (7) 17.3 ± 1.2 (7) 13.0 ± 1.5 (4) 17.0 ± 1.9 (4) APD90 30.7 ± 3.2 (9) 33.5 ± 2.7 (9) 29.0 ± 1.9 (7) 30.9 ± 3.2 (7) 29.7 ± 2.7 (7) 31.2 ± 2.0 (7) 23.8 ± 2.8 (4) 29.6 ± 2.9 (4) Values are mean ± SEM (number of atria). Abbreviations as in Tables 2 and 3. Table 5 Electrophysiological Effects of Reduced Sodium Conductance in a Human Atrial Model

Table 4Paced CL, ms Wild-Type Pitx2c+/– 120 100 120 100 Baseline Sotalol Baseline Sotalol Baseline Sotalol Baseline Sotalol LA ERP, ms 38.7 ± 7.8 (7) 33.9 ± 6.3 (7) 32.2 ± 6.1 (6) 29.2 ± 5.3 (6) 39.3 ± 4.0 (4) 26.8 ± 3.5 (4) 37.0 ± 5.7 (4) 24.0 ± 3.7 (4) LA APD, ms APD50 11.5 ± 1.2 (9) 13.4 ± 1.2 (9) 10.9 ± 2.0 (7) 12.2 ± 1.3 (7) 10.8 ± 1.1 (7) 11.2 ± 1.0 (7) 8.3 ± 0.9 (4) 11.1 ± 1.7 (4) APD70 17.6 ± 2.2 (9) 20.0 ± 1.9 (9) 16.0 ± 1.3 (7) 18.2 ± 2.3 (7) 16.5 ± 1.6 (7) 17.3 ± 1.2 (7) 13.0 ± 1.5 (4) 17.0 ± 1.9 (4) APD90 30.7 ± 3.2 (9) 33.5 ± 2.7 (9) 29.0 ± 1.9 (7) 30.9 ± 3.2 (7) 29.7 ± 2.7 (7) 31.2 ± 2.0 (7) 23.8 ± 2.8 (4) 29.6 ± 2.9 (4) Values are mean ± SEM (number of atria). Abbreviations as in Tables 2 and 3. Table 5 Electrophysiological Effects of Reduced Sodium Conductance in a Human Atrial Model Table 5Paced CL, ms Wild-Type Model Pitx2c Deficiency Model 500 1,000 500 1,000 RMP, mV gNa, % 100 -79.92 -81.28 -77.90 -79.61 50 -79.60 -81.12 -77.33 -79.37 40 -79.43 -81.01 -76.99 -79.23 APD at repolarization to -60 mV, ms gNa, % 100 217 253 206 226 50 239 266 233 239 40 248 273 245 245 ERP, ms gNa, % 100 266 301 261 280 50 308 335 316 320 40 327 352 342 339 PRR, ms gNa, % 100 49 48 55 54 50 69 69 83 81 40 79 79 97 94 Conduction velocity, cm/s gNa, % 100 49.5 50.0 50.3 50.5 50 36.9 37.0 37.5 37.5 40 32.9 32.7 33.0 33.1 gNa = reduced sodium conductance; PRR = post-repolarization refractoriness; RMP = resting membrane potential; other abbreviations as in Table 3. Table 6 Electrical Activation Time and Conduction Velocity in Isolated Atria in the Presence of Flecainide (1 μmol/l)

Table 5Paced CL, ms Wild-Type Model Pitx2c Deficiency Model 500 1,000 500 1,000 RMP, mV gNa, % 100 -79.92 -81.28 -77.90 -79.61 50 -79.60 -81.12 -77.33 -79.37 40 -79.43 -81.01 -76.99 -79.23 APD at repolarization to -60 mV, ms gNa, % 100 217 253 206 226 50 239 266 233 239 40 248 273 245 245 ERP, ms gNa, % 100 266 301 261 280 50 308 335 316 320 40 327 352 342 339 PRR, ms gNa, % 100 49 48 55 54 50 69 69 83 81 40 79 79 97 94 Conduction velocity, cm/s gNa, % 100 49.5 50.0 50.3 50.5 50 36.9 37.0 37.5 37.5 40 32.9 32.7 33.0 33.1 gNa = reduced sodium conductance; PRR = post-repolarization refractoriness; RMP = resting membrane potential; other abbreviations as in Table 3. Table 6 Electrical Activation Time and Conduction Velocity in Isolated Atria in the Presence of Flecainide (1 μmol/l) Table 6 Wild-Type Pitx2c+/– Paced CL, ms 1,000 300 120 100 80 1,000 300 120 100 80 Activation time (isolated left atrium), ms 6 ± 0.3 (22) 6 ± 0.3 (22) 9 ± 0.5 (22) 12 ± 1.0 (22) 16 ± 1.4 (22) 6 ± 0.2 (24) 7 ± 0.3 (24) 12 ± 0.9 (24) 13 ± 0.9 (24) 18 ± 1.2 (24) Conduction velocity (optical mapping), cm/s — 30 ± 1.8 (8) 25 ± 2.4 (8) 25 ± 1.9 (8) 23 ± 1.9 (8) — 29 ± 1.5 (8) 26 ± 1.6 (8) 25 ± 1.6 (8) 23 ± 1.6 (8) Values are mean ± SEM (number of atria). — = not applicable; other abbreviations as in Table 3.

sterol, total cholesterol, atrial fibrillation, inflammatory conditions (autoimmune conditions, chronic obstructive pulmonary disease, or inflammatory bowel disease), cancer, statin use, blood pressure medication, and acute conditions at the time of blood testing. The baseline hazard was stratified by practice and sex. We plotted Schoenfeld residuals to assess the proportional hazards assumption, and split the follow-up time if HRs changed over time. We examined for interactions with age, sex, smoking, and acute clinical state. We handled missing baseline covariate data by multiple imputation using random forest (18), as described in the Online Appendix. We explored additional adjustments for eosinophil and lymphocyte counts, which are also associated with incidence of CVD (14) (each was included as a 5-category variable). We used a Bonferroni correction for 14 comparisons to designate the level of statistical significance as 0.0036, and expressed p values in categories: <0.05 (suggestive of a trend), <0.0036 (statistically significant), and <0.0001 (strong evidence). Results We included 621,052 patients with neutrophil counts while clinically stable and 154,179 patients with neutrophil counts performed during acute illness or treatment (Online Figure 1, Online Table 2). We observed 55,004 initial presentations of CVD over a median follow-up of 3.8 years (interquartile range [IQR]: 1.7 to 6.0 years).

Alterations in lipid metabolism underlie atherosclerotic cardiovascular disease (CVD) (1). The low-density lipoprotein (LDL)–cholesterol axis is an established target in CVD prevention. In contrast to genetic disorders that lead to higher or lower LDL cholesterol (LDL-C), genetic polymorphisms that affect high-density lipoprotein (HDL) cholesterol (HDL-C) have not consistently translated into altered cardiovascular risk. Thus far, explanations for CVD risk have primarily focused on quantities of only a few lipids within established lipoprotein classes, such as LDL-C and very low-density lipoprotein (VLDL) triglycerides (TGs), and, for the most part, ignored other lipid species (1). Yet individual molecular lipid species within the same lipid class display different associations with CVD (1). Although lipoproteins are defined by their flotation properties during ultracentrifugation, their functions and metabolism are principally governed by their apolipoprotein content. However, no comprehensive analysis of plasma apolipoproteins and lipids has been performed in the same cohort to assess their comparative association with future CVD in the general community.

properties during ultracentrifugation, their functions and metabolism are principally governed by their apolipoprotein content. However, no comprehensive analysis of plasma apolipoproteins and lipids has been performed in the same cohort to assess their comparative association with future CVD in the general community. Here, we capitalized on recent advances in mass spectrometry (MS) by applying multiple-reaction monitoring MS (MRM-MS) to the prospective, population-based Bruneck Study. We measured 13 plasma apolipoproteins by use of spiked-in, stable isotope–labeled standards, integrated the apolipoprotein panel with clinical, proteomic, and lipidomic measurements, and analyzed their predictive value for CVD. Unexpectedly, after multivariate analysis, VLDL-associated apolipoproteins and predominant lipids emerged as the strongest determinants of CVD risk. We also show that these VLDL-associated apolipoproteins and their corresponding lipid species can be reduced by targeting apolipoprotein C-III (apoC-III), a central regulator of plasma triglyceride-rich lipoprotein (TRL) metabolism 2, 3. Methods An expanded Methods section is available in the Online Appendix. Associations of 13 apolipoproteins, 135 lipid species, and 211 other plasma proteins with incident CVD were assessed prospectively over a 10-year period in the Bruneck Study. Changes in apolipoprotein and lipid levels following treatment with volanesorsen, a second-generation antisense drug targeting apoC-III, were determined in 2 human intervention trials (IONIS1 and IONIS2), one of which was randomized.

h incident CVD were assessed prospectively over a 10-year period in the Bruneck Study. Changes in apolipoprotein and lipid levels following treatment with volanesorsen, a second-generation antisense drug targeting apoC-III, were determined in 2 human intervention trials (IONIS1 and IONIS2), one of which was randomized. Bruneck study The Bruneck Study is a prospective, population-based survey of the epidemiology and pathogenesis of atherosclerosis and CVD (1). An age- and sex-stratified random sample of all inhabitants of Bruneck, Italy, all of Caucasian descent, was enrolled in 1990. In 2000, 702 subjects were still alive, and participated in the second quinquennial follow-up. Measurements taken in 2000 served as the baseline for the present study. Detailed information on fatal and nonfatal CVD was carefully collected for these subjects until 2010, with follow-up 100% complete for clinical outcomes. Clinical outcomes were adjudicated by 1 senior researcher blinded to baseline data. The study protocol was approved by the ethics committees of Bolzano and Verona, and conformed to the Declaration of Helsinki, and all study subjects gave written informed consent. The composite CVD endpoint included incident fatal and nonfatal myocardial infarction, ischemic stroke, and sudden cardiac death. The presence of myocardial infarction was assessed by World Health Organization criteria (4), and ischemic stroke was classified according to the criteria of the National Survey of Stroke (5).

e composite CVD endpoint included incident fatal and nonfatal myocardial infarction, ischemic stroke, and sudden cardiac death. The presence of myocardial infarction was assessed by World Health Organization criteria (4), and ischemic stroke was classified according to the criteria of the National Survey of Stroke (5). MRM-MS in plasma Citrate plasma samples were stored at −80°C until analysis. Upon thawing, peptide standards were spiked into the samples (PlasmaDive kits, Biognosys AG, Schlieren, Switzerland). The peptide standard for apoB-100 did not overlap with the proximal portion of apoB that would include both apoB-48 and apoB-100. An appropriate apo(a) standard was not available, and thus apo(a) levels were not measured. Plasma samples were processed according to the manufacturer’s instructions. Briefly, 10 μl of plasma samples were denatured, reduced, and alkylated. Proteins (20 μg) were spiked with authentic heavy peptide standards. An in-solution digestion was performed overnight. After solid-phase extraction with C18 spin columns (96-well format, Harvard apparatus, Holliston, Massachusetts), the eluted peptides were dried using a SpeedVac (ThermoFisher Scientific, Woburn, Massachusetts) and resuspended in 40 μl of liquid chromatography solution. The samples were analyzed on an Agilent 1290 liquid chromatography system (Agilent Technologies, Santa Clara, California) interfaced to an Agilent 6495 Triple Quadrupole MS (Agilent Technologies). Samples (10 μl) were directly injected onto a 25-cm column (AdvanceBio Peptide Map 2.1 × 250 mm, Agilent Technologies) and separated over a 23-min gradient at 350 μl/min. Data were analyzed using Skyline software version 3.1 (MacCoss Lab, University of Washington, Seattle, Washington) and protein concentrations were calculated using the heavy/light (H/L) ratio.

a 25-cm column (AdvanceBio Peptide Map 2.1 × 250 mm, Agilent Technologies) and separated over a 23-min gradient at 350 μl/min. Data were analyzed using Skyline software version 3.1 (MacCoss Lab, University of Washington, Seattle, Washington) and protein concentrations were calculated using the heavy/light (H/L) ratio. Statistical analysis Baseline characteristics are presented as count (percentage), mean ± SD, or median (interquartile range). Associations with incident endpoints were examined using Cox regression. The proportional hazards assumption was tested using the correlation of Schoenfeld residuals with survival time, and was not refuted. Cox models were progressively adjusted for age, sex, and statin therapy (model 1), plus diabetes mellitus, current smoking, and systolic blood pressure (model 2), plus HDL-C and non–HDL-C (model 3). Cross-sectional analyses on clinical variables and on lipidomic data employed linear regression, adjusted for age, sex, and statin therapy. Correlation analyses used Pearson correlation partial to age, sex, and statin therapy, and log-transformed proteins if their skewness exceeded 2. Lipid variables were clustered using agglomerative hierarchical clustering on the basis of complete linkage, defining the distance between 2 variables as 1 minus their correlation. Cross-sectional analyses of high-dimensional protein data deemed associations significant according to a false discovery rate q value below 0.05. Other results were not adjusted for multiplicity (6).

cal clustering on the basis of complete linkage, defining the distance between 2 variables as 1 minus their correlation. Cross-sectional analyses of high-dimensional protein data deemed associations significant according to a false discovery rate q value below 0.05. Other results were not adjusted for multiplicity (6). To estimate effects of apoC-III synthesis inhibition, for each subject, measurements at day 1 (baseline) and means of measurements at days 57 and 92 (under treatment) were log-transformed, and the change from baseline was calculated as their difference. The mean change from baseline in each group (IONIS1 treated, IONIS2 treated, IONIS2 placebo) was tested by 1-sample Student t tests against a mean of 0. For presentation of effect sizes, the mean change from baseline was transformed from the log scale to a percent scale. Differential changes from baseline in the IONIS2-treated and placebo groups were tested using Mann-Whitney-Wilcoxon tests. The incremental predictive value provided by apolipoprotein measurements was investigated as described in the Online Appendix. Analyses were conducted using R 3.2.0 (R Project for Statistical Computing, Vienna, Austria). The p values are 2-sided, and an alpha level of 0.05 is used.

ed using Mann-Whitney-Wilcoxon tests. The incremental predictive value provided by apolipoprotein measurements was investigated as described in the Online Appendix. Analyses were conducted using R 3.2.0 (R Project for Statistical Computing, Vienna, Austria). The p values are 2-sided, and an alpha level of 0.05 is used. Results Associations of baseline apolipoproteins and lipids with CVD Associations of apolipoproteins with incident CVD (2000 to 2010) were investigated in the population-based Bruneck Study (N = 688). Baseline clinical characteristics are summarized in Online Table 1. Subjects were on average 66 years old, 52% were female, 6.4% reported prior CVD, and 9% were prescribed statins. Among 13 apolipoproteins quantified by MRM-MS, the most significant associations with incident CVD were detected for apoC-II, apoC-III, and apoE (p < 0.001 each, under adjustment for age, sex, and statin therapy) (Figure 1, model 1), followed by apoL-I, apoB-100, and apoH (p ≤ 0.01 each). Additional adjustment for diabetes, systolic blood pressure, and current smoking did not appreciably alter these associations (Figure 1, model 2), but further adjustment for HDL-C and non–HDL-C rendered apoB-100 and apoH nonsignificant, and weakened the associations obtained for apoC-III, apoC-II, and apoE (Figure 1, model 3). The association of TGs with CVD (p < 0.001) also lost significance after adjustment for HDL-C and non–HDL-C (Figure 1). Similar results were obtained for the individual endpoints of stroke and myocardial infarction (Online Figures 1A and 1B, respectively). ApoL-I displayed a strong association specifically with stroke (Figure 1, Online Figures 1A and 1B). Upon exclusion of subjects with prior CVD (Online Figure 2) or of subjects prescribed statins (Online Figure 3), results did not change appreciably.

ke and myocardial infarction (Online Figures 1A and 1B, respectively). ApoL-I displayed a strong association specifically with stroke (Figure 1, Online Figures 1A and 1B). Upon exclusion of subjects with prior CVD (Online Figure 2) or of subjects prescribed statins (Online Figure 3), results did not change appreciably. When investigating whether apoC-III, apoC-II, and apoE could improve on traditional risk factors in 10-year cardiovascular risk prediction (Online Table 2), no significant change in the c-index was found; however, a significantly positive net reclassification index indicated that 12.3% of subjects could be more appropriately classified into the clinically relevant risk categories of 0.0% to 5.0%, 5.0% to 7.5%, or more than 7.5% when including apolipoproteins.

ine Table 2), no significant change in the c-index was found; however, a significantly positive net reclassification index indicated that 12.3% of subjects could be more appropriately classified into the clinically relevant risk categories of 0.0% to 5.0%, 5.0% to 7.5%, or more than 7.5% when including apolipoproteins. Interrelationships between apolipoproteins Correlations among apolipoproteins and standard lipid measures are shown in Figure 2. ApoC-II, apoC-III, apoE, and TGs formed one set of highly intercorrelated variables along with non–HDL-C (which represents VLDL cholesterol [VLDL-C] and LDL-C) and apoB-100 (primarily representative of LDL) (Figure 2, lower left quadrant). Another cluster comprised variables generally more correlated with apoA-I (Figure 2, upper right quadrant). ApoH and apoJ showed moderate correlations with most other variables. ApoC-I, which is known to primarily associate with VLDL, correlated most strongly with the apoB-100 cluster, and more weakly with apoA-I and HDL-C. Given the interrelationships among apolipoproteins, the hazard ratios for apolipoproteins significantly associated with CVD were adjusted for apoC-II, apoC-III, and apoE (Figure 3). Adjustment for apoE weakened the associations of apolipoproteins with CVD, in particular for apoB-100, whereas adjustment for apoC-II or apoC-III rendered all associations nonsignificant.

s, the hazard ratios for apolipoproteins significantly associated with CVD were adjusted for apoC-II, apoC-III, and apoE (Figure 3). Adjustment for apoE weakened the associations of apolipoproteins with CVD, in particular for apoB-100, whereas adjustment for apoC-II or apoC-III rendered all associations nonsignificant. After adjustment for age, sex, and statin therapy, apoC-II, apoC-III, and apoE were related to several environmental and anthropological parameters known to be correlated with TG values, such as body mass index, waist-hip ratio, blood pressure, and alcohol consumption, and metabolic parameters, such as liver function tests, but surprisingly, only weakly to fasting plasma glucose and HbA1c. As expected, they related strongly to TGs, total cholesterol, LDL-C and non–HDL-C, with apoC-II and apoC-III showing stronger associations than apoE (Online Figure 4). Among 211 plasma proteins, apoC-II, apoC-III, and apoE showed correlations for proteins involved in lipid metabolism, blood coagulation, the complement system, or inflammation and immunity (Online Figures 5A to 5C, respectively), many of which were correlated to the liver-specific microRNA, miR-122, indicating a common hepatic origin.

proteins, apoC-II, apoC-III, and apoE showed correlations for proteins involved in lipid metabolism, blood coagulation, the complement system, or inflammation and immunity (Online Figures 5A to 5C, respectively), many of which were correlated to the liver-specific microRNA, miR-122, indicating a common hepatic origin. Interrelationships between apolipoproteins and plasma lipids The correlations of 135 molecular lipid species with apoC-II, apoC-III, and apoE are presented in Figure 4. Lipid species are represented by circles, with their position in the 2-dimensional lipid class graphs determined by their total acyl chain carbon numbers (x-axis) and double-bond content (y-axis). Circle color represents strength and direction (positive or negative) of correlations, and circle size indicates their level of statistical significance. ApoC-II, apoC-III, and apoE showed strong direct associations with cholesterol esters (CEs), phosphatidylcholines (PCs), phosphatidylethanolamines (PEs), and, particularly, triacylglycerols (TAGs).

h and direction (positive or negative) of correlations, and circle size indicates their level of statistical significance. ApoC-II, apoC-III, and apoE showed strong direct associations with cholesterol esters (CEs), phosphatidylcholines (PCs), phosphatidylethanolamines (PEs), and, particularly, triacylglycerols (TAGs). The pattern of lipid species related to these 3 apolipoproteins resembled the adverse lipid signature of CVD that we recently described in the Bruneck Study (1), where associations with CVD risk were most pronounced for TAGs and CEs of lower carbon number and double-bond content (i.e., saturated and monounsaturated fatty acids), and the risk profile was complemented by PEs/PCs, sphingomyelins (both positive), and lysophosphatidylcholines (negative) (Online Figure 6, middle row). Of note, associations with CVD of most lipid species were substantially attenuated upon adjustment for apoC-II and apoC-III (Online Figures 6A and 6B, bottom row) and, to a lesser extent, upon adjustment for apoE (Online Figure 6C). This is consistent with the notion that apoC-II and apoC-III were more strongly correlated with total cholesterol, non–HDL-C, and TGs than apoE (Online Figure 4).

enuated upon adjustment for apoC-II and apoC-III (Online Figures 6A and 6B, bottom row) and, to a lesser extent, upon adjustment for apoE (Online Figure 6C). This is consistent with the notion that apoC-II and apoC-III were more strongly correlated with total cholesterol, non–HDL-C, and TGs than apoE (Online Figure 4). Effect of lowering apoC-III levels on apolipoproteins and lipids in plasma The data presented thus far support the hypothesis that VLDL and their associated lipids and apolipoproteins are atherogenic. It would be of considerable interest, therefore, to examine the consequences of lowering VLDL levels on plasma apolipoproteins and relevant lipids. Most hypolipidemic agents that have been used to lower VLDL levels directly target multiple lipoprotein levels and metabolism. However, antisense therapy is emerging as a novel lipid-lowering strategy, because it can specifically target the synthesis of a single apolipoprotein, and thus enable an examination of the consequences on the entire apolipoprotein and lipid profile. As noted earlier in the text, apoC-III has emerged as a central regulator of TRL metabolism.

apy is emerging as a novel lipid-lowering strategy, because it can specifically target the synthesis of a single apolipoprotein, and thus enable an examination of the consequences on the entire apolipoprotein and lipid profile. As noted earlier in the text, apoC-III has emerged as a central regulator of TRL metabolism. In 2 recent clinical trials, a generation-2+ single-stranded antisense agent that targets hepatic APOC3 messenger RNA, termed volanesorsen, was used to lower plasma apoC-III levels 2, 3. We obtained plasma samples from cohorts of these studies that had varying degrees of marked hypertriglyceridemia before and after volanesorsen therapy, and measured apolipoproteins and selected lipids by MRM-MS. As expected, inhibition of hepatic apoC-III synthesis substantially reduced plasma apoC-III levels in all subjects (mean decreases >75%). Remarkably, this was associated with ∼50% decreases in both apoC-II and apoE, and modest increases in apoA-I, apoA-II, and apoM, whereas levels of apoB-100 did not change, except in the 3 subjects with familial chylomicronemia syndrome (FCS) (IONIS1), who experienced a marked decrease in TGs (Figure 5A). Consistent with this, apoC-III inhibition lowered plasma concentrations of TAG as expected, but also lowered diacylglycerols (DAGs) (Figure 5B).

hereas levels of apoB-100 did not change, except in the 3 subjects with familial chylomicronemia syndrome (FCS) (IONIS1), who experienced a marked decrease in TGs (Figure 5A). Consistent with this, apoC-III inhibition lowered plasma concentrations of TAG as expected, but also lowered diacylglycerols (DAGs) (Figure 5B). Discussion In an analysis of baseline samples from a prospective population-based study, apoC-III, apoC-II, and apoE were the apolipoproteins most strongly related to CVD (Central Illustration). ApoC-II, apoC-III, and apoE are abundant on TRLs, and profoundly modulate their metabolism (7). Our findings are consistent with a role of TRLs in the pathogenesis of CVD. The concept that TRL remnants penetrate the arterial intima where they promote atherosclerosis, similar to LDL 8, 9, is corroborated by recent Mendelian randomization studies causally implicating TRLs in CVD 10, 11, 12, 13, 14, 15.

abolism (7). Our findings are consistent with a role of TRLs in the pathogenesis of CVD. The concept that TRL remnants penetrate the arterial intima where they promote atherosclerosis, similar to LDL 8, 9, is corroborated by recent Mendelian randomization studies causally implicating TRLs in CVD 10, 11, 12, 13, 14, 15. VLDL-associated apolipoproteins Because samples were taken in the fasting state, the predominant TRL captured in our analysis is VLDL. After hepatic secretion, VLDL contains variable amounts of apoC-II/apoC-III/apoE, which in turn have variable and complex effects on the fate of the various VLDL constituents and on plasma TG levels. ApoE and apoC-III have been reported to stimulate hepatic secretion of VLDL in isolated hepatocyte cultures, but there is no evidence for such effects in vivo in humans 16, 17. However, apoE does play an important role in mediating rapid hepatic removal of TRLs by serving as a ligand to mediate binding to hepatic LDL receptors (LDLR) and LDL receptor-related protein 1 (LRP-1) (18). In rare patients who have complete apoE deficiency and/or in subjects homozygous for the E2/E2 apoE alleles, which have decreased affinity for LDLR, marked hypertriglyceridemia and even chylomicronemia can occur (19).

nd to mediate binding to hepatic LDL receptors (LDLR) and LDL receptor-related protein 1 (LRP-1) (18). In rare patients who have complete apoE deficiency and/or in subjects homozygous for the E2/E2 apoE alleles, which have decreased affinity for LDLR, marked hypertriglyceridemia and even chylomicronemia can occur (19). ApoC-II and apoC-III have opposing effects on plasma TG levels. Lipoprotein lipase (LPL) is primarily responsible for the hydrolysis of TGs in TRLs such as VLDL and chylomicrons. In the absence of LPL activity, as occurs in FCS, marked accumulation of both VLDL and chylomicrons occurs, resulting in massive hypertriglyceridemia, with TG values often exceeding 2,000 to 5,000 mg/dl, causing acute pancreatitis. ApoC-II is an obligate activator of LPL, and its absence can lead to FCS (20). Thus, apoC-II promotes VLDL-TG hydrolysis and the generation of smaller and denser VLDL remnants 18, 21. By contrast, apoC-III interferes with apoC-II–mediated activation of LPL, and thereby promotes hypertriglyceridemia (7). Indeed, it was previously thought that inhibiting LPL activity was the primary mechanism by which apoC-III raised plasma TG. However, the recent observation that lowering apoC-III levels in FCS patients by use of volanesorsen dramatically lowered patients’ plasma TG levels demonstrated conclusively that apoC-III also impaired TRL clearance by an LPL-independent pathway 2, 22, 23. This is thought to be due to inhibiting hepatic clearance of TRL lipoproteins mediated by LDLR or LRP-1 18, 21, 23. Thus, it is now apparent that apoC-III regulates TRL metabolism by both an LPL-dependent and LPL-independent pathway, and is thus a central regulator of plasma TG levels 2, 7, 18, 22.

t pathway 2, 22, 23. This is thought to be due to inhibiting hepatic clearance of TRL lipoproteins mediated by LDLR or LRP-1 18, 21, 23. Thus, it is now apparent that apoC-III regulates TRL metabolism by both an LPL-dependent and LPL-independent pathway, and is thus a central regulator of plasma TG levels 2, 7, 18, 22. ApoC-II, apoC-III, and apoE were associated with obesity, hypertension, impaired glucose metabolism, and most strongly with lipid parameters (Online Figure 4), particularly, CEs and TAGs with shorter and more saturated fatty acid chains (Online Figure 6). This pattern is consistent with hepatic de novo lipogenesis, and resembles the lipid signature of CVD previously observed in the Bruneck Study (1). Following adjustment for apoC-II/apoC-III/apoE, associations of lipid species with CVD were substantially attenuated, further emphasizing the relevance of TRLs for CVD risk.

his pattern is consistent with hepatic de novo lipogenesis, and resembles the lipid signature of CVD previously observed in the Bruneck Study (1). Following adjustment for apoC-II/apoC-III/apoE, associations of lipid species with CVD were substantially attenuated, further emphasizing the relevance of TRLs for CVD risk. Triglycerides and apoC-III Considerable epidemiological, genetic, and now therapeutic data have emerged to suggest that apoC-III is a central regulator of TRL metabolism 2, 3. It would be the logical preferred therapeutic target for lowering VLDL levels, as inhibiting hepatic VLDL release might theoretically lead to steatosis. Similarly, inhibition of apoE would have the net effect of reducing TRL clearance, and apoC-II is necessary for physiological VLDL and chylomicron TG lipolysis. Furthermore, genome-wide association and Mendelian randomization studies suggest that loss-of-function mutations of apoC-III confer cardiovascular protection 12, 13, 24, 25. The Framingham Study has linked apoC-III, as measured by immunoassays, to incident myocardial infarction or angina pectoris (12). We now provide the first data that apoC-III, along with apoC-II and apoE, associates significantly and independently with incident stroke and myocardial infarction on the basis of a direct comparison of apolipoprotein levels by MRM-MS, rather than immunoassays. ApoC-III in VLDL or LDL, as well as total plasma apoC-III, was associated with CVD, but results for HDL-bound apoC-III were ambiguous (26).

tes significantly and independently with incident stroke and myocardial infarction on the basis of a direct comparison of apolipoprotein levels by MRM-MS, rather than immunoassays. ApoC-III in VLDL or LDL, as well as total plasma apoC-III, was associated with CVD, but results for HDL-bound apoC-III were ambiguous (26). Antisense inhibition of apoC-III Antisense therapy that inhibits hepatic apoC-III synthesis results in effective reductions in plasma apoC-III and TG levels. Beneficial effects of apoC-III lowering may extend beyond the impact of lowering plasma TRL levels. ApoC-III loss of function also resulted in decreased LDL-C (12) and apoB-100 (13). Conversely, elevated apoC-III was linked to high levels of the particularly atherogenic small dense LDL (21), HDL dysfunction (27), and promotion of proteoglycan binding of LDL (28). Thus, we examined the effects of apoC-III synthesis inhibition by volanesorsen on apolipoprotein levels in 2 human intervention trials, 1 in FCS subjects (IONIS1), who lack LPL activity (2), and 1 in subjects with prominent hypertriglyceridemia (IONIS2) of varied etiology (Figure 5A) (3). The reduction in apoC-III levels was profound, leading to >75% decreases at the dose of antisense used. This was associated not only with marked reductions in plasma TGs of ∼70%, but there were nearly 50% decreases in both apoC-II and apoE. These changes are consistent with lowering of VLDL and remnant lipoproteins 7, 21, and are in line with observations after apoC-III inhibition in mice, nonhuman primates, and humans 2, 3, 29. However, the disparity between the extent of reduction in apoC-III, apoC-II, and apoE suggests some measures of independence in the metabolism of these 3 apolipoproteins, and it is well known that they may reside on other lipoproteins, such as HDL (30). Indeed, the observed increase in apoA-I, apoA-II, and apoM is consistent with the reported rise of HDL-C following apoC-III inhibition 3, 29 and lower HDL-C levels in apoC-III transgenic mice (31). A potential mechanistic explanation is reduced exchange of HDL-C with VLDL-TG mediated via cholesterol ester transfer protein (CETP) 7, 29. Notably, apoM has been reported to mark an HDL subpopulation that stimulates particularly efficient cholesterol efflux (32).

nd lower HDL-C levels in apoC-III transgenic mice (31). A potential mechanistic explanation is reduced exchange of HDL-C with VLDL-TG mediated via cholesterol ester transfer protein (CETP) 7, 29. Notably, apoM has been reported to mark an HDL subpopulation that stimulates particularly efficient cholesterol efflux (32). Although apoC-III delays clearance of VLDL remnants that contain apoB-100, apoC-III inhibition by volanesorsen did not reduce total apoB-100 levels. This may be explained by the fact that although VLDL-apoB levels were decreased 2, 3, there was a small compensatory increase in LDL. This was likely in part due to CETP-mediated remodeling of the lipoprotein cholesterol content (3), increased conversion of VLDL to LDL mediated by LPL, or changes in so-called metabolic channeling of VLDL to small dense LDL 7, 14, 15, 18. Importantly, total apoB levels did not increase or were slightly decreased in the volanesorsen-treated hypertriglyceridemic subjects who were on other hypolipidemic agents (3). This would explain why only 16% lower apoB-100 levels were reported in carriers of apoC-III loss-of-function mutations with normal TG levels (13), although a recent report of such subjects did not find lower LDL levels (25). Following apoC-III inhibition, TAGs and DAGs were decreased (Figure 5B). Although the former is expected, the latter is notable because DAGs are precursors of TAGs in the last step of TG synthesis (33). Inhibition of hepatic TG synthesis might thus be relevant for the TAG-lowering effects upon apoC-III inhibition, although changes in VLDL-TG secretion were not seen in volanesorsen-treated apoC-III transgenic mice (29). However, lipid metabolism in mice is different from that in humans. For example, mice do not have plasma activity of the CETP that facilitates the exchange of CE for TG between HDL and TG-rich remnant lipoproteins. As a consequence, HDL is the major cholesterol carrying lipoprotein in mice, but not in humans.

ice (29). However, lipid metabolism in mice is different from that in humans. For example, mice do not have plasma activity of the CETP that facilitates the exchange of CE for TG between HDL and TG-rich remnant lipoproteins. As a consequence, HDL is the major cholesterol carrying lipoprotein in mice, but not in humans. The observational and interventional results presented in this study suggest that TG levels are a modifiable CVD risk factor. Although randomized trials testing TG lowering for CVD prevention have reported mixed results (8), a meta-regression of the results shows a dose–effect relation between degree of TG lowering and CVD risk reduction (8). For subjects with high TG levels, this relation was accentuated, and the results in individual studies were also consistently significant (8). Thus, the present study is in line with prior reports suggesting TG lowering as a potential therapeutic approach.

ation between degree of TG lowering and CVD risk reduction (8). For subjects with high TG levels, this relation was accentuated, and the results in individual studies were also consistently significant (8). Thus, the present study is in line with prior reports suggesting TG lowering as a potential therapeutic approach. Study strengths Strengths of this study include its representativeness for the general population, rigorous endpoint evaluation, near-complete follow-up, and comparability of observational and interventional data, facilitated by identical protein measurement methods using MRM-MS, rather than immunoassays. Variability in antibody sensitivity and specificity can hamper direct comparisons of biomarkers using antibody-based assays. At present, MRM-MS is not a high-throughput method, but has the advantage of measuring proteins directly, without a binder. The range of apolipoproteins investigated by MRM-MS for association with incident CVD in a population-based cohort is unprecedented.

r direct comparisons of biomarkers using antibody-based assays. At present, MRM-MS is not a high-throughput method, but has the advantage of measuring proteins directly, without a binder. The range of apolipoproteins investigated by MRM-MS for association with incident CVD in a population-based cohort is unprecedented. Study limitations Weaknesses of this study include that the many correlated tests presented herein were not adjusted for multiplicity, although key results would resist such adjustment, and that use of statins may weaken the association of VLDL- and LDL-associated apolipoproteins with regard to CVD risk; however, <10% of participants in the Bruneck study were on statin therapy, and exclusion of subjects on statin therapy yielded similar results (Online Figure 3). Future studies could extend the present study by measuring apolipoproteins within lipoprotein subfractions. Our findings of strong associations with apolipoproteins, such as apoL-I and apoH, for which published data are lacking, should be considered hypothesis-generating and deserving of further study.

nline Figure 3). Future studies could extend the present study by measuring apolipoproteins within lipoprotein subfractions. Our findings of strong associations with apolipoproteins, such as apoL-I and apoH, for which published data are lacking, should be considered hypothesis-generating and deserving of further study. Conclusions Our data provide strong epidemiological support to the concept that TRLs contribute to atherosclerosis. ApoC-II, apoC-III, and apoE are abundant on VLDL, which may represent an underappreciated risk factor for CVD. The interventional trials with volanesorsen demonstrate that targeting apoC-III favorably affects apolipoprotein and lipid profiles. Thus, lowering VLDL, in addition to LDL and lipoprotein(a), might represent a novel strategy to further reduce CVD risk in the statin era, and could be tested by appropriately designed outcomes trials.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: The apolipoproteins apoC-III, apoC-II, and apoE are found on triglyceride-rich lipoproteins, regulate their metabolism, and outperform other apolipoproteins, including apoB-100 and apoA-I, as predictors of cardiovascular events. TRANSLATIONAL OUTLOOK: Further studies are needed to assess the impact of inhibiting apoC-III synthesis on clinical cardiovascular outcomes. Appendix Online Data

Conclusions Our data provide strong epidemiological support to the concept that TRLs contribute to atherosclerosis. ApoC-II, apoC-III, and apoE are abundant on VLDL, which may represent an underappreciated risk factor for CVD. The interventional trials with volanesorsen demonstrate that targeting apoC-III favorably affects apolipoprotein and lipid profiles. Thus, lowering VLDL, in addition to LDL and lipoprotein(a), might represent a novel strategy to further reduce CVD risk in the statin era, and could be tested by appropriately designed outcomes trials.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: The apolipoproteins apoC-III, apoC-II, and apoE are found on triglyceride-rich lipoproteins, regulate their metabolism, and outperform other apolipoproteins, including apoB-100 and apoA-I, as predictors of cardiovascular events. TRANSLATIONAL OUTLOOK: Further studies are needed to assess the impact of inhibiting apoC-III synthesis on clinical cardiovascular outcomes. Appendix Online Data Dr. Mayr is a Senior Fellow of the British Heart Foundation (BHF). This study was supported by the BHF, Heart Research UK, the Pustertaler Verein zur Prävention von Herz- und Hirngefaesserkrankungen, the Gesundheitsbezirk Bruneck, and the Südtiroler Sanitätsbetrieb, the Province of Bolzano, Italy, and by an excellence initiative (Competence Centers for Excellent Technologies – COMET) of the Austrian Research Promotion Agency FFG: “Research Center of Excellence in Vascular Ageing – Tyrol, VASCage” (K-Project number 843536) funded by the BMVIT, BMWFW, the Wirtschaftsagentur Wien, and the Standortagentur Tirol, and by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust and King’s College London in partnership with King’s College Hospital. The measurements in the interventional trials were funded by Ionis Pharmaceuticals with a research grant to King’s College London. Drs. Tsimikas and Witztum are supported by NIH grants R01-HL119828, P01-HL088093, P01 HL055798, R01-HL106579, R01-HL078610, and R01-HL124174. Medical University Innsbruck and King’s College London have filed a joint patent on cardiometabolic biomarkers. The study sponsors had no role in the collection, analysis, or interpretation of data; in the writing of the report; or in the decision to submit the paper for publication. Drs. Tsimikas and Witztum are co-inventors and receive royalties from patents owned by the University of California at San Diego (UCSD) on oxidation-specific antibodies and on biomarkers related to oxidized lipoproteins. Dr. Tsimikas currently has a dual appointment with UCSD and as an employee of Ionis Pharmaceuticals. Dr. Witztum is a consultant to Ionis Pharmaceuticals, CymaBay, Intercept, and Prometheus; and has stock in Ionis Pharmaceuticals. Dr. Alexander is an employee of Ionis Pharmaceuticals. Dr. Mayr has received research funding from Ionis Pharmaceuticals for measurements related to volanesorsen; and is named inventor on patents for cardiovascular biomarkers, including molecular lipid species.

t, and Prometheus; and has stock in Ionis Pharmaceuticals. Dr. Alexander is an employee of Ionis Pharmaceuticals. Dr. Mayr has received research funding from Ionis Pharmaceuticals for measurements related to volanesorsen; and is named inventor on patents for cardiovascular biomarkers, including molecular lipid species. The other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Pechlaner and Tsimikas contributed equally to this work. Drs. Mayr and Kiechl contributed equally as joint senior authors. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section as well as supplemental figures and tables, please see the online version of this paper. Figure 1 Associations of Apolipoproteins and Lipid Measures With Incident CVD

The other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Pechlaner and Tsimikas contributed equally to this work. Drs. Mayr and Kiechl contributed equally as joint senior authors. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section as well as supplemental figures and tables, please see the online version of this paper. Figure 1 Associations of Apolipoproteins and Lipid Measures With Incident CVD Plasma levels of 13 apolipoproteins and of 4 conventional lipid measures were determined in 688 participants of the Bruneck Study. During 10 years of follow-up, 91 cardiovascular events occurred, comprising stroke, myocardial infarction, and sudden cardiac death. Model 1: Adjustment for age, sex, and statin therapy. Model 2: As in model 1, with additional adjustment for diabetes, systolic blood pressure, and smoking. Model 3: As in model 2, with additional adjustment for HDL-C and non–HDL-C. Quantitatively, for each variable, 1 SD corresponds to: ApoA-I, 607 mg/l; ApoA-II, 6.44 mg/l; ApoA-IV, 15.0 mg/l; ApoB-100, 363 mg/l; ApoC-I, 6.46 mg/l; ApoC-II, 6.30 mg/l; ApoC-III, 25.6 mg/l; ApoD, 7.98 mg/l; ApoE, 9.23 mg/l; ApoH, 38.2 mg/l; ApoL-I, 3.93 mg/l; ApoM, 2.42 mg/l; ApoJ, 23.1 mg/l; HDL-C, 15.2 mg/dl; LDL-C, 36.5 mg/dl; non-HDL-C, 41.4 mg/dl; triglycerides, 77.6 mg/dl. apo = apolipoprotein; CI = confidence interval; CVD = cardiovascular disease; HDL-C = high-density lipoprotein cholesterol; LDL-C = low-density lipoprotein cholesterol.

38.2 mg/l; ApoL-I, 3.93 mg/l; ApoM, 2.42 mg/l; ApoJ, 23.1 mg/l; HDL-C, 15.2 mg/dl; LDL-C, 36.5 mg/dl; non-HDL-C, 41.4 mg/dl; triglycerides, 77.6 mg/dl. apo = apolipoprotein; CI = confidence interval; CVD = cardiovascular disease; HDL-C = high-density lipoprotein cholesterol; LDL-C = low-density lipoprotein cholesterol. Figure 2 Correlations Among Apolipoproteins and Lipids Results are adjusted for age, sex, and statin therapy. Tile color codes for direction and magnitude of correlation, whereas tile text gives its sign and the first 2 decimal digits. Variables are arranged by similarity, as shown in the right-hand dendrogram. Only significant correlations are shown. Clustering gave rise to several groups of highly intercorrelated variables. High-level clusters were characterized by more extensive correlations with apoA-I (top right region) or with apoB-100 (bottom left region), with the latter containing subclusters likely representing VLDL (including apoC-III, apoC-II, apoE, TG, non–HDL-C, and HDL-C) and LDL (including apoB-100 and LDL-C). VLDL = very-low-density lipoprotein; other abbreviations as in Figure 1. Figure 3 Associations of Apolipoproteins With Incident CVD Upon Additional Adjustment for ApoC-II, ApoC-III, or Apo-E

Results are adjusted for age, sex, and statin therapy. Tile color codes for direction and magnitude of correlation, whereas tile text gives its sign and the first 2 decimal digits. Variables are arranged by similarity, as shown in the right-hand dendrogram. Only significant correlations are shown. Clustering gave rise to several groups of highly intercorrelated variables. High-level clusters were characterized by more extensive correlations with apoA-I (top right region) or with apoB-100 (bottom left region), with the latter containing subclusters likely representing VLDL (including apoC-III, apoC-II, apoE, TG, non–HDL-C, and HDL-C) and LDL (including apoB-100 and LDL-C). VLDL = very-low-density lipoprotein; other abbreviations as in Figure 1. Figure 3 Associations of Apolipoproteins With Incident CVD Upon Additional Adjustment for ApoC-II, ApoC-III, or Apo-E Base adjustment consisted of adjustment for age, sex, and statin therapy and is shown for the significant apolipoproteins only in the first column (as in Figure 1). Additional adjustment for apoC-II, apoC-III, or apoE is shown in the other 3 columns, respectively. Note that apoB-100 loses its association with incident CVD upon adjustment for any of the 3 VLDL-associated apolipoproteins. Abbreviations as in Figures 1 and 2. Figure 4 Associations of 135 Plasma Lipid Species With ApoC-II, ApoC-III, or ApoE

Base adjustment consisted of adjustment for age, sex, and statin therapy and is shown for the significant apolipoproteins only in the first column (as in Figure 1). Additional adjustment for apoC-II, apoC-III, or apoE is shown in the other 3 columns, respectively. Note that apoB-100 loses its association with incident CVD upon adjustment for any of the 3 VLDL-associated apolipoproteins. Abbreviations as in Figures 1 and 2. Figure 4 Associations of 135 Plasma Lipid Species With ApoC-II, ApoC-III, or ApoE Individual lipid species are depicted by filled circles and arranged by lipid class in 8 panels according to the number of total carbon atoms (x-axes) and number of double bonds (y-axes). Circle fill color represents the correlation of each lipid species with plasma concentrations of apoC-II, apoC-III, and apoE. Lipids with the same number of carbon atoms and double bonds are pulled apart vertically to increase their visibility. The distinguishing feature in this case is the presence of an alkyl ether linkage, signified in the formula as, for example, PC(O-38:3). Lipids possessing such a linkage are pulled upward, and their alkyl-ether-free counterparts are pulled downward. CE = cholesteryl ester; CI = confidence interval; LPC = lysophosphatidylcholine; LPE = lysophosphatidylethanolamine; PC = phosphatidylcholine; PE = phosphatidylethanolamine; PS = phosphatidylserine; SM = sphingomyelin; TAG = triacylglycerol. Figure 5 Effects of ApoC-III Synthesis Inhibition on Plasma Concentrations of Apolipoproteins and Lipid Classes

Individual lipid species are depicted by filled circles and arranged by lipid class in 8 panels according to the number of total carbon atoms (x-axes) and number of double bonds (y-axes). Circle fill color represents the correlation of each lipid species with plasma concentrations of apoC-II, apoC-III, and apoE. Lipids with the same number of carbon atoms and double bonds are pulled apart vertically to increase their visibility. The distinguishing feature in this case is the presence of an alkyl ether linkage, signified in the formula as, for example, PC(O-38:3). Lipids possessing such a linkage are pulled upward, and their alkyl-ether-free counterparts are pulled downward. CE = cholesteryl ester; CI = confidence interval; LPC = lysophosphatidylcholine; LPE = lysophosphatidylethanolamine; PC = phosphatidylcholine; PE = phosphatidylethanolamine; PS = phosphatidylserine; SM = sphingomyelin; TAG = triacylglycerol. Figure 5 Effects of ApoC-III Synthesis Inhibition on Plasma Concentrations of Apolipoproteins and Lipid Classes In 2 independent intervention trials (IONIS1 and IONIS2), apoC-III synthesis was inhibited with the second-generation antisense oligonucleotide volanesorsen. In the randomized double-blind IONIS2 trial, 11 patients received treatment and 6 received placebo. *p values are for change from baseline in IONIS1, and for differential change from baseline in the treatment and placebo groups in IONIS2. (A) Effect on apolipoproteins. Among the apolipoproteins measured, apoC-III decreased most strongly, followed by apoC-II. An increase in plasma levels was observed for apoA-I, apoA-II, and apoM. (B) Effect on lipid classes. A substantial reduction in plasma levels was observed for TAG and DAG. CER = ceramide; DAG = diacylglycerol; FFA = free fatty acid; HCER = hexosylceramide; LCER = lactosylceramide; other abbreviations as in Figure 4.

. An increase in plasma levels was observed for apoA-I, apoA-II, and apoM. (B) Effect on lipid classes. A substantial reduction in plasma levels was observed for TAG and DAG. CER = ceramide; DAG = diacylglycerol; FFA = free fatty acid; HCER = hexosylceramide; LCER = lactosylceramide; other abbreviations as in Figure 4. Central Illustration Associations of Apolipoproteins With Incident Cardiovascular Disease Plasma levels of 13 apolipoproteins were determined in 688 participants of the Bruneck Study. During 10 years of follow-up, 91 cardiovascular events occurred, comprising stroke, myocardial infarction, and sudden cardiac death. Results are adjusted for age, sex, and statin therapy. Apo = apolipoprotein; CI = confidence interval.

The most numerous type of white blood cell, neutrophils, play a major role in inflammation. Neutrophil count is used routinely as a biomarker of acute infection and inflammation, but not in cardiology. Chronic inflammation contributes to atherosclerosis and cardiovascular diseases (CVD) 1, 2 but, compared with other inflammatory biomarkers such as C-reactive protein (3) and interleukin-6 (4), neutrophil counts have been little studied in relation to long-term CVD risk, even though they are available at scale in clinically collected electronic health record data. Previous studies have shown that high neutrophil counts are associated with an higher incidence of coronary disease (5), heart failure (HF) (6), and stroke (7) (Online Table 1). However, these studies were too small to examine thresholds or shape of the association. No study used a clinically recorded measure of neutrophil count, which is important to understand the relevance of findings to usual practice, or studied associations with peripheral vascular diseases. This study investigated the association of neutrophil counts with initial presentation of 12 CVDs in a large, population-based cohort from a linked electronic health record database: the CALIBER program (Clinical Research Using Linked Bespoke Studies and Electronic Health Records) (8). CALIBER has been extensively validated, replicating known prospective associations of CVDs with sex (9), smoking (10), blood pressure (11), socioeconomic deprivation (12), and type 2 diabetes (13).

h record database: the CALIBER program (Clinical Research Using Linked Bespoke Studies and Electronic Health Records) (8). CALIBER has been extensively validated, replicating known prospective associations of CVDs with sex (9), smoking (10), blood pressure (11), socioeconomic deprivation (12), and type 2 diabetes (13). Methods We used the same study cohort as our study on the association of eosinophil and lymphocyte counts with incidence of CVD (14). The study population was drawn from the CALIBER program (8), which links 4 sources of electronic health data in England: primary care health records (coded diagnoses, clinical measurements, and prescriptions) from 244 general practices contributing to the Clinical Practice Research Datalink; coded hospital discharges (Hospital Episode Statistics); the Myocardial Ischemia National Audit Project (MINAP); and death registrations (Online Appendix). CALIBER includes about 4% of the population of England (15) and is representative in terms of age, sex, ethnicity, and mortality (8). The study period was January 1998 to March 2010, and individuals were eligible for inclusion when they were at least 30 years of age and had been registered for at least 1 year in a practice that met research data recording standards. The study start date (index date) for each participant was the date of the first complete blood count recorded in the primary care record while the participant was eligible. Persons with a prior history of CVD and women with a pregnancy record within 6 months of the start of the study were excluded.

ecording standards. The study start date (index date) for each participant was the date of the first complete blood count recorded in the primary care record while the participant was eligible. Persons with a prior history of CVD and women with a pregnancy record within 6 months of the start of the study were excluded. Approval was granted by the Independent Scientific Advisory Committee of the Medicines and Healthcare Products Regulatory Agency (protocol 12_153R) and the MINAP Academic Group. The main exposure was the neutrophil count (part of the complete blood count) as recorded in primary care. We investigated the neutrophil count initially as a categorical variable to avoid assuming a particular shape for association with CVD. We wished to specifically look at associations with ‘normal’ as well as extreme neutrophil counts; there is no consensus definition for the normal range, but many laboratories quote the range of 2 to 7 × 109/l (16). This lent itself to convenient 5-level categorization within the ‘normal’ range. All category intervals were closed at the lower bound and open at the upper bound (i.e., ‘2 to 3’ includes 2 but not 3).

; there is no consensus definition for the normal range, but many laboratories quote the range of 2 to 7 × 109/l (16). This lent itself to convenient 5-level categorization within the ‘normal’ range. All category intervals were closed at the lower bound and open at the upper bound (i.e., ‘2 to 3’ includes 2 but not 3). White cell counts can be affected by infections, autoimmune diseases, medication, and hematologic conditions. We classified the patient state at the time of the blood test as acute or stable. An acute clinical state was defined as any of the following conditions: in hospital on the date of blood test; vaccination in the previous 7 days; anemia diagnosis within the previous 30 days; symptoms or diagnosis of infection within the previous 30 days; prior diagnosis of myelodysplastic syndrome; hemoglobinopathy, cancer chemotherapy, or injection of granulocyte colony-stimulating factor within the previous 6 months; or the use of drugs affecting the immune system, such as methotrexate or steroids, within the previous 3 months. Patients were considered stable if they did not fulfill the criteria for an acute clinical state. Patients on dialysis, those with human immunodeficiency virus infection, or a history of splenectomy were excluded from this study, because their neutrophil counts may be difficult to interpret. These criteria were based on those proposed by the eMERGE (Electronic Medical Records and Genomics) consortium for studying the genetic determinants of white cell counts (17) (further details in the Online Appendix).

lenectomy were excluded from this study, because their neutrophil counts may be difficult to interpret. These criteria were based on those proposed by the eMERGE (Electronic Medical Records and Genomics) consortium for studying the genetic determinants of white cell counts (17) (further details in the Online Appendix). In secondary analyses, we explored associations between onset of CVD and the mean of the first 2 stable measurements of neutrophil count taken since the start of eligibility. We extracted demographic variables, cardiovascular risk factors, comorbidities, and acute conditions and prescriptions around the time of the blood test from the primary care record. For continuous covariates, we used the most recent value in the year before or up to 1 day after the complete blood count measurement. We also extracted the first measurement after this time window and the last measurement before the time window, along with the timing of these measurements relative to the index date, to use as auxiliary variables for multiple imputation. We also used comorbidity information from hospitalization records.

t measurement. We also extracted the first measurement after this time window and the last measurement before the time window, along with the timing of these measurements relative to the index date, to use as auxiliary variables for multiple imputation. We also used comorbidity information from hospitalization records. Individuals were followed until initial presentation of CVD, death, or transfer out of the practice. The primary endpoint was the first record of 1 of the following 12 initial cardiovascular presentations in any of the data sources: coronary artery disease (stable angina, unstable angina, nonfatal myocardial infarction [MI], unheralded coronary death), HF, transient ischemic attack, ischemic or hemorrhagic stroke, subarachnoid hemorrhage, peripheral arterial disease (PAD), abdominal aortic aneurysm, or a composite of ventricular arrhythmia, implantable cardioverter-defibrillator, cardiac arrest, or sudden cardiac death. Nonspecific coronary artery disease and nonspecific stroke were also included in the analysis as 2 additional endpoints, although they are artefacts of imprecise coding rather than separate disease entities. Any events occurring after the first cardiovascular presentation were ignored. Endpoint definitions are described in the Online Appendix and phenotyping algorithms are available on the CALIBER web portal (website in the Online Appendix). We analyzed all-cause mortality and a composite of all initial cardiovascular presentations as secondary endpoints.

rst cardiovascular presentation were ignored. Endpoint definitions are described in the Online Appendix and phenotyping algorithms are available on the CALIBER web portal (website in the Online Appendix). We analyzed all-cause mortality and a composite of all initial cardiovascular presentations as secondary endpoints. Statistical analysis We examined associations of CVD with neutrophil counts initially as a categorical variable. If the shape of the association was found to be linear, we also performed analyses with neutrophil count as a continuous variable. We generated cumulative incidence curves by category of neutrophil count under a competing risks framework. We used Cox proportional hazards models to estimate cause-specific hazards for each of the 14 cardiovascular endpoints. Hazard ratios (HRs) were adjusted for age (linear and quadratic), sex, age–sex interaction, index of multiple deprivation, ethnicity, smoking status, diabetes, body mass index, systolic blood pressure, estimated glomerular filtration rate, high-density lipoprotein cholesterol, total cholesterol, atrial fibrillation, inflammatory conditions (autoimmune conditions, chronic obstructive pulmonary disease, or inflammatory bowel disease), cancer, statin use, blood pressure medication, and acute conditions at the time of blood testing. The baseline hazard was stratified by practice and sex.

d 621,052 patients with neutrophil counts while clinically stable and 154,179 patients with neutrophil counts performed during acute illness or treatment (Online Figure 1, Online Table 2). We observed 55,004 initial presentations of CVD over a median follow-up of 3.8 years (interquartile range [IQR]: 1.7 to 6.0 years). Patients with higher neutrophil count were more likely to smoke, live in a socioeconomically deprived area, and have comorbidities such as diabetes, asthma, chronic obstructive pulmonary disease, connective tissue diseases, and inflammatory bowel disease (Table 1). Patients with neutrophil counts above or below the limits of normal were more likely to have an acute state at the time of blood testing than those with neutrophil counts within the normal range (27.5% [22,678 of 82,376] vs. 19.0% [131,501 of 692,855]; p < 0.0001). Symptoms or diagnosis of infection were the most frequent reason for the patient’s condition to be classified as acute (Online Table 3). People with neutrophil counts of 6 to 7 × 109/l (at the upper end of the normal range) had a higher incidence of nonfatal MI, unheralded coronary death, HF, PAD, and abdominal aortic aneurysm compared with people with neutrophil counts of 2 to 3 × 109/l (Figure 1). The risk difference appeared to be greatest for the first few months (Online Table 4). Individuals with higher neutrophil counts were relatively more likely to present with unheralded coronary death, HF, or PAD than stable angina (Online Table 5).

ompared with people with neutrophil counts of 2 to 3 × 109/l (Figure 1). The risk difference appeared to be greatest for the first few months (Online Table 4). Individuals with higher neutrophil counts were relatively more likely to present with unheralded coronary death, HF, or PAD than stable angina (Online Table 5). There were strong, specific associations between neutrophil counts and different initial presentations of CVD (Figure 2). Adjusted HRs comparing neutrophil counts 6 to 7 versus 2 to 3 × 109/l showed strong associations with HF (HR: 2.04; 95% confidence interval [CI]: 1.82 to 2.29), unheralded coronary death (HR: 1.78; 95% CI: 1.51 to 2.10), and nonfatal MI (HR: 1.58; 95% CI: 1.42 to 1.76), but not stable angina (HR: 0.97; 95% CI: 0.88 to 1.07) or unstable angina (HR: 1.00; 95% CI: 0.84 to 1.19) (Figure 2). The association with ischemic stroke was weak (HR: 1.36; 95% CI: 1.17 to 1.57) and there was no association with hemorrhagic stroke. There were strong associations with PAD (HR: 1.95; 95% CI: 1.72 to 2.21) and abdominal aortic aneurysm (HR: 1.72; 95% CI: 1.34 to 2.21) (Figure 2). The associations were stronger in models adjusted only for age and sex (Online Figure 2).

1.36; 95% CI: 1.17 to 1.57) and there was no association with hemorrhagic stroke. There were strong associations with PAD (HR: 1.95; 95% CI: 1.72 to 2.21) and abdominal aortic aneurysm (HR: 1.72; 95% CI: 1.34 to 2.21) (Figure 2). The associations were stronger in models adjusted only for age and sex (Online Figure 2). There was a strong association of neutrophil count with noncardiovascular death when comparing neutrophil counts of 6 to 7 versus 2 to 3 × 109/l (HR: 2.01; 95% CI: 1.91 to 2.11), with a higher proportion of deaths due to pneumonia or chronic obstructive pulmonary disease (Online Table 4). There was also an association of higher neutrophil count with the composite of CVD (Online Figure 3) and all-cause mortality (Online Figure 4). Neutrophil counts of <2 × 109/l were associated with greater risk of noncardiovascular death (compared with 2 to 3 × 109/l: HR: 1.52; 95% CI: 1.41 to 1.63), but were not associated with greater risk of any presentation of CVD (Online Figure 3). Because the associations between neutrophil counts and cardiovascular presentations were monotonic and linear, we treated neutrophil count as a linear variable in subsequent modeling. We found stronger associations within the normal range (Online Figure 5), but no interaction with smoking status (Online Figure 6). Additional adjustment for eosinophil and lymphocyte counts did not alter the estimates (Online Figure 7).

d linear, we treated neutrophil count as a linear variable in subsequent modeling. We found stronger associations within the normal range (Online Figure 5), but no interaction with smoking status (Online Figure 6). Additional adjustment for eosinophil and lymphocyte counts did not alter the estimates (Online Figure 7). There was some evidence that associations between neutrophil counts and cardiovascular presentations were stronger for stable compared with acute measurements, particularly for PAD and HF (Figure 3). Associations were further strengthened when we used the mean of 2 consecutive stable neutrophil counts, which were available in 393,543 patients and were taken at a median of 1.4 years apart (IQR: 0.6 to 2.7 years) (Figure 3). There was considerable variability but minimal trend over time between repeat measurements of stable neutrophil counts; the SD of differences between 2 consecutive measurements was 1.67 × 109/l, the correlation coefficient was 0.568, and the mean rate of change was a decrease of 0.014 per year (95% CI: 0.011 to 0.017) (Online Figure 8).

able variability but minimal trend over time between repeat measurements of stable neutrophil counts; the SD of differences between 2 consecutive measurements was 1.67 × 109/l, the correlation coefficient was 0.568, and the mean rate of change was a decrease of 0.014 per year (95% CI: 0.011 to 0.017) (Online Figure 8). We found that hazards were nonproportional for some of the endpoints. We therefore split the follow-up time by 6 months, and found that neutrophil counts were more strongly associated with HF, unheralded coronary death, and ischemic stroke in the first 6 months (Online Figure 9). Associations with coronary endpoints, HF, and PAD were stronger among younger patients (Online Figure 10). The association between neutrophil count and initial presentation with HF was stronger in men than women (HR per 109/l higher neutrophil count: 1.10 vs. 1.07; p = 0.001). There was an association between neutrophil count and initial presentation with transient ischemic attack in women but not men (HR: 1.05 vs. 1.00; p = 0.007) (Online Figure 11).

rophil count and initial presentation with HF was stronger in men than women (HR per 109/l higher neutrophil count: 1.10 vs. 1.07; p = 0.001). There was an association between neutrophil count and initial presentation with transient ischemic attack in women but not men (HR: 1.05 vs. 1.00; p = 0.007) (Online Figure 11). Discussion Neutrophil counts within the range clinicians currently consider normal had strong linear associations with some, but not all, CVDs in a population-based cohort. We found a greater cumulative incidence of unheralded coronary death, nonfatal MI, HF, PAD, and abdominal aortic aneurysm in patients with neutrophil counts at the upper end of the normal range (Figure 1), and these associations were confirmed in multivariable survival models. Our findings were consistent with those of previous, smaller studies that showed a positive association of higher neutrophil count with HF (6) and a moderate association with cerebral infarction (7); a key novel finding of our study is the association with PAD and abdominal aortic aneurysm, which has not been reported previously. In contrast with MI, we found that neutrophil count was not associated with a greater incidence of stable or unstable angina. This study’s large sample size (>700,000 patients) made it possible to investigate less common CVDs; we showed lack of association with intracerebral hemorrhage and subarachnoid hemorrhage.

sly. In contrast with MI, we found that neutrophil count was not associated with a greater incidence of stable or unstable angina. This study’s large sample size (>700,000 patients) made it possible to investigate less common CVDs; we showed lack of association with intracerebral hemorrhage and subarachnoid hemorrhage. Specificity and strength of associations We disaggregated CVD into pathologically diverse initial presentations to help elucidate the mechanistic role of neutrophils. The stronger association seen with MI than angina suggests that as well being involved in inflammation and atherosclerosis, neutrophils also increase the risk of arterial thrombosis (Central Illustration). Possible mechanisms include interactions with the endothelium and platelets, and overactivity of neutrophil extracellular traps (19). We also observed associations of higher neutrophil count with noncardiovascular and overall mortality, suggesting that chronic inflammation has noncardiovascular adverse effects that warrant further study. We examined associations with other leukocyte subtypes and found that monocyte counts had a similar pattern of associations with initial presentations of CVD to neutrophils, but the associations were not as strong (Online Figure 12). Eosinophil and lymphocyte counts have a different pattern of association with initial presentations of CVD (14), but additional adjustment for these leukocyte subtypes did not alter the results for neutrophils (Online Figure 7), suggesting that these associations are independent.

were not as strong (Online Figure 12). Eosinophil and lymphocyte counts have a different pattern of association with initial presentations of CVD (14), but additional adjustment for these leukocyte subtypes did not alter the results for neutrophils (Online Figure 7), suggesting that these associations are independent. Neutrophil count has a similar strength and shape of association with CVD as systolic blood pressure, with no evidence of a threshold effect among higher neutrophil counts. A moderate chronic elevation of systolic blood pressure of 60 mm Hg (e.g., 180 mm Hg instead of 120 mm Hg) is associated with an approximate doubling of the risk of incident HF (scaled from Rapsomaniki et al. [11]), which is comparable with the HR comparing the upper and lower ends of the normal range for neutrophil count (HR: 2.04; 95% CI: 1.82 to 2.29) (Figure 2). We also found that associations were stronger among patients with neutrophil counts taken under stable conditions, and when the mean of 2 neutrophil counts was used; this finding provides further evidence for the relevance of a patient’s chronic inflammatory state.

Neutrophil count has a similar strength and shape of association with CVD as systolic blood pressure, with no evidence of a threshold effect among higher neutrophil counts. A moderate chronic elevation of systolic blood pressure of 60 mm Hg (e.g., 180 mm Hg instead of 120 mm Hg) is associated with an approximate doubling of the risk of incident HF (scaled from Rapsomaniki et al. [11]), which is comparable with the HR comparing the upper and lower ends of the normal range for neutrophil count (HR: 2.04; 95% CI: 1.82 to 2.29) (Figure 2). We also found that associations were stronger among patients with neutrophil counts taken under stable conditions, and when the mean of 2 neutrophil counts was used; this finding provides further evidence for the relevance of a patient’s chronic inflammatory state. Targets and interventions Reducing chronic inflammation could be a potential therapeutic avenue in atherosclerotic disease. Our study could not ascertain whether it is circulating neutrophils per se that confer the additional risk or the underlying inflammatory state of which the neutrophil count is a marker (20). Investigation of causal mechanisms could involve epidemiological studies of upstream determinants of neutrophil counts, such as granulocyte colony-stimulating factor, interleukin-17, and interleukin-23 (21). Mendelian randomization studies using single nucleotide polymorphisms for genes associated with neutrophil count, such as those identified in the 17q21 region (22), might also help to evaluate causal relevance.

of neutrophil counts, such as granulocyte colony-stimulating factor, interleukin-17, and interleukin-23 (21). Mendelian randomization studies using single nucleotide polymorphisms for genes associated with neutrophil count, such as those identified in the 17q21 region (22), might also help to evaluate causal relevance. Colchicine, which has a range of actions on many cell types, including inhibiting microtubule formation in neutrophils, reduced the incidence of acute coronary syndrome and stroke in a trial among patients with stable coronary disease (23). Trials are underway to investigate whether anti-inflammatory agents such as methotrexate (24) or canakinumab (a human monoclonal antibody against interleukin-1-beta) (25) can prevent CVD events in high-risk individuals. Smoking causes an elevation of neutrophil counts and is associated more strongly with MI than stable angina (10), like neutrophil counts. A clinical trial of smoking cessation found that it reduced neutrophil count by 1.0 × 109/l (26). We adjusted for smoking in the main analysis, but if an increased neutrophil count (or the underlying chronic inflammation it represents) is on a causal pathway linking smoking to CVD, we might be overadjusting for smoking, thereby underestimating the component of cardiovascular risk conveyed by inflammation or neutrophils. Other modifiable factors that can increase the level of chronic low-grade inflammation include air pollution (27), obesity (28), lack of exercise (29), and periodontal disease (30).

Smoking causes an elevation of neutrophil counts and is associated more strongly with MI than stable angina (10), like neutrophil counts. A clinical trial of smoking cessation found that it reduced neutrophil count by 1.0 × 109/l (26). We adjusted for smoking in the main analysis, but if an increased neutrophil count (or the underlying chronic inflammation it represents) is on a causal pathway linking smoking to CVD, we might be overadjusting for smoking, thereby underestimating the component of cardiovascular risk conveyed by inflammation or neutrophils. Other modifiable factors that can increase the level of chronic low-grade inflammation include air pollution (27), obesity (28), lack of exercise (29), and periodontal disease (30). Study limitations Although our study had strengths—its large size, population base, and extensive adjustment for potential confounders—it also had important limitations. Our distinction between acute and stable patients was crude, but the similarity of findings in both groups was reassuring. As with any observational study, the results cannot be taken to imply causation due to the possibility of residual confounding. The measurement of neutrophil counts was undertaken in usual clinical care, and by different laboratories without study-wide protocols. Heterogeneity of measurement methods and heterogeneity among the study population itself may have led to biased estimates of association, but these would tend to be underestimates. The ascertainment of endpoints was in coded clinical data without manual endpoint adjudication. All the data sources used in this study missed some events (15); however, any errors in endpoint recording were likely to be nondifferential in relation to the neutrophil count. Because our study was based on electronic health records, some values of baseline variables were missing for some patients. However, we obtained similar results by imputing missing data using 2 different methods of multiple imputation.

int recording were likely to be nondifferential in relation to the neutrophil count. Because our study was based on electronic health records, some values of baseline variables were missing for some patients. However, we obtained similar results by imputing missing data using 2 different methods of multiple imputation. Clinical implications Our results suggest that the current clinical practice of labeling the range of neutrophil counts 2 to 7 × 109/l as normal—and ignoring any risk information conveyed by the actual value—should be reconsidered. Neutrophil counts are a measure of a patient’s chronic inflammatory state, which relates to cardiovascular risk and can be modified. Clinicians should look out for treatable causes of chronic inflammation in such patients, such as periodontal disease (30). Further research on the associations of neutrophil counts with CVD is warranted, including investigation of diurnal and seasonal variations and utility in risk prediction (20). Other biomarkers of inflammation are associated with greater risk of coronary disease (3), and U.S. guidelines recommend that high-sensitivity C-reactive protein can be measured to inform decisions on cardiovascular risk management in uncertain cases (31). An advantage of using the neutrophil count for this purpose, over other inflammatory biomarkers, is that it is already measured and would not incur any additional costs for testing.

nd that high-sensitivity C-reactive protein can be measured to inform decisions on cardiovascular risk management in uncertain cases (31). An advantage of using the neutrophil count for this purpose, over other inflammatory biomarkers, is that it is already measured and would not incur any additional costs for testing. Third, neutrophil counts could be considered as a means of monitoring a patient’s progress or as a surrogate endpoint in trials investigating anti-inflammatory interventions to reduce cardiovascular risk. Conclusions Clinically recorded neutrophil counts were strongly associated with the incidence of specific CVDs, even within the normal range. The neutrophil count should be further evaluated as an inflammatory biomarker relevant to CVD.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: Peripheral blood neutrophil counts within the normal range in people without prior CVD range are associated linearly with risk of developing MI, ischemic stroke, HF, peripheral artery disease, and abdominal aortic aneurysm. TRANSLATIONAL OUTLOOK: Future investigations, such as Mendelian randomization studies, should seek to understand the causal relevance of neutrophils to various forms of CVD, and ascertain whether therapies targeting neutrophils have clinical value in preventing these diseases. Appendix Online Appendix Online Data

Conclusions Clinically recorded neutrophil counts were strongly associated with the incidence of specific CVDs, even within the normal range. The neutrophil count should be further evaluated as an inflammatory biomarker relevant to CVD.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: Peripheral blood neutrophil counts within the normal range in people without prior CVD range are associated linearly with risk of developing MI, ischemic stroke, HF, peripheral artery disease, and abdominal aortic aneurysm. TRANSLATIONAL OUTLOOK: Future investigations, such as Mendelian randomization studies, should seek to understand the causal relevance of neutrophils to various forms of CVD, and ascertain whether therapies targeting neutrophils have clinical value in preventing these diseases. Appendix Online Appendix Online Data This study was supported by the National Institute for Health Research (RP-PG-0407-10314, PI HH); Wellcome Trust (WT 086091/Z/08/Z, PI HH); the Medical Research Prognosis Research Strategy Partnership (G0902393/99558, PI HH) and the Farr Institute of Health Informatics Research, funded by the Medical Research Council (K006584/1, PI HH), in partnership with Arthritis Research UK, the British Heart Foundation, Cancer Research UK, the Economic and Social Research Council, the Engineering and Physical Sciences Research Council, the National Institute of Health Research, the National Institute for Social Care and Health Research (Welsh Assembly Government), the Chief Scientist Office (Scottish Government Health Directorates) and the Wellcome Trust. ADS was supported by a Wellcome Trust clinical research training fellowship (0938/30/Z/10/Z). Dr. Denaxas was supported by a UCL Provost's Strategic Development Fund fellowship. The funding organizations had no involvement in the design of the study; the collection, analysis, and interpretation of the data; or the decision to approve publication of the finished manuscript.

training fellowship (0938/30/Z/10/Z). Dr. Denaxas was supported by a UCL Provost's Strategic Development Fund fellowship. The funding organizations had no involvement in the design of the study; the collection, analysis, and interpretation of the data; or the decision to approve publication of the finished manuscript. The authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For supplemental tables and figures, please see the online version of this article. Figure 1 Cumulative Incidence Curves For cardiovascular presentations among people without prior cardiovascular disease (CVD), crude cumulative incidence curves are shown for the highest and lowest categories of neutrophil count within the normal range. An artefact of imprecise coding rather than a clinical diagnosis, ‘nonspecific coronary disease’ was combined with unstable angina. Similarly, nonspecific stroke was combined with ischemic stroke. The plots show that, for myocardial infarction, heart failure, ischemic stroke, peripheral arterial disease (PAD), and abdominal aortic aneurysm, the incidence was greater among people with higher neutrophil counts. Figure 2 Association of Neutrophil Count With Initial CVD Presentation

For cardiovascular presentations among people without prior cardiovascular disease (CVD), crude cumulative incidence curves are shown for the highest and lowest categories of neutrophil count within the normal range. An artefact of imprecise coding rather than a clinical diagnosis, ‘nonspecific coronary disease’ was combined with unstable angina. Similarly, nonspecific stroke was combined with ischemic stroke. The plots show that, for myocardial infarction, heart failure, ischemic stroke, peripheral arterial disease (PAD), and abdominal aortic aneurysm, the incidence was greater among people with higher neutrophil counts. Figure 2 Association of Neutrophil Count With Initial CVD Presentation Neutrophil count categories influenced cause-specific adjusted hazard ratios for cardiovascular presentations among people without prior cardiovascular disease (CVD). Hazard ratios were adjusted for age, sex, deprivation, ethnicity, smoking, diabetes, systolic blood pressure (SBP), blood pressure medication, body mass index (BMI), total cholesterol, high-density lipoprotein cholesterol (HDL-C), statin use, estimated glomerular filtration rate (eGFR), atrial fibrillation (AF), autoimmune conditions, inflammatory bowel disease (IBD), chronic obstructive pulmonary disease (COPD), cancer, and acute conditions at the time of blood testing. Shaded = normal range. *p < 0.05; **p < 0.0036; ***p < 0.0001. CI = confidence interval; other abbreviations as in Figure 1. Figure 3 CVD and Neutrophil Counts by Clinical State at Blood Sampling

Neutrophil count categories influenced cause-specific adjusted hazard ratios for cardiovascular presentations among people without prior cardiovascular disease (CVD). Hazard ratios were adjusted for age, sex, deprivation, ethnicity, smoking, diabetes, systolic blood pressure (SBP), blood pressure medication, body mass index (BMI), total cholesterol, high-density lipoprotein cholesterol (HDL-C), statin use, estimated glomerular filtration rate (eGFR), atrial fibrillation (AF), autoimmune conditions, inflammatory bowel disease (IBD), chronic obstructive pulmonary disease (COPD), cancer, and acute conditions at the time of blood testing. Shaded = normal range. *p < 0.05; **p < 0.0036; ***p < 0.0001. CI = confidence interval; other abbreviations as in Figure 1. Figure 3 CVD and Neutrophil Counts by Clinical State at Blood Sampling Hazard ratios for initial presentation of CVDs by neutrophil count varied by clinical state at the time of blood sampling. ‘Mean of 2 stable’ refers to the mean of 2 consecutive neutrophil counts performed in a stable clinical state. Hazard ratios are adjusted for age, sex, socioeconomic deprivation, ethnicity, smoking, diabetes, SBP, blood pressure medication, BMI, total cholesterol, HDL-C, statin use, eGFR, AF, autoimmune conditions, IBD, COPD, and cancer. *p < 0.05; **p < 0.0036; ***p < 0.0001. Abbreviations as in Figures 1 and 2. Central Illustration Neutrophil Counts and CVDs

Hazard ratios for initial presentation of CVDs by neutrophil count varied by clinical state at the time of blood sampling. ‘Mean of 2 stable’ refers to the mean of 2 consecutive neutrophil counts performed in a stable clinical state. Hazard ratios are adjusted for age, sex, socioeconomic deprivation, ethnicity, smoking, diabetes, SBP, blood pressure medication, BMI, total cholesterol, HDL-C, statin use, eGFR, AF, autoimmune conditions, IBD, COPD, and cancer. *p < 0.05; **p < 0.0036; ***p < 0.0001. Abbreviations as in Figures 1 and 2. Central Illustration Neutrophil Counts and CVDs Potential causal pathways are depicted linking chronic inflammation, neutrophil counts, and onset of cardiovascular diseases (CVD). Environmental and behavioral risk factors such as smoking, air pollution, and physical inactivity contribute to chronic inflammation. An inflammatory state results in a higher neutrophil count, which may be causally linked with increased risk of certain cardiovascular conditions. Table 1 Patient Characteristics∗

Potential causal pathways are depicted linking chronic inflammation, neutrophil counts, and onset of cardiovascular diseases (CVD). Environmental and behavioral risk factors such as smoking, air pollution, and physical inactivity contribute to chronic inflammation. An inflammatory state results in a higher neutrophil count, which may be causally linked with increased risk of certain cardiovascular conditions. Table 1 Patient Characteristics∗ Neutrophil Count p Value for Trend Below Normal Range Within Normal Range Above Normal Range Neutrophils, × 109/l <2 2–3 3–6 6–7 ≥7 N 26,588 154,863 489,143 48,849 55,788 Female 16,592 (62.4) 89,956 (58.1) 288,221 (58.9) 31,465 (64.4) 35,722 (64.0) <0.0001 Age, yrs 51.3 (40.6–61.8) 53 (42.2–63.1) 52.9 (42.0–65.3) 49.8 (38.8–65) 48.2 (36.9–64.7) <0.0001 Most deprived quintile 4,513 (17.0) 21,722 (14.1) 84,740 (17.4) 10,158 (20.9) 11,838 (21.3) <0.0001 Acute condition at time of blood test 5,670 (21.3) 26,932 (17.4) 92,969 (19.0) 11,600 (23.7) 17,008 (30.5) <0.0001 Ethnicity White 12,945 (80.3) 82,468 (90.8) 283,495 (93.5) 30,174 (94.5) 35,952 (95.4) <0.0001 South Asian 327 (2.0) 2,482 (2.7) 9,083 (3.0) 813 (2.5) 745 (2.0) <0.0001 Black 2,229 (13.8) 3,224 (3.6) 3,936 (1.3) 310 (1.0) 275 (0.7) <0.0001 Other 626 (3.9) 2,624 (2.9) 6,740 (2.2) 620 (1.9) 710 (1.9) <0.0001 Missing 10,461 (39.3) 64,065 (41.4) 185,889 (38.0) 16,932 (34.7) 18,106 (32.5) <0.0001 Full blood count parameters on index date Eosinophils, × 109/l 0.1 (0.09–0.2) 0.13 (0.1–0.2) 0.2 (0.1–0.27) 0.2 (0.1–0.3) 0.16 (0.1–0.27) <0.0001 Lymphocytes, × 109/l 1.7 (1.36–2.07) 1.8 (1.5–2.2) 2.0 (1.61–2.49) 2.1 (1.69–2.7) 2.05 (1.58–2.63) <0.0001 Monocytes, × 109/l 0.34 (0.29–0.42) 0.4 (0.3–0.5) 0.5 (0.4–0.6) 0.6 (0.5–0.74) 0.7 (0.5–0.9) <0.0001 Basophils, × 109/l 0.01 (0–0.03) 0.02 (0–0.05) 0.03 (0–0.08) 0.03 (0–0.1) 0.03 (0–0.1) <0.0001 Hemoglobin, g/dl 13.5 (12.6–14.5) 13.9 (13–14.8) 14.0 (13.1–15) 13.8 (12.8–14.9) 13.7 (12.4–14.8) <0.0001 Platelets, × 109/l 225 (190–263) 241 (207–279) 262 (224–306) 286 (242–338) 298 (249–360) <0.0001 Smoking status Never 16,657 (66.1) 91,447 (62.1) 240,087 (51.7) 18,758 (40.7) 19,753 (37.7) <0.0001 Former 5,798 (23.0) 36,839 (25.0) 114,129 (24.6) 9,736 (21.1) 10,697 (20.4) <0.0001 Current 2,754 (10.9) 18,938 (12.9) 109,796 (23.7) 17,615 (38.2) 21,918 (41.9) <0.0001 Missing 1,379 (5.2) 7,639 (4.9) 25,131 (5.1) 2,740 (5.6) 3,420 (6.1) <0.0001 Most recent value within 1 yr before the index date Systolic BP 131 (120–145) 135 (120–148) 138 (123–150) 135 (120–150) 130 (120–146) <0.0001 BMI 25.4 (22.7–28.7) 26.3 (23.4–29.8) 27.4 (24.1–31.5) 27.1 (23.4–32.1) 26.4 (22.9–31.1) <0.00

0.0001 Missing 1,379 (5.2) 7,639 (4.9) 25,131 (5.1) 2,740 (5.6) 3,420 (6.1) <0.0001 Most recent value within 1 yr before the index date Systolic BP 131 (120–145) 135 (120–148) 138 (123–150) 135 (120–150) 130 (120–146) <0.0001 BMI 25.4 (22.7–28.7) 26.3 (23.4–29.8) 27.4 (24.1–31.5) 27.1 (23.4–32.1) 26.4 (22.9–31.1) <0.00 01 Total cholesterol 5.4 (4.7–6.2) 5.5 (4.8–6.3) 5.5 (4.8–6.2) 5.3 (4.6–6.1) 5.3 (4.5–6.1) <0.0001 HDL cholesterol 1.53 (1.24–1.9) 1.47 (1.20–1.79) 1.35 (1.10–1.62) 1.30 (1.08–1.60) 1.3 (1.07–1.59) <0.0001 eGFR 84.4 (71.9–97.4) 82.8 (70.7–95.2) 81.4 (68.3–94.5) 82.4 (67.7–96.7) 83.1 (67.3–97.5) <0.0001 Diagnoses on or before index date Atrial fibrillation 176 (0.7) 1,044 (0.7) 5,169 (1.1) 602 (1.2) 731 (1.3) <0.0001 Cancer 2,089 (7.9) 9,343 (6.0) 29,468 (6.0) 2,863 (5.9) 3,758 (6.7) 0.2 Diabetes 860 (3.2) 5,322 (3.4) 25,081 (5.1) 2,896 (5.9) 2,968 (5.3) <0.0001 Asthma 2,721 (10.2) 17,046 (11.0) 61,521 (12.6) 7,165 (14.7) 8,344 (15.0) <0.0001 COPD 178 (0.7) 1,329 (0.9) 9,545 (2.0) 1,614 (3.3) 2,355 (4.2) <0.0001 Connective tissue disease 596 (2.2) 3,057 (2.0) 13,585 (2.8) 2,007 (4.1) 2,614 (4.7) <0.0001 IBD 207 (0.8) 1,316 (0.8) 5,377 (1.1) 717 (1.5) 1,029 (1.8) <0.0001 Medication use in the year before index date Antihypertensives 5,342 (20.1) 34,000 (22.0) 130,791 (26.7) 13,255 (27.1) 13,893 (24.9) <0.0001 Statins 1,141 (4.3) 8,120 (5.2) 31,713 (6.5) 3,010 (6.2) 2,822 (5.1) <0.0001 Values are n (%) or median (interquartile range) unless otherwise indicated.

(1.5) 1,029 (1.8) <0.0001 Medication use in the year before index date Antihypertensives 5,342 (20.1) 34,000 (22.0) 130,791 (26.7) 13,255 (27.1) 13,893 (24.9) <0.0001 Statins 1,141 (4.3) 8,120 (5.2) 31,713 (6.5) 3,010 (6.2) 2,822 (5.1) <0.0001 Values are n (%) or median (interquartile range) unless otherwise indicated. BMI = body mass index; BP = blood pressure; COPD = chronic obstructive pulmonary disease; eGFR = estimated glomerular filtration rate; HDL = high-density lipoprotein; IBD = inflammatory bowel disease. ∗ Completeness of recording of continuous variables at any time (as used for imputation) was 98.8% for BP, 90.6% for body mass index, 58.9% for HDL and total cholesterol, and 57.1% for eGFR. Completeness of recording of these variables within 1 year before the index date was 65.4%, 31.3%, 31.6%, and 47.0%, respectively. Diagnoses of comorbid conditions were considered completely recorded (i.e., absence of a diagnosis code was interpreted as absence of the condition).

Vascular imaging with 18F-fluorodeoxyglucose positron emission tomography (18FDG-PET) provides a noninvasive surrogate of inflammation and has been used to test novel drug treatments for atherosclerosis (1). Hypoxia exists in atherosclerosis (2) and may contribute to the measured FDG signal (3). 18F-fluoromisonidazole (FMISO) PET can quantify hypoxia in tumors; it has not been applied to human atherosclerosis. Our hypotheses were: 1) carotid plaques that cause transient ischemic attack (TIA) or stroke are more hypoxic than contralateral asymptomatic plaques; 2) carotid FDG and FMISO PET signals are positively correlated; and 3) carotid FMISO PET signal positively correlates with plaque hypoxia-inducible factor-1α (HIF-1α) staining. Sixteen participants with carotid atherosclerosis (mean age 70 ± 7 years; 69% male; 8 recently symptomatic) underwent computed tomography (CT) angiography and PET/CT imaging with FDG (250 MBq) and FMISO (300 MBq). FMISO PET imaging occurred within 2 ± 1 days of FDG PET imaging, and symptomatic individuals underwent imaging 16 ± 11 days after TIA/stroke.

therosclerosis (mean age 70 ± 7 years; 69% male; 8 recently symptomatic) underwent computed tomography (CT) angiography and PET/CT imaging with FDG (250 MBq) and FMISO (300 MBq). FMISO PET imaging occurred within 2 ± 1 days of FDG PET imaging, and symptomatic individuals underwent imaging 16 ± 11 days after TIA/stroke. Dynamic imaging of FMISO (120 to 180 min after tracer injection) allowed estimation of both the mean Ki (net influx rate constant) and mean of maximum target-to-background ratio (mean max TBR) (derived from the final 3 acquisition frames 165 to 180 min post-injection). Carotid endarterectomy was performed a median of 6 days (range: 3 to 15 days) after imaging. Eleven plaques were processed for immunohistochemistry: mean percentage area CD68 and α-smooth muscle actin staining; mean CD31 staining/mm2; and numbers of HIF-1α nuclei.

al 3 acquisition frames 165 to 180 min post-injection). Carotid endarterectomy was performed a median of 6 days (range: 3 to 15 days) after imaging. Eleven plaques were processed for immunohistochemistry: mean percentage area CD68 and α-smooth muscle actin staining; mean CD31 staining/mm2; and numbers of HIF-1α nuclei. Mean max TBR of FMISO was significantly greater in symptomatic plaques than that in contralateral lesions (1.11 ± 0.07 vs. 1.05 ± 0.06; p <0.05) (Figure 1). Mean Ki in symptomatic plaques was also significantly higher than that in asymptomatic plaques (3.6 × 10−4 ± 2.9 × 10−4 min−1 vs. 1.6 × 10−4 ± 1.6 × 10−4 min−1; p = 0.03). FMISO uptake correlated positively with FDG (TBR: r = 0.51; p < 0.01 and mean Ki: r = 0.43; p = 0.02). FDG uptake and plaque macrophage content were strongly related (r = 0.67; p = 0.02). There was a trend to a positive correlation between FMISO and plaque macrophages (r = 0.52; p = 0.10). There was no relationship between FMISO uptake and risk factors for atherosclerosis, plaque smooth muscle content, or numbers of positively stained nuclei for HIF-1α. There was a trend to a negative correlation between the FMISO signal and CD31 staining (r = −0.62; p = 0.08). In summary, our first and second hypotheses were proven: 1) culprit carotid plaques after TIA or stroke are more hypoxic than asymptomatic lesions, and 2) plaque hypoxia makes a significant contribution to carotid artery FDG PET signals. The third hypothesis, linking imaging and a tissue marker of hypoxia, was not proven.

In summary, our first and second hypotheses were proven: 1) culprit carotid plaques after TIA or stroke are more hypoxic than asymptomatic lesions, and 2) plaque hypoxia makes a significant contribution to carotid artery FDG PET signals. The third hypothesis, linking imaging and a tissue marker of hypoxia, was not proven. Hypoxia has been linked with adverse features of plaque biology, including inflammation and intraplaque hemorrhage (2). Furthermore, inflammatory stimuli increase glycolytic flux in macrophages, and this effect is amplified in hypoxic conditions (4). We now suggest that hypoxia is more common in symptomatic lesions. The robust correlation between FDG and FMISO signals suggests that hypoxia contributes to the FDG signal in FDG PET studies of atherosclerosis. In a pilot study using a related PET hypoxia tracer, HX-4, van der Valk et al. (5) drew similar conclusions (r = 0.75; p = 0.03). We extended this result using a better validated tracer and a larger patient cohort; the congruence of both static and dynamic measures of hypoxia further supports our conclusions.

clerosis. In a pilot study using a related PET hypoxia tracer, HX-4, van der Valk et al. (5) drew similar conclusions (r = 0.75; p = 0.03). We extended this result using a better validated tracer and a larger patient cohort; the congruence of both static and dynamic measures of hypoxia further supports our conclusions. The principal study limitation was the sample size. Caution should be used in the interpretation of data other than those regarding our a priori hypotheses; no adjustment was made for multiple observations within individuals. Tissue fixation methods might have adversely affected HIF-1α immunohistochemistry, although cancer studies also reported conflicting results with respect to correlations between hypoxia imaging and HIF-1α. Nevertheless, we hope these data stimulate further study into hypoxia and markers of plaque destabilization. We reported the first prospective human study to quantify hypoxia in atherosclerosis using the validated PET tracer, FMISO. Symptomatic carotid plaques were more hypoxic than asymptomatic lesions, perhaps identifying a novel target for drug therapy. The correlation between FDG and FMISO suggested that hypoxia contributes to the FDG signal in atherosclerosis PET studies.

ntify hypoxia in atherosclerosis using the validated PET tracer, FMISO. Symptomatic carotid plaques were more hypoxic than asymptomatic lesions, perhaps identifying a novel target for drug therapy. The correlation between FDG and FMISO suggested that hypoxia contributes to the FDG signal in atherosclerosis PET studies. Please note: This study was funded by a programme grant (RG/10/007/28300) from the British Heart Foundation (BHF). Dr. Joshi was supported by a BHF Clinical Research Training Fellowship (FS/12/29/29463), a British Atherosclerosis Society Binks Trust Travel Award, and a Raymond and Beverly Sackler PhD Studentship. Dr. Manavaki is funded by the NIHR Cambridge Biomedical Research Centre. Dr. Rudd is partially supported by the NIHR Cambridge Biomedical Research Centre, the BHF, The Wellcome Trust, and the EPSRC Cambridge Centre for Mathematical Imaging in Healthcare. The authors have reported that they have no relationships relevant to the contents of this paper to disclose. Figure 1 Imaging of Hypoxia in Carotid Atherosclerosis 1) Positron emission tomography/computed tomography (PET/CT) imaging of a culprit carotid stenosis after stroke. (A) CT angiography. Stenosis in the left internal carotid artery (arrow). (B) Fused 18F-fluorodeoxyglucose (FDG) PET/CT. Intense uptake in the culprit lesion (arrow). (C) Fused 18F-fluoromisonidazole (FMISO) PET/CT. Corresponding uptake is indicative of intraplaque hypoxia (arrow). 2) Carotid uptake of FDG and FMISO are positively correlated. Mean max TBR = mean of maximum target-to-background ratio.

Systemic inflammation triggers culprit pathogenic mechanisms, relating clinical cardiovascular disease (CVD) risk factors to atherosclerotic plaque progression and rupture (1). Quantifying vascular inflammation in atherosclerosis may reveal mechanistic pathways, allow efficacy testing of new drugs, and improve CVD risk prediction. Carotid, aortic, and peripheral arterial inflammation can be measured by fluorine-18-labeled fluorodeoxyglucose positron emission tomography/computed tomography ([18F]FDG PET/CT) (2). However, myocardial [18F]FDG signal spillover occurs due to myocardial muscle [18F]FDG uptake, often hampering coronary artery signal quantification (3). Lack of cell specificity and the influence of hypoxia on [18F]FDG uptake within macrophages and other cells (4) are further limitations of [18F]FDG imaging. Up-regulation of the G-protein-coupled receptor somatostatin receptor subtype-2 (SST2) occurs on the surface of activated macrophages (5). Pre-clinical 6, 7 and retrospective 8, 9, 10 studies suggest that gallium-68-labeled [1,4,7,10-tetraazacyclododecane-N,N',N'',N'''-tetraacetic acid]-d-Phe1, Tyr3-octreotate (DOTATATE), a PET ligand with high-specificity binding affinity for SST2 (11), may be superior to [18F]FDG in marking macrophage activity, particularly in the coronary arteries. However, robust evaluation of 68Ga-DOTATATE in atherosclerosis is lacking. We present a prospective clinical study evaluating 68Ga-DOTATATE PET for imaging coronary, carotid, and aortic inflammation in patients with CVD.

Up-regulation of the G-protein-coupled receptor somatostatin receptor subtype-2 (SST2) occurs on the surface of activated macrophages (5). Pre-clinical 6, 7 and retrospective 8, 9, 10 studies suggest that gallium-68-labeled [1,4,7,10-tetraazacyclododecane-N,N',N'',N'''-tetraacetic acid]-d-Phe1, Tyr3-octreotate (DOTATATE), a PET ligand with high-specificity binding affinity for SST2 (11), may be superior to [18F]FDG in marking macrophage activity, particularly in the coronary arteries. However, robust evaluation of 68Ga-DOTATATE in atherosclerosis is lacking. We present a prospective clinical study evaluating 68Ga-DOTATATE PET for imaging coronary, carotid, and aortic inflammation in patients with CVD. Methods RNA sequencing To determine target specificity of 68Ga-DOTATATE imaging in atherosclerosis, expression of the SSTR1–5 genes within in vitro- differentiated macrophage subtypes and other blood-derived cells relevant to atherosclerosis were characterized using population-based “next generation” RNA sequencing data from the European BLUEPRINT (a BLUEPRINT of haematopoietic epigenomes) project, for which all data are publicly available (12). The expression levels of glucose transporter 1 (GLUT1) and glucose transporter 3 (GLUT3) genes were also analyzed from the dataset; these genes encode the main glucose transporters that facilitate uptake of [18F]FDG in atherosclerotic plaques.

poietic epigenomes) project, for which all data are publicly available (12). The expression levels of glucose transporter 1 (GLUT1) and glucose transporter 3 (GLUT3) genes were also analyzed from the dataset; these genes encode the main glucose transporters that facilitate uptake of [18F]FDG in atherosclerotic plaques. Clinical study In the VISION (Vascular Inflammation imaging using Somatostatin receptor positron emissION tomography; NCT02021188) study, an unselected “real-world” cohort of patients with wide-ranging severity of stable (n = 18) and unstable (n = 24) CVD was prospectively enrolled from Addenbrooke’s Hospital, Cambridge, United Kingdom (Figure 1). “Stable” patients had stable angina or asymptomatic atherosclerosis and at least a 30% stenosis of a major epicardial coronary artery or an internal carotid artery. “Unstable” patients had experienced a clinical event (acute coronary syndrome [ACS] or carotid territory transient ischemic attack [TIA]/stroke) within the 3 months before imaging. Baseline cardiovascular risk factors were noted, including measurement of serum lipids and high-sensitivity C-reactive protein. Patients were older than 40 years of age and provided written, informed consent. The study protocol approved by the local research ethics committee (REC 14/EE/0019) was in accordance with the Declaration of Helsinki.

risk factors were noted, including measurement of serum lipids and high-sensitivity C-reactive protein. Patients were older than 40 years of age and provided written, informed consent. The study protocol approved by the local research ethics committee (REC 14/EE/0019) was in accordance with the Declaration of Helsinki. PET-CT imaging Patients underwent 68Ga-DOTATATE PET-CT and [18F]FDG PET-CT imaging, using established methods (13) for vascular PET imaging on a Discovery 690 combined PET-CT system model (GE Healthcare, Little Chalfont, United Kingdom; extended Methods are detailed in the Online Appendix). 68Ga-DOTATATE had an average radiochemical purity of 99% on quality control testing performed by the manufacturer (Mallinckrodt, St. Louis, Missouri). Patients fasted for 6 h prior to [18F]FDG imaging; capillary blood glucose concentration was confirmed as <7.0 mmol/l in nondiabetic patients prior to scanning. Patients with diabetes mellitus were instructed to take their antidiabetic medications as usual prior to [18F]FDG scanning but to hold insulin within 4 h of the scan; if glucose level was >11.0 mmol/l, the scan was rescheduled according to our standard clinical practice. Insulin was not administered to any patient prior to [18F]FDG PET imaging. The mean injected dose of 68Ga-DOTATATE was 147.8 ± 31.6 MBq and 248.1 ± 22.3 MBq for [18F]FDG. Electrocardiography (ECG)-gated CT coronary angiography plus calcium scanning and carotid angiography were also performed.

tandard clinical practice. Insulin was not administered to any patient prior to [18F]FDG PET imaging. The mean injected dose of 68Ga-DOTATATE was 147.8 ± 31.6 MBq and 248.1 ± 22.3 MBq for [18F]FDG. Electrocardiography (ECG)-gated CT coronary angiography plus calcium scanning and carotid angiography were also performed. Image analysis Static PET images were reconstructed using 3-dimensional (3D) iterative time-of-flight ordered-subset expectation maximization with point spread function modeling to reduce partial volume error. ECG-gated coronary PET images were reconstructed in diastole (50% to 75% of the R–R interval). PET-CT images were coregistered and analyzed by experienced observers masked to the clinical details, using OsiriX imaging software (version 7.0; Pixmeo, Bernex, Switzerland). CT angiography was used as the anatomical reference standard; 2D regions of interest were drawn on consecutive, fused PET-CT slices to quantify the maximum arterial radioactivity concentration, normalized by mean blood pool activity in the superior vena cava (maximum tissue-to-blood ratio [TBRmax]). Mean (m) and most diseased segment (mds) TBRmax values were measured for each coronary segment, carotid artery, and thoracic aorta. Reproducibility of 68Ga-DOTATATE TBR measurements were tested by 2 independent observers using 10% of the coronary and carotid scans (n = 4 for both) selected at random, with 1 week between intraobserver readings. Coronary artery PET data were deemed uninterpretable if the maximum myocardial standardized uptake value was >5.0.

ibility of 68Ga-DOTATATE TBR measurements were tested by 2 independent observers using 10% of the coronary and carotid scans (n = 4 for both) selected at random, with 1 week between intraobserver readings. Coronary artery PET data were deemed uninterpretable if the maximum myocardial standardized uptake value was >5.0. Coronary lesions were classified according to established CT criteria for plaque composition (calcified, noncalcified, or mixed plaque) and the presence of high-risk morphological features (spotty calcification [<3.0 mm], low attenuation [<30 HU], and positive remodeling [cross-sectional area >10% of a reference segment]) (14). “Culprit” lesions were defined in patients with ACS or TIA/stroke by the attending cardiologist or stroke physician before PET imaging, with no involvement of the VISION study team. Assignment of culprit artery status took clinical data into consideration (e.g., ECG, angiographic and echocardiographic findings, site of any neurological deficit at time of clinical presentation, and carotid artery or brain imaging). Arteries targeted for intervention (with percutaneous coronary intervention or carotid endarterectomy surgery) were presumed to be culprit arteries. In patients who were managed medically, if the culprit lesion was uncertain, the relevant data were excluded from this part of the analysis.

otid artery or brain imaging). Arteries targeted for intervention (with percutaneous coronary intervention or carotid endarterectomy surgery) were presumed to be culprit arteries. In patients who were managed medically, if the culprit lesion was uncertain, the relevant data were excluded from this part of the analysis. Quantitative polymerase chain reaction The pattern of SSTR2 gene expression observed using population-based RNA-sequenced data was confirmed in lipopolysaccharide-stimulated macrophages from subjects in our imaging cohort by using quantitative real-time polymerase chain reaction assay results and compared to those of age- and sex-matched healthy volunteers (n = 3 for both). SSTR2 and CD68 mRNA levels were measured in excised carotid plaques, and compared with 68Ga-DOTATATE signals in PET images obtained prior to surgery.

imaging cohort by using quantitative real-time polymerase chain reaction assay results and compared to those of age- and sex-matched healthy volunteers (n = 3 for both). SSTR2 and CD68 mRNA levels were measured in excised carotid plaques, and compared with 68Ga-DOTATATE signals in PET images obtained prior to surgery. Autoradiography and histology To confirm specific ligand binding in atherosclerotic plaques, 68Ga-DOTATATE autoradiography was performed in carotid tissue sections adjacent to those used for quantitative polymerase chain reaction. After the radioactivity decayed, sections were stained with antibodies for SST2, the panmacrophage marker CD68, and Movat’s pentachrome stain for anatomic characterization. Autoradiography and immunostaining were similarly tested in cultured macrophages. Colocalization of SST2 and CD68 staining in macrophages within carotid plaque sections were assessed by immunofluorescence, with isotype and concentration-matched immunoglobulin G (IgG) as the negative control. The retention, storage, and use of tissue sections and blood samples were compliant with the UK Human Tissue Act of 2004.

on of SST2 and CD68 staining in macrophages within carotid plaque sections were assessed by immunofluorescence, with isotype and concentration-matched immunoglobulin G (IgG) as the negative control. The retention, storage, and use of tissue sections and blood samples were compliant with the UK Human Tissue Act of 2004. Statistical analysis The primary outcome was comparison of culprit versus nonculprit coronary and carotid artery 68Ga-DOTATATE mTBRmax in patients with ACS or TIA/stroke. Pre-specified secondary outcomes included comparisons of vascular 68Ga-DOTATATE mTBRmax values versus clinical CVD risk factors, CT plaque morphology, [18F]FDG mTBRmax, and SSTR2/CD68 gene expression levels in excised carotid plaques. Primary and secondary outcome data expressed as medians (interquartile range [IQR]) were compared using Wilcoxon signed rank test or Mann-Whitney U test, as appropriate, with differences of medians derived for paired data. For comparisons between more than 2 groups, the Kruskal-Wallis test was used.

ed carotid plaques. Primary and secondary outcome data expressed as medians (interquartile range [IQR]) were compared using Wilcoxon signed rank test or Mann-Whitney U test, as appropriate, with differences of medians derived for paired data. For comparisons between more than 2 groups, the Kruskal-Wallis test was used. Based on 68Ga-DOTATATE TBR values from our pilot work and previously published data (9), our sample size (n = 42) was chosen to detect differences in mTBRmax of ≥1.13 between high- and low-risk arteries, with 80% power and a 2-sided p value of <0.05. Patients with stable and unstable CVD were not formally matched as our primary comparison used “within patient” data (culprit versus nonculprit artery) rather than stable versus unstable patients. We anticipated that if one-third of patients had TIA/stroke, this would yield a comparable number of explanted carotid specimens to similar PET validation work performed by our group (15). Spearman’s correlation and simple linear regression were used to identify statistically significant clinical and biochemical predictors of 68Ga-DOTATATE mTBRmax that were then evaluated together using multivariate analysis. In the regression analysis, mean arterial values were used to mitigate the problem of multiple observations, as each patient contributed an equal number of arteries. Two-sided p values of <0.05 were considered significant. Statistical analysis was performed using Prism version 6.0 software (GraphPad Software, Redwood, California) and Stata version 14.1 software (StataCorp, Cary, North Carolina).

blem of multiple observations, as each patient contributed an equal number of arteries. Two-sided p values of <0.05 were considered significant. Statistical analysis was performed using Prism version 6.0 software (GraphPad Software, Redwood, California) and Stata version 14.1 software (StataCorp, Cary, North Carolina). Results Population-based validation of SSTR2 gene expression in macrophages Prior to clinical PET imaging, we tested the target expression of SSTR2 in blood-derived macrophages compared to other relevant cell types by using data from a large-scale population study. High levels of SSTR2 mRNA were detected exclusively in proinflammatory M1 macrophages and no other macrophage phenotype. This pattern and degree of expression was not seen for any other SST receptor subtype or cell line (Figure 2). SSTR3 was expressed by CD4+ T lymphocytes to a lesser extent, as is known to occur (16). Very low levels of SSTR2 mRNA were detected in unstimulated M0 macrophages and alternatively activated M2 macrophages, but SSTR2 was not expressed by any of the following cells: monocytes, T or B lymphocytes, natural killer cells, platelets, neutrophils, and endothelial cells. GLUT1 and GLUT3 were highly expressed by all cell types, demonstrating that SSTR2 offers greater cell specificity as an inflammation imaging target than glucose metabolism. Clinical study Baseline clinical data are summarized in Table 1. The median time interval between ACS and PET imaging was 35 days (IQR: 21 to 66 days) and 18 days (IQR: 11 to 25 days) for TIA/stroke.

Very low levels of SSTR2 mRNA were detected in unstimulated M0 macrophages and alternatively activated M2 macrophages, but SSTR2 was not expressed by any of the following cells: monocytes, T or B lymphocytes, natural killer cells, platelets, neutrophils, and endothelial cells. GLUT1 and GLUT3 were highly expressed by all cell types, demonstrating that SSTR2 offers greater cell specificity as an inflammation imaging target than glucose metabolism. Clinical study Baseline clinical data are summarized in Table 1. The median time interval between ACS and PET imaging was 35 days (IQR: 21 to 66 days) and 18 days (IQR: 11 to 25 days) for TIA/stroke. The reproducibility of 68Ga-DOTATATE TBRmax measurements was excellent for both intraobserver observations (coronary artery intraclass coefficient value [ICC]: 0.90; 95% confidence interval [CI]: 0.85 to 0.94; carotid artery ICC: 0.96; 95% CI: 0.95 to 0.97) and interobserver observations (coronary artery ICC: 0.96; 95% CI: 0.94 to 0.97; carotid artery ICC: 0.91; 95% CI: 0.88 to 0.94).

ent for both intraobserver observations (coronary artery intraclass coefficient value [ICC]: 0.90; 95% confidence interval [CI]: 0.85 to 0.94; carotid artery ICC: 0.96; 95% CI: 0.95 to 0.97) and interobserver observations (coronary artery ICC: 0.96; 95% CI: 0.94 to 0.97; carotid artery ICC: 0.91; 95% CI: 0.88 to 0.94). 68Ga-DOTATATE identifies culprit ACS lesions in coronary arteries Myocardial binding of 68Ga-DOTATATE was sufficiently low in all patients to allow unimpeded coronary artery PET signal measurement (Central Illustration, Online Figure 1). In patients with ACS, culprit 68Ga-DOTATATE uptake was consistently greater than the highest nonculprit coronary segment within the same individual (median difference mTBRmax: 0.69; IQR: 0.22 to 1.15; p = 0.008; median difference mdsTBRmax: 1.17; IQR: 0.45 to 1.70; p = 0.02), regardless of whether the lesion had been stented prior to imaging (culprit stented mTBRmax: 2.91; IQR: 2.66 to 4.63 vs. stable stented mTBRmax: 2.00; IQR: 1.51 to 2.70; p = 0.006) (Online Figure 2). Using receiver operator characteristic (ROC) analysis, coronary 68Ga-DOTATATE mTBRmax values >2.66 had 87.5% (95% CI: 47.4 to 99.7) sensitivity and 78.4% (95% CI: 72.4 to 83.6) specificity to detect a culprit coronary segment (ROC area under the curve [AUC]: 0.86; 95% CI: 0.78 to 0.93; p = 0.0006).

68Ga-DOTATATE identifies culprit ACS lesions in coronary arteries Myocardial binding of 68Ga-DOTATATE was sufficiently low in all patients to allow unimpeded coronary artery PET signal measurement (Central Illustration, Online Figure 1). In patients with ACS, culprit 68Ga-DOTATATE uptake was consistently greater than the highest nonculprit coronary segment within the same individual (median difference mTBRmax: 0.69; IQR: 0.22 to 1.15; p = 0.008; median difference mdsTBRmax: 1.17; IQR: 0.45 to 1.70; p = 0.02), regardless of whether the lesion had been stented prior to imaging (culprit stented mTBRmax: 2.91; IQR: 2.66 to 4.63 vs. stable stented mTBRmax: 2.00; IQR: 1.51 to 2.70; p = 0.006) (Online Figure 2). Using receiver operator characteristic (ROC) analysis, coronary 68Ga-DOTATATE mTBRmax values >2.66 had 87.5% (95% CI: 47.4 to 99.7) sensitivity and 78.4% (95% CI: 72.4 to 83.6) specificity to detect a culprit coronary segment (ROC area under the curve [AUC]: 0.86; 95% CI: 0.78 to 0.93; p = 0.0006). 68Ga-DOTATATE identifies high-risk stable lesions in coronary arteries Data from 6 ± 2 coronary segments were analyzed from each patient. Increased 68Ga-DOTATATE signals were often observed in nonculprit (bystander) lesions in ACS patients, particularly in low-attenuation plaques defined by CT (Figure 3). 68Ga-DOTATATE mTBRmax values were higher in nonculprit coronary segments in patients with both stable and unstable CVD, with either noncalcified/mixed plaque morphology or with high-risk CT features (spotty calcification, low attenuation, or positive remodeling) versus heavily calcified or normal arteries with no high-risk features (p < 0.0001) (Figure 4).

er in nonculprit coronary segments in patients with both stable and unstable CVD, with either noncalcified/mixed plaque morphology or with high-risk CT features (spotty calcification, low attenuation, or positive remodeling) versus heavily calcified or normal arteries with no high-risk features (p < 0.0001) (Figure 4). Coronary 68Ga-DOTATATE mTBRmax >2.12 had 83.3% (95% CI: 67.2% to 93.6%) sensitivity and 71.7% (95% CI: 64.6% to 78.0%) specificity (ROC AUC: 0.86; 95% CI: 0.80 to 0.92; p < 0.0001) to detect a segment with at least 1 high-risk CT feature. 68Ga-DOTATATE identifies culprit TIA/stroke lesions in carotid arteries In patients with TIA or stroke, increased 68Ga-DOTATATE inflammatory signals reliably differentiated between culprit carotid plaques and contralateral nonculprit carotid arteries (median difference mTBRmax: 0.13; IQR: 0.07 to 0.32; p = 0.003; median difference mdsTBRmax: 0.34; IQR: −0.01 to 0.53; p = 0.005) (Figure 5). Contralateral carotid 68Ga-DOTATATE mdsTBRmax in patients with TIA/stroke was also greater than in diseased (p = 0.01) or normal (p = 0.0001) carotids from patients with stable CVD (i.e., those without TIA/stroke or ACS). Nonculprit carotid 68Ga-DOTATATE mTBRmax was also higher in patients with unstable CVD (either TIA/stroke or ACS) versus stable CVD (p = 0.02).

in patients with TIA/stroke was also greater than in diseased (p = 0.01) or normal (p = 0.0001) carotids from patients with stable CVD (i.e., those without TIA/stroke or ACS). Nonculprit carotid 68Ga-DOTATATE mTBRmax was also higher in patients with unstable CVD (either TIA/stroke or ACS) versus stable CVD (p = 0.02). Aortic 68Ga-DOTATATE signals are related to coronary 68Ga-DOTATATE signals 68Ga-DOTATATE mTBRmax values in the coronary arteries and neighboring aorta showed a moderate correlation (r = 0.43; 95% CI: 0.11 to 0.66; p = 0.008). Aortic 68Ga-DOTATATE mTBRmax was negatively correlated with coronary calcium scores in patients with a total score of <400 (r = −0.66; 95% CI: −0.87 to 0.26; p = 0.003).

68Ga-DOTATATE mTBRmax values in the coronary arteries and neighboring aorta showed a moderate correlation (r = 0.43; 95% CI: 0.11 to 0.66; p = 0.008). Aortic 68Ga-DOTATATE mTBRmax was negatively correlated with coronary calcium scores in patients with a total score of <400 (r = −0.66; 95% CI: −0.87 to 0.26; p = 0.003). Vascular 68Ga-DOTATATE signals are related to clinical CVD risk factors Relationships between vascular 68Ga-DOTATATE signals and clinical CVD risk factors were evaluated to explore possible mechanistic links between 68Ga-DOTATATE and underlying disease pathology. Age (r = 0.44; 95% CI: 0.20 to 0.62; p = 0.0004), total cholesterol (r = 0.51; 95% CI: 0.30 to 0.68]; p < 0.0001), and Framingham risk score (r = 0.53; 95% CI: 0.32 to 0.69; p <0.0001) showed significant correlations with carotid 68Ga-DOTATATE mTBRmax (Figure 6). Carotid 68Ga-DOTATATE mTBRmax also differed significantly across patients grouped according to Framingham risk score (p < 0.0001). Body mass index (BMI) was positively correlated with aortic 68Ga-DOTATATE mTBRmax (r = 0.38; 95% CI: 0.06 to 0.64; p = 0.017). When age, total cholesterol, and BMI were evaluated with other relevant clinical factors using multivariate linear regression, they remained significant predictors of 68Ga-DOTATATE mTBRmax (Online Table 1). Carotid 68Ga-DOTATATE TBRmax values also varied significantly in patients without TIA/stroke who were taking statins, with lower values seen in patients taking high-intensity statins compared to those taking moderate or low dosages (p = 0.004) (Figure 7).

Vascular 68Ga-DOTATATE signals are related to clinical CVD risk factors Relationships between vascular 68Ga-DOTATATE signals and clinical CVD risk factors were evaluated to explore possible mechanistic links between 68Ga-DOTATATE and underlying disease pathology. Age (r = 0.44; 95% CI: 0.20 to 0.62; p = 0.0004), total cholesterol (r = 0.51; 95% CI: 0.30 to 0.68]; p < 0.0001), and Framingham risk score (r = 0.53; 95% CI: 0.32 to 0.69; p <0.0001) showed significant correlations with carotid 68Ga-DOTATATE mTBRmax (Figure 6). Carotid 68Ga-DOTATATE mTBRmax also differed significantly across patients grouped according to Framingham risk score (p < 0.0001). Body mass index (BMI) was positively correlated with aortic 68Ga-DOTATATE mTBRmax (r = 0.38; 95% CI: 0.06 to 0.64; p = 0.017). When age, total cholesterol, and BMI were evaluated with other relevant clinical factors using multivariate linear regression, they remained significant predictors of 68Ga-DOTATATE mTBRmax (Online Table 1). Carotid 68Ga-DOTATATE TBRmax values also varied significantly in patients without TIA/stroke who were taking statins, with lower values seen in patients taking high-intensity statins compared to those taking moderate or low dosages (p = 0.004) (Figure 7). In the 1.6 ± 0.2 years following PET imaging, 2 patients attended the emergency department with nonanginal chest pain, and there were 2 out-of-hospital deaths; our study was not powered to assess the ability of PET imaging to predict clinical events.

Carotid 68Ga-DOTATATE TBRmax values also varied significantly in patients without TIA/stroke who were taking statins, with lower values seen in patients taking high-intensity statins compared to those taking moderate or low dosages (p = 0.004) (Figure 7). In the 1.6 ± 0.2 years following PET imaging, 2 patients attended the emergency department with nonanginal chest pain, and there were 2 out-of-hospital deaths; our study was not powered to assess the ability of PET imaging to predict clinical events. Comparison of 68Ga-DOTATATE versus [18F]FDG-defined inflammation The time interval between 68Ga-DOTATATE and [18F]FDG imaging was a median of 2 days (IQR: 1 to 7 days). Coronary, carotid, and aortic 68Ga-DOTATATE and [18F]FDG mTBRmax values were strongly correlated with each other (r = 0.73; 95% CI: 0.64 to 0.81; p < 0.0001), although coronary artery [18F]FDG data were uninterpretable in 27 (64%) patients because of high myocardial spillover. Of 2 ACS patients with interpretable coronary [18F]FDG data, culprit mTBRmax values were numerically higher than the highest nonculprit segment in 1 patient.

= 0.73; 95% CI: 0.64 to 0.81; p < 0.0001), although coronary artery [18F]FDG data were uninterpretable in 27 (64%) patients because of high myocardial spillover. Of 2 ACS patients with interpretable coronary [18F]FDG data, culprit mTBRmax values were numerically higher than the highest nonculprit segment in 1 patient. [18F]FDG mTBRmax but not mdsTBRmax differentiated culprit from contralateral carotids (median difference: 0.12; IQR: 0.00 to 0.23; p = 0.008). Comparisons between coronary [18F]FDG mTBRmax values and CT morphology are shown in Figure 4 and clinical risk factors in Figure 6. Coronary 68F-FDG mTBRmax of >2.05 had 53.3% (95% CI: 26.6% to 78.7%) sensitivity and 92.4% (95% CI: 83.2% to 97.5%) specificity (ROC AUC: 0.76; 95% CI: 0.62 to 0.91; p = 0.002) for high-risk CT features. 68Ga-DOTATATE demonstrated higher TBR values and superior ability to discriminate high-risk versus low-risk coronary atherosclerotic lesions than [18F]FDG. Target validation in macrophages from CVD patients In CVD patients, macrophage SSTR2 mRNA was increased a median 91-fold (IQR: 56 to 104) above baseline versus 13-fold (IQR: 4.0 to 25) in age- and sex-matched healthy volunteers (p = 0.01), after stimulation with lipopolysaccharide. Presence of SST2 receptors was confirmed by immunostaining and specific binding of 68Ga-DOTATATE to SST2 in cultured macrophages shown by autoradiography.

old (IQR: 56 to 104) above baseline versus 13-fold (IQR: 4.0 to 25) in age- and sex-matched healthy volunteers (p = 0.01), after stimulation with lipopolysaccharide. Presence of SST2 receptors was confirmed by immunostaining and specific binding of 68Ga-DOTATATE to SST2 in cultured macrophages shown by autoradiography. Autoradiographic and histological target validation in carotid plaques Following PET-CT imaging, 8 patients underwent carotid endarterectomy. The PET scan-to-surgery time interval was a median of 9 (IQR: 3 to 35) days. Ex vivo 68Ga-DOTATATE carotid autoradiography showed high levels of specific ligand binding to SST2 receptors in all specimens (n = 8). A small degree of nonspecific binding was seen in relation to freshly cut calcium and as a result of edge artifact, which occurs when tissue edges curl causing the ligand to remain trapped during the experiment. 68Ga-DOTATATE binding within carotid plaques occurred mainly in the necrotic cores and shoulder regions, where there was strong colocalization of CD68 and SST2 staining (Figure 8, Online Figure 3). Neither 68Ga-DOTATATE binding nor SST2 staining was observed within thick fibrous cap regions consisting mainly of vascular smooth muscle cells.

within carotid plaques occurred mainly in the necrotic cores and shoulder regions, where there was strong colocalization of CD68 and SST2 staining (Figure 8, Online Figure 3). Neither 68Ga-DOTATATE binding nor SST2 staining was observed within thick fibrous cap regions consisting mainly of vascular smooth muscle cells. Carotid SSTR2/CD68 mRNA versus 68Ga-DOTATATE activity SSTR2 and CD68 mRNA levels were highly correlated within carotid plaque (r = 0.93; 95% CI: 0.49 to 0.99; p = 0.007) (Figure 9). Carotid SSTR2 and CD68 mRNA levels also showed strong correlation with in vivo 68Ga-DOTATATE TBRmax values measured at the corresponding level in clinical PET images, orientated at the bifurcation (SSTR2 r = 0.89; 95% CI: 0.28 to 0.99; p = 0.02; CD68 r = 0.84; 95% CI: 0.09 to 0.98; p = 0.04). Moreover, immunofluorescence demonstrated high cell specificity of colocalized of SST2 and CD68 staining in carotid plaque macrophages. These data provided both histological and molecular validation of 68Ga-DOTATATE as a specific marker of atherosclerotic inflammation. Discussion There have been previous reports of 68Ga-DOTATATE imaging in atherosclerosis, but they have been pre-clinical or retrospective studies, with the exception of 1 report limited to the carotid arteries. We provide the first definitive prospective validation of 68Ga-DOTATATE imaging as a marker of atherosclerotic inflammation.

e have been previous reports of 68Ga-DOTATATE imaging in atherosclerosis, but they have been pre-clinical or retrospective studies, with the exception of 1 report limited to the carotid arteries. We provide the first definitive prospective validation of 68Ga-DOTATATE imaging as a marker of atherosclerotic inflammation. Which cells express SSTR2 in atherosclerosis? We confirmed that high target SSTR2 gene expression occurs exclusively among activated proinflammatory M1 macrophages in atherosclerosis and demonstrated the presence of SST2 receptors in macrophages from patients with CVD. As a glucose analog, [18F]FDG lacks cell specificity, but there is some evidence that [18F]FDG accumulates more in M1 macrophages than in other macrophage subtypes because of differing glycolytic activity between these cells (17).

is and demonstrated the presence of SST2 receptors in macrophages from patients with CVD. As a glucose analog, [18F]FDG lacks cell specificity, but there is some evidence that [18F]FDG accumulates more in M1 macrophages than in other macrophage subtypes because of differing glycolytic activity between these cells (17). Binding of 68Ga-DOTATATE within atherosclerotic plaques We observed specific 68Ga-DOTATATE ligand binding to SST2 receptors within CD68+ macrophage-rich carotid plaque regions and strong correlations between carotid SSTR2 mRNA and in vivo 68Ga-DOTATATE activity. Although low levels of SST2 expression have been previously reported in vascular smooth muscle cells, we did not observe 68Ga-DOTATATE binding nor SST2 staining within the thick fibrous cap regions where these cells are abundant, suggesting that the synthetic atherosclerotic vascular smooth muscle cell phenotype is unlikely to express SST2 to a degree that would be detectable by clinical imaging. These laboratory-based findings provide robust histological and molecular validation of 68Ga-DOTATATE as a specific marker of atherosclerotic inflammation.

sting that the synthetic atherosclerotic vascular smooth muscle cell phenotype is unlikely to express SST2 to a degree that would be detectable by clinical imaging. These laboratory-based findings provide robust histological and molecular validation of 68Ga-DOTATATE as a specific marker of atherosclerotic inflammation. Culprit and high-risk plaque inflammation In clinical imaging, we found that 68Ga-DOTATATE PET correctly identified culprit coronary and carotid arteries in individuals with ACS or TIA/stroke. The median difference between culprit and nonculprit carotid arteries was less pronounced than in coronary arteries, but these 2 regions are not necessarily directly comparable because of the high prevalence of asymptomatic contralateral carotid disease, differing imaging time points affecting tracer kinetics, and local factors determining tracer delivery and clearance. 68Ga-DOTATATE demonstrated reliable diagnostic accuracy to detect stable yet inflamed coronary lesions with high-risk CT morphological features.

prevalence of asymptomatic contralateral carotid disease, differing imaging time points affecting tracer kinetics, and local factors determining tracer delivery and clearance. 68Ga-DOTATATE demonstrated reliable diagnostic accuracy to detect stable yet inflamed coronary lesions with high-risk CT morphological features. Systemic inflammation The ability of 68Ga-DOTATATE to detect generalized vascular inflammation was shown by the close relationship between PET signals in neighboring coronary and aortic vasculature and increased inflammatory signals arising from nonculprit carotids in patients with unstable CVD. Both of these features have been previously demonstrated using [18F]FDG (2). Moreover, significant correlations were observed between clinical CVD risk factors and generalized vascular 68Ga-DOTATATE inflammatory signals, which were overall lower in patients receiving high-intensity statins and with increasing coronary calcium scores up to 400. The inverse relationship between statin dosages and signal intensity provide anecdotal evidence that 68Ga-DOTATATE PET may provide a useful imaging platform for monitoring the anti-inflammatory effects of atherosclerosis drugs.

ents receiving high-intensity statins and with increasing coronary calcium scores up to 400. The inverse relationship between statin dosages and signal intensity provide anecdotal evidence that 68Ga-DOTATATE PET may provide a useful imaging platform for monitoring the anti-inflammatory effects of atherosclerosis drugs. 68Ga-DOTATATE outperforms [18F]FDG Although 68Ga-DOTATATE signals were strongly correlated with [18F]FDG-defined inflammation in multiple vascular territories, disparity between these 2 tracers reflects the fact that 68Ga-DOTATATE is a specific macrophage marker in atherosclerosis, whereas [18F]FDG provides a nonspecific measurement of glucose metabolism within plaque cells. Superiority of 68Ga-DOTATATE compared with [18F]FDG was shown by better power to discriminate high-risk versus low-risk coronary atherosclerotic lesions, higher signal-to-blood ratios, and consistently lower myocardial activity, affording clear coronary signal interpretation.

of glucose metabolism within plaque cells. Superiority of 68Ga-DOTATATE compared with [18F]FDG was shown by better power to discriminate high-risk versus low-risk coronary atherosclerotic lesions, higher signal-to-blood ratios, and consistently lower myocardial activity, affording clear coronary signal interpretation. [18F]FDG imaging is notoriously unreliable in coronaries; in contrast, myocardial 68Ga-DOTATATE binding was sufficiently low to allow coronary artery inflammation imaging in all patients. 68Ga-DOTATATE inflammatory signals differentiated culprit from contralateral carotids, using both “mean of the whole artery” and “most-diseased segment” methods, but [18F]FDG only detected a difference in mean carotid uptake, hinting that 68Ga-DOTATATE may offer a more focal approach. 68Ga-DOTATATE signals also appeared more discretely localized than [18F]FDG signals in clinical images (Figures 3 and 5). Given the higher cost of 68Ga-DOTATATE than [18F]FDG, its use for noncoronary vascular imaging may not be justified, although in the context of research, increased macrophage specificity of 68Ga-DOTATATE potentially holds significant advantage for detection of subtle changes in vascular biology that may not be as clearly appreciated using a blunter imaging tool such as [18F]FDG.

for noncoronary vascular imaging may not be justified, although in the context of research, increased macrophage specificity of 68Ga-DOTATATE potentially holds significant advantage for detection of subtle changes in vascular biology that may not be as clearly appreciated using a blunter imaging tool such as [18F]FDG. A small number of previous studies have investigated SST2 PET imaging in CVD. Two studies demonstrated autoradiographic 68Ga-DOTATATE binding within macrophage-rich aortic atherosclerotic plaques in mice 6, 7. Five retrospective analyses of PET scans from patients who underwent imaging for oncological indications reported significant statistical relationships between vascular SST2 signals and clinical CVD factors, including older age, male sex, hypercholesterolemia, presence of calcified coronary plaque, prior CVD events, and Framingham risk score calculated using BMI 8, 9, 10, 18, 19. In 1 study, a strong correlation was observed between 68Ga-DOTATATE and [18F]FDG vascular TBR values, although signals from the 2 tracers did not colocalize at the sites of highest tracer uptake (9). In another study of 11 patients who underwent 3 serial 68Ga-DOTATATE scans following peptide receptor radionuclide therapy with lutetium-177-labeled DOTATATE, good interscan reproducibility of 68Ga-DOTATATE TBR measurements prior to radionuclide therapy was observed, as well as significant signal reduction 1 month after, which was most pronounced in relation to noncalcified plaques (10). These studies, although retrospective and without CT angiography or ECG-gating, are consistent with our findings.

ility of 68Ga-DOTATATE TBR measurements prior to radionuclide therapy was observed, as well as significant signal reduction 1 month after, which was most pronounced in relation to noncalcified plaques (10). These studies, although retrospective and without CT angiography or ECG-gating, are consistent with our findings. Our finding that SST2 PET can differentiate culprit from contralateral carotid arteries is supported by another study of 64Cu-DOTATATE PET cardiac magnetic resonance in 10 patients with carotid TIA/stroke (20). However, in that study, correlation between carotid copper-64-labeled DOTATATE signals and gene expression of the monocyte/macrophage marker CD163 was observed using a multivariate regression model, leading the authors to conclude that this tracer reports on alternatively activated M2 macrophages. As hemoglobin-haptoglobin scavenging by CD163 in the setting of intraplaque hemorrhage directs monocyte differentiation toward an atheroprotective M2 phenotype (21), the finding of increased CD163 mRNA within advanced ruptured plaques is unsurprising. However, there is no current evidence to indicate that significant SSTR2 expression occurs in M2 macrophages. Our findings agree with those of previous work indicating that 68Ga-DOTATATE signals in atherosclerosis occur because of intracellular tracer accumulation following cell surface binding and receptor internalization among dense clusters of classically activated M1 macrophages (22).

2 expression occurs in M2 macrophages. Our findings agree with those of previous work indicating that 68Ga-DOTATATE signals in atherosclerosis occur because of intracellular tracer accumulation following cell surface binding and receptor internalization among dense clusters of classically activated M1 macrophages (22). Next steps involve testing in a larger, longitudinal study with clinical outcomes, similar to the ongoing BioImage (NCT00738725) and PESA (Progression of Early Subclinical Atherosclerosis; NCT01410318) [18F]FDG studies. Study limitations Limitations of our study include inherent technical challenges of vascular PET imaging, namely low spatial resolution (∼5 mm) and image artifacts created by cardiorespiratory motion that are confounded by the high positron energy of 68Ga (Emax: 1.9 MeV; average positron range: 2.4 mm). To overcome these difficulties, we used CT angiography for precise anatomical PET signal localization (spatial resolution: 0.5 to 0.6 mm), ECG-gated PET reconstruction to reduce the impact of motion, and iterative time-of-flight reconstruction with point spread function modeling to provide resolution recovery and reduce partial volume error. Coronary signal-to-noise ratio could potentially be improved further by motion correction methods in active development (23). We did not attempt myocardial suppression of [18F]FDG using dietary manipulation or prolonged fasting, because in our experience, these methods are ineffective in ∼50% of cases (3) and are inconvenient for patients. Nevertheless, others have reported success using these methods.

Study limitations Limitations of our study include inherent technical challenges of vascular PET imaging, namely low spatial resolution (∼5 mm) and image artifacts created by cardiorespiratory motion that are confounded by the high positron energy of 68Ga (Emax: 1.9 MeV; average positron range: 2.4 mm). To overcome these difficulties, we used CT angiography for precise anatomical PET signal localization (spatial resolution: 0.5 to 0.6 mm), ECG-gated PET reconstruction to reduce the impact of motion, and iterative time-of-flight reconstruction with point spread function modeling to provide resolution recovery and reduce partial volume error. Coronary signal-to-noise ratio could potentially be improved further by motion correction methods in active development (23). We did not attempt myocardial suppression of [18F]FDG using dietary manipulation or prolonged fasting, because in our experience, these methods are ineffective in ∼50% of cases (3) and are inconvenient for patients. Nevertheless, others have reported success using these methods. Most of the ACS patients underwent stenting prior to PET imaging and persistence of procedure-related inflammation could have, in theory, augmented inflammatory signals detected in culprit coronary lesions. Clinical identification of culprit arteries can be challenging, particularly in the coronary arteries; although intravascular imaging can be used to confirm plaque rupture, this investigation was not performed in any of the patients in this study. Last, although the novel finding of increased SSTR2 expression in LPS-stimulated macrophages from patients with CVD versus healthy volunteers is intriguing, further testing in a larger patient cohort is needed.

be used to confirm plaque rupture, this investigation was not performed in any of the patients in this study. Last, although the novel finding of increased SSTR2 expression in LPS-stimulated macrophages from patients with CVD versus healthy volunteers is intriguing, further testing in a larger patient cohort is needed. Conclusions We provide gene-, cell-, plaque-, and patient-level data demonstrating that SST2 PET imaging using 68Ga-DOTATATE provides a quantifiable, cell-specific marker of atherosclerotic inflammation that outperforms [18F]FDG in the coronary arteries. Further work is needed to confirm these findings in a larger patient population and to compare imaging with clinical outcomes. 68Ga-DOTATATE PET offers measurement of both generalized atherosclerotic disease activity and detailed information about local plaque functional phenotype to complement multimodal assessments of anatomic, morphologic, and hemodynamic disease severity. This approach, in selected patient populations, has the potential to improve CVD risk prediction, allowing personalized tailoring of therapies aimed to improve clinical outcomes.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: 68Ga-DOTATATE binds to somatostatin receptor-2 (SST2) in activated inflammatory macrophages, and the tissue-to-blood ratios of 68Ga-DOTATATE distinguishes culprit from nonculprit coronary and carotid arteries in patients with ACS, stroke, or TIA. Although [18F]FDG also differentiates these types of arterial lesions, myocardial spillover renders coronary [18F]FDG PET scans uninterpretable in a high proportion of patients.

sue-to-blood ratios of 68Ga-DOTATATE distinguishes culprit from nonculprit coronary and carotid arteries in patients with ACS, stroke, or TIA. Although [18F]FDG also differentiates these types of arterial lesions, myocardial spillover renders coronary [18F]FDG PET scans uninterpretable in a high proportion of patients. TRANSLATIONAL OUTLOOK: Future research should explore the utility of 68Ga-DOTATATE PET imaging of inflammation to classify patients for more aggressive therapeutic intervention and explore potential application to other inflammatory cardiovascular diseases. Appendix Online Data

sue-to-blood ratios of 68Ga-DOTATATE distinguishes culprit from nonculprit coronary and carotid arteries in patients with ACS, stroke, or TIA. Although [18F]FDG also differentiates these types of arterial lesions, myocardial spillover renders coronary [18F]FDG PET scans uninterpretable in a high proportion of patients. TRANSLATIONAL OUTLOOK: Future research should explore the utility of 68Ga-DOTATATE PET imaging of inflammation to classify patients for more aggressive therapeutic intervention and explore potential application to other inflammatory cardiovascular diseases. Appendix Online Data This study was funded by the Wellcome Trust and supported by the National Institute for Health Research (NIHR) Cambridge Biomedical Research Centre and the Cambridge Clinical Trials Unit. Dr. Tarkin is supported by a Wellcome Trust research training fellowship (104492/Z/14/Z). Dr. Evans is supported by a Dunhill Medical Trust fellowship (RTF44/0114). Dr. Chowdhury is supported by Royal College of Surgeons of England and British Heart Foundation (BHF) fellowships (FS/16/29/31957). Drs. Manavaki and Warburton are supported by the NIHR Biomedical Research Centres. Drs. Yu and Frontini are supported by the BHF (RE/13/6/30180). Dr. Fryer is supported by Higher Education Funding Council for England (HEFCE). Dr. Groves is supported by the University College London Hospital NIHR Biomedical Research Centre; and has received grant support from GlaxoSmithKline. Dr. Ouwehand’s laboratory is funded by EU-FP7 project Blueprint (Health-F5-2011-282510), BHF (PG-0310-1002 and RG/09/12/28096), and National Health Service Blood and Transplant. Dr. Bennett is supported by NIHR and BHF. Dr. Davenport is supported by research grants from Wellcome Trust (107715/Z/15/Z), Medical Research Council (MC_PC_14116), and BHF (RE-13-6-3180). Dr. Rudd is supported by the NIHR, BHF, Wellcome Trust, and HEFCE. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

port is supported by research grants from Wellcome Trust (107715/Z/15/Z), Medical Research Council (MC_PC_14116), and BHF (RE-13-6-3180). Dr. Rudd is supported by the NIHR, BHF, Wellcome Trust, and HEFCE. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section and supplemental figures and a table, please see the online version of this article. Figure 1 The VISION Study Patient (A) and procedure (B) flowcharts. ∗Did not meet study criteria, n = 8; other clinical factors, n = 3; declined/cancelled, n = 49. †Coronary artery PET data excluded in ACS patients with ambiguous culprit arteries (n = 2). ‡Carotid artery PET data excluded in patients with prior carotid surgery (n = 2). §[18F]-FDG PET imaging not completed because of timing of surgery (n = 1). ‖Tissue samples excluded owing to insufficient mRNA extracted for quantitative PCR (n = 2). ¶CT scans not completed (calcium scan, n = 1; coronary angiogram, n = 5; carotid angiogram, n = 2). ACS = acute coronary syndrome; CT = computed tomography; CVD = cardiovascular disease; FDG = fluorodeoxyglucose; PCR = polymerase chain reaction; PET = positron emission tomography; TIA = transient ischemic attack; VISION = Vascular Inflammation imaging using Somatostatin receptor positron emissION tomography. Figure 2 Target SSRT2 Expression in Proinflammatory Macrophages

Patient (A) and procedure (B) flowcharts. ∗Did not meet study criteria, n = 8; other clinical factors, n = 3; declined/cancelled, n = 49. †Coronary artery PET data excluded in ACS patients with ambiguous culprit arteries (n = 2). ‡Carotid artery PET data excluded in patients with prior carotid surgery (n = 2). §[18F]-FDG PET imaging not completed because of timing of surgery (n = 1). ‖Tissue samples excluded owing to insufficient mRNA extracted for quantitative PCR (n = 2). ¶CT scans not completed (calcium scan, n = 1; coronary angiogram, n = 5; carotid angiogram, n = 2). ACS = acute coronary syndrome; CT = computed tomography; CVD = cardiovascular disease; FDG = fluorodeoxyglucose; PCR = polymerase chain reaction; PET = positron emission tomography; TIA = transient ischemic attack; VISION = Vascular Inflammation imaging using Somatostatin receptor positron emissION tomography. Figure 2 Target SSRT2 Expression in Proinflammatory Macrophages Heatmap of population-based RNA sequencing data (A) showing high SSTR2 expression in proinflammatory M1 macrophages (n = 4), very low levels of SSTR2 expression in unstimulated M0 macrophages (n = 4), and alternatively activated M2 macrophages (n = 5). For comparison, a heatmap of GLUT1 and GLUT3 shows significant gene expression in all cell types (note, different scales for SSRT and GLUT genes; mean values are log2 fragments per kilobase of transcript for million mapped reads [FPKM+1]). SSTR2 expression in LPS-stimulated macrophages from CVD patients versus age- and sex-matched healthy volunteers (n = 3 for both) using quantitative PCR (B). Photomicrograph shows green fluorescent immunoreactive SST2 staining in macrophages (C), with blue nuclear DAPI-stained ([inset] concentration and isotype-matched IgG negative control). Brightfield photomicrograph shows brown immunoreactive SST2-stained cultured macrophages, with nuclear counterstain (D). Phosphor autoradiographic image shows total binding of 68Ga-DOTATATE (E) in clusters of cultured macrophages ([inset] parallel incubation with 68Ga-DOTATATE and cold competing ligand showing very low levels of nonspecific binding). IgG = immunoglobulin G; LPS = lipopolysaccharide; other abbreviations as in Figure 1.

). Phosphor autoradiographic image shows total binding of 68Ga-DOTATATE (E) in clusters of cultured macrophages ([inset] parallel incubation with 68Ga-DOTATATE and cold competing ligand showing very low levels of nonspecific binding). IgG = immunoglobulin G; LPS = lipopolysaccharide; other abbreviations as in Figure 1. Figure 3 Coronary PET Inflammation Imaging: ACS Culprit Versus Bystander Lesions

). Phosphor autoradiographic image shows total binding of 68Ga-DOTATATE (E) in clusters of cultured macrophages ([inset] parallel incubation with 68Ga-DOTATATE and cold competing ligand showing very low levels of nonspecific binding). IgG = immunoglobulin G; LPS = lipopolysaccharide; other abbreviations as in Figure 1. Figure 3 Coronary PET Inflammation Imaging: ACS Culprit Versus Bystander Lesions X-ray angiography images from a 59-year old man with ACS, showing a culprit first obtuse marginal lesion ([A] hatched oval) and nonculprit (bystander) right coronary artery disease ([E] circle). Identification of a culprit artery was aided by electrocardiographic findings of lateral T-wave inversion. Corresponding CT angiography images (B, F) show stented culprit lesion (*) and native bystander lesion with high-risk plaque morphology ([inset] low attenuation, cross-section of artery with outer wall boundary marked by dotted outline). In both lesions, intense inflammation (arrows) detected by 68Ga-DOTATATE PET (C, G) is reproduced by [18F]FDG PET (D, H). Graphs of culprit versus highest nonculprit coronary 68Ga-DOTATATE TBRmax values in patients (n = 8) with ACS and stented culprit ACS lesions (n = 6) versus stable stented (n = 18) lesions (I). ROC analysis demonstrates good diagnostic accuracy of 68Ga-DOTATATE for culprit coronary lesions (J). Note stable stented lesions are coronary stents that were inserted >3 months prior to PET imaging in all but 1 patient. AUC = area under curve; mTBRmax = mean of maximum tissue-to-blood ratios; ROC = receiver operating characteristic; other abbreviations as in Figures 1 and 2.

DOTATATE for culprit coronary lesions (J). Note stable stented lesions are coronary stents that were inserted >3 months prior to PET imaging in all but 1 patient. AUC = area under curve; mTBRmax = mean of maximum tissue-to-blood ratios; ROC = receiver operating characteristic; other abbreviations as in Figures 1 and 2. Figure 4 Coronary PET Inflammation Imaging: High-Risk CT Features (A) X-ray and (D) CT coronary angiograms of a 67-year-old man with stable angina, showing minor LCx atheroma (hatched oval) with spotty calcification ([inset] *calcium scan) and calcified plaque in the LAD artery. Although 68Ga-DOTATATE PET (B, E) allows unimpeded interpretation of inflammation in the LCx lesion (B, arrow), and lack of signal in the LAD, coronary [18F]FDG imaging is obscured by patchy myocardial tracer uptake (C). Graphs compare 68Ga-DOTATATE (F) with [18F]FDG (G) coronary TBRmax values by CT plaque morphology in coronary segments (68Ga-DOTATATE: NCP or MP, n = 86; normal, n = 45; spotty calcium, n = 30; large calcium, n = 72; LA or PR, n = 11; no high-risk CT, n = 186; [18F]FDG: NCP or MP, n = 43; normal, n = 13; spotty calcium, n = 15; large calcium n = 14; LA or PR, n = 4; no high-risk CT, n = 66), and ROC analysis demonstrating good diagnostic accuracy for high-risk coronary lesions. LA = low attenuation; LAD = left anterior descending; LCx = left circumflex; NCP = noncalcified plaque; MP = mixed plaque; PR = positive remodeling; other abbreviations in Figures 1, 2, and 3. Figure 5 Carotid PET Inflammation Imaging: TIA/Stroke

(A) X-ray and (D) CT coronary angiograms of a 67-year-old man with stable angina, showing minor LCx atheroma (hatched oval) with spotty calcification ([inset] *calcium scan) and calcified plaque in the LAD artery. Although 68Ga-DOTATATE PET (B, E) allows unimpeded interpretation of inflammation in the LCx lesion (B, arrow), and lack of signal in the LAD, coronary [18F]FDG imaging is obscured by patchy myocardial tracer uptake (C). Graphs compare 68Ga-DOTATATE (F) with [18F]FDG (G) coronary TBRmax values by CT plaque morphology in coronary segments (68Ga-DOTATATE: NCP or MP, n = 86; normal, n = 45; spotty calcium, n = 30; large calcium, n = 72; LA or PR, n = 11; no high-risk CT, n = 186; [18F]FDG: NCP or MP, n = 43; normal, n = 13; spotty calcium, n = 15; large calcium n = 14; LA or PR, n = 4; no high-risk CT, n = 66), and ROC analysis demonstrating good diagnostic accuracy for high-risk coronary lesions. LA = low attenuation; LAD = left anterior descending; LCx = left circumflex; NCP = noncalcified plaque; MP = mixed plaque; PR = positive remodeling; other abbreviations in Figures 1, 2, and 3. Figure 5 Carotid PET Inflammation Imaging: TIA/Stroke Views from a 66-year-old man ([top] axial plane) and a 70-year-old man ([bottom] sagittal plane), both of whom had TIAs resulting from right internal carotid artery lesions, shown on CT (A [hatched circle], D [∗]), with intense culprit plaque inflammation (hatched circles/arrows) detected by 68Ga-DOTATATE (B, E) and reproduced by [18F]FDG (C, F). Graphs compare culprit versus nonculprit 68Ga-DOTATATE (G) and [18F]FDG (H) TBRmax values (68Ga-DOTATATE: culprit n = 14; contralateral n = 14; nonculprit [unstable CVD] n = 31; nonculprit [stable CVD] n = 24; contralateral [TIA/stroke] n = 14; diseased stable CVD n = 19; [18F]FDG: culprit n = 13; contralateral n = 13; nonculprit [unstable CVD] n = 31; nonculprit [stable CVD] n = 24; contralateral [TIA/stroke] n = 13; diseased [stable CVD] n = 19). TIA = transient ischemic attack; other abbreviations as in Figures 1, 2, and 3.

ntralateral [TIA/stroke] n = 14; diseased stable CVD n = 19; [18F]FDG: culprit n = 13; contralateral n = 13; nonculprit [unstable CVD] n = 31; nonculprit [stable CVD] n = 24; contralateral [TIA/stroke] n = 13; diseased [stable CVD] n = 19). TIA = transient ischemic attack; other abbreviations as in Figures 1, 2, and 3. Figure 6 Vascular Inflammation Versus Clinical Risk Factors Graphs show correlations of vascular inflammation detected by 68Ga-DOTATATE (A to C) and [18F]FDG (D) versus clinical cardiovascular disease risk factors. (Carotid arteries n = 62; aortas, n = 38; note data from patients not taking statins [n = 4] were excluded to control for this variable). Abbreviations as in Figure 1. Figure 7 68Ga-DOTATATE Versus [18F]FDG-Defined Inflammation and Other Clinical Factors

Graphs show correlations of vascular inflammation detected by 68Ga-DOTATATE (A to C) and [18F]FDG (D) versus clinical cardiovascular disease risk factors. (Carotid arteries n = 62; aortas, n = 38; note data from patients not taking statins [n = 4] were excluded to control for this variable). Abbreviations as in Figure 1. Figure 7 68Ga-DOTATATE Versus [18F]FDG-Defined Inflammation and Other Clinical Factors Graphs show (A) the strong correlations among coronary, carotid, and aortic 68Ga-DOTATATE mTBRmax versus [18F]FDG mTBRmax (n = 123 mean arterial values per tracer); (B) carotid 68Ga-DOTATATE mTBRmax grouped by FRS (<8%, n = 16; 8% to 16%, n = 14; >16%, n = 32); (C) negative correlation of coronary aortic mTBRmax versus calcium score in patients with CAC <400 (n = 19); and (D) carotid 68Ga-DOTATATE TBRmax ROI values in non TIA/stroke patients grouped by statin dosages (n = 20 patients [14 ROIs per artery]; low-dose n = 4; moderate dose, n = 9; high-dose, n = 7). Log transformed CAC values used to account for the non-normal distribution in the general population. Median difference between low-dose vs. high-dose: −0.13; (95% CI: −0.17 to −0.016); p = 0.02; %Δ −8.65. Low dose: simvastatin, 20 mg or lower; high-dose: atorvastatin, 80 mg or equivalent; moderate dose: all other dosages. CAC = coronary artery calcium score; FRS = %10-year Framingham risk score; ROI = region of interest; TIA = transient ischemic attack. Figure 8 68Ga-DOTATATE Ligand Binding to Macrophage SST2 in Carotid Plaque

Graphs show (A) the strong correlations among coronary, carotid, and aortic 68Ga-DOTATATE mTBRmax versus [18F]FDG mTBRmax (n = 123 mean arterial values per tracer); (B) carotid 68Ga-DOTATATE mTBRmax grouped by FRS (<8%, n = 16; 8% to 16%, n = 14; >16%, n = 32); (C) negative correlation of coronary aortic mTBRmax versus calcium score in patients with CAC <400 (n = 19); and (D) carotid 68Ga-DOTATATE TBRmax ROI values in non TIA/stroke patients grouped by statin dosages (n = 20 patients [14 ROIs per artery]; low-dose n = 4; moderate dose, n = 9; high-dose, n = 7). Log transformed CAC values used to account for the non-normal distribution in the general population. Median difference between low-dose vs. high-dose: −0.13; (95% CI: −0.17 to −0.016); p = 0.02; %Δ −8.65. Low dose: simvastatin, 20 mg or lower; high-dose: atorvastatin, 80 mg or equivalent; moderate dose: all other dosages. CAC = coronary artery calcium score; FRS = %10-year Framingham risk score; ROI = region of interest; TIA = transient ischemic attack. Figure 8 68Ga-DOTATATE Ligand Binding to Macrophage SST2 in Carotid Plaque In vivo CT angiography views of culprit carotid artery (hatched oval = internal carotid artery) in axial (A) and sagittal (E) views, with corresponding fused 68Ga-DOTATATE PET-CT (B). Ex vivo views of macrographic images of the explanted carotid specimen (I, hatched line signifies location of carotid section); phosphor autoradiographic image shows the total binding of 68Ga-DOTATATE to SST2 receptors in macrophages within a transverse carotid section (C) corresponding to the level shown in clinical images. Adjacent section was incubated with 68Ga-DOTATATE and cold competing ligand (D) showing very low levels of nonspecific binding. Brightfield photomicrographs show brown immunoreactive SST2 staining (G, J, M) of macrophages identified with the panmacrophage marker CD68 (H, K, N), colocalized SST2(brown), and CD68 (blue) staining in the same section (L); Movat’s pentachrome stain (F).

ting ligand (D) showing very low levels of nonspecific binding. Brightfield photomicrographs show brown immunoreactive SST2 staining (G, J, M) of macrophages identified with the panmacrophage marker CD68 (H, K, N), colocalized SST2(brown), and CD68 (blue) staining in the same section (L); Movat’s pentachrome stain (F). Figure 9 Carotid SSTR2/CD68 mRNA Versus In Vivo 68Ga-DOTATATE Activity Graphs show correlations of SSTR2 versus CD68 mRNA within ex vivo carotid plaques measured by quantitative PCR (A); SSTR2(B) and CD68(C) mRNA versus corresponding in vivo 68Ga-DOTATATE TBRmax values measured from clinical images (n = 6). Representative photomicrograph shows red SST2 and green CD68 fluorescent immunoreactive staining of macrophages within carotid plaque (D), with blue nuclear DAPI staining. Note presence of both double positive (+) and double negative (−) staining indicating high cell specificity (E). Central Illustration Comparison Between 68Ga-DOTATATE and [18F]FDG Coronary PET Inflammation Imaging

Graphs show correlations of SSTR2 versus CD68 mRNA within ex vivo carotid plaques measured by quantitative PCR (A); SSTR2(B) and CD68(C) mRNA versus corresponding in vivo 68Ga-DOTATATE TBRmax values measured from clinical images (n = 6). Representative photomicrograph shows red SST2 and green CD68 fluorescent immunoreactive staining of macrophages within carotid plaque (D), with blue nuclear DAPI staining. Note presence of both double positive (+) and double negative (−) staining indicating high cell specificity (E). Central Illustration Comparison Between 68Ga-DOTATATE and [18F]FDG Coronary PET Inflammation Imaging Images from a 57-year old man with acute coronary syndrome who presented with deep anterolateral T-wave inversion (arrow) on electrocardiogram (A) and serum troponin-I concentration elevated at 4,650 ng/l (NR: <17 ng/l). Culprit left anterior descending artery stenosis (dashed oval) was identified by X-ray angiography (B). After the patient underwent percutaneous coronary stenting (C), residual coronary plaque (*inset) with high-risk morphology (low attenuation and spotty calcification) is evident on CT angiography (D, E). Use of 68Ga-DOTATATE PET (F, H, I) clearly detected intense inflammation in this high-risk atherosclerotic plaque/distal portion of the stented culprit lesion (arrow) and recently infarcted myocardium (*). In contrast, using [18F]FDG PET (G, J), myocardial spillover completely obscures the coronary arteries. CT = computed tomography; [18F]FDG = fluorine-18-labeled fluorodeoxyglucose; 68Ga-DOTATATE = gallium-68-labeled DOTATATE; PET = positron emission tomography.

prit lesion (arrow) and recently infarcted myocardium (*). In contrast, using [18F]FDG PET (G, J), myocardial spillover completely obscures the coronary arteries. CT = computed tomography; [18F]FDG = fluorine-18-labeled fluorodeoxyglucose; 68Ga-DOTATATE = gallium-68-labeled DOTATATE; PET = positron emission tomography. Table 1 Baseline Clinical Factors

prit lesion (arrow) and recently infarcted myocardium (*). In contrast, using [18F]FDG PET (G, J), myocardial spillover completely obscures the coronary arteries. CT = computed tomography; [18F]FDG = fluorine-18-labeled fluorodeoxyglucose; 68Ga-DOTATATE = gallium-68-labeled DOTATATE; PET = positron emission tomography. Table 1 Baseline Clinical Factors Stable CVD (n = 18) Unstable CVD∗ (n = 24) All (n = 42) Age, yrs 67 ± 10 71 ± 7 69 ± 9 Male 14 (78) 20 (83) 34 (81) Body mass index, kg/m2 29 ± 5 27 ± 4 28 ± 5 Heart rate, beats/min 57 ± 9 58 ± 6 57 ± 8 Systolic blood pressure, mm Hg 141 ± 22 144 ± 24 143 ± 21 Diastolic blood pressure, mm Hg 74 ± 9 76 ± 10 75 ± 9 Occurrences of previous cardiovascular history Angina 8 (44) 4 (17) 12 (29) Myocardial infarction 3 (17) 9 (38) 12 (29) Coronary stenting 5 (28) 3 (13) 8 (19) Coronary artery bypass surgery 1 (6) 3 (13) 4 (10) Transient ischemia attack or stroke 4 (22) 3 (13) 7 (17) Carotid endarterectomy surgery 2 (11) 0 (0) 2 (5) Occurrences of cardiovascular risk factors Hypertension 9 (50) 16 (67) 25 (60) Hypercholesterolemia 17 (94) 17 (71) 34 (81) Noninsulin dependent diabetes 2 (11) 6 (25) 8 (19) Smoking habit (ex or current) 10 (56) 18 (75) 28 (67) Family history of coronary heart disease† 7 (39) 12 (50) 19 (45) Occurrences of cardiovascular medications Aspirin 15 (83) 15 (63) 30 (71) Clopidogrel 6 (33) 21 (88) 27 (64) Statin 16 (89) 22 (92) 38 (91) β-Adrenergic receptor blocker 11 (61) 14 (58) 25 (60) Angiotensin converting enzyme inhibitor/receptor blocker 8 (44) 17 (71) 25 (60) Calcium-channel blocker 3 (17) 9 (38) 12 (29) Other antihypertensive 3 (17) 4 (17) 7 (17) Oral nitrates 4 (22) 3 (13) 7 (17) Random lipid profile Total cholesterol, mmol/l 4.0 ± 1.1 3.6 ± 0.9 3.8 ± 0.8 HDL cholesterol, mmol/l 1.1 ± 0.2 1.1 ± 0.3 1.1 ± 0.2 LDL cholesterol, mmol/l 2.1 ± 0.6 1.9 ± 0.7 2.0 ± 0.7 Triglycerides, mmol/l 1.8 ± 0.9 1.3 ± 0.5 1.5 ± 0.7 HDL cholesterol, mmol/l 3.8 ± 1.0 3.4 ± 0.7 3.6 ± 0.9 Median high-sensitivity CRP, mg/l 2.5 (0.8–3.7) 2.1 (0.7–5.8) 2.4 (0.7–4.6) Median peak serum troponin-I concentration, ng/l‡ – 573 (59.5–3,957) – Median %10-year Framingham risk score 9 (8–21) 18 (11–26) 16 (8–26) Median coronary artery calcium score, Agatston units 177 (96–680) 756 (255–1,419) 433 (120–1,314) Values are mean ± SD, n (%), or mean (interquartile range).

(0.7–5.8) 2.4 (0.7–4.6) Median peak serum troponin-I concentration, ng/l‡ – 573 (59.5–3,957) – Median %10-year Framingham risk score 9 (8–21) 18 (11–26) 16 (8–26) Median coronary artery calcium score, Agatston units 177 (96–680) 756 (255–1,419) 433 (120–1,314) Values are mean ± SD, n (%), or mean (interquartile range). ACS = acute coronary syndrome; CRP = C-reactive protein; CVD = cardiovascular disease; HDL = high-density lipoprotein; LDL = low-density lipoprotein; TIA = transient ischemic attack. ∗ Unstable CVD = ACS or TIA/stroke within the previous 3 months. † <65 years of age. ‡ ACS patients.

Dilated cardiomyopathy (DCM) has a population prevalence of ∼1 in 500 and is associated with prognostically adverse arrhythmias at initial disease presentation in up to one-third of patients (1). While increasing age, male sex, and impaired ventricular function are established arrhythmic risk factors, arrhythmias also occur in patients without known risk factors. Recently, the advent of high throughput sequencing technologies has enabled new insights into the genetic predisposition of DCM. In particular, titin-truncating variants (TTNtv) are now known to occur in ∼15% of cases of DCM and represent the commonest genetic cause of DCM 2, 3. We evaluated whether genetic information can be used as an additional tool to identify patients at risk for arrhythmias by exploring whether there is an association between TTNtv and the occurrence of arrhythmias at the time of first diagnosis in a large cohort of patients with DCM.

commonest genetic cause of DCM 2, 3. We evaluated whether genetic information can be used as an additional tool to identify patients at risk for arrhythmias by exploring whether there is an association between TTNtv and the occurrence of arrhythmias at the time of first diagnosis in a large cohort of patients with DCM. In total, 572 prospectively recruited patients fulfilling diagnostic criteria for DCM by cardiovascular magnetic resonance were recruited between 2009 and 2015 (68% men, mean age 53.5 ± 14.4 years). All patients had detailed clinical assessment and sequencing for novel or rare (Exome Aggregation Consortium frequency <0.001) truncating variants in constitutively expressed TTN exons. Focusing on early arrhythmic risk, data on arrhythmia history (atrial fibrillation [AF], nonsustained ventricular tachycardia [VT], and sustained VT) on recruitment to the study were collated from hospital and primary care notes. Multivariable logistic regression was used to evaluate variables associated with arrhythmias at presentation. In the cohort, mean left ventricular ejection fraction was 39.0 ± 12.6% (median = 40%; interquartile range: 29% to 49%). Midwall late gadolinium enhancement (LGE) myocardial fibrosis was detected in 198 patients (35%). A family history of DCM was found in 82 patients (14%) and a family history of sudden cardiac death in 76 patients (13%).

left ventricular ejection fraction was 39.0 ± 12.6% (median = 40%; interquartile range: 29% to 49%). Midwall late gadolinium enhancement (LGE) myocardial fibrosis was detected in 198 patients (35%). A family history of DCM was found in 82 patients (14%) and a family history of sudden cardiac death in 76 patients (13%). Arrhythmias prior to recruitment were documented in 196 (34%) patients. Specifically, 139 (24%) patients had confirmed AF, 69 (12%) patients had confirmed nonsustained VT and 11 (2%) patients had confirmed sustained VT. Of these, 22 patients had more than 1 type of arrhythmia: 15 had both AF and nonsustained VT; 1 had both AF and sustained VT; 5 had both sustained VT and nonsustained VT; and 1 had AF, nonsustained VT, and sustained VT. Patients with arrhythmia were more likely to be older, be men, and have worse biventricular function (age 58.7 ± 12.2 years vs. 50.8 ± 14.7 years; 161 [82.1%] men vs. 227 [60.4%] men; left ventricular ejection fraction 36.1 ± 12.1% vs. 40.4 ± 12.7%; right ventricular ejection fraction 33.9 ± 13.7% vs. 40.7 ± 13.8%; p < 0.0001 for all). Although LGE is associated with arrhythmia later in established disease, there was no significant difference in the proportion of patients with LGE between the arrhythmia positive and negative groups at presentation (122 [32.4%] vs. 76 [38.8%]; p = 0.16).

ion fraction 33.9 ± 13.7% vs. 40.7 ± 13.8%; p < 0.0001 for all). Although LGE is associated with arrhythmia later in established disease, there was no significant difference in the proportion of patients with LGE between the arrhythmia positive and negative groups at presentation (122 [32.4%] vs. 76 [38.8%]; p = 0.16). TTNtv were observed in 13.3% (n = 26) of patients with a history of arrhythmia compared to 8% (n = 30) of patients without a history of arrhythmia (p = 0.05). Conversely, an arrhythmia was documented in 26 patients (46%) with TTNtv compared to 170 patients (33%) without TTNtv (p = 0.05). In exploratory univariable analysis, the presence of a TTNtv was predictive of baseline arrhythmia in DCM patients (unadjusted odds ratio: 1.76; 95% confidence interval: 1.01 to 3.08; p = 0.05) (Table 1). This association was stronger in multivariable regression analyses, adjusting for variables associated with baseline arrhythmia in this cohort (age, gender, indexed left atrial volume, and ventricular function). TTNtv independently predicted early arrhythmias in DCM (adjusted odds ratio: 2.90; 95% confidence interval: 1.48 to 5.78; p = 0.002) (Table 1).

ltivariable regression analyses, adjusting for variables associated with baseline arrhythmia in this cohort (age, gender, indexed left atrial volume, and ventricular function). TTNtv independently predicted early arrhythmias in DCM (adjusted odds ratio: 2.90; 95% confidence interval: 1.48 to 5.78; p = 0.002) (Table 1). Variants in LMNA are found in up to 4% of DCM cases and are strongly associated with an arrhythmic phenotype. In sensitivity analyses to control for potential LMNA effects, TTNtv remained predictive of arrhythmia after 12 patients with rare (Exome Aggregation Consortium frequency <0.001), protein-altering LMNA variants were excluded from analysis (adjusted odds ratio: 2.88; 95% confidence interval: 1.44 to 5.81; p = 0.003). Putative DCM variants in other genes were not evaluated due to the small number of affected individuals and no prior associations with arrhythmia. Our data demonstrate that TTNtv are associated with early arrhythmic risk in patients with DCM, independent of conventional arrhythmic risk factors. Although all patients were identified prospectively, baseline arrhythmia data were collected retrospectively and we have consolidated ventricular and atrial arrhythmias into 1 arrhythmia category, with a modest absolute increase in arrhythmic risk (13%). However, these findings have relevance for all DCM cases with TTNtv, representing ∼15% of all DCM. This study provides insights into the arrhythmic burden associated with TTNtv and highlights additional genetic tools for the stratification of high-risk DCM patients.

a modest absolute increase in arrhythmic risk (13%). However, these findings have relevance for all DCM cases with TTNtv, representing ∼15% of all DCM. This study provides insights into the arrhythmic burden associated with TTNtv and highlights additional genetic tools for the stratification of high-risk DCM patients. Please note: This project was funded by the Medical Research Council UK; the Rosetrees Foundation; the Jansons Foundation; the National Institute for Health Research Cardiovascular Biomedical Research Unit of Royal Brompton and Harefield National Health Service Foundation Trust and Imperial College London; a Health Innovation Challenge Fund (HICF-R6-373); and the Wellcome Trust and Department of Health, UK; and the British Heart Foundation. This publication includes independent research commissioned by the Health Innovation Challenge Fund (HICF-R6-373), a parallel funding partnership between the Department of Health and the Wellcome Trust. The views expressed in this work are those of the authors and not necessarily those of the Department of Health or the Wellcome Trust. Dr. Cook has served as a consultant for Illumina. Dr. Prasad has received honoraria for speaking from Bayer-Schering. The other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Cook and Prasad contributed equally to this work and are joint senior authors. Table 1 Results of Logistic Regression Analysis

Please note: This project was funded by the Medical Research Council UK; the Rosetrees Foundation; the Jansons Foundation; the National Institute for Health Research Cardiovascular Biomedical Research Unit of Royal Brompton and Harefield National Health Service Foundation Trust and Imperial College London; a Health Innovation Challenge Fund (HICF-R6-373); and the Wellcome Trust and Department of Health, UK; and the British Heart Foundation. This publication includes independent research commissioned by the Health Innovation Challenge Fund (HICF-R6-373), a parallel funding partnership between the Department of Health and the Wellcome Trust. The views expressed in this work are those of the authors and not necessarily those of the Department of Health or the Wellcome Trust. Dr. Cook has served as a consultant for Illumina. Dr. Prasad has received honoraria for speaking from Bayer-Schering. The other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Cook and Prasad contributed equally to this work and are joint senior authors. Table 1 Results of Logistic Regression Analysis Variable Unadjusted Analysis Adjusted Analysis OR p Value 95% Confidence Interval OR p Value 95% Confidence Interval Age (per 10 yrs) 1.53 <0.0001 1.33–1.76 1.60 <0.0001 1.36–1.89 Male 3.02 <0.0001 2.00–4.65 2.33 <0.0001 1.44–3.83 LVEF (per 10%) 0.76 <0.001 0.66–0.87 — — — RVEF (per 10%) 0.71 <0.0001 0.62–0.80 0.79 <0.0001 0.67–0.92 TTNtv positive 1.76 0.05 1.01–3.08 2.90 0.002 1.48–5.78 LAVi (per 1 ml/m2) 1.03 <0.0001 1.02–1.04 1.03 <0.0001 1.02–1.04 LGE present 1.32 0.13 0.92–1.89 — — — FHx DCM 0.59 0.06 0.34–1.00 — — — FHx SCD 0.62 0.09 0.35–1.06 — — — Results of logistic regression model of predictors of early arrhythmias in dilated cardiomyopathy (DCM). Variables with p < 0.10 from the univariable analysis were considered for inclusion in an optimized multivariable model, created using backward stepwise selection until only significant variables remained. Truncating variant in titin (TTNtv) was added to this optimized model.

rrhythmias in dilated cardiomyopathy (DCM). Variables with p < 0.10 from the univariable analysis were considered for inclusion in an optimized multivariable model, created using backward stepwise selection until only significant variables remained. Truncating variant in titin (TTNtv) was added to this optimized model. FHx = family history; LAVi = indexed left atrial volume; LGE = midwall fibrosis late gadolinium enhancement; LVEF = left ventricular ejection fraction; RVEF = right ventricular ejection fraction; SCD = sudden cardiac death.

Cardiovascular diseases (CVDs) are a leading cause of death in the world and a major barrier to sustainable human development (1). In 2011, the United Nations formally recognized noncommunicable diseases, including CVDs, as a major concern for global health and set out an ambitious plan to dramatically reduce the effect of these diseases in all regions (2). An increased awareness of these global noncommunicable disease goals has expanded attempts to track and benchmark national efforts at reducing CVD and other noncommunicable diseases 3, 4. The third Sustainable Development Goal recognized the importance of CVD by targeting a one-third reduction in premature mortality due to noncommunicable diseases (5). Countries that take the SDG goals seriously will have to contend with a wide range of barriers limiting their ability to improve health care and reduce CVD risks. In many regions of the world, the relative position of CVD as a health problem remains unclear or is based on limited data. Many low- and middle-income countries have implemented health examination surveys that have improved measurement of CVD and its associated risk factors (6).

prove health care and reduce CVD risks. In many regions of the world, the relative position of CVD as a health problem remains unclear or is based on limited data. Many low- and middle-income countries have implemented health examination surveys that have improved measurement of CVD and its associated risk factors (6). Systematic evaluation of data collected in death certificates, verbal autopsy, health surveys, prospective cohort studies, health system administrative data, and disease registries is needed to appropriately guide efforts to reduce the health burden of CVD. The GBD (Global Burden of Disease) study is an effort to continuously improve our understanding of the burden of diseases by integrating the available data on disease incidence, prevalence, and mortality to produce consistent, transparent, and up-to-date global, regional, and national estimates (7). The global number of CVD deaths and regional patterns of total CVD mortality were previously reported from the GBD 2013 study (8). The GBD 2015 study results provide a completely new mortality time-series estimated from 1990 forward and updated through 2015. We now also report national estimates of mortality to provide results relevant to specific countries at the level of each underlying CVD condition. In addition, the study addresses the nonfatal burden of CVD by reporting global, regional, and national estimates of prevalence, years lived with disability, and disability-adjusted life years.

tional estimates of mortality to provide results relevant to specific countries at the level of each underlying CVD condition. In addition, the study addresses the nonfatal burden of CVD by reporting global, regional, and national estimates of prevalence, years lived with disability, and disability-adjusted life years. Methods GBD estimation framework The Global Burden of Diseases, Injuries, and Risk Factors 2015 study is a multinational collaborative research project with the goal of producing consistent estimates of health loss due to over 310 diseases and injuries. A wide range of data sources and methods were used to produce age-, sex-, and country-specific results for the years 1990 to 2015. Results are updated annually for the entire time series, and these results supersede previous versions of the GBD study. Methods have been reported in detail previously and are summarized here and in the Online Appendix 9, 10, 11.

were used to produce age-, sex-, and country-specific results for the years 1990 to 2015. Results are updated annually for the entire time series, and these results supersede previous versions of the GBD study. Methods have been reported in detail previously and are summarized here and in the Online Appendix 9, 10, 11. Changes since the GBD 2013 study There have been numerous changes to data collection and methods in the current study to update and improve upon the results of the GBD 2013 study. Mortality data has been updated through 2014 using newly-identified data sources released or collected since GBD 2013. Because of their available vital registration data, 7 territories have been added: American Samoa, Bermuda, Greenland, Guam, the Northern Mariana Islands, Puerto Rico, and the Virgin Islands. Disease burden from these territories are now included in the national totals for the United States, United Kingdom, and Denmark. A new approach to estimating uncertainty for countries with long (>95% complete) time series of vital registration data has been used uniformly for all CVD causes so that results are not affected by uncertainty in regions with less complete vital registration. Models of disease incidence and prevalence now uniformly include estimates of excess mortality and, for stroke, cause-specific mortality, so that they are better informed by the available mortality data. For each incidence or prevalence data point, we matched the age-sex-location-year-cause–specific mortality rate to produce a ratio conceptually equivalent to an excess mortality rate. Because of implausibly rapid increases in deaths reported due to atrial fibrillation, we have developed a unified model of atrial fibrillation that makes use of prevalence, case fatality, and mortality data to estimate both the nonfatal and fatal burden due to this condition.

lly equivalent to an excess mortality rate. Because of implausibly rapid increases in deaths reported due to atrial fibrillation, we have developed a unified model of atrial fibrillation that makes use of prevalence, case fatality, and mortality data to estimate both the nonfatal and fatal burden due to this condition. Defining disease categories CVD was estimated overall and separately for the 10 most common global causes of CVD-related death. These causes were ischemic heart disease (IHD), ischemic stroke, hemorrhagic and other stroke, atrial fibrillation, peripheral arterial disease (PAD), aortic aneurysm, cardiomyopathy and myocarditis, hypertensive heart disease, endocarditis, rheumatic heart disease (RHD), and a category for other CVD conditions. The GBD cause list is a hierarchical, mutually-exclusive, and collectively exhaustive list of causes of death. The 3 level 1 GBD causes consist of communicable, maternal, neonatal, and nutritional disorders; noncommunicable diseases; and injuries. Level 2 causes consist of 21 cause groups, such as neoplasms and CVD. Levels 3 and 4 consist of disaggregated subcauses (Online Methods Appendix Table 1).

ive list of causes of death. The 3 level 1 GBD causes consist of communicable, maternal, neonatal, and nutritional disorders; noncommunicable diseases; and injuries. Level 2 causes consist of 21 cause groups, such as neoplasms and CVD. Levels 3 and 4 consist of disaggregated subcauses (Online Methods Appendix Table 1). Cause of death was defined by international standards governing the reporting of death certificates, in which a single underlying cause is assigned by a physician. For example, IHD was defined as an underlying cause of death across International Classification of Diseases (ICD) revisions (most recently ICD-10 I20 to I25, ICD-9 410 to 414) (12). The leading causes included as “other cardiovascular and circulatory diseases” were nonrheumatic valvular disorders and pulmonary embolism. A proportion of deaths that were assigned on death certificates to nonfatal, undefined, or intermediate causes (e.g., cardiac arrest, heart failure, or hypertension) were redistributed using statistical regression methods or fixed proportions (9). Redistribution of deaths coded to heart failure was accomplished using a regression model that accounted for the variable use of these codes by age, sex, and location. This approach improves upon methods that either exclude deaths coded to an intermediate cause or reassign them using a fixed proportion that ignores variation by age, sex, or location. Deaths due to unspecified types of stroke (ICD-10 I64) were distributed using the ratio of ischemic to hemorrhagic stroke deaths in a country’s region or, for South Asia, the global ratio, stratified by age. A Bayesian noise reduction algorithm was applied to death data to improve estimation of the underlying mortality rate (see the Online Appendix for details). This noise reduction algorithm was adopted to improve upon prior methods in which 0 counts were excluded, an approach that leads to an upward bias in estimates. Verbal autopsy, a method in which a standardized interview collects information from household members on symptoms preceding death, was included as a data input only for total CVD, ischemic heart disease, and stroke deaths, and was excluded for other CVD causes of death.

roach that leads to an upward bias in estimates. Verbal autopsy, a method in which a standardized interview collects information from household members on symptoms preceding death, was included as a data input only for total CVD, ischemic heart disease, and stroke deaths, and was excluded for other CVD causes of death. Disease prevalence was estimated at a more granular level of specific disease sequelae, using input data from systematic reviews of the published scientific reports, unpublished registry data, and health system administrative data. A regression equation was used to adjust data in the direction of the gold-standard case definition. Detailed nonfatal modeling methods are included in the Online Appendix. IHD was the summation of 4 distinct disease sequelae: acute myocardial infarction, chronic stable angina, chronic IHD, and heart failure due to IHD. Myocardial infarction was defined according to the Third Universal Definition of Myocardial Infarction and the case-finding approach from the MONICA (Multinational MONItoring of trends and determinants in CArdiovascular disease) studies, which accounts for out-of-hospital sudden cardiac death 13, 14. Adjustments were made for the advent of troponin-testing technology for diagnosis of acute coronary syndromes during the years covered by the study using meta-analysis of its increased sensitivity (compared with prior markers) to adjust pre-2000 incidence rates upward by 56%. Stable angina was defined according to the Rose Angina Questionnaire, which was adjusted to account for the observed differences in survey and administrative data found in the United States. Cerebrovascular disease relied on a case definition developed by the World Health Organization and was estimated separately for 2 subcategories: 1) ischemic stroke; and 2) hemorrhagic or other nonischemic stroke (15). Stroke data was adjusted to match our case definition of subtype-specific first-ever incident events, and was used to separately estimate acute and chronic stroke. PAD was defined by an ankle brachial index (ABI) <0.9, and symptomatic PAD was defined as self-report of claudicatory symptoms among those with ABI <0.9 (16). Atrial fibrillation was defined by electrocardiogram and included atrial flutter. The prevalence of symptomatic heart failure was estimated using both health system administrative and population-based registry data, and was then attributed to specific underlying heart failure etiologies (some of which were not CVD).

). Atrial fibrillation was defined by electrocardiogram and included atrial flutter. The prevalence of symptomatic heart failure was estimated using both health system administrative and population-based registry data, and was then attributed to specific underlying heart failure etiologies (some of which were not CVD). Hypertensive heart disease was defined as symptomatic heart failure due to the direct and long-term effects of hypertension, with its nonfatal burden derived from the model of heart failure. Cardiomyopathy was defined as symptomatic heart failure due to primary myocardial disease or toxic exposures, such as alcohol, with its nonfatal burden derived from the model of heart failure (17). Acute myocarditis was estimated as an acute and time-limited condition due to myocardial inflammation using health system administrative data. Endocarditis and RHD were defined by their clinical diagnosis. Estimates of RHD include cases identified by clinical history and physical examination, including auscultation or standard echocardiographic criteria for definite disease.

ted condition due to myocardial inflammation using health system administrative data. Endocarditis and RHD were defined by their clinical diagnosis. Estimates of RHD include cases identified by clinical history and physical examination, including auscultation or standard echocardiographic criteria for definite disease. Data sources and analytic methods A map of data availability for each country are included in Online Figures 1A and 1B (9). The GBD 2015 study used country-level surveillance data, verbal autopsy, vital registration, published and unpublished disease registries, and published scientific reports. Table 1 summarizes data sources used to estimate CVD burden. Table 1 also shows the data representativeness index for nonfatal estimates, which is the proportion of age-sex-location strata with available data for nonfatal modeling shown by cause and over time. Online Methods Appendix Tables 2 and 3 are tables of all data sources. Data sources for models are also available online from the Global Health Data Exchange (18). National income, metabolic and nutritional risk factors, and other country-level covariates were estimated from surveys and published systematic reviews. Analysis of mortality used Cause of Death Ensemble modeling (CODEm), an approach that incorporates country-level covariates, including age-sex-country-year–specific estimates of CVD risk factors, national income, and other causal factors (Online Appendix). CODEm borrows strength across space, time, and age groups using a variety of geospatial model types, and weighs the results using tests of out-of-sample predictive validity. Analysis of disease prevalence used epidemiological state-transition–based disease modeling software, DisMod-MR, which accounts for study-level differences in measurement method (9). Disease-specific incidence, prevalence, case fatality, and mortality rates were integrated to produce consistent estimates of prevalence of all geographies in the study (19). Estimates were considered significantly different if there was no overlap in their 95% uncertainty intervals (UIs). The cause-specific mortality rate for atrial fibrillation was also estimated using DisMod because of implausible increases in the rate when derived only from death certificates. Prevalent cases of each disease’s sequelae are assigned specific levels of severity based on the U.S.

n their 95% uncertainty intervals (UIs). The cause-specific mortality rate for atrial fibrillation was also estimated using DisMod because of implausible increases in the rate when derived only from death certificates. Prevalent cases of each disease’s sequelae are assigned specific levels of severity based on the U.S. Medical Expenditure Panel Survey 2000 to 2011, a population-based survey with data on functional health that also provides linkage to respondent medical records. DisMod-MR models were run separately by sex, country, and year.Table 1 Data Representativeness in GBD 2015 Fatal and Nonfatal Modeling by CVD Cause Cause Number of Site-Years of Mortality Data Percentage of Geographies With Data for Nonfatal Modeling Vital Registration Verbal Autopsy Before 2005 2005–2015 Total Cardiovascular diseases 10,446 964 81 74 85 Rheumatic heart disease 10,417 0 29 27 37 Ischemic heart disease 10,652 734 47 30 51 Cerebrovascular disease 10,660 692 64 67 74 Ischemic stroke 9,207 0 63 61 68 Hemorrhagic stroke 9,211 0 63 61 68 Hypertensive heart disease 10,039 0 13 6 16 Cardiomyopathy and myocarditis 10,020 0 25 22 31 Atrial fibrillation and flutter 8,104 0 22 24 27 Aortic aneurysm∗ 9,215 0 N/A N/A N/A Peripheral vascular disease 8,087 0 19 20 23 Endocarditis 9,274 0 18 19 21 Other cardiovascular and circulatory diseases 10,340 0 1 1 1 CVD = cardiovascular disease; GBD = Global Burden of Disease; N/A = not available. ∗ Nonfatal estimates are not produced for aortic aneurysm.

Cause Number of Site-Years of Mortality Data Percentage of Geographies With Data for Nonfatal Modeling Vital Registration Verbal Autopsy Before 2005 2005–2015 Total Cardiovascular diseases 10,446 964 81 74 85 Rheumatic heart disease 10,417 0 29 27 37 Ischemic heart disease 10,652 734 47 30 51 Cerebrovascular disease 10,660 692 64 67 74 Ischemic stroke 9,207 0 63 61 68 Hemorrhagic stroke 9,211 0 63 61 68 Hypertensive heart disease 10,039 0 13 6 16 Cardiomyopathy and myocarditis 10,020 0 25 22 31 Atrial fibrillation and flutter 8,104 0 22 24 27 Aortic aneurysm∗ 9,215 0 N/A N/A N/A Peripheral vascular disease 8,087 0 19 20 23 Endocarditis 9,274 0 18 19 21 Other cardiovascular and circulatory diseases 10,340 0 1 1 1 CVD = cardiovascular disease; GBD = Global Burden of Disease; N/A = not available. ∗ Nonfatal estimates are not produced for aortic aneurysm. Disability-adjusted life years Disability-adjusted life-years (DALYs) combine information regarding premature death (years of life lost [YLL]) and disability caused by the condition (years lived with disability [YLD]) to provide a summary measure of health lost due to that condition. YLL was calculated by multiplying observed deaths for a specific age in the year of interest by the age-specific reference life expectancy estimated using life table methods. The normative standard life expectancy at birth is 86.59 years, based on the lowest observed death rates for each 5-year age group in populations larger than 5 million. YLD was calculated by multiplying disease prevalence (in number of cases) by a health-state–specific disability weight representing a degree of lost functional capacity. A detailed explanation of the process of disability weight estimation has been reported separately 10, 11. Briefly, disability weights were developed using household surveys in multiple countries that asked members of the general public to choose between lay descriptions of health states 20, 21. Adjustment was made for comorbidity by simulating 40,000 individuals in each age-sex-country-year stratum exposed to the independent probability of acquiring each condition based on disease prevalence.

The relationship between SDI and the age-standardized CVD death rate at the global level is shown in Figure 3. As SDI increases beyond 0.25, the highest CVD mortality rates shift from women to men. Cerebrovascular mortality rates begin to decline among women above an SDI of 0.3, although still increasing for men. IHD and cerebrovascular mortality rates continue to increase with greater SDI among men, peaking at an SDI of 0.5 to 0.75 (compared with 0.25 among women). CVD mortality decreases sharply for both sexes in countries with an SDI >0.75.Figure 3 Relationship Between Age-Standardized Mortality Rate, CVD Cause, and SDI, by Sex This figure displays the distribution of age-standardized death rate by causes of CVD mortality by SDI. Abbreviations as in Figure 1. Prevalence and mortality for CVD by cause groups Ischemic heart disease In 2015, IHD was the leading cause of all health loss globally, as well as in each world region (Online Figure 3). There were an estimated 7.29 million acute myocardial infarctions (95% UI: 6.80 to 7.81 million acute myocardial infarctions) and 110.55 million prevalent cases of IHD (95% UI: 100.68 to 121.80 million cases) in 2015. Prevalent cases of IHD began accounting for a large proportion of prevalent cases of CVD after 40 years of age, and the prevalence rose steeply with older age categories (Online Figure 4A). There were an estimated 10.88 million prevalent cases of IHD (95% UI: 8.82 to 13.25 million cases) among persons 50 to 54 years of age, which is more than 3-fold the number of cases for persons 40 to 44 years of age. The IHD prevalence rose from an estimated 290 cases per 100,000 (95% UI: 255 to 328 cases per 100,000) for those 40 to 44 years of age to 11,203 cases per 100,000 (95% UI: 9,610 to 13,178 cases per 100,000) for those 75 to 79 years of age, declining slightly for those 80 years of age and over to a rate of 9,700 cases per 100,000 (95% UI: 8,773 to 10,738 cases per 100,000).

ultiple countries that asked members of the general public to choose between lay descriptions of health states 20, 21. Adjustment was made for comorbidity by simulating 40,000 individuals in each age-sex-country-year stratum exposed to the independent probability of acquiring each condition based on disease prevalence. The 95% UIs reported for each estimate used 1,000 samples from the posterior distribution from the respective step in the modeling process, reported as the 2.5th and 97.5th values of the distribution. Age standardization was via the direct method, applying a global age structure. Sociodemographic index Instead of using the categories of national socioeconomic status developed for the GBD 2013 study, we have produced a new continuous measure of sociodemographic status. The sociodemographic index (SDI) was estimated to examine changes in CVD burden as a function of the global epidemiological transition (9). Similar to the method used to compute the human development index, SDI was calculated for each country or territory in each year from 1990 to 2015. SDI was the equally-weighted geometric mean of income per capita, educational attainment, and total fertility rate. Least squares regression of death rates on SDI was used with a smoothing spline and dummy variables for outlier regions that skewed fit to capture the average relationship for each age-sex-cause group.

5. SDI was the equally-weighted geometric mean of income per capita, educational attainment, and total fertility rate. Least squares regression of death rates on SDI was used with a smoothing spline and dummy variables for outlier regions that skewed fit to capture the average relationship for each age-sex-cause group. Results All results of the GBD 2015 study, including prevalence, mortality, YLL, YLD, and DALYs, for all country-years are available for download from the GBD results tool webpage (22) and can be explored visually at the GBD Compare visualization hub (23). Regional variation in CVD Prevalence Globally, there were an estimated 422.7 million prevalent cases of CVD (95% UI: 415.53 to 427.87 million cases) in 2015 (Table 2). The age-standardized prevalence of CVD varied significantly by country (Figure 1). Countries with the lowest age-standardized prevalence in 2015, all with <5,000 cases per 100,000 individuals, included Singapore, Japan, South Korea, Chile, Argentina, Uruguay, Canada, Australia, New Zealand, Ireland, Cyprus, Malta, Italy, Greece, and Israel. Countries in Western Europe, as well as the United States, the United Arab Emirates, and Nepal, all had only slightly higher prevalence. Countries with the highest age-standardized prevalence in 2015, all >9,000 cases per 100,000 persons, included most countries in West Africa, Morocco, Iran, Oman, Zambia, Mozambique, and Madagascar.Figure 1 Global Map, Age-Standardized Prevalence of CVD in 2015

rates, and Nepal, all had only slightly higher prevalence. Countries with the highest age-standardized prevalence in 2015, all >9,000 cases per 100,000 persons, included most countries in West Africa, Morocco, Iran, Oman, Zambia, Mozambique, and Madagascar.Figure 1 Global Map, Age-Standardized Prevalence of CVD in 2015 Choropleth map showing the estimated age-standardized prevalence of total CVD in 2015 for each country. ATG = Antigua and Barbuda; BRB = Barbados; COM = Comoros; CVD = cardiovascular diseases; DMA = Dominica; E Med = Eastern Mediterranean; FJI = Fiji; FSM = Federated States of Micronesia; GRD = Grenada; KIR = Kiribati; KS = Kaposi sarcoma; LCA = Saint Lucia; MDV = Maldives; MHL = Marshall Islands; MLT = Malta; MUS = Mauritius; NMSC = nonmelanoma skin cancer; SGP = Singapore; SLB = Solomon Islands; SYC = Seychelles; TLS = Timor-Leste; TON = Tonga; TTO = Trinidad and Tobago; VCT = Saint Vincent and the Grenadines; VUT = Vanuatu; W Africa = West Africa; WSM = Samoa. Table 2 Global and Regional All-Age Deaths and Age-Standardized Death Rates in 2015, by Sex, for Selected Causes of CVD Mortality∗

Choropleth map showing the estimated age-standardized prevalence of total CVD in 2015 for each country. ATG = Antigua and Barbuda; BRB = Barbados; COM = Comoros; CVD = cardiovascular diseases; DMA = Dominica; E Med = Eastern Mediterranean; FJI = Fiji; FSM = Federated States of Micronesia; GRD = Grenada; KIR = Kiribati; KS = Kaposi sarcoma; LCA = Saint Lucia; MDV = Maldives; MHL = Marshall Islands; MLT = Malta; MUS = Mauritius; NMSC = nonmelanoma skin cancer; SGP = Singapore; SLB = Solomon Islands; SYC = Seychelles; TLS = Timor-Leste; TON = Tonga; TTO = Trinidad and Tobago; VCT = Saint Vincent and the Grenadines; VUT = Vanuatu; W Africa = West Africa; WSM = Samoa. Table 2 Global and Regional All-Age Deaths and Age-Standardized Death Rates in 2015, by Sex, for Selected Causes of CVD Mortality∗ Death All Ages Age-Standardized (per 100,000) Total Female Male Total Female Male Cardiovascular diseases Global 17,921,047 (17,590,537–18,276,848) 8,501,409 (8,301,355–8,722,665) 9,419,637 (9,199,720–9,648,088) 286 (280–291) 242 (236–248) 335 (327–342) Andean Latin America 63,861 (59,748–68,356) 32,495 (29,222–35,927) 31,366 (28,782–34,204) 157 (146–168) 144 (130–160) 170 (156–185) Australasia 67,481 (65,263–69,507) 34,620 (33,129–36,147) 32,861 (31,754–34,076) 147 (143–151) 127 (122–132) 168 (163–175) Caribbean 126,769 (121,035–132,439) 65,947 (61,476–70,829) 60,822 (57,446–64,113) 293 (280–306) 274 (254–294) 314 (298–331) Central Asia 304,212 (296,495–311,855) 148,071 (142,407–153,246) 156,141 (151,207–161,193) 545 (532–558) 451 (433–466) 674 (654–693) Central Europe 666,173 (654,844–676,711) 355,129 (347,583–362,138) 311,044 (305,556–317,045) 338 (333–344) 278 (272–283) 419 (411–427) Central Latin America 337,507 (328,984–345,456) 167,760 (162,327–173,252) 169,747 (164,317–175,158) 198 (193–203) 176 (171–182) 223 (216–229) Central sub-Saharan Africa 147,629 (100,125–205,190) 85,290 (54,094–123,452) 62,339 (40,084–92,009) 418 (291–560) 455 (298–633) 366 (244–519) East Asia 3,953,300 (3,805,196–4,117,647) 1,651,066 (1,566,414–1,740,137) 2,302,234 (2,183,097–2,432,661) 295 (284–307) 237 (225–249) 359 (341–377) Eastern Europe 1,774,861 (1,740,489–1,811,091) 993,829 (969,164–1,020,806) 781,033 (761,821–801,141) 532 (522–543) 423 (413–435) 701 (685–718) Eastern sub-Saharan Africa 424,364 (353,978–507,026) 218,704 (169,208–277,600) 205,661 (163,706–260,525) 349 (295–414) 346 (272–433) 352 (285–437) High-income Asia Pacific 498,622 (485,719–511,659) 270,969 (261,880–280,189) 227,653 (222,117–233,411) 112 (110–115) 93 (90–96) 135 (131–138) High-income North America 946,416 (924,685–967,818) 474,764 (460,316–489,759) 471,652 (461,763–481,077) 171 (168–175) 143 (139–147) 204 (200–208) North Africa and Middle East 1,079,493 (1,028,619–1,134,703) 508,366 (475,397–543,603) 571,127 (537,220–607,652) 361 (344–376) 326 (306–347) 398 (376–421) Oceania 27,503 (20,884–36,700) 13,649 (10,261–18,281) 13,854 (10,540–18,619) 525 (416–664) 506 (392–648) 540 (432–677) South Asia 3,61

5) 143 (139–147) 204 (200–208) North Africa and Middle East 1,079,493 (1,028,619–1,134,703) 508,366 (475,397–543,603) 571,127 (537,220–607,652) 361 (344–376) 326 (306–347) 398 (376–421) Oceania 27,503 (20,884–36,700) 13,649 (10,261–18,281) 13,854 (10,540–18,619) 525 (416–664) 506 (392–648) 540 (432–677) South Asia 3,61 0,666 (3,473,581–3,755,833) 1,509,355 (1,420,049–1,599,458) 2,101,312 (1,993,733–2,221,549) 369 (355–383) 314 (296–332) 424 (404–447) Southeast Asia 1,351,557 (1,238,336–1,455,239) 632,078 (565,558–694,348) 719,479 (633,386–798,803) 321 (296–344) 274 (247–300) 377 (335–411) Southern Latin America 164,667 (160,162–169,048) 87,224 (83,804–90,842) 77,443 (74,888–79,792) 218 (212–224) 178 (172–186) 269 (260–277) Southern sub-Saharan Africa 136,002 (123,737–150,420) 78,333 (69,335–88,846) 57,669 (51,859–64,908) 338 (309–372) 321 (285–363) 349 (317–387) Tropical Latin America 435,272 (418,494–455,826) 205,462 (195,338–218,739) 229,811 (218,571–242,203) 256 (247–269) 211 (201–225) 316 (301–332) Western Europe 1,483,792 (1,444,804–1,521,399) 798,509 (772,187–825,482) 685,283 (668,957–701,669) 157 (154–161) 132 (128–135) 187 (183–192) Western sub-Saharan Africa 320,897 (274,658–384,354) 169,791 (136,039–224,498) 151,106 (125,014–187,897) 285 (247–335) 298 (244–386) 266 (226–324) Ischemic heart disease Global 8,916,964 (8,751,617–9,108,850) 4,035,936 (3,941,319–4,146,339) 4,881,028 (4,747,381–5,022,975) 142 (140–145) 115 (112–118) 173 (168–178) Andean Latin America 34,041 (31,629–36,640) 16,786 (14,973–18,783) 17,255 (15,646–18,940) 84 (78–91) 75 (67–84) 94 (86–103) Australasia 38,507 (36,922–40,107) 18,382 (17,258–19,556) 20,125 (19,203–21,078) 84 (81–88) 67 (63–71) 103 (99–108) Caribbean 65,422 (62,394–68,337) 31,991 (29,892–34,192) 33,431 (31,703–35,282) 151 (144–158) 132 (123–142) 173 (164–182) Central Asia 185,521 (179,788–191,429) 88,069 (84,158–91,666) 97,452 (93,405–101,577) 336 (326–347) 271 (259–282) 425 (409–442) Central Europe 357,073 (350,311–364,381) 182,915 (177,942–187,778) 174,158 (170,286–178,258) 181 (177–184) 141 (138–145) 234 (229–240) Central Latin America 202,329 (196,619–207,546) 95,669 (92,132–99,212) 106,660 (103,143–110,376) 119 (116–122) 101 (97–105) 140 (135–144) Central sub-Saharan Africa 47,589 (31,139–67,216) 24,686 (15,044–36,067) 22,902 (14,369–35,115) 143 (94–196) 143 (89–205) 139 (91–204) East Asia 1,507,596 (1,443,996–1,579,379) 636,714 (599,802–674,258) 870,883 (817,224–928,402) 114 (109–119) 92 (87–98) 137 (129–145) Eastern Europe 1,

16–122) 101 (97–105) 140 (135–144) Central sub-Saharan Africa 47,589 (31,139–67,216) 24,686 (15,044–36,067) 22,902 (14,369–35,115) 143 (94–196) 143 (89–205) 139 (91–204) East Asia 1,507,596 (1,443,996–1,579,379) 636,714 (599,802–674,258) 870,883 (817,224–928,402) 114 (109–119) 92 (87–98) 137 (129–145) Eastern Europe 1, 093,600 (1,070,126–1,117,719) 599,344 (580,595–618,307) 494,256 (480,975–507,720) 326 (319–333) 252 (245–260) 445 (433–456) Eastern sub-Saharan Africa 143,019 (113,946–175,641) 64,860 (46,988–86,834) 78,159 (60,498–102,024) 122 (98–147) 108 (79–142) 137 (109–174) High-income Asia Pacific 197,492 (190,870–203,632) 103,780 (98,289–108,644) 93,712 (90,345–97,136) 45 (44–46) 35 (34–37) 56 (54–58) High-income North America 583,761 (565,503–600,239) 276,513 (265,351–286,639) 307,247 (297,884–315,053) 106 (102–108) 83 (79–86) 133 (129–136) North Africa and Middle East 599,360 (565,847–631,996) 257,621 (239,786–276,947) 341,738 (319,275–366,768) 201 (190–210) 168 (157–179) 236 (222–252) Oceania 12,707 (9,738–16,864) 5,309 (3,976–7,101) 7,399 (5,637–9,955) 240 (192–304) 205 (160–262) 275 (219–345) South Asia 2,073,496 (1,985,218–2,169,575) 805,483 (754,450–868,996) 1,268,013 (1,197,852–1,346,206) 212 (204–221) 171 (160–184) 254 (240–268) Southeast Asia 558,700 (510,069–601,308) 236,125 (213,084–261,064) 322,575 (284,187–356,407) 131 (121–140) 103 (93–113) 166 (148–182) Southern Latin America 83,253 (80,419–86,198) 41,921 (39,779–44,220) 41,332 (39,592–42,881) 110 (107–114) 85 (81–89) 143 (137–148) Southern sub-Saharan Africa 56,521 (51,368–62,508) 31,275 (27,621–35,669) 25,246 (22,593–28,281) 142 (130–157) 130 (115–148) 154 (140–171) Tropical Latin America 201,510 (193,188–211,658) 90,837 (85,863–97,235) 110,673 (103,939–117,172) 118 (113–124) 94 (89–100) 149 (140–158) Western Europe 745,878 (722,801–767,056) 365,207 (350,185–380,330) 380,670 (369,601–391,158) 80 (78–82) 60 (58–62) 105 (102–108) Western sub-Saharan Africa 129,590 (111,345–153,647) 62,449 (50,092–82,307) 67,141 (54,806–84,115) 123 (107–144) 119 (98–151) 127 (106–155) Ischemic strokes Global 2,977,980 (2,880,779–3,068,756) 1,550,557 (1,477,734–1,619,514) 1,427,423 (1,369,627–1,484,115) 49 (47–50) 44 (42–46) 54 (52–56) Andean Latin America 9,701 (8,824–10,726) 5,215 (4,501–5,941) 4,485 (3,970–5,121) 24 (22–27) 23 (20–27) 25 (23–29) Australasia 8,726 (8,048–9,498) 5,110 (4,560–5,752) 3,616 (3,264–3,997) 18 (17–20) 18 (16–20) 18 (16–20) Caribbean 20,504 (19,174–22,125) 11,679 (10,509–13,181) 8,824 (8,159–9,61

–50) 44 (42–46) 54 (52–56) Andean Latin America 9,701 (8,824–10,726) 5,215 (4,501–5,941) 4,485 (3,970–5,121) 24 (22–27) 23 (20–27) 25 (23–29) Australasia 8,726 (8,048–9,498) 5,110 (4,560–5,752) 3,616 (3,264–3,997) 18 (17–20) 18 (16–20) 18 (16–20) Caribbean 20,504 (19,174–22,125) 11,679 (10,509–13,181) 8,824 (8,159–9,61 0) 48 (45–51) 48 (43–54) 47 (43–51) Central Asia 39,172 (37,193–41,353) 20,028 (18,461–21,835) 19,144 (17,907–20,352) 73 (69–77) 62 (57–68) 89 (83–95) Central Europe 125,872 (122,372–129,393) 74,220 (71,301–77,020) 51,652 (49,753–53,800) 63 (61–64) 57 (55–59) 70 (67–73) Central Latin America 37,869 (36,541–39,267) 20,762 (19,776–21,833) 17,107 (16,369–17,959) 23 (22–24) 22 (21–23) 24 (23–26) Central sub-Saharan Africa 22,654 (14,109–32,102) 14,551 (8,419–21,794) 8,103 (4,927–12,473) 77 (49–108) 89 (53–133) 60 (37–89) East Asia 785,226 (703,812–827,827) 330,404 (271,544–357,625) 454,821 (418,276–485,389) 61 (54–64) 48 (40–52) 74 (68–79) Eastern Europe 385,151 (372,405–399,825) 245,598 (234,439–257,431) 139,553 (132,831–146,054) 113 (109–117) 102 (97–107) 131 (124–137) Eastern sub-Saharan Africa 60,759 (46,770–76,049) 34,978 (22,707–49,063) 25,781 (19,130–34,120) 57 (43–70) 61 (40–83) 52 (39–67) High-income Asia Pacific 132,454 (127,229–137,646) 74,185 (70,362–77,931) 58,269 (55,682–60,833) 28 (27–29) 24 (22–25) 33 (31–34) High-income North America 123,894 (118,687–129,215) 73,295 (69,314–77,624) 50,599 (48,173–53,465) 21 (21–22) 21 (20–22) 22 (21–23) North Africa and Middle East 149,264 (136,741–159,778) 80,022 (71,508–88,363) 69,242 (62,597–75,881) 55 (51–59) 55 (49–60) 56 (51–61) Oceania 3,364 (2,427–4,652) 1,843 (1,269–2,891) 1,521 (1,088–2,178) 77 (58–103) 78 (55–118) 76 (57–102) South Asia 500,203 (457,812–552,667) 229,925 (198,486–274,331) 270,278 (245,928–297,888) 57 (53–63) 52 (45–62) 63 (57–69) Southeast Asia 215,754 (192,311–239,386) 116,663 (100,381–133,400) 99,091 (83,978–113,695) 58 (52–64) 55 (47–62) 62 (53–70) Southern Latin America 23,765 (22,597–25,049) 13,896 (12,952–14,931) 9,869 (9,256–10,486) 31 (29–32) 27 (25–29) 35 (33–38) Southern sub-Saharan Africa 21,684 (19,588–23,974) 14,081 (12,432–15,964) 7,603 (6,822–8,508) 58 (53–65) 60 (53–68) 54 (48–60) Tropical Latin America 48,553 (46,004–53,803) 23,527 (21,698–27,136) 25,026 (23,488–26,988) 31 (29–34) 25 (23–29) 39 (37–43) Western Europe 208,929 (199,938–219,001) 126,842 (119,343–134,411) 82,087 (78,376–86,524) 21 (20–22) 20 (19–21) 22 (21–23) Western sub-Saharan Africa 54,484 (45,545–66,759)

60 (53–68) 54 (48–60) Tropical Latin America 48,553 (46,004–53,803) 23,527 (21,698–27,136) 25,026 (23,488–26,988) 31 (29–34) 25 (23–29) 39 (37–43) Western Europe 208,929 (199,938–219,001) 126,842 (119,343–134,411) 82,087 (78,376–86,524) 21 (20–22) 20 (19–21) 22 (21–23) Western sub-Saharan Africa 54,484 (45,545–66,759) 33,733 (26,361–44,685) 20,752 (16,778–26,538) 60 (51–72) 70 (56–91) 47 (39–58) Hemorrhagic and other strokes Global 3,348,155 (3,240,908–3,500,094) 1,544,379 (1,472,186–1,644,306) 1,803,776 (1,728,916–1,888,569) 52 (50–54) 44 (42–47) 62 (59–65) Andean Latin America 9,090 (8,336–10,212) 4,756 (4,169–5,506) 4,334 (3,827–4,957) 21 (19–24) 21 (18–24) 22 (19–25) Australasia 9,007 (8,338–9,723) 5,345 (4,818–5,960) 3,661 (3,329–4,043) 20 (18–21) 20 (18–22) 19 (17–21) Caribbean 19,121 (17,586–20,601) 10,401 (9,105–11,825) 8,720 (7,952–9,529) 44 (41–47) 44 (38–50) 44 (40–48) Central Asia 46,416 (44,225–49,783) 24,013 (22,074–27,101) 22,403 (21,174–23,800) 80 (76–86) 71 (66–80) 92 (86–99) Central Europe 75,563 (73,070–78,708) 40,535 (38,502–42,524) 35,028 (33,720–37,208) 40 (38–41) 34 (32–36) 47 (45–50) Central Latin America 47,484 (45,863–49,166) 24,888 (23,802–26,120) 22,596 (21,627–23,654) 27 (26–28) 26 (24–27) 28 (27–29) Central sub-Saharan Africa 38,990 (25,248–55,651) 22,364 (12,823–33,985) 16,626 (10,509–25,575) 99 (62–139) 107 (59–159) 89 (57–135) East Asia 1,168,477 (1,109,258–1,263,491) 457,939 (424,166–529,581) 710,537 (662,234–763,372) 83 (79–91) 63 (59–74) 105 (98–112) Eastern Europe 135,182 (127,552–143,922) 74,236 (68,208–80,933) 60,946 (56,905–65,141) 41 (39–44) 34 (31–37) 53 (49–56) Eastern sub-Saharan Africa 93,683 (73,892–115,030) 49,175 (33,474–65,910) 44,509 (33,637–57,836) 69 (54–85) 69 (45–92) 69 (53–89) High-income Asia Pacific 84,648 (81,230–88,822) 42,987 (40,471–45,896) 41,661 (39,815–43,536) 21 (20–22) 17 (16–18) 26 (25–27) High-income North America 58,576 (56,053–61,567) 31,967 (30,125–34,035) 26,609 (25,175–28,227) 11 (11–12) 11 (10–11) 12 (11–12) North Africa and Middle East 153,741 (141,926–171,810) 77,195 (68,269–88,362) 76,546 (70,712–87,959) 48 (44–53) 46 (41–53) 50 (46–57) Oceania 6,272 (4,551–8,805) 3,664 (2,538–5,333) 2,608 (1,822–3,673) 114 (82–152) 127 (87–180) 96 (70–126) South Asia 594,675 (542,901–645,401) 269,859 (223,569–308,772) 324,816 (290,521–360,005) 56 (51–62) 52 (42–60) 61 (54–68) Southeast Asia 409,791 (366,137–451,121) 192,227 (165,661–219,609) 217,564 (188,214–247,832) 93 (84–102) 80 (69–92) 107 (93–121) Southern Lati

608 (1,822–3,673) 114 (82–152) 127 (87–180) 96 (70–126) South Asia 594,675 (542,901–645,401) 269,859 (223,569–308,772) 324,816 (290,521–360,005) 56 (51–62) 52 (42–60) 61 (54–68) Southeast Asia 409,791 (366,137–451,121) 192,227 (165,661–219,609) 217,564 (188,214–247,832) 93 (84–102) 80 (69–92) 107 (93–121) Southern Lati n America 18,630 (17,766–19,510) 9,436 (8,777–10,113) 9,194 (8,677–9,734) 26 (24–27) 21 (20–23) 31 (29–33) Southern sub-Saharan Africa 25,307 (22,691–28,908) 14,474 (12,620–16,999) 10,833 (9,475–12,614) 60 (54–69) 58 (51–68) 61 (54–71) Tropical Latin America 99,629 (94,975–106,647) 49,108 (46,136–52,691) 50,521 (47,601–56,066) 57 (55–61) 50 (47–53) 67 (63–75) Western Europe 174,106 (166,453–182,116) 98,706 (92,998–104,421) 75,400 (71,667–79,634) 19 (18–20) 17 (16–18) 21 (20–22) Western sub-Saharan Africa 79,764 (67,506–97,152) 41,101 (32,116–55,046) 38,662 (31,506–49,231) 59 (50–71) 60 (48–80) 57 (47–71) Values are n (95% uncertainty intervals). Abbreviations as in Table 1. ∗ Results for all causes are available online via the GBD Compare web visualization tool (23).

n America 18,630 (17,766–19,510) 9,436 (8,777–10,113) 9,194 (8,677–9,734) 26 (24–27) 21 (20–23) 31 (29–33) Southern sub-Saharan Africa 25,307 (22,691–28,908) 14,474 (12,620–16,999) 10,833 (9,475–12,614) 60 (54–69) 58 (51–68) 61 (54–71) Tropical Latin America 99,629 (94,975–106,647) 49,108 (46,136–52,691) 50,521 (47,601–56,066) 57 (55–61) 50 (47–53) 67 (63–75) Western Europe 174,106 (166,453–182,116) 98,706 (92,998–104,421) 75,400 (71,667–79,634) 19 (18–20) 17 (16–18) 21 (20–22) Western sub-Saharan Africa 79,764 (67,506–97,152) 41,101 (32,116–55,046) 38,662 (31,506–49,231) 59 (50–71) 60 (48–80) 57 (47–71) Values are n (95% uncertainty intervals). Abbreviations as in Table 1. ∗ Results for all causes are available online via the GBD Compare web visualization tool (23). Many countries had no significant change in the estimated age-standardized prevalence of CVD from 1990 to 2015, often reflecting low data availability (Online Figures 1B and 2A). A number of countries showed a significant decline in the age-standardized prevalence of CVD, including the United States, Canada, Western Europe, Brazil, Australia, New Zealand, Japan, South Korea, Kenya, Cambodia, Laos, and India. There was a significant but small (<1%) increase in the age-standardized prevalence of CVD in Mexico, Venezuela, Saudi Arabia, and Mongolia.

the age-standardized prevalence of CVD, including the United States, Canada, Western Europe, Brazil, Australia, New Zealand, Japan, South Korea, Kenya, Cambodia, Laos, and India. There was a significant but small (<1%) increase in the age-standardized prevalence of CVD in Mexico, Venezuela, Saudi Arabia, and Mongolia. Deaths There were 12.59 million deaths (95% UI: 12.38 to 12.80 million deaths) due to CVD in 1990, increasing to 17.92 million deaths (95% UI: 17.59 to 18.28 million deaths) in 2015. There was broad variation in the age-standardized CVD mortality rate among countries (Central Illustration). Significant declines in the age-standardized death rate due to CVD occurred between 1990 and 2015 in all high-income and some middle-income countries, but no significant changes were detected over this time period for most of sub-Saharan Africa and multiple countries in Oceania and Southeast Asia, as well as Pakistan, Afghanistan, Kyrgyzstan, and Mongolia (Online Figure 2B). Bangladesh and the Philippines had significant increases in the age-standardized death rate due to CVD.Central Illustration Global Map, Age-Standardized Death Rate of CVD in 2015

n Africa and multiple countries in Oceania and Southeast Asia, as well as Pakistan, Afghanistan, Kyrgyzstan, and Mongolia (Online Figure 2B). Bangladesh and the Philippines had significant increases in the age-standardized death rate due to CVD.Central Illustration Global Map, Age-Standardized Death Rate of CVD in 2015 Choropleth map showing the estimated age-standardized mortality rate of total CVD in 2015 for each country. ATG = Antigua and Barbuda; BRB = Barbados; COM = Comoros; CVD = cardiovascular diseases; DMA = Dominica; E Med = Eastern Mediterranean; FJI = Fiji; FSM = Federated States of Micronesia; GRD = Grenada; KIR = Kiribati; KS = Kaposi sarcoma; LCA = Saint Lucia; MDV = Maldives; MHL = Marshall Islands; MLT = Malta; MUS = Mauritius; NMSC = nonmelanoma skin cancer; SGP = Singapore; SLB = Solomon Islands; SYC = Seychelles; TLS = Timor-Leste; TON = Tonga; TTO = Trinidad and Tobago; VCT = Saint Vincent and the Grenadines; VUT = Vanuatu; W Africa = West Africa; WSM = Samoa.

nt Lucia; MDV = Maldives; MHL = Marshall Islands; MLT = Malta; MUS = Mauritius; NMSC = nonmelanoma skin cancer; SGP = Singapore; SLB = Solomon Islands; SYC = Seychelles; TLS = Timor-Leste; TON = Tonga; TTO = Trinidad and Tobago; VCT = Saint Vincent and the Grenadines; VUT = Vanuatu; W Africa = West Africa; WSM = Samoa. CVD deaths and sociodemographic transition When each world region’s age-standardized CVD death rate for each year from 1990 to 2015 is plotted against an index of that region’s sociodemographic status in the same year, distinct patterns of epidemiological change are observed (Figure 2). Both an increase in SDI and a decline in cardiovascular mortality occurred in many regions. Despite marked improvement in SDI, CVD mortality did not decrease for men in South Asia and for both sexes in much of sub-Saharan Africa. A smaller rise and fall were seen for women in Oceania and men in East Asia. CVD mortality declined among high-income regions, but has plateaued in recent years. The lack of further decline in CVD mortality is particularly notable in high-income North America, Australasia, Western Europe, and the Caribbean. The maximum likelihood estimate relationship between SDI and CVD mortality, shown as a black line in Figure 2, is a gradual decline for women as SDI increases, with a more rapid decline at the highest levels of SDI. For men, CVD mortality increases as regions move from the lowest to middle levels of the SDI, with a significant decrease in mortality only observed for Central and Eastern Europe, high-income regions, and in Central Asia during the past decade.Figure 2 Age-Standardized CVD Mortality Rate, From 1990 to 2015, of 21 GBD World Regions by SDI

ty increases as regions move from the lowest to middle levels of the SDI, with a significant decrease in mortality only observed for Central and Eastern Europe, high-income regions, and in Central Asia during the past decade.Figure 2 Age-Standardized CVD Mortality Rate, From 1990 to 2015, of 21 GBD World Regions by SDI Relationship between age-standardized mortality rate for CVD and SDI over time. Each colored line represents a time trend of the relationship for the specified region. Each point represents a specific year for that region. The black line represents the overall global trend for age-standardized death rate of CVD in relation to SDI. CVD = cardiovascular diseases; GBD = Global Burden of Disease; SDI = sociodemographic index. The relationship between SDI and the age-standardized CVD death rate at the global level is shown in Figure 3. As SDI increases beyond 0.25, the highest CVD mortality rates shift from women to men. Cerebrovascular mortality rates begin to decline among women above an SDI of 0.3, although still increasing for men. IHD and cerebrovascular mortality rates continue to increase with greater SDI among men, peaking at an SDI of 0.5 to 0.75 (compared with 0.25 among women). CVD mortality decreases sharply for both sexes in countries with an SDI >0.75.Figure 3 Relationship Between Age-Standardized Mortality Rate, CVD Cause, and SDI, by Sex This figure displays the distribution of age-standardized death rate by causes of CVD mortality by SDI. Abbreviations as in Figure 1.

100,000 (95% UI: 255 to 328 cases per 100,000) for those 40 to 44 years of age to 11,203 cases per 100,000 (95% UI: 9,610 to 13,178 cases per 100,000) for those 75 to 79 years of age, declining slightly for those 80 years of age and over to a rate of 9,700 cases per 100,000 (95% UI: 8,773 to 10,738 cases per 100,000). Eastern Europe had the highest estimated age-standardized prevalence of IHD in 2015 (4,140 cases per 100,000; 95% UI: 3,811 to 4,499 cases per 100,000), followed by Central Asia and then Central Europe (Online Figure 5A). IHD accounted for almost one-half of all CVD cases in Central Asia and Eastern Europe, but a smaller proportion in Central Europe, where other cardiovascular and circulatory diseases made up a larger proportion of total cases (Figure 4). Eastern sub-Saharan Africa, the Middle East/North Africa region, and South Asia all had similar estimated rates of just over 2,000 prevalent cases per 100,000. More than one-quarter of cases in sub-Saharan Africa were due to other cardiovascular and circulatory diseases. The lowest estimated age-standardized prevalence of IHD was in Central sub-Saharan Africa (622 per 100,000; 95% UI: 578 to 677 per 100,000), with similar low rates estimated in the southern Latin America and high-income Asia Pacific regions.Figure 4 Percent of Age-Standardized Prevalent Cases per 100,000 for CVD Causes, 2015 This figure displays the relative distribution of age-standardized prevalence by CVD cause for 21 GBD world regions. Abbreviations as in Figure 1.

Eastern Europe had the highest estimated age-standardized prevalence of IHD in 2015 (4,140 cases per 100,000; 95% UI: 3,811 to 4,499 cases per 100,000), followed by Central Asia and then Central Europe (Online Figure 5A). IHD accounted for almost one-half of all CVD cases in Central Asia and Eastern Europe, but a smaller proportion in Central Europe, where other cardiovascular and circulatory diseases made up a larger proportion of total cases (Figure 4). Eastern sub-Saharan Africa, the Middle East/North Africa region, and South Asia all had similar estimated rates of just over 2,000 prevalent cases per 100,000. More than one-quarter of cases in sub-Saharan Africa were due to other cardiovascular and circulatory diseases. The lowest estimated age-standardized prevalence of IHD was in Central sub-Saharan Africa (622 per 100,000; 95% UI: 578 to 677 per 100,000), with similar low rates estimated in the southern Latin America and high-income Asia Pacific regions.Figure 4 Percent of Age-Standardized Prevalent Cases per 100,000 for CVD Causes, 2015 This figure displays the relative distribution of age-standardized prevalence by CVD cause for 21 GBD world regions. Abbreviations as in Figure 1. There were an estimated 8.92 million deaths (95% UI: 8.75 to 9.12 million deaths) due to IHD in 2015, making IHD the leading cause of death in the world. The death rate due to IHD rose steeply above age 40 years, increasing from an estimated 33 deaths per 100,000 (95% UI: 32 to 35 per 100,000) for those 40 to 44 years of age to 1,050 per 100,000 (95% UI: 1,025 to 1,076 per 100,000) by ages 75 to 79 years (Online Figure 4B). Above 80 years of age, the IHD death rate was estimated to be more than twice that rate (2,671 per 100,000; 95% UI: 2,600 to 2,738 per 100,000) and was by far the leading global cause of death.

ose 40 to 44 years of age to 1,050 per 100,000 (95% UI: 1,025 to 1,076 per 100,000) by ages 75 to 79 years (Online Figure 4B). Above 80 years of age, the IHD death rate was estimated to be more than twice that rate (2,671 per 100,000; 95% UI: 2,600 to 2,738 per 100,000) and was by far the leading global cause of death. The estimated age-standardized IHD death rate was highest in Central Asia (336 per 100,000; 95% UI: 326 to 347 per 100,000) and Eastern Europe (326 per 100,000; 95% UI: 319 to 333 per 100,000), followed by Oceania, South Asia, and the Middle East/North Africa (Table 3). These regions, as well as high-income North America and Latin America, had particularly high proportions of total CVD deaths that were due to IHD (Figure 5). The estimated age-standardized IHD death rate was similar, from 100 to 150 deaths per 100,000, across a wide range of regions including high-income North America, Southeast and East Asia, and the Caribbean. The high-income Asia Pacific region had a much lower estimated age-standardized IHD death rate than any other region (45 per 100,000; 95% UI: 44 to 46 per 100,000) (Online Figure 6A).Figure 5 Percent of Age-Standardized Deaths per 100,000 for CVD Causes, 2015 This figure displays the relative distribution of age-standardized prevalence by CVD cause for 21 GBD world regions. Abbreviations as in Figure 1. Table 3 Global and Regional All-Age Prevalence and Age-Standardized Prevalence Rates in 2015, by Sex, for Selected Causes of CVD Mortality∗

The estimated age-standardized IHD death rate was highest in Central Asia (336 per 100,000; 95% UI: 326 to 347 per 100,000) and Eastern Europe (326 per 100,000; 95% UI: 319 to 333 per 100,000), followed by Oceania, South Asia, and the Middle East/North Africa (Table 3). These regions, as well as high-income North America and Latin America, had particularly high proportions of total CVD deaths that were due to IHD (Figure 5). The estimated age-standardized IHD death rate was similar, from 100 to 150 deaths per 100,000, across a wide range of regions including high-income North America, Southeast and East Asia, and the Caribbean. The high-income Asia Pacific region had a much lower estimated age-standardized IHD death rate than any other region (45 per 100,000; 95% UI: 44 to 46 per 100,000) (Online Figure 6A).Figure 5 Percent of Age-Standardized Deaths per 100,000 for CVD Causes, 2015 This figure displays the relative distribution of age-standardized prevalence by CVD cause for 21 GBD world regions. Abbreviations as in Figure 1. Table 3 Global and Regional All-Age Prevalence and Age-Standardized Prevalence Rates in 2015, by Sex, for Selected Causes of CVD Mortality∗ All-Age Prevalence Age-Standardized Rate (per 100,000) Total Female Male Total Female Male Cardiovascular diseases Global 422,738,396 (415,534,458–427,870,820) 205,821,777 (202,134,194–208,585,279) 216,916,608 (213,418,653–219,875,383) 6,304 (6,196–6,383) 5,812 (5,711–5,892) 6,833 (6,717–6,913) Andean Latin America 2,527,168 (2,440,089–2,632,508) 1,172,826 (1,114,050–1,250,584) 1,354,342 (1,295,192–1,433,551) 5,666 (5,472–5,913) 5,049 (4,798–5,397) 6,330 (6,052–6,719) Australasia 1,761,024 (1,697,613–1,854,157) 806,618 (770,407–861,873) 954,406 (911,026–1,015,916) 4,472 (4,305–4,705) 3,816 (3,637–4,093) 5,171 (4,938–5,496) Caribbean 2,822,711 (2,751,853–2,904,703) 1,374,060 (1,325,923–1,427,497) 1,448,651 (1,404,659–1,502,391) 6,477 (6,314–6,664) 5,943 (5,731–6,178) 7,071 (6,859–7,335) Central Asia 4,555,309 (4,461,547–4,654,438) 2,356,480 (2,299,260–2,433,326) 2,198,828 (2,137,653–2,268,905) 7,147 (6,989–7,309) 6,576 (6,413–6,787) 7,880 (7,675–8,107) Central Europe 13,560,336 (13,304,071–13,830,352) 6,901,529 (6,747,945–7,104,379) 6,658,807 (6,512,843–6,843,977) 7,552 (7,418–7,704) 6,610 (6,465–6,798) 8,656 (8,480–8,886) Central Latin America 11,826,535 (11,582,577–12,074,503) 5,897,220 (5,737,257–6,055,563) 5,929,315 (5,785,197–6,072,577) 6,225 (6,095–6,359) 5,842 (5,685–6,007) 6,653 (6,502–6,813) Central sub-Saharan Africa 3,570,609 (3,420,593–3,723,098) 1,923,654 (1,820,220–2,025,073) 1,646,955 (1,545,442–1,747,781) 6,852 (6,590–7,144) 7,070 (6,672–7,494) 6,621 (6,321–7,010) East Asia 93,345,590 (91,674,478–95,227,641) 43,135,337 (42,131,579–44,032,727) 50,210,253 (49,254,116–51,093,487) 6,146 (6,019–6,259) 5,543 (5,409–5,684) 6,725 (6,594–6,848) Eastern Europe 23,898,622 (23,231,343–24,726,919) 13,142,602 (12,664,583–13,765,752) 10,756,019 (10,338,916–11,263,598) 7,482 (7,270–7,739) 6,441 (6,192–6,736) 8,982 (8,648–9,382) Eastern sub-Saharan Africa 13,924,061 (13,569,314–14,266,069) 7,357,728 (7,135,795–7,570,055) 6,566,333 (6,356,066–6,795,777) 8,444 (8,237–8,654) 8,542 (8,290–8,806) 8,317 (8,084–8,624) High-income Asia Pacific 13,619,192 (13,383,086–13,853,346) 6,987,560 (6,835,073–7,141,021) 6,631,632 (6,506,293–6,750,103) 3,831 (3,756–3,900) 3,433 (3,35

frica 13,924,061 (13,569,314–14,266,069) 7,357,728 (7,135,795–7,570,055) 6,566,333 (6,356,066–6,795,777) 8,444 (8,237–8,654) 8,542 (8,290–8,806) 8,317 (8,084–8,624) High-income Asia Pacific 13,619,192 (13,383,086–13,853,346) 6,987,560 (6,835,073–7,141,021) 6,631,632 (6,506,293–6,750,103) 3,831 (3,756–3,900) 3,433 (3,35 5–3,515) 4,268 (4,182–4,344) High-income North America 26,512,238 (26,323,905–26,683,623) 12,993,521 (12,883,697–13,104,758) 13,518,717 (13,407,429–13,637,389) 5,302 (5,268–5,338) 4,732 (4,693–4,776) 5,953 (5,906–6,003) North Africa and Middle East 31,044,596 (30,460,856–31,635,075) 14,248,387 (13,908,987–14,665,663) 16,796,211 (16,442,389–17,219,725) 8,017 (7,880–8,150) 7,425 (7,263–7,609) 8,627 (8,461–8,820) Oceania 515,811 (499,389–532,290) 256,344 (245,973–265,385) 259,468 (248,274–271,865) 7,057 (6,866–7,253) 6,908 (6,646–7,118) 7,243 (7,006–7,539) South Asia 71,649,779 (69,838,392–73,308,893) 33,572,668 (32,670,754–34,519,966) 38,077,110 (37,050,891–39,261,090) 6,006 (5,854–6,120) 5,567 (5,420–5,700) 6,460 (6,313–6,605) Southeast Asia 31,811,965 (31,108,918–32,506,116) 15,689,372 (15,295,087–16,147,415) 16,122,592 (15,672,125–16,657,696) 6,293 (6,153–6,427) 5,860 (5,714–6,040) 6,772 (6,595–6,992) Southern Latin America 2,907,935 (2,789,462–3,052,480) 1,514,205 (1,432,185–1,625,489) 1,393,730 (1,326,067–1,486,534) 4,092 (3,921–4,305) 3,668 (3,459–3,962) 4,631 (4,409–4,935) Southern sub-Saharan Africa 3,451,145 (3,361,155–3,547,451) 1,891,089 (1,841,419–1,949,982) 1,560,056 (1,508,904–1,619,296) 7,668 (7,495–7,844) 7,386 (7,205–7,600) 7,936 (7,698–8,188) Tropical Latin America 15,369,478 (15,061,429–15,652,622) 7,474,127 (7,279,837–7,700,746) 7,895,350 (7,725,078–8,091,713) 8,102 (7,946–8,252) 7,233 (7,038–7,445) 9,110 (8,922–9,328) Western Europe 38,123,584 (37,400,667–38,901,652) 18,947,709 (18,414,908–19,464,239) 19,175,875 (18,772,625–19,677,426) 5,106 (5,005–5,214) 4,479 (4,355–4,607) 5,778 (5,649–5,927) Western sub-Saharan Africa 15,940,704 (15,547,008–16,301,023) 8,178,745 (7,950,866–8,396,914) 7,761,959 (7,528,316–8,000,790) 9,475 (9,269–9,679) 9,753 (9,489–9,996) 9,179 (8,938–9,423) Ischemic heart disease Global 110,550,305 (100,675,890–121,798,660) 46,102,255 (42,134,035–50,800,684) 64,448,047 (58,685,079–71,017,092) 1,663 (1,519–1,828) 1,312 (1,198–1,447) 2,045 (1,864–2,251) Andean Latin America 433,378 (392,679–480,935) 191,058 (172,479–211,737) 242,320 (218,848–270,674) 999 (904–1,110) 836 (755–928) 1,179 (1,064–1,318) Australasia 374,391 (356,1

0–121,798,660) 46,102,255 (42,134,035–50,800,684) 64,448,047 (58,685,079–71,017,092) 1,663 (1,519–1,828) 1,312 (1,198–1,447) 2,045 (1,864–2,251) Andean Latin America 433,378 (392,679–480,935) 191,058 (172,479–211,737) 242,320 (218,848–270,674) 999 (904–1,110) 836 (755–928) 1,179 (1,064–1,318) Australasia 374,391 (356,1 54–394,637) 141,494 (133,557–150,318) 232,897 (220,558–245,473) 951 (903–1,003) 659 (620–703) 1,261 (1,193–1,328) Caribbean 649,913 (594,603–712,186) 291,676 (266,641–320,463) 358,237 (326,370–393,695) 1,499 (1,367–1,645) 1,265 (1,155–1,391) 1,763 (1,609–1,933) Central Asia 1,959,856 (1,792,534–2,148,360) 938,381 (850,809–1,037,029) 1,021,475 (934,886–1,115,021) 3,158 (2,886–3,464) 2,683 (2,432–2,970) 3,778 (3,473–4,102) Central Europe 4,887,250 (4,421,084–5,396,339) 2,054,737 (1,855,428–2,273,925) 2,832,513 (2,556,599–3,130,360) 2,723 (2,462–3,006) 1,977 (1,786–2,192) 3,634 (3,292–4,009) Central Latin America 2,743,458 (2,506,477–3,022,604) 1,314,991 (1,201,625–1,444,890) 1,428,466 (1,302,990–1,579,648) 1,465 (1,341–1,617) 1,313 (1,200–1,449) 1,640 (1,499–1,811) Central sub-Saharan Africa 262,008 (242,179–285,446) 116,600 (107,838–126,833) 145,408 (133,789–159,226) 622 (577–677) 533 (492–577) 721 (666–787) East Asia 20,261,326 (18,299,731–22,520,859) 8,319,850 (7,520,800–9,229,590) 11,941,475 (10,714,137–13,265,553) 1,276 (1,158–1,414) 1,046 (945–1,158) 1,511 (1,368–1,674) Eastern Europe 13,228,694 (12,169,855–14,388,371) 6,092,839 (5,547,898–6,669,667) 7,135,855 (6,530,196–7,789,228) 4,140 (3,812–4,499) 2,973 (2,710–3,261) 5,903 (5,419–6,416) Eastern sub-Saharan Africa 3,098,872 (2,735,642–3,532,901) 1,264,989 (1,131,992–1,434,792) 1,833,883 (1,598,281–2,126,672) 2,060 (1,812–2,348) 1,607 (1,431–1,821) 2,578 (2,229–2,961) High-income Asia Pacific 2,596,645 (2,457,585–2,752,645) 1,216,369 (1,151,727–1,286,526) 1,380,276 (1,300,678–1,469,956) 708 (669–752) 562 (531–598) 873 (824–928) High-income North America 6,278,342 (6,098,359–6,483,078) 2,789,029 (2,708,288–2,870,653) 3,489,312 (3,371,951–3,618,277) 1,226 (1,191–1,267) 972 (945–1,000) 1,517 (1,466–1,572) North Africa and Middle East 7,072,468 (6,265,905–8,007,140) 2,975,936 (2,637,853–3,378,580) 4,096,532 (3,629,716–4,639,947) 2,055 (1,825–2,315) 1,705 (1,505–1,931) 2,424 (2,156–2,723) Oceania 62,813 (56,704–69,835) 27,472 (24,802–30,653) 35,341 (31,670–39,554) 1,103 (995–1,226) 947 (858–1,057) 1,269 (1,141–1,411) South Asia 23,148,018 (20,930,975–25,698,774) 8,781,644 (7,924,913–9,746,863) 14,366,374 (12,94

) 4,096,532 (3,629,716–4,639,947) 2,055 (1,825–2,315) 1,705 (1,505–1,931) 2,424 (2,156–2,723) Oceania 62,813 (56,704–69,835) 27,472 (24,802–30,653) 35,341 (31,670–39,554) 1,103 (995–1,226) 947 (858–1,057) 1,269 (1,141–1,411) South Asia 23,148,018 (20,930,975–25,698,774) 8,781,644 (7,924,913–9,746,863) 14,366,374 (12,94 6,829–16,030,805) 2,052 (1,861–2,271) 1,576 (1,423–1,750) 2,533 (2,290–2,808) Southeast Asia 7,471,104 (6,551,421–8,655,836) 3,246,055 (2,871,795–3,703,755) 4,225,049 (3,637,737–4,990,540) 1,520 (1,339–1,742) 1,261 (1,116–1,442) 1,801 (1,571–2,089) Southern Latin America 488,329 (458,416–516,912) 178,843 (167,464–189,854) 309,486 (289,061–330,028) 690 (647–731) 425 (398–451) 1,021 (954–1,088) Southern sub-Saharan Africa 653,589 (584,539–736,284) 283,992 (255,828–318,332) 369,597 (328,378–419,248) 1,415 (1,264–1,595) 1,091 (981–1,230) 1,826 (1,627–2,065) Tropical Latin America 3,480,148 (3,159,912–3,841,271) 1,414,323 (1,296,129–1,549,721) 2,065,825 (1,858,218–2,300,500) 1,830 (1,666–2,018) 1,370 (1,258–1,507) 2,384 (2,155–2,641) Western Europe 9,054,333 (8,208,626–10,087,602) 3,670,540 (3,286,094–4,154,807) 5,383,793 (4,897,589–5,957,280) 1,244 (1,125–1,388) 892 (793–1,012) 1,634 (1,485–1,809) Western sub-Saharan Africa 2,345,370 (2,057,647–2,677,442) 791,437 (707,199–894,197) 1,553,933 (1,341,625–1,795,108) 1,513 (1,327–1,724) 1,038 (928–1,172) 2,024 (1,759–2,321) Ischemic strokes Global 24,929,026 (24,362,226–25,609,959) 11,902,196 (11,625,159–12,233,046) 13,026,829 (12,718,240–13,382,922) 376 (367–386) 340 (331–349) 416 (406–428) Andean Latin America 72,045 (69,967–74,245) 37,542 (36,413–38,744) 34,503 (33,492–35,599) 165 (160–170) 163 (158–168) 167 (162–173) Australasia 97,732 (95,064–100,797) 48,401 (46,706–50,165) 49,331 (47,986–50,972) 252 (245–260) 235 (227–243) 271 (263–280) Caribbean 117,888 (114,875–121,193) 59,348 (57,746–61,088) 58,539 (56,879–60,294) 274 (267–282) 259 (252–267) 290 (281–299) Central Asia 344,977 (337,341–353,209) 189,937 (185,553–194,749) 155,040 (151,404–159,070) 552 (539–566) 535 (522–549) 575 (560–590) Central Europe 796,434 (775,591–820,957) 424,906 (412,679–439,291) 371,527 (361,686–382,321) 445 (434–458) 415 (404–429) 483 (470–497) Central Latin America 194,157 (187,314–201,435) 99,012 (95,566–102,898) 95,145 (91,835–98,817) 105 (102–110) 100 (97–104) 111 (107–115) Central sub-Saharan Africa 187,635 (182,346–192,989) 107,722 (104,532–111,025) 79,912 (77,566–82,174) 440 (427–453) 476 (461–492) 394 (382–407) East Asia 7,423,862

4–429) 483 (470–497) Central Latin America 194,157 (187,314–201,435) 99,012 (95,566–102,898) 95,145 (91,835–98,817) 105 (102–110) 100 (97–104) 111 (107–115) Central sub-Saharan Africa 187,635 (182,346–192,989) 107,722 (104,532–111,025) 79,912 (77,566–82,174) 440 (427–453) 476 (461–492) 394 (382–407) East Asia 7,423,862 (7,228,513–7,652,139) 3,161,924 (3,077,657–3,259,358) 4,261,938 (4,146,738–4,395,442) 474 (461–490) 403 (392–416) 548 (532–566) Eastern Europe 1,939,626 (1,886,897–1,993,437) 1,201,817 (1,169,100–1,236,376) 737,809 (713,336–763,066) 614 (597–630) 603 (588–620) 625 (604–647) Eastern sub-Saharan Africa 772,966 (750,967–795,839) 418,379 (406,070–431,313) 354,587 (344,928–364,964) 506 (490–523) 520 (503–538) 489 (473–505) High-income Asia Pacific 1,193,488 (1,162,144–1,229,286) 530,971 (516,893–546,687) 662,517 (645,074–683,051) 358 (350–369) 294 (287–303) 434 (423–447) High-income North America 1,167,328 (1,133,548–1,205,267) 581,089 (563,548–599,833) 586,239 (569,088–605,367) 236 (229–244) 216 (209–223) 259 (252–268) North Africa and Middle East 1,312,093 (1,285,007–1,341,997) 672,113 (658,117–688,072) 639,980 (626,008–654,639) 355 (347–364) 357 (349–366) 353 (344–362) Oceania 32,854 (31,947–33,759) 16,173 (15,656–16,722) 16,681 (16,256–17,125) 535 (519–551) 504 (488–523) 570 (554–587) South Asia 3,932,151 (3,831,984–4,047,704) 1,714,307 (1,667,742–1,765,753) 2,217,843 (2,160,750–2,281,985) 333 (323–343) 289 (280–298) 377 (367–390) Southeast Asia 2,106,975 (2,059,929–2,157,906) 1,011,731 (986,323–1,038,279) 1,095,244 (1,068,060–1,121,515) 424 (413–435) 383 (373–394) 470 (457–482) Southern Latin America 206,767 (200,751–213,651) 107,942 (104,638–111,837) 98,824 (95,589–102,483) 296 (287–306) 275 (266–285) 324 (313–336) Southern sub-Saharan Africa 190,018 (185,419–195,150) 103,090 (100,211–106,287) 86,928 (84,803–89,122) 419 (408–431) 396 (384–410) 444 (432–456) Tropical Latin America 356,275 (338,561–375,658) 172,946 (164,866–182,222) 183,329 (173,927–194,059) 193 (183–203) 169 (161–179) 221 (210–235) Western Europe 1,865,822 (1,817,012–1,917,658) 924,111 (899,640–951,652) 941,712 (915,999–969,970) 255 (248–262) 227 (222–234) 286 (278–294) Western sub-Saharan Africa 617,935 (603,712–633,755) 318,735 (310,971–327,448) 299,200 (292,051–306,700) 388 (378–399) 397 (386–410) 377 (367–387) Hemorrhagic and other strokes Global 18,669,622 (18,258,729–19,124,482) 8,900,797 (8,700,081–9,121,277) 9,768,824 (9,543,517–10,012,685) 268 (262–275) 249 (243–255) 288 (282–296) Ande

b-Saharan Africa 617,935 (603,712–633,755) 318,735 (310,971–327,448) 299,200 (292,051–306,700) 388 (378–399) 397 (386–410) 377 (367–387) Hemorrhagic and other strokes Global 18,669,622 (18,258,729–19,124,482) 8,900,797 (8,700,081–9,121,277) 9,768,824 (9,543,517–10,012,685) 268 (262–275) 249 (243–255) 288 (282–296) Ande an Latin America 82,248 (79,877–84,631) 44,824 (43,468–46,162) 37,424 (36,298–38,588) 164 (160–169) 174 (169–179) 155 (150–159) Australasia 46,455 (45,253–47,851) 25,097 (24,270–25,923) 21,358 (20,810–22,010) 124 (121–128) 129 (125–133) 119 (116–123) Caribbean 114,404 (111,412–117,288) 60,894 (59,271–62,556) 53,511 (52,084–54,945) 258 (251–264) 266 (258–273) 250 (243–257) Central Asia 285,031 (279,091–291,260) 155,973 (152,515–159,672) 129,058 (126,281–131,955) 404 (395–413) 402 (393–412) 407 (398–416) Central Europe 482,420 (470,995–495,973) 278,486 (271,057–287,367) 203,934 (199,058–209,297) 299 (292–307) 319 (310–329) 279 (272–286) Central Latin America 180,268 (173,792–187,229) 93,592 (90,155–97,387) 86,676 (83,670–90,019) 86 (83–89) 85 (82–89) 87 (84–90) Central sub-Saharan Africa 181,821 (177,249–186,526) 105,226 (102,578–108,171) 76,596 (74,630–78,729) 332 (323–340) 372 (362–382) 285 (278–293) East Asia 4,933,114 (4,809,787–5,071,809) 2,009,705 (1,958,387–2,067,784) 2,923,409 (2,850,909–3,004,011) 307 (299–316) 254 (248–262) 360 (350–370) Eastern Europe 1,129,261 (1,103,084–1,156,455) 683,094 (666,554–701,155) 446,168 (433,515–459,717) 377 (369–386) 379 (370–389) 377 (367–389) Eastern sub-Saharan Africa 840,950 (816,966–866,461) 461,124 (448,039–475,939) 379,825 (368,337–392,136) 405 (395–417) 433 (422–446) 373 (363–383) High-income Asia Pacific 587,292 (572,980–603,859) 294,104 (286,904–302,836) 293,188 (285,828–301,761) 202 (197–208) 194 (189–200) 213 (207–219) High-income North America 403,245 (391,490–416,047) 208,999 (202,897–215,686) 194,246 (188,361–200,619) 85 (82–88) 82 (79–85) 88 (85–91) North Africa and Middle East 1,093,535 (1,069,924–1,119,497) 594,477 (581,671–608,840) 499,059 (487,623–511,343) 253 (248–259) 275 (269–281) 232 (227–237) Oceania 38,961 (38,070–39,939) 20,324 (19,764–20,923) 18,637 (18,175–19,116) 493 (481–505) 507 (492–522) 479 (468–492) South Asia 3,970,983 (3,865,439–4,086,075) 1,734,618 (1,687,573–1,786,688) 2,236,364 (2,178,258–2,300,244) 286 (279–294) 254 (247–261) 318 (310–327) Southeast Asia 2,040,033 (1,995,560–2,086,665) 973,642 (950,221–998,350) 1,066,391 (1,041,873–1,090,836) 357 (350–365) 331 (323–339) 385 (376–394

7 (492–522) 479 (468–492) South Asia 3,970,983 (3,865,439–4,086,075) 1,734,618 (1,687,573–1,786,688) 2,236,364 (2,178,258–2,300,244) 286 (279–294) 254 (247–261) 318 (310–327) Southeast Asia 2,040,033 (1,995,560–2,086,665) 973,642 (950,221–998,350) 1,066,391 (1,041,873–1,090,836) 357 (350–365) 331 (323–339) 385 (376–394 ) Southern Latin America 126,534 (122,934–130,607) 68,749 (66,763–71,058) 57,786 (56,064–59,706) 185 (180–191) 186 (181–192) 186 (180–192) Southern sub-Saharan Africa 186,863 (182,132–192,225) 95,226 (92,741–98,041) 91,637 (89,262–94,206) 332 (324–341) 316 (308–325) 349 (341–358) Tropical Latin America 292,725 (279,182–307,749) 149,531 (142,631–157,232) 143,194 (136,362–150,570) 146 (139–153) 138 (132–146) 156 (149–164) Western Europe 926,435 (903,480–950,683) 482,432 (470,375–495,455) 444,003 (432,460–456,481) 141 (137–145) 138 (135–142) 145 (141–149) Western sub-Saharan Africa 727,043 (709,295–746,947) 360,683 (352,133–370,683) 366,360 (356,969–377,288) 323 (316–331) 332 (324–340) 312 (305–320) Values are n (95% uncertainty intervals). ∗ Results for all causes are available online via the GBD Compare web visualization tool (23).

) Southern Latin America 126,534 (122,934–130,607) 68,749 (66,763–71,058) 57,786 (56,064–59,706) 185 (180–191) 186 (181–192) 186 (180–192) Southern sub-Saharan Africa 186,863 (182,132–192,225) 95,226 (92,741–98,041) 91,637 (89,262–94,206) 332 (324–341) 316 (308–325) 349 (341–358) Tropical Latin America 292,725 (279,182–307,749) 149,531 (142,631–157,232) 143,194 (136,362–150,570) 146 (139–153) 138 (132–146) 156 (149–164) Western Europe 926,435 (903,480–950,683) 482,432 (470,375–495,455) 444,003 (432,460–456,481) 141 (137–145) 138 (135–142) 145 (141–149) Western sub-Saharan Africa 727,043 (709,295–746,947) 360,683 (352,133–370,683) 366,360 (356,969–377,288) 323 (316–331) 332 (324–340) 312 (305–320) Values are n (95% uncertainty intervals). ∗ Results for all causes are available online via the GBD Compare web visualization tool (23). Stroke Globally, hemorrhagic and other strokes, and ischemic strokes were the second- and third-largest CVD causes of DALYs in 2015, and the 4th- and 13th-largest overall, respectively. Ischemic stroke DALY outranked hemorrhagic and other strokes only in Central and Eastern Europe and high-income North America. There were an estimated 5.39 million acute first-ever ischemic strokes (95% UI: 5.02 to 5.73 million), 3.58 million acute first-ever hemorrhagic and other strokes (95% UI: 3.34 to 3.82 million), and 42.43 million prevalent cases of cerebrovascular disease (95% UI: 42.07 to 42.77 million) overall in 2015. The prevalence of stroke began increasing above 40 years of age, reaching the highest rate for any 5-year age group for those 74 to 79 years of age (4,201 cases per 100,000; 95% UI: 4,140 to 4,258 per 100,000), with the rate declining by one-half this amount among those over 80 years of age.

7 million) overall in 2015. The prevalence of stroke began increasing above 40 years of age, reaching the highest rate for any 5-year age group for those 74 to 79 years of age (4,201 cases per 100,000; 95% UI: 4,140 to 4,258 per 100,000), with the rate declining by one-half this amount among those over 80 years of age. Oceania was the region with the highest prevalence of stroke (1,003 per 100,000; 95% UI: 985 to 1,025 per 100,000), followed by Eastern Europe, Central Asia, and Southeast Asia. The lowest stroke prevalence in the world, less than one-fifth that of the highest regions, was in Central Latin America (177 per 100,000; 95% UI: 174 to 180 per 100,000). The estimated age-standardized prevalence of ischemic stroke was greater than that of hemorrhagic stroke in most regions, but was the same for Andean Latin America. There were 6.33 million deaths due to stroke in 2015 (95% UI: 6.18 to 6.49 million deaths), with 57% of these stroke deaths due to ischemic stroke. The stroke mortality rate began rising above 50 years of age, increasing for each older age group until it reached 1,812 per 100,000 (95% UI: 1,764 to 1,864 per 100,000) among those over age 80 years.

hs due to stroke in 2015 (95% UI: 6.18 to 6.49 million deaths), with 57% of these stroke deaths due to ischemic stroke. The stroke mortality rate began rising above 50 years of age, increasing for each older age group until it reached 1,812 per 100,000 (95% UI: 1,764 to 1,864 per 100,000) among those over age 80 years. The estimated stroke age-standardized mortality rate in 2015 was also greatest in Oceania (191 per 100,000; 95% UI: 148 to 248 per 100,000), followed by Central sub-Saharan Africa. The lowest estimated rates were in high-income North America, Australasia, Western Europe, Andean Latin America, high-income Asia Pacific, Central Latin America, and southern Latin America. The ischemic stroke death rate dominated the hemorrhagic stroke death rate in Eastern and Central Europe, Middle East/North Africa, the Caribbean, and Southern Latin America, with similar rates found in Western sub-Saharan Africa, high-income North America, Australasia, South Asia, and Western Europe.

rn Latin America. The ischemic stroke death rate dominated the hemorrhagic stroke death rate in Eastern and Central Europe, Middle East/North Africa, the Caribbean, and Southern Latin America, with similar rates found in Western sub-Saharan Africa, high-income North America, Australasia, South Asia, and Western Europe. Hypertensive heart disease Hypertensive heart disease was the fourth-highest ranked CVD cause for DALYs in 2015 globally; however, it ranked lower in high-income regions such as Australasia, high-income Asia Pacific, and Western Europe. There were 6.09 million (95% UI: 5.73 to 6.43 million) prevalent cases of hypertensive heart disease in 2015. The prevalence rose continuously for each age group, from 2.0 per 100,000 (95% UI: 1.7 to 2.1 per 100,000) at ages 20 to 24 years until reaching 1,360 per 100,000 (95% UI: 1,248 to 1,502 per 100,000) for those >80 years of age. The prevalence was highest in Western sub-Saharan Africa, followed by Central and Eastern sub-Saharan Africa, tropical Latin America, and the Caribbean. The lowest rates were estimated for Western and Eastern Europe. There were 962,400 deaths (95% UI: 873,600 to 1,024,500 deaths) due to hypertensive heart disease in 2015. The mortality rate rose for ages >60 years, peaking at 296 per 100,000 (95% UI: 257 to 315 per 100,000) for age >80 years. Death rates due to hypertensive heart disease followed a similar pattern as the condition's prevalence.

2,400 deaths (95% UI: 873,600 to 1,024,500 deaths) due to hypertensive heart disease in 2015. The mortality rate rose for ages >60 years, peaking at 296 per 100,000 (95% UI: 257 to 315 per 100,000) for age >80 years. Death rates due to hypertensive heart disease followed a similar pattern as the condition's prevalence. Cardiomyopathy Cardiomyopathies and acute myocarditis were a higher-ranked cause of CVD DALYs in the regions of Central and Eastern Europe than in other world regions. There were an estimated 2.54 million (95% UI: 2.41 to 2.66 million) prevalent cases of cardiomyopathy and myocarditis in 2015. There was a slightly higher prevalence among children 1 to 4 years of age (29 per 100,000; 95% UI: 26 to 32 per 100,000), which then decreased for older children, increasing slowly throughout adulthood. A very large increase in cases of cardiomyopathy and acute myocarditis was estimated above 80 years of age (628 per 100,000; 95% UI: 569 to 692 per 100,000), more than 6-fold higher than for the next youngest age group of 75 to 79 years of age. This condition accounted for a relatively small proportion of CVD cases overall, with the greatest age-standardized prevalence estimated for Southern sub-Saharan Africa, followed by tropical Latin America, high-income North America, and other regions of sub-Saharan Africa.

t youngest age group of 75 to 79 years of age. This condition accounted for a relatively small proportion of CVD cases overall, with the greatest age-standardized prevalence estimated for Southern sub-Saharan Africa, followed by tropical Latin America, high-income North America, and other regions of sub-Saharan Africa. There were 353,700 (95% UI: 339,500 to 370,600) deaths due to cardiomyopathy and myocarditis in 2015. The mortality rate was as high as 47 per 100,000 persons (95% UI: 33 to 59 per 100,000) within the first week of life. The mortality rate decreased by 5 years of age, and then increased steadily throughout adulthood before increasing more than 300% after 80 years of age, to a peak rate of 90 per 100,000 (95% UI: 85 to 94 per 100,000). CVD deaths due to cardiomyopathy and myocarditis were most common in regions with high prevalence, including Southern sub-Saharan Africa and tropical Latin America.

ased steadily throughout adulthood before increasing more than 300% after 80 years of age, to a peak rate of 90 per 100,000 (95% UI: 85 to 94 per 100,000). CVD deaths due to cardiomyopathy and myocarditis were most common in regions with high prevalence, including Southern sub-Saharan Africa and tropical Latin America. Aortic aneurysm Aortic aneurysm was the fifth leading cause of CVD DALYs in the high-income Asia Pacific region, but ranked lower in other world regions. Globally, there were an estimated 168,200 deaths (95% UI: 163,500 to 172,800 deaths) due to aortic aneurysm in 2015. The age-specific mortality rate due to aortic aneurysm rose most quickly after 60 years of age, reaching a peak of 50 per 100,000 (95% UI: 48 to 53 per 100,000) above 80 years of age. Tropical Latin America had the highest death rate due to aortic aneurysm, followed by Southern Latin America, high-income Asia Pacific, Australasia, and Oceania. Death rates due to aortic aneurysm were lowest in Western sub-Saharan Africa and East Asia.

per 100,000 (95% UI: 48 to 53 per 100,000) above 80 years of age. Tropical Latin America had the highest death rate due to aortic aneurysm, followed by Southern Latin America, high-income Asia Pacific, Australasia, and Oceania. Death rates due to aortic aneurysm were lowest in Western sub-Saharan Africa and East Asia. Atrial fibrillation Atrial fibrillation ranked higher as a CVD cause of DALYs for Western Europe (fifth), Australasia (sixth), and South Asia (sixth). Globally there were an estimated 33.3 million (95% UI: 30.0 to 37.2 million) prevalent cases of atrial fibrillation in 2015. The age-specific prevalence of atrial fibrillation increased steadily for each 5-year age group above 30 years of age, reaching a rate of 5,544 per 100,000 (95% UI: 4,863 to 6,307 per 100,000) above 80 years of age. Age-standardized prevalence was highest in high-income North America (1,224 per 100,000; 95% UI: 1,157 to 1,297 per 100,000), followed by Western and Central Europe. There were an estimated 195,300 (95% UI: 159,519 to 236,176) deaths due to atrial fibrillation in 2015. This death rate was almost 5× higher, 116 per 100,000 (95% UI: 90 to 145 per 100,000) above 80 years of age than below that age. Atrial fibrillation mortality rates were highest in Western and Central Europe and high-income North America, and lowest in sub-Saharan Africa.

176) deaths due to atrial fibrillation in 2015. This death rate was almost 5× higher, 116 per 100,000 (95% UI: 90 to 145 per 100,000) above 80 years of age than below that age. Atrial fibrillation mortality rates were highest in Western and Central Europe and high-income North America, and lowest in sub-Saharan Africa. Rheumatic heart disease RHD was the fifth highest-ranked CVD cause of DALYs for South Asia and Central Asia and the sixth highest-ranked globally, although it was among the lowest-ranked causes in high-income regions. It accounted for substantially larger proportions of prevalent CVD and mortality in Oceania, sub-Saharan Africa, South Asia, the Caribbean, and Central Asia compared with other regions. Endocarditis Although a relatively less-common cause of CVD, endocarditis DALYs were greater in sub-Saharan Africa compared with other regions. There were an estimated 115,700 prevalent cases of endocarditis in 2015 (95% UI: 108,000 to 125,000 cases). There was an increase and then decline in the mortality rate among children, which peaked among children 28 days to 1 year of age (0.22 cases per 100,000; 95% UI: 0.18 to 0.26 per 100,000). The prevalence rose quickly above 60 years of age, reaching 28 per 100,000 (95% UI: 24 to 34 per 100,000) above 80 years of age. The highest prevalence of endocarditis was estimated for Oceania, 14 per 100,000 (95% UI: 13 to 15 per 100,000), which was twice the rate of the next-highest region, Southern Latin America. The lowest rates were estimated for East Asia and Andean Latin America.

,000 (95% UI: 24 to 34 per 100,000) above 80 years of age. The highest prevalence of endocarditis was estimated for Oceania, 14 per 100,000 (95% UI: 13 to 15 per 100,000), which was twice the rate of the next-highest region, Southern Latin America. The lowest rates were estimated for East Asia and Andean Latin America. There were 84,900 deaths due to endocarditis in 2015. Mortality rates were highest at the extremes of the age range: 7 per 100,000 (95% UI: 4 to 9 per 100,000) among those in the first week of life; 8 per 100,000 (95% UI: 7 to 9 per 100,000) among those 75 to 79 years of age; and 22 per 100,000 (95% UI 21 to 24 per 100,000) older than 80 years of age. Age-standardized mortality rates were greatest in Oceania, 6 per 100,000 (95% UI: 4 to 8 per 100,000), more than twice that of the next 2 regions, Central sub-Saharan Africa and Middle East/North Africa.

000) among those 75 to 79 years of age; and 22 per 100,000 (95% UI 21 to 24 per 100,000) older than 80 years of age. Age-standardized mortality rates were greatest in Oceania, 6 per 100,000 (95% UI: 4 to 8 per 100,000), more than twice that of the next 2 regions, Central sub-Saharan Africa and Middle East/North Africa. Peripheral arterial disease PAD was among the lowest-ranked CVD cause of DALYs in most world regions, but accounted for the largest proportion of cases of prevalent CVD in most world regions, reflecting high estimates of prevalence (based on surveys using ABI) and low estimates of symptomatic PAD (based on self-reported claudication among those with reduced ABI) and PAD death. There were an estimated 154.7 million (95% UI: 136.3 to 176.2 million) cases of PAD in 2015. The age-specific prevalence increased steadily after 40 years of age, reaching 23,913 per 100,000 (95% UI: 20,555 to 27,853 per 100,000) after 80 years of age. Age-standardized PAD prevalence was as high as 4,286 per 100,000 (95% UI: 3,773 to 4,897 per 100,000) in tropical Latin America, followed by Southern sub-Saharan Africa. The lowest prevalence of PAD, less than one-half of that rate, was estimated for high-income Asia Pacific and high-income North America.

years of age. Age-standardized PAD prevalence was as high as 4,286 per 100,000 (95% UI: 3,773 to 4,897 per 100,000) in tropical Latin America, followed by Southern sub-Saharan Africa. The lowest prevalence of PAD, less than one-half of that rate, was estimated for high-income Asia Pacific and high-income North America. There were 52,500 deaths (95% UI: 49,700 to 55,700 deaths) due to PAD in 2015. Death rates were highest above 80 years of age (24 per 100,000; 95% UI: 22 to 26 per 100,000). The Southern sub-Saharan region had the greatest age-standardized mortality rate, 3 per 100,000 (95% UI: 2 to 3 per 100,000), followed by Australasia, Eastern Europe, and the Caribbean, whereas the lowest mortality rates were estimated for Western sub-Saharan Africa, Andean Latin America, and Southeast Asia. Discussion CVD accounted for one-third of all deaths in 2015, and there were an estimated 422 million prevalent cases. The prevalence of CVD varied widely among countries, and when age-standardized, was declining in many high-income countries. Our analysis of mortality and sociodemographic change demonstrates a global disease gradient dominated by atherosclerotic vascular diseases, such as IHD and stroke, and with the most rapid decline occurring only at the highest levels of development. An alarming finding is that trends in CVD mortality have plateaued and are no longer declining for high-income regions. Overall, these results demonstrate the importance of increased investment in prevention and treatment of CVD for all regions of the world.

decline occurring only at the highest levels of development. An alarming finding is that trends in CVD mortality have plateaued and are no longer declining for high-income regions. Overall, these results demonstrate the importance of increased investment in prevention and treatment of CVD for all regions of the world. It is notable that very high age-standardized death rates due to CVD were not limited to any single region of the world, occurring among a subset of countries throughout Eastern Europe, Central Asia, the Middle East, South America, sub-Saharan Africa, and Oceania. There were also extremely large differences in estimated country-level age-standardized prevalence of CVD, ranging from <4,000 to >11,000 prevalent cases per 100,000 persons in 2015.

ing among a subset of countries throughout Eastern Europe, Central Asia, the Middle East, South America, sub-Saharan Africa, and Oceania. There were also extremely large differences in estimated country-level age-standardized prevalence of CVD, ranging from <4,000 to >11,000 prevalent cases per 100,000 persons in 2015. Sociodemographic transition Our regional analysis of SDI over time shows that Oceania, Central and Eastern Europe, Central Asia, and high-income North America have experienced levels of CVD far higher than would be expected given global patterns of disease. Over the same time period, tropical Latin America and the Middle East/North Africa have experienced steady declines during periods of continuous socioeconomic development. Western sub-Saharan Africa and other areas of Latin America have maintained levels well below regions with similar SDI levels. No decline at all is estimated for South and Southeast Asia. Recent continuous and rapid decline in CVD mortality remains a phenomenon limited to only a subset of high-SDI regions. Regional differences in CVD are likely a result of variation in exposure to modifiable risk factors, as well as access to effective health care interventions 24, 25, 26, 27, 28.

th and Southeast Asia. Recent continuous and rapid decline in CVD mortality remains a phenomenon limited to only a subset of high-SDI regions. Regional differences in CVD are likely a result of variation in exposure to modifiable risk factors, as well as access to effective health care interventions 24, 25, 26, 27, 28. Changes in the decline of CVD mortality Of particular concern is that CVD age-standardized mortality shows less decline in the past 5 years than over the past 25 years. This trend, which is most obvious for IHD and aortic aneurysm, is observed not only in high-income countries, but also in Central Latin America for men. Regions with very high rates of CVD that have declined rapidly, such as Central Asia and Eastern Europe, also see moderation in that decline. Our use of the most recently available mortality data (through 2013 in many high-income countries) may explain why our findings differ from a recent analysis of CVD trends (29). Although an explanation of stagnation in declining CVD mortality is beyond the scope of this analysis, several possibilities can be considered. Rising rates of obesity may be increasing CVD risk over a short period of time (30). Interventions that reduce CVD mortality rates may have maximally diffused to the population able to access them, whereas interventions to address obesity are more challenging to implement. Some CVD risk factors, in particular air pollution or changes in average temperature, may account for larger increases of CVD mortality than previously suspected 31, 32. Improving methods for estimating the most likely future trajectories for CVD is an important area for further research.

more challenging to implement. Some CVD risk factors, in particular air pollution or changes in average temperature, may account for larger increases of CVD mortality than previously suspected 31, 32. Improving methods for estimating the most likely future trajectories for CVD is an important area for further research. Discontinuities in cardiovascular mortality Any broad conclusions on the global influence of socioeconomic development must be tempered by the fact that rapid increases in CVD burden have occurred due to a diverse and evolving set of health risks. Economic crises in Eastern Europe in the 1990s and the resulting rapid changes in CVD mortality rates have only been partially described and require more investigation. There is strong evidence that the hazardous use of alcohol was a major contributor to this pattern 33, 34. South Africa experienced a similar mortality crisis, peaking around the year 2000 due to “colliding” epidemics of human immunodeficiency virus/acquired immunodeficiency syndrome and noncommunicable diseases (35). These striking examples of abrupt discontinuities in the number of deaths due to CVD (“fatal discontinuities”) demonstrate how political and social unrest may lead not only to outbreaks of communicable disease, but also to dramatic changes in cardiovascular health. Further attention is needed to understand how CVD is influenced by and would be best treated during rapid changes in material living conditions caused by war, natural disasters, and mass migration from lower to higher SDI regions. Patterns of CVD during times of large-scale migration require particular attention, due to the potential unfavorable effects of immigration on atherosclerotic risk factors (when moving to higher-SDI regions) as well as in reducing access to basic health care (36).

s, and mass migration from lower to higher SDI regions. Patterns of CVD during times of large-scale migration require particular attention, due to the potential unfavorable effects of immigration on atherosclerotic risk factors (when moving to higher-SDI regions) as well as in reducing access to basic health care (36). Cause-specific variation in the epidemiological transition IHD and stroke mortality rates increase and then fall as SDI increases, supporting the theory of an epidemiological transition for CVD 37, 38. The initial increases in IHD and stroke mortality at low levels of SDI reflect the particularly high mortality rates estimated for Oceania, tropical Latin America, and the Middle East/North Africa region in the 1990s. Almost all other CVD burdens decrease continuously at higher levels of SDI. Declines in IHD and stroke mortality lag for men compared with women as SDI increases. Differences in CVD risk exposures may explain this differential by sex, such as exposure to tobacco smoking (39). The decline in stroke mortality among women at lower levels of SDI ranging from 0.3 to 0.6, driven by trends in South Asia, North Africa/Middle East, and Oceania, suggests that countries need not achieve the highest levels of development to reduce the burden of stroke. Results of the Prospective Urban Rural Epidemiological study show that women have lower use of the most common medications used to treat CVD compared with men, suggesting that this differential outcome is not the result of higher use of secondary medications among women (40).

development to reduce the burden of stroke. Results of the Prospective Urban Rural Epidemiological study show that women have lower use of the most common medications used to treat CVD compared with men, suggesting that this differential outcome is not the result of higher use of secondary medications among women (40). Peripheral arterial disease We estimated that PAD is the most prevalent cardiovascular condition globally, although low estimated rates of claudication and mortality made it a minor contributor to DALYs. The high prevalence of PAD in comparison to IHD is a notable finding that may reflect the ease of its diagnosis using ABI, compared with more complex diagnostic testing required for IHD. Further attention should be paid to the use of ABI or palpation of foot pulses as a screening tool for overall vascular risk in low-income settings 41, 42. There is some evidence that even those with asymptomatic PAD would benefit significantly from inexpensive medications, such as an angiotensin-converting enzyme inhibitor or antiplatelet agents 43, 44.

the use of ABI or palpation of foot pulses as a screening tool for overall vascular risk in low-income settings 41, 42. There is some evidence that even those with asymptomatic PAD would benefit significantly from inexpensive medications, such as an angiotensin-converting enzyme inhibitor or antiplatelet agents 43, 44. Stroke We estimated that hemorrhagic and other stroke accounted for more CVD DALYs than ischemic stroke globally and for almost all regions. Although the predominance of hemorrhagic and other strokes over ischemic subtypes may reflect variable use of unspecified stroke ICD-10 codes in some regions, this ranking also results from the predominance of hemorrhagic and other stroke deaths over ischemic stroke deaths for all ages <75 years, even among high-income countries. Both stroke subtypes share a common set of modifiable risk factors, suggesting that public health and health system investments in their shared risks could reduce overall burden due to all stroke (45).

orrhagic and other stroke deaths over ischemic stroke deaths for all ages <75 years, even among high-income countries. Both stroke subtypes share a common set of modifiable risk factors, suggesting that public health and health system investments in their shared risks could reduce overall burden due to all stroke (45). The GBD 2015 study offers improved access to its data sources, research methods, and results. Citations for all data used in the study can be found via the publicly available Global Health Data Exchange (18). The new GATHER (Guidelines for Accurate and Transparent Health Estimates Reporting) guidelines for reporting estimates of public health have been fully implemented (Online Appendix). Intermediate model results from DisMod and all final results have been made publicly available using web-based data visualization tools (46). Finally, software code and data tables have been posted online and are freely available for download (47).

timates of public health have been fully implemented (Online Appendix). Intermediate model results from DisMod and all final results have been made publicly available using web-based data visualization tools (46). Finally, software code and data tables have been posted online and are freely available for download (47). Study limitations CVD estimates of the GBD 2015 study have several important limitations. Misclassification bias due to miscoding of death certificates remains likely 48, 49. The GBD study takes several steps to improve the reliability and comparability of vital registration data, including redistribution of garbage codes, but some systematic bias due to regional patterns in the use of diagnosis codes may remain. For example, the relatively large number of deaths coded to cardiomyopathy in Balkan countries may lead to an underestimate of the true number of IHD deaths. Many stroke deaths are coded to non–subtype-specific stroke codes, which, when combined with lack of access to computed tomography scanners to aid in the acute management of stroke, adds additional uncertainty to the breakdown of stroke into ischemic and hemorrhagic subtypes. Uncertainty regarding stroke subtype is of particular concern in South Asia, where no subtype-specific mortality data were found, and the global proportion of ischemic and hemorrhagic stroke was used instead. The rapid rise, globally, in death certificates coded to atrial fibrillation is implausible in the face of cohort data showing stable age-standardized prevalence and case-fatality rates, and led to our adopting a new natural history method for estimating its global burden that incorporates this cohort data. Estimation of the burden of atrial fibrillation is limited by variable coding of this condition as an underlying cause of death, especially in the setting of stroke, and very little data on its prevalence in low- and middle-income countries. Similar increases in PAD may also reflect changes in coding practice among physicians.

on of the burden of atrial fibrillation is limited by variable coding of this condition as an underlying cause of death, especially in the setting of stroke, and very little data on its prevalence in low- and middle-income countries. Similar increases in PAD may also reflect changes in coding practice among physicians. Health data on CVD remains extremely limited for some regions of the world, such as India and sub-Saharan Africa. In sub-Saharan Africa, a much larger proportion of prevalent cases of CVD were estimated to come from other cardiovascular and circulatory diseases. This category of other causes includes pericardial disease, nonrheumatic valvular diseases, unspecified lymphatic and vascular conditions, and pulmonary embolism, as well as deaths coded to left and right heart failure and other nonunderlying or nonspecific causes of death. Lack of data is 1 reason why no significant change in prevalence could be detected in many regions, for example, Western sub-Saharan Africa. There is only limited data available in some countries with both high all-cause mortality and high estimated CVD mortality. For example, estimates of CVD mortality in Afghanistan are based on a single verbal autopsy survey performed in 2008 that reported CVD as the cause of 14.4% of all male deaths and 20.7% of female deaths.

s only limited data available in some countries with both high all-cause mortality and high estimated CVD mortality. For example, estimates of CVD mortality in Afghanistan are based on a single verbal autopsy survey performed in 2008 that reported CVD as the cause of 14.4% of all male deaths and 20.7% of female deaths. Data on the severity distribution of CVD is also particularly limited, and future iterations of GBD will benefit from additional sources from more regions. Limited data presents a special challenge in India, where GBD 2015 relied on 3 major sample vital registration sources for estimating mortality: Medical Certification of Causes of Death, primarily an urban-area program; the Survey of Causes of Death and Survey of Causes of Death-Rural; and the Sample Registration System. The latter source, in particular, has not been released in detail for specific causes of CVD. Expanded access to this kind of data in India and other countries that have invested in sample death registration systems could significantly improve our ability to forecast trends in atherosclerotic vascular diseases. The GBD study has reported subnational estimates for a growing number of countries; however, the current report is limited to country-level estimates 50, 51, 52. There is substantial small-area variation in CVD burden within countries, and our national estimates represent only an average level for an entire country. Even these national estimates are important starting points for improving evidence available to policymakers.

port is limited to country-level estimates 50, 51, 52. There is substantial small-area variation in CVD burden within countries, and our national estimates represent only an average level for an entire country. Even these national estimates are important starting points for improving evidence available to policymakers. The GBD study accounts for comorbidity using a simulation method that assumes an independent probability of having any disease state. A sensitivity analysis found that taking dependent and independent comorbidity into account changed overall estimated YLDs by a small amount that increased with age, ranging from 0.6% to 3.4%. Because CVD is more common at older ages, the assumption of independence may have a larger effect on this group of causes. Unfortunately, data on the full correlation structure of prevalent CVD conditions remains limited. GBD includes an estimate of measurement error, reported as a 95% UI, for each result. Our ability to detect significant trends over time is limited for those regions where UIs are wide, as seen on our global map of change in prevalence over time. Some countries may have experienced a rise or fall in CVD burden that we cannot detect because of limited data.

t error, reported as a 95% UI, for each result. Our ability to detect significant trends over time is limited for those regions where UIs are wide, as seen on our global map of change in prevalence over time. Some countries may have experienced a rise or fall in CVD burden that we cannot detect because of limited data. Although inclusion of measurement error is a strength of the GBD study, nonsampling error has not been quantified. GBD uses a wide range of validation methods, but relies on in-sample and out-of-sample validity testing to guide model selection. Additional sources of error in the GBD study may include regional patterns of clinical diagnosis, death code redistribution, selection of data sources, covariate selection for selected models, and measurement of the SDI. For example, measures of wealth, rather than income per capita, could potentially capture additional aspects of the epidemiological transition for some countries. Limited data on PAD with increased ABI due to noncompressible arteries may lead to underestimates of its true burden. The burden due to Chagas cardiomyopathy has not been included, but is estimated by the GBD study as a sequelae of Chagas disease. Chronic kidney disease and congenital heart disease are also estimated by the GBD study and have been reported separately. Overall, the GBD study is most likely to underestimate uncertainty for those geographies where few data sources are available.

ed, but is estimated by the GBD study as a sequelae of Chagas disease. Chronic kidney disease and congenital heart disease are also estimated by the GBD study and have been reported separately. Overall, the GBD study is most likely to underestimate uncertainty for those geographies where few data sources are available. Conclusions CVD remain a major cause of premature death and chronic disability for all regions of the world. IHD and stroke account for the majority of health lost to CVD. Sociodemographic change over the past 25 years has been associated with dramatic declines in age-standardized rates of CVD mortality in regions with very high SDI, but only a gradual decrease or no change at all in most regions. It is concerning that large reductions in atherosclerotic vascular disease mortality, a crowning achievement for public health, are no longer apparent in many world regions, despite impressive advances in technical capacity for preventing and treating CVD.

ut only a gradual decrease or no change at all in most regions. It is concerning that large reductions in atherosclerotic vascular disease mortality, a crowning achievement for public health, are no longer apparent in many world regions, despite impressive advances in technical capacity for preventing and treating CVD. The GBD study offers a unique platform for tracking rapidly evolving patterns in CVD epidemiology and their relationship to demographic and socioeconomic change. Specific causes of CVD can be examined within the broader context of global health. Countries should consider further investment in CVD surveillance and population-based registries to benchmark their efforts toward reducing the burden of CVD. Future updates of the GBD study can be used to guide policymakers who are focused on reducing the overall burden of noncommunicable disease and achieving specific global health targets.Perspectives COMPETENCY IN SYSTEMS-BASED PRACTICE: Ischemic heart disease and stroke account for most of the global burden of cardiovascular disease, but there is wide regional variation. Improvements in mortality related to cardiovascular disease appear to be slowing for regions of the world characterized by a high sociodemographic index based on per capita income, educational attainment, and fertility. TRANSLATIONAL OUTLOOK: Development of evidence-based public health policies requires incorporation of comparative health status and burden of disease estimates between and within countries. Appendix Online Tables 1–4 and Online Figures 1–9 Online Data

The GBD study offers a unique platform for tracking rapidly evolving patterns in CVD epidemiology and their relationship to demographic and socioeconomic change. Specific causes of CVD can be examined within the broader context of global health. Countries should consider further investment in CVD surveillance and population-based registries to benchmark their efforts toward reducing the burden of CVD. Future updates of the GBD study can be used to guide policymakers who are focused on reducing the overall burden of noncommunicable disease and achieving specific global health targets.Perspectives COMPETENCY IN SYSTEMS-BASED PRACTICE: Ischemic heart disease and stroke account for most of the global burden of cardiovascular disease, but there is wide regional variation. Improvements in mortality related to cardiovascular disease appear to be slowing for regions of the world characterized by a high sociodemographic index based on per capita income, educational attainment, and fertility. TRANSLATIONAL OUTLOOK: Development of evidence-based public health policies requires incorporation of comparative health status and burden of disease estimates between and within countries. Appendix Online Tables 1–4 and Online Figures 1–9 Online Data Acknowledgments The authors express their appreciation to Ben Zipkin and Minh Nguyen for assistance in producing tables and figures; and to Rachel Woodbrook and Adrienne Chew for their assistance in writing and technical editing of the manuscript.

TRANSLATIONAL OUTLOOK: Development of evidence-based public health policies requires incorporation of comparative health status and burden of disease estimates between and within countries. Appendix Online Tables 1–4 and Online Figures 1–9 Online Data Acknowledgments The authors express their appreciation to Ben Zipkin and Minh Nguyen for assistance in producing tables and figures; and to Rachel Woodbrook and Adrienne Chew for their assistance in writing and technical editing of the manuscript. The Institute for Health Metrics and Evaluation received funding from the Bill & Melinda Gates Foundation. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Dr. Ärnlöv has received lecturing fees from AstraZeneca. Dr. Fowkes has served on the advisory boards of AstraZeneca, Merck, and Bayer. Dr. Hankey has received honoraria from Bayer for lecturing at sponsored scientific symposia about stroke prevention in atrial fibrillation. Dr. Khubchandani has received research funding from Merck Laboratories. Dr. Lotufo has received honoraria from Amgen Brazil and AbbVie Brazil. Dr. Rahimi’s research is supported by grants from the National Institute for Health Research Oxford Biomedical Research Centre, National Institute for Health Research Career Development Fellowship, and Oxford Martin School. Dr. Schutte is a speaker to general practitioners and cardiologists on hypertension guidelines (funded by Boehringer Ingelheim), to cardiologists on health behaviors and cardiovascular diseases (funded by Novartis), and to nurses and pharmacists on arterial stiffness and blood pressure (funded by Omron Healthcare and I.E.M.). Dr. Watkins is funded by Medtronic Foundation through support to RhEACH and RHD Action, and by the Disease Control Priorities Network grant from the Bill and Melinda Gates Foundation. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

lthcare and I.E.M.). Dr. Watkins is funded by Medtronic Foundation through support to RhEACH and RHD Action, and by the Disease Control Priorities Network grant from the Bill and Melinda Gates Foundation. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript's audio summary by JACC Editor-in-Chief Dr. Valentin Fuster. For an expanded Methods section and supplemental figures and tables, please see the online version of this article.

both recent MI (SUVmax 1.60 [IQR: 1.45 to 2.11] vs. 1.33 [IQR: 1.25 to 1.52]; p = 0.03; TBRmax 2.33 [IQR: 1.55 to 2.71] vs. 1.80 [IQR: 1.32 to 2.22]; p = 0.03) and old MI (SUVmax 2.22 [IQR: 2.03 to 2.50] vs. 1.78 [IQR: 1.63 to 2.13]; p < 0.0001; TBRmax 2.79 [IQR: 2.47 to 3.23] vs. 1.89 [IQR: 1.52 to 2.36]; p < 0.0001). Unlike 68Ga-DOTATATE, which exhibited very low background myocardial binding in all patients, avid myocardial 18F-FDG uptake (basal inferoseptum SUVmax >5) rendered 5 (42%) scans uninterpretable despite 6-h pre-scan fasting. In the readable scans, the 2 tracers showed reasonable agreement in the myocardium (r = 0.38, 95% confidence interval [CI]: 0.20 to 0.53; p < 0.0001). Despite high liver and spleen 68Ga-DOTATATE activity, focal myocardial signals were clearly distinguishable in all 5 patients with inferior infarcts. Bone marrow 68Ga-DOTATATE signals were highly correlated with both infarct-related myocardial inflammation detected by 68Ga-DOTATATE (r = 0.83 [95% CI: 0.48 to 0.95]; p = 0.001), and metabolic bone marrow activity measured by 18F-FDG (r = 0.64 [95% CI: 0.08 to 0.89]; p = 0.03).

Unlike 68Ga-DOTATATE, which exhibited very low background myocardial binding in all patients, avid myocardial 18F-FDG uptake (basal inferoseptum SUVmax >5) rendered 5 (42%) scans uninterpretable despite 6-h pre-scan fasting. In the readable scans, the 2 tracers showed reasonable agreement in the myocardium (r = 0.38, 95% confidence interval [CI]: 0.20 to 0.53; p < 0.0001). Despite high liver and spleen 68Ga-DOTATATE activity, focal myocardial signals were clearly distinguishable in all 5 patients with inferior infarcts. Bone marrow 68Ga-DOTATATE signals were highly correlated with both infarct-related myocardial inflammation detected by 68Ga-DOTATATE (r = 0.83 [95% CI: 0.48 to 0.95]; p = 0.001), and metabolic bone marrow activity measured by 18F-FDG (r = 0.64 [95% CI: 0.08 to 0.89]; p = 0.03). We found that 68Ga-DOTATATE identified active inflammation in recently infarcted myocardium, as well as old ischemic injury. Our observations agree with existing clinical data (3), but contradict findings in mice (4). 68Ga-DOTATATE binding in chronically damaged myocardium, particularly at the infarct border (Figure 1), likely reflects residual macrophage-driven inflammation; however, histological validation is needed. While tracer binding to myocytes and/or fibroblasts are possible alternative explanations, transcriptomic data from infarcted mouse hearts (5) indicates that SSTR2 is not expressed in these cell types.Figure 1 Post-Infarction Myocardial Inflammation Identified by 68Ga-DOTATATE PET

er, histological validation is needed. While tracer binding to myocytes and/or fibroblasts are possible alternative explanations, transcriptomic data from infarcted mouse hearts (5) indicates that SSTR2 is not expressed in these cell types.Figure 1 Post-Infarction Myocardial Inflammation Identified by 68Ga-DOTATATE PET (A)68Ga-DOTATATE positron emission tomography (PET)-computed tomography image (scale bar: standardized uptake values) demonstrating residual inflammation (arrow) in (B) partially viable myocardium with subendocardial infarct (dashed arrow), bordering full-thickness scarring (asterisk) confirmed by late gadolinium enhancement magnetic resonance imaging, 4 years after a left anterior descending artery myocardial infarction. 18F-FDG positron emission tomography imaging reproduced a near-identical pattern of abnormal myocardial tracer uptake. Stress magnetic resonance imaging was negative for ischemia. Residual myocardial inflammation detected by 68Ga-DOTATATE could represent an important prognostic biomarker to study disease mechanisms and test novel therapies for the inflamed, failing heart.

Myocardial infarction (MI) healing occurs in 2 phases: first an inflammatory phase, where clearance of necrotic debris occurs, followed by a reparative phase characterized by angiogenesis, granulation tissue formation, and attempts to repair the extracellular matrix. While efficient healing relies on coordinated mobilization of monocytes to the damaged myocardium, with resolution of the acute inflammatory response by ∼10 to 14 days, excessive inflammation impairs myocardial salvage and promotes adverse cardiac remodeling. In ischemic heart failure, pro-inflammatory macrophages persist long after the formation of healed scar in remote and border zones of the infarcted, remodeled heart because of maladaptive changes in the mononuclear phagocytic network and spleen (1). An accurate means of diagnosing harmful inflammation after an MI is urgently needed. We previously demonstrated that 68Ga-DOTATATE, a somatostatin receptor subtype-2 positron emission tomography (PET) ligand, could identify pro-inflammatory macrophages within atherosclerotic plaques (2). Here, in this substudy of our original prospective observational study, we examined whether 68Ga-DOTATATE could reveal residual post-infarction myocardial inflammation. Patients with an MI within 3 months treated by percutaneous coronary intervention (“recent MI,” n = 6), and patients with a past history of MI and echocardiography data available from after their event (“old MI,” n = 6), were included. Patients with equivocal culprit arteries, and those managed medically or with coronary artery bypass grafting surgery, were excluded.

rcutaneous coronary intervention (“recent MI,” n = 6), and patients with a past history of MI and echocardiography data available from after their event (“old MI,” n = 6), were included. Patients with equivocal culprit arteries, and those managed medically or with coronary artery bypass grafting surgery, were excluded. ECG-gated PET imaging was performed as previously described (2). Maximum standardized uptake values (SUVmax) and tissue-to-blood ratios (TBRmax), normalized for blood pool activity in the superior vena cava, were derived blinded to clinical details in each of the 16 myocardial segments. Myocardial 68Ga-DOTATATE and 18F-FDG PET signals were compared: 1) within infarcted and noninfarcted segments; 2) to each other; and 3) to tracer activity in the thoracic vertebral bone marrow as an experimental marker of systemic inflammation, using standard nonparametric statistical tests (all data median [interquartile range (IQR)] unless stated). Recently infarcted myocardial segments were defined by clinically adjudicated (treated) culprit artery territories, with individual anatomical variation verified by angiography. In patients with old MI, infarcted myocardium was determined by echocardiographic wall motion abnormalities (hypokinesia/akinesia), assessed independently of the study and prior to enrollment.

d by clinically adjudicated (treated) culprit artery territories, with individual anatomical variation verified by angiography. In patients with old MI, infarcted myocardium was determined by echocardiographic wall motion abnormalities (hypokinesia/akinesia), assessed independently of the study and prior to enrollment. Demographics were similar for recent MI (age 74 years [IQR: 64 to 78 years], 83% male) and old MI (age 59 years [IQR: 56 to 72 years], all male) patients. There were 3 ST-segment elevation MIs, which were all old MIs. PET imaging occurred 35 days (range 21 to 80 days) after recent MIs, and 7 years (range 1.8 to 22 years) after old MIs, with 2 days (range 1 to 21 days) in between 68Ga-DOTATATE and 18F-FDG scans. 68Ga-DOTATATE signals were higher in infarcted compared with noninfarcted myocardium in patients with both recent MI (SUVmax 1.60 [IQR: 1.45 to 2.11] vs. 1.33 [IQR: 1.25 to 1.52]; p = 0.03; TBRmax 2.33 [IQR: 1.55 to 2.71] vs. 1.80 [IQR: 1.32 to 2.22]; p = 0.03) and old MI (SUVmax 2.22 [IQR: 2.03 to 2.50] vs. 1.78 [IQR: 1.63 to 2.13]; p < 0.0001; TBRmax 2.79 [IQR: 2.47 to 3.23] vs. 1.89 [IQR: 1.52 to 2.36]; p < 0.0001).

(A)68Ga-DOTATATE positron emission tomography (PET)-computed tomography image (scale bar: standardized uptake values) demonstrating residual inflammation (arrow) in (B) partially viable myocardium with subendocardial infarct (dashed arrow), bordering full-thickness scarring (asterisk) confirmed by late gadolinium enhancement magnetic resonance imaging, 4 years after a left anterior descending artery myocardial infarction. 18F-FDG positron emission tomography imaging reproduced a near-identical pattern of abnormal myocardial tracer uptake. Stress magnetic resonance imaging was negative for ischemia. Residual myocardial inflammation detected by 68Ga-DOTATATE could represent an important prognostic biomarker to study disease mechanisms and test novel therapies for the inflamed, failing heart. Please note: Dr. Tarkin has been supported by the Wellcome Trust (104492/Z/14/Z, 211100/Z/18/Z) and the National Institute for Health Research. Dr. Calcagno has been supported by the National Institutes of Health (NIH) (P01 HL131478, R01 HL071021, R01HL135878) and the American Heart Association (16SDG27250090). Dr. Evans has been supported by the National Institute for Health Research and Dunhill Trust (RTF44/0114). Dr. Chowdhury has been supported by the British Heart Foundation (FS/16/29/31957). Dr. Fayad has been supported by the NIH (P01 HL131478, R01 HL071021, R01 HL128056, R01HL135878, NBIB R01 EB009638) and the American Heart Association (14SFRN20780005). Dr. Rudd has been supported by the Higher Education Funding Council for England, the National Institute for Health Research Cambridge Biomedical Research Centre, the Wellcome Trust, the British Heart Foundation, and the Cambridge Center for Mathematics in Healthcare. This research was conducted in accordance with the study protocol approved by the local research ethics committee (14/EE/0019), Good Clinical Practice, and the Declaration of Helsinki. (Vascular Inflammation Imaging Using Somatostatin Receptor Positron Emission Tomography [VISION]; NCT02021188). Damini Dey, PhD, served as Guest Associate Editor for this letter.

Adamson PD, Anderson JA, Brook RD, Calverley PMA, Celli BR, Cowans NJ, Crim C, Dixon IJ, Martinez FJ, Newby DE, Vestbo J, Yates JC, Mills NL Cardiac Troponin I and Cardiovascular Risk in Patients With Chronic Obstructive Pulmonary Disease J Am Coll Cardiol 2018;72:1126–37. On page 1129, the n values in the table column headings were incorrect. The corrected Table 1 is shown below:Table 1 Patient Characteristics in the SUMMIT Study Population, the Biomarker Substudy Population, and Split by Cardiac Troponin I Quintile

Cardiac Troponin I and Cardiovascular Risk in Patients With Chronic Obstructive Pulmonary Disease J Am Coll Cardiol 2018;72:1126–37. On page 1129, the n values in the table column headings were incorrect. The corrected Table 1 is shown below:Table 1 Patient Characteristics in the SUMMIT Study Population, the Biomarker Substudy Population, and Split by Cardiac Troponin I Quintile Troponin Quintile 1 (<2.3 ng/l) (n = 307) Troponin Quintile 2 (≥2.3 to <3.4 ng/l) (n = 325) Troponin Quintile 3 (≥3.4 to <4.8 ng/l) (n = 319) Troponin Quintile 4 (≥4.8 to <7.7 ng/l) (n = 330) Troponin Quintile 5 (≥7.7 ng/l) (n = 318) Biomarker Substudy∗ (n = 1,673) SUMMIT ITT-E Population (n = 16,485) Median troponin 1.7 2.8 4.0 5.8 12.0 4.0 - Age, yrs 63 ± 8 65 ± 8 67 ± 8 68 ± 7 68 ± 7 66 ± 8 65 ± 8 Female 172 (56) 153 (47) 107 (34) 98 (30) 80 (25) 635 (38) 4,196 (25) BMI, kg/m2 30 ± 6 31 ± 7 31 ± 6 31 ± 7 31 ± 7 31 ± 7 28 ± 6 Systolic blood pressure, mm Hg 128 ± 14 129 ± 16 132 ± 16 131 ± 16 134 ± 19 131 ± 16 135 ± 15 Heart rate, beats/min 75 ± 10 74 ± 11 73 ± 11 73 ± 11 73 ± 11 73 ± 11 76 ± 10 Estimated GFR, ml/min/1.73 m2 101.2 ± 33.6 100.5 ± 37.1 100.5 ± 38.4 95.2 ± 36.1 89.3 ± 34.9 97.7 ± 36.7 97.3 ± 36.6 CRP, mg/l 5.3 ± 6.9 6.5 ± 8.3 5.6 ± 7.0 6.5 ± 10.3 6.8 ± 8.7 6.2 ± 8.3 6.2 ± 8.3 Past medical history Prior myocardial infarction or coronary revascularization 71 (23) 89 (27) 112 (35) 145 (44) 162 (51) 601 (36) 3,436 (21) Coronary artery disease 113 (37) 132 (41) 148 (46) 186 (56) 199 (63) 818 (49) 8,379 (51) Congestive heart failure 15 (5) 11 (3) 21 (7) 34 (10) 61 (19) 146 (9) 3,456 (21) Hypercholesterolemia 243 (79) 280 (86) 283 (89) 295 (89) 292 (92) 1458 (87) 11,518 (70) Hypertension 258 (84) 285 (88) 300 (94) 307 (93) 302 (95) 1519 (91) 14,851 (90) Diabetes mellitus 108 (35) 116 (36) 116 (36) 138 (42) 140 (44) 642 (38) 4,997 (30) Family history of CVD 128 (42) 128 (39) 116 (36) 146 (44) 145 (46) 691 (41) 3,429 (21) Respiratory history Former smoker 141 (46) 152 (47) 170 (53) 178 (54) 169 (53) 845 (51) 8,807 (53) Post-bronchodilator FEV1, l 1.7 ± 0.4 1.7 ± 0.4 1.7 ± 0.4 1.7 ± 0.4 1.7 ± 0.4 1.7 ± 0.4 1.7 ± 0.4 Predicted post-bronchodilator FEV1, % of predicted 59.7 ± 6.9 59.4 ± 6.7 59.2 ± 6.9 59.5 ± 6.6 59.3 ± 7.0 59.4 ± 6.8 59.7 ± 6.1 Exacerbations in 12 months before study 0 220 (72) 233 (72) 228 (71) 249 (75) 236 (74) 1,215 (73) 10,021 (61) 1 55 (18) 57 (18) 58 (18) 47 (14) 55 (17) 290 (17) 4,020 (24) 2+ 32 (10) 35 (11) 33 (10) 34 (10) 27 (8) 168 (10) 2,444 (15) Concomitant cardiovascular therapy Antiplatelet therapy 176 (57) 196 (60) 197 (62) 231

xacerbations in 12 months before study 0 220 (72) 233 (72) 228 (71) 249 (75) 236 (74) 1,215 (73) 10,021 (61) 1 55 (18) 57 (18) 58 (18) 47 (14) 55 (17) 290 (17) 4,020 (24) 2+ 32 (10) 35 (11) 33 (10) 34 (10) 27 (8) 168 (10) 2,444 (15) Concomitant cardiovascular therapy Antiplatelet therapy 176 (57) 196 (60) 197 (62) 231 (70) 238 (75) 1,081 (65) 8,517 (52) Statin therapy 207 (67) 251 (77) 238 (75) 269 (82) 245 (77) 1,263 (75) 10,721 (65) Antiplatelet and statin therapy 137 (45) 167 (51) 157 (49) 198 (60) 191 (60) 886 (53) 6151 (37) Treatment allocation Placebo 83 (27) 87 (27) 81 (25) 78 (24) 92 (29) 439 (26) 4,111 (25) Fluticasone furoate 69 (22) 87 (27) 83 (26) 76 (23) 74 (23) 415 (25) 4,135 (25) Vilanterol 87 (28) 84 (26) 70 (22) 80 (24) 77 (24) 416 (25) 4,118 (25) Combination therapy 68 (22) 67 (21) 85 (27) 96 (29) 75 (24) 403 (24) 4,121 (25) Values are mean ± SD or n (%). BMI = body mass index; CVD = cardiovascular disease; FEV1 = forced expiratory volume in 1 s; GFR = glomerular filtration rate; ITT-E = intention-to-treat efficacy. ∗ Of the 1,673 patients in the biomarker population, 74 did not have baseline cardiac troponin I measured and are therefore not included in the cardiac troponin I quintiles and analyses. The authors apologize for these errors. The online version of the article has been corrected to reflect these changes.

Shah AD, Denaxas S, Nicholas O, Hingorani AD, Hemingway H Neutrophil Counts and Initial Presentation of 12 Cardiovascular Diseases: A CALIBER Cohort Study J Am Coll Cardiol 2017;69:1160–9. In Figure 2, in the Transient Ischemic Attack panel, “0.32 (1.16-1.49)” should have read “1.32 (1.16-1.49).” The corrected Figure 2 is printed below.Figure 2 Association of Neutrophil Count With Initial CVD Presentation Neutrophil count categories influenced cause-specific adjusted hazard ratios for cardiovascular presentations among people without prior cardiovascular disease (CVD). Hazard ratios were adjusted for age, sex, deprivation, ethnicity, smoking, diabetes, systolic blood pressure (SBP), blood pressure medication, body mass index (BMI), total cholesterol, high-density lipoprotein cholesterol (HDL-C), statin use, estimated glomerular filtration rate (eGFR), atrial fibrillation (AF), autoimmune conditions, inflammatory bowel disease (IBD), chronic obstructive pulmonary disease (COPD), cancer, and acute conditions at the time of blood testing. Shaded = normal range. *p < 0.05; **p < 0.0036; ***p < 0.0001. CI = confidence interval; other abbreviations as in Figure 1. The authors apologize for these errors. The online version of the article has been corrected to reflect these changes.

Electronic cigarettes or E-cigarettes (EC) are gaining popularity worldwide as an alternative to smoking tobacco cigarettes (TC) with a 55% increase in users between 2013 and 2015 with growth in the United Kingdom occurring fastest in Europe (1). The prevalence of EC use in the United Kingdom and United States is around 6% (2), and 51% of users did so because they believed it to be less harmful than regular cigarettes (3). Observational data in the 2014 and 2016 U.S. National Health Interview Surveys revealed that although the risk of myocardial infarction remains higher with TC (odds ratio [OR]: 2.72; 95% confidence interval [CI]: 2.29 to 3.24), daily EC use was also associated with an increased myocardial infarction risk (OR: 1.79; 95% CI: 1.20 to 2.66) (4). Despite this, there remains little good quality evidence on the short- and long-term safety of these devices. Furthermore, conflicting advice from various public health bodies worldwide on the use of these devices has resulted in lack of clarity for policymakers as well as the public at large (5,6).

1.20 to 2.66) (4). Despite this, there remains little good quality evidence on the short- and long-term safety of these devices. Furthermore, conflicting advice from various public health bodies worldwide on the use of these devices has resulted in lack of clarity for policymakers as well as the public at large (5,6). TC contain >7,000 chemicals, including exposing smokers to high levels of nicotine, carbon monoxide (CO), acrolein, and pro-oxidant compounds. Data from chemical analysis and toxicology studies suggest that exposure to toxic chemicals from EC is lower compared with exposure from TC (7,8). However, other studies have shown that there remains the presence of potentially harmful tobacco-specific alkaloids such as anabasine, myosmine, and β-nicotyrine in EC liquid cartridge samples tested (9). The impact of nicotine on vascular health is also unclear. Nicotine may accelerate the atherogenic process by binding to high-affinity nicotinic acetylcholine receptor cell surface receptors (10). However, longer-term nicotine use appears not to accelerate atherogenesis but may contribute to acute cardiovascular events in the presence of cardiovascular (CV) disease (11). The early vascular impact of switching from TC to EC--nicotine versus EC- nicotine-free is not known.

ine receptor cell surface receptors (10). However, longer-term nicotine use appears not to accelerate atherogenesis but may contribute to acute cardiovascular events in the presence of cardiovascular (CV) disease (11). The early vascular impact of switching from TC to EC--nicotine versus EC- nicotine-free is not known. Endothelial dysfunction is the earliest detectable change in vascular health, and, importantly, it has consistently been shown to be associated with CV risk and long-term outcomes (12,13). We measured endothelial function using flow-mediated dilatation (FMD) and arterial stiffness by pulse wave velocity (PWV), 2 validated and independent predictors of CV risk above and beyond traditional risk factors (14,15). We conducted the current trial to address specific questions on the early CV effects of switching from TC to EC and the impact of nicotine itself on any early vascular changes that might be seen.

se wave velocity (PWV), 2 validated and independent predictors of CV risk above and beyond traditional risk factors (14,15). We conducted the current trial to address specific questions on the early CV effects of switching from TC to EC and the impact of nicotine itself on any early vascular changes that might be seen. Methods The VESUVIUS (Vascular Effects of Regular Cigarettes Versus Electronic Cigarette Use) trial (NCT02878421) was a prospective, randomized controlled trial with a parallel, nonrandomized preference cohort and blinded endpoint of smokers ≥18 years of age who had smoked ≥15 cigarettes/day for at least 2 years; were free from established CV disease, diabetes, and chronic kidney disease; and were not on medication for those conditions. The trial was conducted between August 2016 and July 2018 in a single tertiary research center. Participants were recruited from local advertisements, smoking cessation databases, and visits to local businesses, as well as via the Scottish Primary Care Research Network. Consented participants who were willing to quit smoking were randomized to one of the EC arms in a 1:1 fashion using a centrally controlled web-based good clinical practices–compliant randomization system (TrusT, Health Informatics Centre, University of Dundee) to either: 1) EC containing 16 mg nicotine (Vapourlites Starter Kit with XR5 16 mg nicotine cartomizer; Vapourlites, Peterlee, United Kingdom); or 2) nicotine-free EC plus nicotine flavoring (Vapourlites Starter Kit with 0 mg nicotine cartomizer) because it was considered by the institutional ethics committee as ethically unacceptable to randomize those who were willing to quit smoking into a smoking arm. Those unwilling to consider quitting smoking continued in the parallel preference TC cohort. Participants in the TC arm continued their usual daily smoking habits and did not use EC for the 4-week period of the trial. The study was approved by the Tayside Research Ethics Committee and was carried out in accordance with the Declaration of Helsinki. Exhaled CO breath test was measured as an indicator of treatment allocation adherence to EC, as in previous trials (16), and was added to the primary model to assess the effect of adherence. Studies have previously shown that CO levels fall significantly when switching completely from TC to EC (17).

ration of Helsinki. Exhaled CO breath test was measured as an indicator of treatment allocation adherence to EC, as in previous trials (16), and was added to the primary model to assess the effect of adherence. Studies have previously shown that CO levels fall significantly when switching completely from TC to EC (17). The primary efficacy endpoint was change in FMD among the TC group and the EC-nicotine and EC-nicotine-free groups. Flow-mediated dilation Endothelial function was assessed by measuring FMD of the brachial artery using a Sequoia 512 (Siemens, Camberley, United Kingdom) and an 8-MHz linear array ultrasound probe as described previously (18). Patients fasted overnight and measurements were conducted at baseline and 1 month according to the International Brachial Artery Reactivity Task Force guidelines (19) by a single operator (M.H.) blinded to study allocation at a single site. All participants were required to refrain from smoking TC or EC for 4 h before each FMD test. Pulse wave velocity Pulse wave velocity and augmentation index were measured at baseline and 1 month by a single operator (M.H.) blinded to study allocation. Measurements were recorded with a SphygmoCor (AtCor, Sydney, Australia) machine using a high-fidelity micromanometer. Biomarkers We measured oxidized low-density lipoprotein, high-sensitivity C-reactive protein, tissue plasminogen activator, and platelet activation inhibitor-1 at baseline and at 1 month. All biomarkers were measured by enzyme-linked immunosorbent assay at the Immunoassay Biomarker Core Laboratory, University of Dundee.

Biomarkers We measured oxidized low-density lipoprotein, high-sensitivity C-reactive protein, tissue plasminogen activator, and platelet activation inhibitor-1 at baseline and at 1 month. All biomarkers were measured by enzyme-linked immunosorbent assay at the Immunoassay Biomarker Core Laboratory, University of Dundee. Outcomes The primary outcome was defined as the change in FMD among the TC group and the EC-nicotine and EC-nicotine-free arms as a linear contrast. Secondary outcomes included change in FMD, PWV, augmentation index at 75 beats/min, heart rate, blood pressure, and biomarkers (oxidized low-density lipoprotein, high-sensitivity C-reactive protein, tissue plasminogen activator, and platelet activation inhibitor-1) for the TC, EC-nicotine, and EC-nicotine-free arms. Statistical analysis The primary endpoint of change in brachial artery FMD is expressed as the maximum FMD percentage change from baseline. Using linear contrast tests, a sample size of 36 subjects in each group would have 80% power to detect an improvement in FMD of 2.0% and 1.0% in the EC-nicotine-free and EC-nicotine arms, respectively, compared with the TC group at 5% significance. In this explanatory trial, all dropouts were replaced to achieve 36 completed subjects in each group. The primary analysis was performed on a per-protocol basis.

er to detect an improvement in FMD of 2.0% and 1.0% in the EC-nicotine-free and EC-nicotine arms, respectively, compared with the TC group at 5% significance. In this explanatory trial, all dropouts were replaced to achieve 36 completed subjects in each group. The primary analysis was performed on a per-protocol basis. Descriptive statistics in the form of mean ± SD for continuous variables and percentages and denominators for categorical variables are tabulated for baseline and at the 1-month visit. The dependent variable was assessed for approximation to a normal distribution and transformed if necessary. The FMD response relationship was assessed by a linear contrast test (TC, EC-nicotine, EC-nicotine-free) in a multiple linear regression on FMD at 4 weeks including the baseline FMD level and experimental group as covariates. The model also included the minimization variables: baseline age (≤40 years, >40 years); sex (male, female); and smoking pack-years (≤20 pack-years, >20 pack-years). Pre-specified subgroup analyses were completed by fitting the appropriate interaction term in the regression model and, if significant, outcomes were analyzed separately by level of subgroup. All comparisons were performed among treatment arms (TC vs. EC-nicotine vs. EC-nicotine-free) at the final visit (4 weeks) and adjusted for the baseline measure of the outcome.

y fitting the appropriate interaction term in the regression model and, if significant, outcomes were analyzed separately by level of subgroup. All comparisons were performed among treatment arms (TC vs. EC-nicotine vs. EC-nicotine-free) at the final visit (4 weeks) and adjusted for the baseline measure of the outcome. As the parallel control arm expressed a preference to not be randomized, a propensity score was created with the binary outcome of randomized versus nonrandomized using logistic regression and subsequently used as an adjustment covariate in the regression models to allow for potential bias. Variables included in the propensity score included demographic data, blood pressure, CO levels, all measured biomarkers, FMD and vascular stiffness parameters, and smoking history (Online Table 1). All analyses were conducted in SAS version 9.4 (SAS Institute Inc., Cary, North Carolina). Results A total of 145 patients were recruited into the trial (Figure 1). A final number of 114 patients (40 TC, 37 EC-nicotine, 37 EC-nicotine-free) completed both visits. Baseline demographic data and smoking history among the 3 arms were comparable and are shown in Table 1. There were no serious adverse events reported during the trial.Figure 1 CONSORT Diagram Flow chart showing patient involvement in the study. CONSORT = Consolidated Standards of Reporting Trials; EC = electronic cigarettes; FMD = flow-mediated dilation; TC = tobacco cigarettes. Table 1 Demography of the Evaluable Dataset by Study Arm

Results A total of 145 patients were recruited into the trial (Figure 1). A final number of 114 patients (40 TC, 37 EC-nicotine, 37 EC-nicotine-free) completed both visits. Baseline demographic data and smoking history among the 3 arms were comparable and are shown in Table 1. There were no serious adverse events reported during the trial.Figure 1 CONSORT Diagram Flow chart showing patient involvement in the study. CONSORT = Consolidated Standards of Reporting Trials; EC = electronic cigarettes; FMD = flow-mediated dilation; TC = tobacco cigarettes. Table 1 Demography of the Evaluable Dataset by Study Arm TC (n = 40, 35.0%) EC-Nicotine (n = 37, 32.5%) EC-Nicotine-Free (n = 37, 32.5%) Male 13 (32.5) 14 (37.8) 12 (33.4) Age, yrs 44.2 (40.4–47.9) 48.0 (44.7–51.3) 48.4 (43.5–53.3) Weekly alcohol intake, U 0.0 [0.0–11.0] 0.0 [0.0–10.0] 4.0 [0.0–12.0] BMI 26.7 (25.0–28.5) 28.1 (25.8–30.4) 27.1 (25.4–28.8) Employment status FT 23 (57.5) 14 (37.8) 16 (43.2) PT 3 (7.5) 5 (13.5) 7 (18.9) Unemployed 7 (17.5) 10 (27.0) 7 (18.9) Other 7 (17.5) 8 (21.6) 7 (18.9) CO, ppm 12.0 [7.3–20.8] 12.0 [7.5–16.0] 11.0 [7.0–14.0] CO% COHb 2.6 [1.8–4.0] 2.6 [1.9–3.2] 2.4 [1.8–2.9] Age started smoking, yrs 15.0 [13.0–16.5] 14.0 [13.0–16.0] 16.0 [13.0–18.0] Cigarettes per day 20 [15–20] 18 [15–20] 18 [15–20] Years smoked 29.0 [19.5–36.5] 36.0 [25.0–41.0] 32.0 [22.0–40.0] Pack-year history 25.4 [15.5–36.5] 33.3 [21.8–44.0] 27 [19.9–36.8] Parents smoked No 8 (20.0) 6 (16.2) 10 (27.0) Yes 32 (80.0) 31 (83.8) 27 (73.0) Other smokers in the home 0 23 (57.5) 24 (64.9) 29 (78.4) 1 15 (37.5) 13 (35.1) 8 (21.6) 2 2 (5.0) 0 (0.0) 0 (0.0) Values are n (%), mean (95% confidence interval), or median [interquartile range]. Analysis of variance used for age, height, weight, BMI, systolic BP, diastolic BP and heart rate. Chi-square test used for categorical variables, sex, and employment status. Kruskal-Wallis test used for age started smoking, cigarettes per day, years smoked, pack-year history weekly alcohol intake, CO ppm, and CO% COHb.

Analysis of variance used for age, height, weight, BMI, systolic BP, diastolic BP and heart rate. Chi-square test used for categorical variables, sex, and employment status. Kruskal-Wallis test used for age started smoking, cigarettes per day, years smoked, pack-year history weekly alcohol intake, CO ppm, and CO% COHb. BMI = body mass index; BP = blood pressure; CO = carbon monoxide; CO% COHb = percentage of CO in carboxyhemoglobin; EC = electronic cigarettes; FT = full time; ppm = parts per million; PT = part time; TC = total cigarettes. Primary outcome The primary outcome of change in FMD of the brachial artery showed a significant trend in the difference among arms from TC to EC-nicotine to EC-nicotine-free (linear trend β for TC, EC-nicotine, EC-nicotine-free = 0.73%; 95% CI: 0.41 to 1.05; p < 0.0001). Within 1 month of switching from TC to EC, FMD significantly improved among TC and combined EC arms (1.49%; 95% CI: 0.93 to 2.04; p < 0.0001) and separately between TC and EC-nicotine and between TC and EC-nicotine-free (Table 2). There was no statistically significant difference in FMD change between the EC-nicotine and EC-nicotine-free arms (Table 2, Central Illustration).Table 2 Regression Analysis of Outcomes for FMD—Linear Contrast With Higher Arm Less Nicotine

rately between TC and EC-nicotine and between TC and EC-nicotine-free (Table 2). There was no statistically significant difference in FMD change between the EC-nicotine and EC-nicotine-free arms (Table 2, Central Illustration).Table 2 Regression Analysis of Outcomes for FMD—Linear Contrast With Higher Arm Less Nicotine Difference Between Arms in Change p Value Primary outcome* Change in FMD (+1 group, 1 = TC, 2 = EC-nicotine, 3 = EC-nicotine-free) 0.73 (0.41 to 1.05) <0.0001 Secondary outcomes* Change in FMD, EC-nicotine-free vs. TC (ref) 1.52 (0.90 to 2.15) <0.0001 Change in FMD, EC-nicotine vs. TC (ref) 1.44 (0.78 to 2.09) <0.0001 Change in FMD, all EC vs. TC (ref) 1.49 (0.93 to 2.04) <0.0001 Change in FMD, EC-nicotine-free vs. EC-nicotine (ref) 0.09 (−0.52 to 0.69) 0.78 Values are regression coefficient (95% CI). CI = confidence interval; FMD = flow-mediated dilatation; ref = reference; other abbreviations as in Table 1. ∗ Adjusted for baseline of the outcome, baseline age (≤40 years, >40 years), sex (male, female), and smoking pack-years (≤20 pack-years, >20 pack-years). Central Illustration Change in Mean Flow-Mediated Dilation Among Tobacco Cigarettes and Electronic Cigarettes With and Without Nicotine Adjusted mean percentage change in forearm flow-mediated dilation with 95% confidence intervals for subjects on electronic cigarettes (EC), EC-nicotine, and EC-nicotine-free.

∗ Adjusted for baseline of the outcome, baseline age (≤40 years, >40 years), sex (male, female), and smoking pack-years (≤20 pack-years, >20 pack-years). Central Illustration Change in Mean Flow-Mediated Dilation Among Tobacco Cigarettes and Electronic Cigarettes With and Without Nicotine Adjusted mean percentage change in forearm flow-mediated dilation with 95% confidence intervals for subjects on electronic cigarettes (EC), EC-nicotine, and EC-nicotine-free. The interaction term between treatment and sex for the primary outcome of FMD change was statistically significant (p = 0.009), therefore the subgroup analyses was performed by sex. The improvement in FMD was seen in both males and females for TC versus EC comparisons but significantly greater improvement in vascular function was seen in females who switched from TC to EC (Table 3).Table 3 Regression Analysis of FMD Primary Outcomes by Sex Subgroup—Linear Contrast With Higher Arm Less Nicotine Difference Between Arms in Change p Value Change in FMD (+1 group, 1 = TC, 2 = EC-nicotine, 3 = EC-nicotine-free) Male 0.213 (−0.248 to 0.675) 0.351 Female 1.049 (0.617 to 1.480) <0.0001 Change in FMD, EC-nicotine-free vs. TC (ref) Male 0.448 (−0.451 to 1.347) 0.315 Female 2.183 (1.336 to 3.030) <0.0001 Change in FMD, EC-nicotine vs. TC (ref) Male 0.822 (−0.067 to 1.710) 0.069 Female 1.824 (0.942 to 2.706) <0.0001 Change in FMD, EC-nicotine-free vs. EC-nicotine (ref) Male −0.374 (−1.239 to 0.492) 0.384 Female 0.359 (−0.449 to 1.167) 0.377 Values are regression coefficient (95% CI).

emale 2.183 (1.336 to 3.030) <0.0001 Change in FMD, EC-nicotine vs. TC (ref) Male 0.822 (−0.067 to 1.710) 0.069 Female 1.824 (0.942 to 2.706) <0.0001 Change in FMD, EC-nicotine-free vs. EC-nicotine (ref) Male −0.374 (−1.239 to 0.492) 0.384 Female 0.359 (−0.449 to 1.167) 0.377 Values are regression coefficient (95% CI). Adjusted for baseline of the outcome, baseline age (≤40 years, >40 years) and smoking pack-years (≤20 pack-years, >20 pack-years). Abbreviations as in Tables 1 and 2. As expected, exhaled CO levels were high at baseline and comparable among the 3 arms of the study (Table 1). However, at the end of study, those with the lowest tertile of CO (best compliance with EC and least dual use) had the greatest gain in vascular function improvement. In the lowest tertile of CO, once again, females who switched from TC to EC had a much greater gain in vascular function improvement than did males. Females who complied less well with allocated therapy (dual use with TC) at the middle and high CO tertiles, still benefited from switching to EC more than males did (Table 4). Data on noncompliant subjects are shown in Online Table 2.Table 4 Change in FMD—Mean and 95% CI by CO Tertiles, Sex, and Group

than did males. Females who complied less well with allocated therapy (dual use with TC) at the middle and high CO tertiles, still benefited from switching to EC more than males did (Table 4). Data on noncompliant subjects are shown in Online Table 2.Table 4 Change in FMD—Mean and 95% CI by CO Tertiles, Sex, and Group CO Tertile Sex TC EC-Nicotine EC-Nicotine-Free Low, 0–5 ppm Male 0.28∗ [1] 1.23 (0.02 to 2.44) [6] 0.79 (0.38 to 1.21) [6] Female 0.29∗ [1] 1.58 (0.50 to 2.66) [12] 2.26 (1.31 to 3.21) [11] Both 0.29 (0.22 to 0.35) 1.46 (0.71 to 2.22) 1.74 (1.05 to 2.43) Middle, 6–11 ppm Male 0.17 (−0.57 to 0.91) [6] 0.81 (−5.39 to 7.00) [2] −0.23 (−3.13 to 2.68) [3] Female −0.64 (−1.76 to 0.47) [9] 0.87 (0.02 to 1.72) [7] 1.43 (0.71 to 2.15) [10] Both −0.32 (−1.01 to 0.37) 0.86 (0.22 to 1.50) 1.05 (0.31 to 1.79) High, 12–32 ppm Male 0.43 (−0.40 to 1.25) [6] 0.83 (−0.40 to 2.07) [6] 0.51 (−3.81 to 4.83) [3] Female 0.16 (−0.30 to 0.62) [17] 1.74 (−0.77 to 4.25) [4] 1.55 (0.59 to 2.52) [4] Both 0.23 (−0.14 to 0.60) 1.20 (0.23 to 2.16) 1.11 (−0.03 to 2.24) Values are mean (95% CI) [n]. Abbreviations as in Tables 1 and 2. ∗ 95 CI% not estimable. Data from our lab for age- and sex-matched nonsmoking healthy volunteers indicate a mean FMD of 7.7%. To put this into context, over a 4-week switch, chronic smokers who switched from TC to EC-nicotine showed improved mean FMD from 5.5% to 6.7% and those who switched from TC to EC-nicotine-free showed improved mean FMD from 5.3% to 6.6%.

or age- and sex-matched nonsmoking healthy volunteers indicate a mean FMD of 7.7%. To put this into context, over a 4-week switch, chronic smokers who switched from TC to EC-nicotine showed improved mean FMD from 5.5% to 6.7% and those who switched from TC to EC-nicotine-free showed improved mean FMD from 5.3% to 6.6%. Secondary outcomes There was no significant trend in difference among the 3 arms for other secondary outcomes including PWV, heart rate, and biomarkers of inflammation and platelet reactivity (Table 5). However, the interaction terms between treatment and smoking pack-years were significant for PWV (p = 0.016) and heart rate (p = 0.003). Therefore, a subgroup analysis was done for these outcomes by smoking pack-years.Table 5 Regression Analysis of Secondary Outcomes—Linear Contrast With Higher Arm Less Nicotine

le 5). However, the interaction terms between treatment and smoking pack-years were significant for PWV (p = 0.016) and heart rate (p = 0.003). Therefore, a subgroup analysis was done for these outcomes by smoking pack-years.Table 5 Regression Analysis of Secondary Outcomes—Linear Contrast With Higher Arm Less Nicotine Difference Between Arms in Change p Value Carotid femoral pulse wave velocity −0.167 (−0.402 to 0.069) 0.164 ≤20 pack-years, n = 27 −0.471 (−0.834 to −0.107) 0.014 >20 pack-years, n = 70 0.031 (−0.271 to 0.332) 0.839 Heart rate −1.190 (−3.050 to 0.670) 0.207 ≤20 pack-years, n = 31 2.647 (0.278 to 5.016) 0.030 >20 pack-years, n = 82 −2.825 (−5.223 to −0.426) 0.022 Augmentation index,75 beats/min 0.112 (−1.833 to 2.058) 0.909 Oxidized LDL −1.113 (−5.458 to 3.232) 0.612 High-sensitivity CRP∗ 0.039 (−0.221 to 0.299) 0.769 Tissue plasminogen activator∗ −0.036 (−0.123 to 0.052) 0.425 Platelet activation inhibitor-1∗ −0.007 (−0.131 to 0.116) 0.906 Systolic blood pressure −2.158 (−4.789 to 0.472) 0.107 Diastolic blood pressure −1.126 (−2.624 to 0.372) 0.139 Values are regression coefficient (95% CI). Change in parameters adjusted for baseline of the outcome, baseline age (≤40 years, >40 years), sex (male, female), and smoking pack-years (≤20 pack-years, >20 pack-years). CI = confidence interval; CRP = C-reactive protein; LDL = low-density lipoprotein. ∗ Log-transformed.

Difference Between Arms in Change p Value Carotid femoral pulse wave velocity −0.167 (−0.402 to 0.069) 0.164 ≤20 pack-years, n = 27 −0.471 (−0.834 to −0.107) 0.014 >20 pack-years, n = 70 0.031 (−0.271 to 0.332) 0.839 Heart rate −1.190 (−3.050 to 0.670) 0.207 ≤20 pack-years, n = 31 2.647 (0.278 to 5.016) 0.030 >20 pack-years, n = 82 −2.825 (−5.223 to −0.426) 0.022 Augmentation index,75 beats/min 0.112 (−1.833 to 2.058) 0.909 Oxidized LDL −1.113 (−5.458 to 3.232) 0.612 High-sensitivity CRP∗ 0.039 (−0.221 to 0.299) 0.769 Tissue plasminogen activator∗ −0.036 (−0.123 to 0.052) 0.425 Platelet activation inhibitor-1∗ −0.007 (−0.131 to 0.116) 0.906 Systolic blood pressure −2.158 (−4.789 to 0.472) 0.107 Diastolic blood pressure −1.126 (−2.624 to 0.372) 0.139 Values are regression coefficient (95% CI). Change in parameters adjusted for baseline of the outcome, baseline age (≤40 years, >40 years), sex (male, female), and smoking pack-years (≤20 pack-years, >20 pack-years). CI = confidence interval; CRP = C-reactive protein; LDL = low-density lipoprotein. ∗ Log-transformed. Vascular stiffness and blood pressure Smokers who smoked ≤20 pack-years also demonstrated an improvement in vascular stiffness within 1 month of switching from TC to EC with an improvement in PWV (−0.471 m/s; 95% CI: −0.834 to −0.107; p = 0.014), whereas those who smoked >20 pack-years showed no change within this time frame (Table 5). For the whole cohort, when both EC arms were combined, there was a significant improvement in PWV in this combined EC group compared with in the TC group (−0.529 m/s; 95% CI: −0.946 to −0.112; p = 0.014). When both EC groups were combined, there was a greater reduction in systolic blood pressure in the EC group than in the TC group, both in smokers of ≤20 pack-years (EC: −4.41 mm Hg; 95% CI: −7.91 to −0.91 vs. TC: −2.86 mm Hg; 95% CI: −8.09 to 2.38; p = 0.59) and >20 pack-years (EC: −7.75 mm Hg; 95% CI: −11.56 to −3.93 vs. TC: −1.37 mm Hg; 95% CI: −5.32 to 2.59; p = 0.04).

reduction in systolic blood pressure in the EC group than in the TC group, both in smokers of ≤20 pack-years (EC: −4.41 mm Hg; 95% CI: −7.91 to −0.91 vs. TC: −2.86 mm Hg; 95% CI: −8.09 to 2.38; p = 0.59) and >20 pack-years (EC: −7.75 mm Hg; 95% CI: −11.56 to −3.93 vs. TC: −1.37 mm Hg; 95% CI: −5.32 to 2.59; p = 0.04). Using analysis of variance, there was a significant difference for the mean change of systolic blood pressure among the 3 arms: TC (−1.89 mm Hg; 95% CI: −4.91 to 1.14); EC-nicotine (−4.27 mm Hg; 95% CI: −7.73 to −0.81); EC-nicotine-free (−9.69 mm Hg; 95% CI: −14.67 to −4.71), p = 0.01. When adjusted for baseline variables, the trend remained toward lower systolic blood pressures among arms from TC to EC-nicotine to EC-nicotine-free but was not statistically significant (β = −2.2 mm Hg; 95% CI: −4.8 to 0.5; p = 0.11). The greatest difference in systolic blood pressure was seen in the TC versus EC-nicotine-free arms (−4.3 mm Hg; 95% CI: –9.6 to 1.0; p = 0.11) followed by TC versus EC-nicotine arms (−2.0 mm Hg; 95% CI: –7.6 to 3.5; p = 0.47).

e-free but was not statistically significant (β = −2.2 mm Hg; 95% CI: −4.8 to 0.5; p = 0.11). The greatest difference in systolic blood pressure was seen in the TC versus EC-nicotine-free arms (−4.3 mm Hg; 95% CI: –9.6 to 1.0; p = 0.11) followed by TC versus EC-nicotine arms (−2.0 mm Hg; 95% CI: –7.6 to 3.5; p = 0.47). Heart rate For smokers who smoked ≤20 pack-years (n = 31), resting heart rate significantly increased by 2.6 beats/min (95% CI: 0.3 to 5.0) for EC-nicotine compared with the rate for TC and increased by 5.2 beats/min (95% CI: 0.6 to 10.0) for EC-nicotine-free compared with the rate for TC (p = 0.03). However, for smokers who smoked >20 pack-years (n = 82), resting heart rate decreased by 2.8 beats/min (95% CI: −5.2 to −0.4) for EC-nicotine compared with the rate for TC and decreased further by 5.6 beats/min (95% CI: −10.4 to −0.8) for EC-nicotine-free compared with the rate for TC (p = 0.02). Discussion The main findings from this present study are that within 1 month of switching from TC to EC, smokers demonstrate a significant improvement in vascular function. The data from this present trial on the early CV impact of switching from TC to EC has yielded several clinically important findings.

Heart rate For smokers who smoked ≤20 pack-years (n = 31), resting heart rate significantly increased by 2.6 beats/min (95% CI: 0.3 to 5.0) for EC-nicotine compared with the rate for TC and increased by 5.2 beats/min (95% CI: 0.6 to 10.0) for EC-nicotine-free compared with the rate for TC (p = 0.03). However, for smokers who smoked >20 pack-years (n = 82), resting heart rate decreased by 2.8 beats/min (95% CI: −5.2 to −0.4) for EC-nicotine compared with the rate for TC and decreased further by 5.6 beats/min (95% CI: −10.4 to −0.8) for EC-nicotine-free compared with the rate for TC (p = 0.02). Discussion The main findings from this present study are that within 1 month of switching from TC to EC, smokers demonstrate a significant improvement in vascular function. The data from this present trial on the early CV impact of switching from TC to EC has yielded several clinically important findings. First, there is an early benefit to vascular function from switching from TC to EC. Within the switching time frame of 1 month, chronic smokers demonstrated significant improvements in vascular endothelial function. This is consistent with the recent review by Benowitz and Fraiman (20) that switching from TC to EC might result in overall benefit to public health. Previous meta-analysis of FMD studies have demonstrated that the pooled, adjusted relative risks of CV events was 13% lower with every 1% improvement in FMD (12). As we have demonstrated with age- and sex-matched healthy volunteer FMD data from our lab, otherwise healthy but chronic TC smokers who switched to EC improved their vascular function, approaching values seen in healthy nonsmokers. Within 1 month of switching from TC, we found a 1.5% improvement between TC and EC-nicotine-free arms, 1.4% improvement between TC and EC-nicotine arms, and a 1.5% improvement between TC and combined EC arms (Table 2).

s who switched to EC improved their vascular function, approaching values seen in healthy nonsmokers. Within 1 month of switching from TC, we found a 1.5% improvement between TC and EC-nicotine-free arms, 1.4% improvement between TC and EC-nicotine arms, and a 1.5% improvement between TC and combined EC arms (Table 2). Second, vascular stiffness was also significantly reduced within 1 month of switching in smokers of ≤20 pack-years compared with in those who smoked >20 pack-years, suggesting that the trend toward lower blood pressure in the EC arms could be important. Longer term studies are required to detect whether there are statistically and clinically significant reductions in blood pressure when switching from TC to EC as a result of improvements in vascular stiffness. Third, switching to EC from TC may benefit females more than males and this is also seen in females who were less compliant (dual use). However, this was a subgroup analysis of our data and should be interpreted with caution. Nevertheless, female smokers face more health risks than male smokers do; they are more likely to develop lung cancer (21) and are almost twice as likely to have a myocardial infarction as a result of their smoking (22). The worrying trend worldwide of increased TC prevalence among women (23) suggests that further measures are urgently required to reduce harms associated with TC. Therefore, the switch to EC may be considered a vascular harms reduction measure for both sexes but particularly for the 200 million women worldwide who currently smoke TC (24).

worldwide of increased TC prevalence among women (23) suggests that further measures are urgently required to reduce harms associated with TC. Therefore, the switch to EC may be considered a vascular harms reduction measure for both sexes but particularly for the 200 million women worldwide who currently smoke TC (24). Fourth, those who complied best with allocated therapy, as indicated by exhaled CO levels, benefitted the most in terms of improvement in endothelial function. Our data shows that those who avoided dual use and had lowest CO levels derived greater vascular benefit from switching. Dual use of EC is a highly prevalent reality worldwide (25). The benefits of total switching may have been even larger if subjects fully complied with the switch. This finding could be used to encourage smokers who dual use to minimize TC exposure. Finally, there was no difference observed between the 2 EC arms (with and without nicotine) for this short-term study. Early improvement appears to be unrelated to the abstinence from nicotine but rather from other toxic material produced by combustion in TC. Further investigation is required to understand the impact of nicotine itself on longer term vascular function.

he 2 EC arms (with and without nicotine) for this short-term study. Early improvement appears to be unrelated to the abstinence from nicotine but rather from other toxic material produced by combustion in TC. Further investigation is required to understand the impact of nicotine itself on longer term vascular function. In addition to these findings, we found a reduction in resting heart rate in the >20 pack-years cohort who switched to EC. The association between resting heart rate and CV events is well known (26,27) and the link between smoking cessation and reduction in heart rate has been previously demonstrated in other studies (28). However, this present study suggests that a switch from TC to EC might also achieve this early on in chronic smokers. A reduction in resting heart rate as seen in this cohort of high CV risk, chronic, heavy smokers would yield the greatest benefit, further supporting the benefits of these cohorts switching from TC to EC. Whether this might be a transient phenomenon or translates to more sustained benefits requires further investigation. We stress that whereas this study provides new evidence for the rapid improvement of vascular function when switching from TC to EC and therefore suggests that from a vascular perspective, EC may be a less harmful alternative to TC, there is no justification nor evidence from our work to state that EC are safe per se and therefore should never be viewed by nonsmokers as harmless devices to try.

ment of vascular function when switching from TC to EC and therefore suggests that from a vascular perspective, EC may be a less harmful alternative to TC, there is no justification nor evidence from our work to state that EC are safe per se and therefore should never be viewed by nonsmokers as harmless devices to try. Comparison with other studies The vascular impact of EC is a new and evolving field and as such there remains a significant paucity of research in this area. Carnevale et al. (29) reported a small (n = 40) single-use crossover study that demonstrated that although both TC and EC had unfavorable effects on markers of oxidative stress and FMD, EC had a lesser impact than TC did. This result of our present study is consistent with this finding. Hajek et al. (16) recently reported that EC was more effective for smoking cessation than nicotine replacement therapy was when both products were accompanied by behavioral support.

markers of oxidative stress and FMD, EC had a lesser impact than TC did. This result of our present study is consistent with this finding. Hajek et al. (16) recently reported that EC was more effective for smoking cessation than nicotine replacement therapy was when both products were accompanied by behavioral support. Study limitations This was a single-center study. We could not perform a full 3-arm randomized controlled design as it was unethical for participants who wished to quit smoking to be allocated to the smoking arm. We created a propensity score as an adjustment covariate in the regression models to allow for any potential bias and the results remained consistent. Baseline characteristics of the cohorts were also comparable. The duration of effect tested was deliberately short as the primary purpose of the study was to investigate whether there were early vascular benefits from switching from TC to EC and the results are reassuring. However, longer follow-up is required to determine whether males also benefit to the same level as females do and whether these changes seen are sustained and to assess the impact of nicotine in EC. There are many different EC devices available in the market, and we tested only 1 device for consistency of effect. Future comparative studies among different devices are required. Finally, we used endothelial dysfunction as an early indicator of CV disease and a surrogate for CV events. However, endothelial dysfunction has consistently been shown to correlate well with longer term CV outcomes (12,13,30).

for consistency of effect. Future comparative studies among different devices are required. Finally, we used endothelial dysfunction as an early indicator of CV disease and a surrogate for CV events. However, endothelial dysfunction has consistently been shown to correlate well with longer term CV outcomes (12,13,30). Conclusions Smokers, particularly females, who switch from TC to EC derive significant benefits in terms of vascular health, and this improvement is seen early on. From a vascular health perspective, recommendations of switching from TC to EC could be considered a vascular harms reduction measure. Further investigation is required on the long-term CV and non-CV effects of these devices.Perspectives COMPETENCY IN MEDICAL KNOWLEDGE: Smoking tobacco cigarettes is known to be harmful. In theory, EC contain fewer harmful substances, but the health risks of EC are currently not fully known. COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Patients who wish to stop smoking TC should be offered less harmful options including switching to EC. TRANSLATIONAL OUTLOOK: This study demonstrates the early vascular impact of switching from TC to EC. Therefore, switching to EC may be considered a vascular harms reduction measure. Appendix Online Tables 1 and 2

COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Patients who wish to stop smoking TC should be offered less harmful options including switching to EC. TRANSLATIONAL OUTLOOK: This study demonstrates the early vascular impact of switching from TC to EC. Therefore, switching to EC may be considered a vascular harms reduction measure. Appendix Online Tables 1 and 2 The VESUVIUS (Vascular Effects of Regular Cigarettes Versus Electronic Cigarette Use) trial was funded by the British Heart Foundation (grant PG/15/64/31681); and supported by Immunoassay Biomarker Core Laboratory, University of Dundee, the Tayside Medical Sciences Centre, and the NHS Tayside Smoking Cessation Service. The funder had no role in the study design, data collection, data analysis, data interpretation, writing of the report, or in the decision to submit for publication. Dr. Donnan has received research grants from AbbVie, Shire, and Gilead Sciences. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript's audio summary by Editor-in-Chief Dr. Valentin Fuster onJACC.org. Appendix For supplemental tables, please see the online version of this paper.

Primary percutaneous coronary intervention (PPCI) to emergently reopen the occluded coronary artery, restore blood flow, and secure vessel patency with a stent is the evidence-based standard of care for ST-segment elevation myocardial infarction (STEMI) (1). However, the success of PPCI is limited by failed microvascular reperfusion, which occurs in one-half of all treated patients (2,3). This complication, described as microvascular obstruction (MVO), is associated with adverse left ventricular remodeling and reduced left ventricular function and is independently predictive of cardiac prognosis (4). During PPCI, distal embolization of thrombus within the lumen of the infarct-related coronary artery and microvascular thrombosis (5, 6, 7, 8, 9), notably of fibrin-rich microthrombi (6), contribute to MVO. Myocardial hemorrhage is closely related to MVO (3) and occurs when endothelial cell injury compromises capillary integrity leading to the extravasation of blood into the extracellular space. T2*-weighted cardiac magnetic resonance (CMR) is the established method to identify and evaluate myocardial hemorrhage in vivo, accumulation of paramagnetic hemoglobin breakdown products leads to a shortening of T2* relaxation times, resulting in a hypointense zone on imaging that represents tissue hemorrhage (9,10). Late gadolinium-enhanced CMR imaging is used to identify MVO, a dark area representing failed perfusion at the core of the bright infarct. Validation in swine established that the hypointense core on T2* imaging corresponds with severe capillary loss and destruction resulting in tissue hemorrhage, with excellent anatomical correlation between the localization and extent of MVO and myocardial hemorrhage (9).

led perfusion at the core of the bright infarct. Validation in swine established that the hypointense core on T2* imaging corresponds with severe capillary loss and destruction resulting in tissue hemorrhage, with excellent anatomical correlation between the localization and extent of MVO and myocardial hemorrhage (9). Fibrinolytic therapy is an effective treatment for acute coronary thrombosis (11). A facilitated PCI strategy involving full- or half-dose adjunctive fibrinolytic therapy given before PCI with stenting improves coronary flow acutely (12,13). Similarly, in patients with an expected PCI-related delay, half-dose alteplase and timely PCI improves epicardial and myocardial flow when compared with PPCI alone. However, combination-facilitated PCI involving either full-dose (14) or half-dose lytic therapy (15) causes paradoxical activation of thrombin, clot formation, and bleeding. In T-TIME (A Trial of Low-Dose Adjunctive Alteplase During Primary PCI), we hypothesized that a therapeutic strategy involving low-dose intracoronary fibrinolytic therapy with alteplase infused early after coronary reperfusion would reduce MVO. Patients with acute STEMI presenting <6 h after symptom onset and a large thrombus burden evident at initial coronary angiography were enrolled in a 3-arm dose-ranging design (10 or 20 mg of alteplase or placebo). The primary analysis determined that alteplase did not reduce the amount of MVO revealed by CMR imaging 2 to 7 days post-MI (primary outcome) and the secondary outcomes were consistent with this result (16).

initial coronary angiography were enrolled in a 3-arm dose-ranging design (10 or 20 mg of alteplase or placebo). The primary analysis determined that alteplase did not reduce the amount of MVO revealed by CMR imaging 2 to 7 days post-MI (primary outcome) and the secondary outcomes were consistent with this result (16). Infarct size is influenced by ischemic time (17), as are the efficacies of primary reperfusion therapies, including systemic fibrinolysis (18) and primary PCI (19). In this pre-specified analysis, we hypothesized that the effects of adjunctive intracoronary administration of low-dose alteplase during PPCI could be associated with ischemic time. We assessed the associations among ischemic time, treatment group (placebo, alteplase 10 mg, alteplase 20 mg), and the primary and secondary outcomes in this clinical trial. Methods Trial design We performed a randomized, double-blind, placebo-controlled, parallel group phase 2 clinical trial of low-dose adjunctive alteplase during PPCI, the main results of which have been published previously (16).

Infarct size is influenced by ischemic time (17), as are the efficacies of primary reperfusion therapies, including systemic fibrinolysis (18) and primary PCI (19). In this pre-specified analysis, we hypothesized that the effects of adjunctive intracoronary administration of low-dose alteplase during PPCI could be associated with ischemic time. We assessed the associations among ischemic time, treatment group (placebo, alteplase 10 mg, alteplase 20 mg), and the primary and secondary outcomes in this clinical trial. Methods Trial design We performed a randomized, double-blind, placebo-controlled, parallel group phase 2 clinical trial of low-dose adjunctive alteplase during PPCI, the main results of which have been published previously (16). Participants and eligibility criteria Patients with a clinical diagnosis of acute STEMI with a symptom onset to reperfusion time of 6 h or less were eligible for randomization. Radial artery access was required, angiographic criteria included a proximal-mid coronary artery occlusion (TIMI [Thrombolysis In Myocardial Infarction] flow grade 0/1) or impaired coronary flow (TIMI flow grade 2) in the presence of definite angiographic evidence of thrombus (TIMI flow grade 2+) in a major coronary artery. Exclusion criteria included any contraindication to fibrinolysis or CMR and lack of informed consent. Full inclusion and exclusion criteria are described in the Supplemental Appendix.

flow (TIMI flow grade 2) in the presence of definite angiographic evidence of thrombus (TIMI flow grade 2+) in a major coronary artery. Exclusion criteria included any contraindication to fibrinolysis or CMR and lack of informed consent. Full inclusion and exclusion criteria are described in the Supplemental Appendix. Setting The participants were enrolled in 11 hospitals in the United Kingdom and guideline-based medical and invasive management was recommended (1). Enrollment started on March 17, 2016, and ended on December 21, 2017. Informed consent and study protocol Screening, witnessed verbal informed consent, study drug administration, and acute assessments of efficacy took place during the standard-of-care PPCI. The protocol is included in the Supplemental Appendix. The trial had ethics committee approval, adhered to Guidelines for Good Clinical Practice in Clinical Trials (20), and complied with the Declaration of Helsinki (21). Randomization, implementation, and blinding Participants were randomized by staff in the catheter laboratory using an interactive voice response–based randomization system. The randomization sequence was created using the method of randomized permuted blocks of length 6, with stratification by location of STEMI and study site. The allocation sequence was on a 1:1:1 basis among the placebo and alteplase (10 mg, 20 mg) groups and the sequence was concealed electronically. The participants, staff, and researchers were blinded to the treatment group allocation. Standard care PPCI followed contemporary practice guidelines (1) (Supplemental Appendix).

Randomization, implementation, and blinding Participants were randomized by staff in the catheter laboratory using an interactive voice response–based randomization system. The randomization sequence was created using the method of randomized permuted blocks of length 6, with stratification by location of STEMI and study site. The allocation sequence was on a 1:1:1 basis among the placebo and alteplase (10 mg, 20 mg) groups and the sequence was concealed electronically. The participants, staff, and researchers were blinded to the treatment group allocation. Standard care PPCI followed contemporary practice guidelines (1) (Supplemental Appendix). Interventions After successful reperfusion of the infarct-related artery, the participants immediately received the allocated intervention. The study drug (placebo, alteplase 10 mg, or alteplase 20 mg) was manually infused before stent implantation. Further details are provided in the Supplemental Appendix. Outcomes Primary outcome The primary outcome was the amount of MVO (percentage of left ventricular mass) revealed by late gadolinium-enhanced CMR 10 to 15 min after administration of gadolinium-based contrast media. CMR at 1.5-T was scheduled during the index hospitalization, 2 to 7 days after enrollment. MVO was defined as a dark zone on early gadolinium enhancement imaging 1, 3, 5, and 7 min post-contrast injection that remained present within an area of late gadolinium enhancement at 15 min. The myocardial mass of the dark zone was quantified by manual delineation and expressed as percentage of left ventricular mass.

MVO was defined as a dark zone on early gadolinium enhancement imaging 1, 3, 5, and 7 min post-contrast injection that remained present within an area of late gadolinium enhancement at 15 min. The myocardial mass of the dark zone was quantified by manual delineation and expressed as percentage of left ventricular mass. Secondary outcomes Infarct definition and size The presence of acute infarction was established based on abnormalities in cine wall motion, rest first-pass myocardial perfusion, and late gadolinium-enhancement imaging in 2 imaging planes. The myocardial mass of late gadolinium was quantified using computer-assisted planimetry and the territory of infarction was delineated using a 5-SD semi-automated method and expressed as percentage of total left ventricular mass. Myocardial hemorrhage On the T2* parametric maps, a threshold of 20 ms was applied. A region of reduced signal intensity within the infarcted area, with a T2* value of <20 ms (3,22) was considered to confirm the presence of myocardial hemorrhage. The area was manually delineated and expressed as percentage of left ventricular mass. Other outcomes Additional CMR secondary outcomes included myocardial salvage index, left ventricular end-diastolic volume, left ventricular end-systolic volume, and left ventricular ejection fraction at 2 to 7 days and 3 months, these are described in the Supplemental Appendix. Biochemistry Troponin T (ng/l) area under the curve (AUC) was measured from blood samples obtained immediately before reperfusion (0 h) and then again at 2 and 24 h.

Other outcomes Additional CMR secondary outcomes included myocardial salvage index, left ventricular end-diastolic volume, left ventricular end-systolic volume, and left ventricular ejection fraction at 2 to 7 days and 3 months, these are described in the Supplemental Appendix. Biochemistry Troponin T (ng/l) area under the curve (AUC) was measured from blood samples obtained immediately before reperfusion (0 h) and then again at 2 and 24 h. Safety Fibrinogen and other parameters of coagulation and hemostasis served as surrogate measures of bleeding and safety (23,24). These parameters were measured in blood samples when site logistics permitted blood sample collection. The sampling time points were at baseline before reperfusion (0 h) and 2 and 24 h post-reperfusion. Trial coordination An independent Data and Safety Monitoring Committee and a Trial Steering Committee had oversight of the trial and liaised with the sponsor. Each committee had a charter that was established before enrollment started.

Safety Fibrinogen and other parameters of coagulation and hemostasis served as surrogate measures of bleeding and safety (23,24). These parameters were measured in blood samples when site logistics permitted blood sample collection. The sampling time points were at baseline before reperfusion (0 h) and 2 and 24 h post-reperfusion. Trial coordination An independent Data and Safety Monitoring Committee and a Trial Steering Committee had oversight of the trial and liaised with the sponsor. Each committee had a charter that was established before enrollment started. Sample size and statistical methods The sample size and statistical methods are described in detail in the Supplemental Appendix. To summarize, outcomes were analyzed using linear or logistic regression models. Continuous outcomes were transformed when necessary to improve model fit. Analyses treating randomized treatment as a 3-level or as a 2-level categorical variable (active vs. placebo) were performed. On the assumption that any treatment effects would manifest themselves as dose-dependent trends, randomized treatment was modeled as a linear trend across dose groups (0 mg, 10 mg, 20 mg) in an attempt to maximize power. The decision to model as a linear trend across treatment groups was made post hoc with knowledge of the data. All models were adjusted for the location of MI (anterior/nonanterior), as per the stratification of the randomization schedule. Models for coagulation and hemostasis parameters included an additional adjustment for baseline value (transformed in the same way as the outcome measurement). Models included ischemic time categorized in 3 groups (<2 h, ≥2 but <4 h, ≥4 to 6 h), and an interaction between ischemic time and randomized treatment.

hedule. Models for coagulation and hemostasis parameters included an additional adjustment for baseline value (transformed in the same way as the outcome measurement). Models included ischemic time categorized in 3 groups (<2 h, ≥2 but <4 h, ≥4 to 6 h), and an interaction between ischemic time and randomized treatment. All tests were 2-tailed, and p values <0.05 were considered significant. All statistical analyses were carried out with R version 3.2.4 (R Development Core Team 2015, Vienna, Austria) (25) according to a pre-specified statistical analysis plan. No adjustments have been made for multiple testing in these analyses, which should be viewed as exploratory rather than definitive.

red significant. All statistical analyses were carried out with R version 3.2.4 (R Development Core Team 2015, Vienna, Austria) (25) according to a pre-specified statistical analysis plan. No adjustments have been made for multiple testing in these analyses, which should be viewed as exploratory rather than definitive. Results On the recommendation of the Data and Safety Monitoring Committee, recruitment was discontinued on December 21, 2017, based on a pre-specified futility analysis. Specifically, the conditional power for an analysis on the primary efficacy outcome based on 40% of the randomized population (n = 267) with follow-up to 3 months was <30% in both treatment arms. The committee noted that there were no safety concerns. By that time, 1,527 patients undergoing PPCI for acute STEMI had been screened (Figure 1) and 440 patients (mean age 60.5 years, 85% male) had been randomized (151 placebo, 144 alteplase 10 mg, 145 alteplase 20 mg) (Table 1 and Supplemental Table 1). The distribution of randomized participants by ischemic time was as follows: <2 h, 107 (24.3%); ≥2 but <4 h, 235 (53.4%); ≥4 to 6 h, 98 (22.3%). All of the randomized participants were included in those analyses for which they had data available. Seventeen patients (3.9%) withdrew from the study during follow-up.Figure 1 T-TIME Flow Diagram

ndomized participants by ischemic time was as follows: <2 h, 107 (24.3%); ≥2 but <4 h, 235 (53.4%); ≥4 to 6 h, 98 (22.3%). All of the randomized participants were included in those analyses for which they had data available. Seventeen patients (3.9%) withdrew from the study during follow-up.Figure 1 T-TIME Flow Diagram The participants are grouped by treatment group and ischemic time. Two patients (1 randomized to placebo and 1 randomized to 10 mg alteplase) received 20 mg alteplase because an incorrect treatment pack had been selected. Four patients were unable to complete the CMR examination meaning evaluable data for the primary outcome was not available: placebo group (n = 1); 10 mg–alteplase group (n = 2); 20 mg–alteplase group (n = 1). CMR = cardiac magnetic resonance; T-TIME = A Trial of Low-Dose Adjunctive Alteplase During Primary PCI. Table 1 Baseline Clinical Characteristics, Ischemic Time of the Randomized Participants (n = 440)

The participants are grouped by treatment group and ischemic time. Two patients (1 randomized to placebo and 1 randomized to 10 mg alteplase) received 20 mg alteplase because an incorrect treatment pack had been selected. Four patients were unable to complete the CMR examination meaning evaluable data for the primary outcome was not available: placebo group (n = 1); 10 mg–alteplase group (n = 2); 20 mg–alteplase group (n = 1). CMR = cardiac magnetic resonance; T-TIME = A Trial of Low-Dose Adjunctive Alteplase During Primary PCI. Table 1 Baseline Clinical Characteristics, Ischemic Time of the Randomized Participants (n = 440) <2 h (n = 151) ≥2 But <4 h (n = 144) ≥4 to 6 h (n = 145) p Value Clinical Age, yrs 58.8 ± 8.8 61.8 ± 10.8 59.5 ± 10.6 0.027∗ Male 97 (90.7) 195 (83.0) 82 (83.7) 0.160† Race, white 97 (90.7) 219 (93.2) 97 (99.0) 0.023† Body mass index, kg/m2 28.4 ± 5.3 28.5 ± 4.8 27.8 ± 4.4 0.704∗ Presenting characteristics Heart rate, beats/min 72.8 ± 25.0 71.6 ± 15.9 76.1 ± 17.6 0.136∗ Systolic blood pressure, mm Hg 128 ± 24 134 ± 26 138 ± 25 0.026∗ Diastolic blood pressure, mm Hg 79 ± 15 80 ± 16 83 ± 16 0.146∗ Infarct location Anterior 54 (50.5) 103 (43.8) 34 (34.7) 0.073† Inferior 46 (43.0) 107 (45.5) 54 (55.1) 0.181† Lateral 3 (2.8) 0 (0.0) 0 (0.0) 0.025† Posterior 4 (3.7) 21 (8.9) 8 (8.2) 0.223† Other 0 (0.0) 4 (1.7) 2 (2.0) 0.422† Medical history Hypertension‡ 29 (27.1) 81 (34.5) 31 (31.6) 0.417† Diabetes mellitus‡ 11 (10.3) 34 (14.5) 11 (11.2) 0.565† Hypercholesterolemia‡ 22 (20.6) 59 (25.1) 21 (21.4) 0.615† Smoking‡ Current 52 (48.6) 106 (45.1) 51 (52.0) 0.490† Former, stopped >3 months 14 (13.1) 51 (21.7) 19 (19.4) 0.168† Never 41 (38.3) 78 (33.2) 28 (28.6) 0.344† Percutaneous coronary intervention 4 (3.7) 9 (3.8) 7 (7.1) 0.401† Coronary artery bypass graft surgery 0 (0.0) 0 (0.0) 0 (0.0) – Angina 0 (0.0) 13 (5.5) 4 (4.1) 0.024† Myocardial infarction 3 (2.8) 12 (5.1) 5 (5.1) 0.652† Stroke or transient ischemic attack‡ 2 (1.9) 2 (0.9) 1 (1.0) 0.833† Peripheral vascular disease‡ 1 (0.9) 8 (3.4) 3 (3.1) 0.459† Pre-existing maintenance medication Aspirin 14 (13.1) 37 (15.7) 15 (15.3) 0.848† P2Y12 inhibitor Clopidogrel 1 (0.9) 0 (0.0) 1 (1.0) 0.217† Ticagrelor or prasugrel 6 (5.6) 11 (4.7) 3 (3.1) 0.750† Glycoprotein IIb/IIIa inhibitor 27 (25.7) 30 (12.9) 16 (17.6) 0.017† Statin 23 (21.5) 52 (22.1) 22 (22.4) 1.0† Beta-blocker 6 (5.6) 20 (8.5) 16 (16.3) 0.030† ACE inhibitor or ARB 15 (14.0) 46 (19.6) 17 (17.3) 0.457† Mineralocorticoid receptor antagonist 3 (2.8) 1 (0.4) 0 (0.0) 0.10† Symptom onset to arrival at PPCI center, h:min 1:16 (0:59–1:26) 2:12 (1:53–2:44) 4:20 (3:51–4:58) <0.001§ Arrival at PPCI center to reperfusion, min 22 (17–32) 25 (20–36) 26 (19–38) 0.054§ Symptom onset to reperfusion, h:min 1:42 (1:28–1:52) 2:44 (2:22–3:15) 4:47 (4:21–5:31) <0.001§ Initial blood

1 (0.4) 0 (0.0) 0.10† Symptom onset to arrival at PPCI center, h:min 1:16 (0:59–1:26) 2:12 (1:53–2:44) 4:20 (3:51–4:58) <0.001§ Arrival at PPCI center to reperfusion, min 22 (17–32) 25 (20–36) 26 (19–38) 0.054§ Symptom onset to reperfusion, h:min 1:42 (1:28–1:52) 2:44 (2:22–3:15) 4:47 (4:21–5:31) <0.001§ Initial blood results on admission Hemoglobin, g/l 145.8 ± 12.1 145.1 ± 14.2 146.0 ± 12.3 0.795∗ Platelet count, ×109 l 260 ± 55 259 ± 61 265 ± 73 0.777∗ Creatinine, μmol/l 83.8 ± 17.5 82.0 ± 20.4 75.6 ± 16.0 0.012∗ Troponin, ng/l 37 (18–69) 57 (29–102) 129 (66–246) <0.001§ Values are mean ± SD, n (%), or median (interquartile range). ACE = angiotensin-converting enzyme; ARB = angiotensin receptor blocker; IQR = interquartile range; PPCI = primary percutaneous coronary intervention. ∗ The p value was derived using analysis of variance. † The p value was derived using Fisher exact test. ‡ At least 1 risk factor for coronary artery disease was required for eligibility. Diabetes mellitus was defined as a history of diet-controlled or treated diabetes. § The p value was derived using Kruskal-Wallis test. Study intervention Adjunctive study treatment was administered to 435 patients (98.9%); 5 patients did not receive any drug (Figure 1). Two patients (1 randomized to placebo and 1 randomized to 10 mg alteplase) received 20 mg alteplase because an incorrect treatment pack had been selected.

§ The p value was derived using Kruskal-Wallis test. Study intervention Adjunctive study treatment was administered to 435 patients (98.9%); 5 patients did not receive any drug (Figure 1). Two patients (1 randomized to placebo and 1 randomized to 10 mg alteplase) received 20 mg alteplase because an incorrect treatment pack had been selected. Primary and secondary outcomes CMR was performed in 400 patients (90.9%) at 2 to 7 days and 367 patients (83.4%) at 3 months. The median (interquartile range [IQR]) times to CMR at these time points were 4 days (IQR: 3 to 6 days) and 91 days (IQR: 86 to 97 days), respectively. Primary outcome The mean MVO (percentage of left ventricular mass) revealed by CMR 2 to 7 days post-STEMI (primary outcome) was 2.32 ± 4.31% in the placebo group, 2.61 ± 4.49% in the 10 mg alteplase group, and 3.48 ± 5.83% in the 20 mg alteplase group. A linear regression analysis of square root–transformed MVO found no evidence of a treatment effect (mean increase in square root–transformed MVO per 10-mg increase in alteplase dose: 0.15; 95% confidence interval [CI]: −0.12 to 0.42; p = 0.28) (16).

10 mg alteplase group, and 3.48 ± 5.83% in the 20 mg alteplase group. A linear regression analysis of square root–transformed MVO found no evidence of a treatment effect (mean increase in square root–transformed MVO per 10-mg increase in alteplase dose: 0.15; 95% confidence interval [CI]: −0.12 to 0.42; p = 0.28) (16). There was a significant interaction between ischemic time and randomized treatment with respect to the primary outcome (mean increase in square root–transformed MVO per 10-mg increase in alteplase dose: 0.56 (95% CI: 0.21 to 0.91; p = 0.009) (Table 2). There was no evidence of a treatment effect on the extent of MVO at 2 to 7 days for patients with ischemic times <4 h. In those with ischemic times of 4 h or more, the amount of MVO (mean percentage of left ventricular mass) at 2 to 7 days increased from 1.14% in those treated with placebo to 3.11% (10 mg) and 5.20% (20 mg) in those treated with alteplase (Central Illustration). Similar results were observed when analyzing treatment as a 3-level or 2-level categorical variable (Table 3).Table 2 Pre-Specified Analyses of the Primary and Secondary Outcomes, Adjusting for Location of MI, by Subgroups of Ischemic Time Randomized Treatment Group Treatment Effect (Trend per 10-mg Dose Increase) n (Missing) Placebo (n = 151) Alteplase 10 mg (n = 144) Alteplase 20 mg (n = 145) Estimate (95% CI), p Value Interaction p Value Primary Outcome: Extent of MVO (% of LV Mass) at 2–7 Days

There was a significant interaction between ischemic time and randomized treatment with respect to the primary outcome (mean increase in square root–transformed MVO per 10-mg increase in alteplase dose: 0.56 (95% CI: 0.21 to 0.91; p = 0.009) (Table 2). There was no evidence of a treatment effect on the extent of MVO at 2 to 7 days for patients with ischemic times <4 h. In those with ischemic times of 4 h or more, the amount of MVO (mean percentage of left ventricular mass) at 2 to 7 days increased from 1.14% in those treated with placebo to 3.11% (10 mg) and 5.20% (20 mg) in those treated with alteplase (Central Illustration). Similar results were observed when analyzing treatment as a 3-level or 2-level categorical variable (Table 3).Table 2 Pre-Specified Analyses of the Primary and Secondary Outcomes, Adjusting for Location of MI, by Subgroups of Ischemic Time Randomized Treatment Group Treatment Effect (Trend per 10-mg Dose Increase) n (Missing) Placebo (n = 151) Alteplase 10 mg (n = 144) Alteplase 20 mg (n = 145) Estimate (95% CI), p Value Interaction p Value Primary Outcome: Extent of MVO (% of LV Mass) at 2–7 Days Summaries of data on original scale (% of LV mass). Overall 396 (44) 2.32 ± 4.31 2.61 ± 4.49 3.48 ± 5.83 Ischemic time <2 h 98 (9) 1.35 ± 2.67 1.49 ± 2.71 2.73 ± 5.03 ≥2 but <4 h 215 (20) 3.01 ± 4.99 3.11 ± 5.28 3.16 ± 5.69 ≥4 to 6 h 83 (15) 1.14 ± 2.62 3.11 ± 4.58 5.20 ± 6.86 Summaries of data on square root–transformed scale, with treatment effect estimates (change in √MVO per 10-mg increase in alteplase dose); estimates reported for all patients, and by subgroups of ischemic time, with interaction test p value. Overall 396 (44) 0.91 ± 1.22 0.99 ± 1.28 1.15 ± 1.48 0.12 (−0.04 to 0.28), 0.128 Ischemic time 0.018 <2 h 98 (9) 0.63 ± 0.99 0.71 ± 1.01 0.93 ± 1.39 0.12 (−0.21 to 0.46), 0.470 ≥2 but <4 h 215 (20) 1.12 ± 1.33 1.10 ± 1.39 1.06 ± 1.44 −0.03 (−0.23 to 0.18), 0.791 ≥4 to 6 h 83 (15) 0.54 ± 0.94 1.14 ± 1.37 1.64 ± 1.61 0.56 (0.21 to 0.91), 0.009

1.22 0.99 ± 1.28 1.15 ± 1.48 0.12 (−0.04 to 0.28), 0.128 Ischemic time 0.018 <2 h 98 (9) 0.63 ± 0.99 0.71 ± 1.01 0.93 ± 1.39 0.12 (−0.21 to 0.46), 0.470 ≥2 but <4 h 215 (20) 1.12 ± 1.33 1.10 ± 1.39 1.06 ± 1.44 −0.03 (−0.23 to 0.18), 0.791 ≥4 to 6 h 83 (15) 0.54 ± 0.94 1.14 ± 1.37 1.64 ± 1.61 0.56 (0.21 to 0.91), 0.009 Secondary Outcomes

1.22 0.99 ± 1.28 1.15 ± 1.48 0.12 (−0.04 to 0.28), 0.128 Ischemic time 0.018 <2 h 98 (9) 0.63 ± 0.99 0.71 ± 1.01 0.93 ± 1.39 0.12 (−0.21 to 0.46), 0.470 ≥2 but <4 h 215 (20) 1.12 ± 1.33 1.10 ± 1.39 1.06 ± 1.44 −0.03 (−0.23 to 0.18), 0.791 ≥4 to 6 h 83 (15) 0.54 ± 0.94 1.14 ± 1.37 1.64 ± 1.61 0.56 (0.21 to 0.91), 0.009 Secondary Outcomes MVO present at 2–7 days. Treatment effect reported as odds ratio per 10-mg increase in alteplase dose. Overall 396 (44) 59 (43.4) 58 (45.0) 59 (45.0) 1.04 (0.82 to 1.33), 0.733 Ischemic time 0.076 <2 h 98 (9) 9 (33.3) 16 (40.0) 11 (35.5) 1.01 (0.59 to 1.73), 0.966 ≥2 but <4 h 215 (20) 42 (50.6) 28 (47.5) 32 (43.8) 0.88 (0.64 to 1.20), 0.411 ≥4 to 6 h 85 (15) 8 (30.8) 14 (46.7) 16 (59.3) 1.84 (1.04 to 3.24), 0.036 Myocardial hemorrhage (% of LV mass) at 2–7 days. Treatment effect reported as mean change per 10-mg increase in alteplase dose. Overall 360 (80) 1.56 ± 3.78 1.98 ± 3.68 2.45 ± 4.80 0.46 (−0.005 to 0.97), 0.075 Ischemic time 0.038 <2 h 90 (17) 0.26 ± 0.71 1.21 ± 2.60 1.37 ± 2.48 0.42 (−0.67 to 1.52), 0.449 ≥2 but <4 h 196 (39) 2.32 ± 4.62 2.34 ± 4.10 2.38 ± 4.92 0.04 (−0.62 to 0.70), 0.903 ≥4 to 6 h 74 (24) 0.48 ± 1.27 2.39 ± 4.08 3.95 ± 6.19 1.74 (0.61 to 2.87), 0.003 Myocardial hemorrhage present at 2–7 days. Treatment effect reported as odds ratio per 10-mg increase in alteplase dose. Overall 378 (62) 52 (40.6) 54 (44.6) 56 (43.4) 1.07 (0.83 to 1.37), 0.603 Ischemic time 0.044 <2 h 96 (11) 7 (26.9) 15 (38.5) 11 (35.5) 1.16 (0.67 to 2.02), 0.597 ≥2 but <4 h 202 (33) 38 (49.4) 25 (46.3) 29 (40.8) 0.85 (0.61 to 1.18), 0.324 ≥4 to 6 h 80 (18) 7 (28.0) 14 (50.0) 16 (59.3) 1.93 (1.09 to3.45), 0.025 Infarct size (% of LV mass) at 2–7 days. Data analyzed on original scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Overall 396 (44) 26.3 ± 13.7 27.3 ± 12.4 26.7 ± 13.4 0.19 (−1.23 to 1.62), 0.7921 Ischemic time 0.527 <2 h 98 (9) 22.9 ± 15.4 25.9 ± 13.5 24.3 ± 15.0 −0.18 (−3.25 to 2.89), 0.908 ≥2 but <4 h 215 (20) 28.0 ± 13.9 27.3 ± 11.9 27.3 ± 13.5 −0.23 (−2.10 to 1.63), 0.807 ≥4 to 6 h 83 (15) 24.5 ± 10.6 29.1 ± 12.1 27.6 ± 11.5 1.85 (−1.36 to 5.05), 0.258 LV ejection fraction at 2–7 days. Treatment effect reported as mean change per 10-mg increase in alteplase dose.

4.3 ± 15.0 −0.18 (−3.25 to 2.89), 0.908 ≥2 but <4 h 215 (20) 28.0 ± 13.9 27.3 ± 11.9 27.3 ± 13.5 −0.23 (−2.10 to 1.63), 0.807 ≥4 to 6 h 83 (15) 24.5 ± 10.6 29.1 ± 12.1 27.6 ± 11.5 1.85 (−1.36 to 5.05), 0.258 LV ejection fraction at 2–7 days. Treatment effect reported as mean change per 10-mg increase in alteplase dose. Overall 400 (40) 44.5 ± 8.8 43.6 ± 8.1 44.2 ± 8.4 −0.2 (−1.1 to 0.8), 0.748 Ischemic time 0.105 <2 h 99 (8) 45.2 ± 8.3 45.1 ± 7.3 45.2 ± 7.1 0.4 (−1.6 to 2.5), 0.664 ≥2 but <4 h 216 (19) 43.4 ± 9.5 44.2 ± 8.0 44.2 ± 8.8 0.3 (−0.9 to 1.5), 0.617 ≥4 to 6 h 85 (13) 47.0 ± 6.2 40.7 ± 8.8 42.9 ± 8.7 −2.2 (−4.3 to −0.1), 0.041 LV end-systolic volume at 2–7 days. Data analyzed on a logarithmic scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Overall 400 (40) 95.8 ± 29.8 104.1 ± 33.0 96.6 ± 30.8 1.00 (0.97 to 1.04), 0.897 Ischemic time 0.277 <2 h 99 (8) 87.3 ± 23.7 102.0 ± 29.2 96.1 ± 30.4 1.03 (0.95 to 1.11), 0.470 ≥2 but <4 h 216 (19) 100.5 ± 32.4 101.2 ± 34.8 95.9 ± 31.1 0.98 (0.93 to 1.02), 0.359 ≥4 to 6 h 85 (13) 90.2 ± 24.2 112.1 ± 33.6 99.2 ± 31.3 1.05 (0.97 to 1.13), 0.269 LV end-diastolic volume at 2–7 days. Data analyzed on a logarithmic scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Overall 400 (40) 171.1 ± 36.5 182.5 ± 40.8 171.4 ± 40.1 1.00 (0.97 to 1.03), 0.960 Ischemic time 0.332 <2 h 99 (8) 160.1 ± 38.8 183.8 ± 36.2 173.7 ± 43.4 1.03 (0.98 to 1.10), 0.245 ≥2 but <4 h 216 (19) 175.5 ± 35.6 178.9 ± 43.0 170.3 ± 39.3 0.98 (0.95 to 1.02), 0.366 ≥4 to 6 h 85 (13) 169.0 ± 35.7 187.4 ± 42.6 171.8 ± 39.5 1.01 (0.95 to 1.07), 0.831 Myocardial salvage (% LV) at 2–7 days. Data analyzed on original scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Overall 396 (44) 14.12 ± 9.17 14.60 ± 9.78 14.35 ± 10.17 0.04 (−1.10 to 1.17), 0.948 Ischemic time 0.125 <2 h 98 (9) 17.84 ± 11.16 15.32 ± 8.48 20.16 ± 11.11 1.09 (−1.35 to 3.52), 0.382 ≥2 but <4 h 215 (20) 12.75 ± 8.16 14.71 ± 10.13 13.51 ± 9.14 0.43 (−1.05 to 1.92), 0.566 ≥4 to 6 h 83 (15) 14.62 ± 9.22 13.41 ± 10.88 9.96 ± 9.00 −2.27 (−4.81 to 0.27), 0.080 Area under the troponin T (mg/l) curve, 0–24 h. Data analyzed on a logarithmic scale; treatment effect reported as ratios per 10-mg increase in alteplase dose.

5 (20) 12.75 ± 8.16 14.71 ± 10.13 13.51 ± 9.14 0.43 (−1.05 to 1.92), 0.566 ≥4 to 6 h 83 (15) 14.62 ± 9.22 13.41 ± 10.88 9.96 ± 9.00 −2.27 (−4.81 to 0.27), 0.080 Area under the troponin T (mg/l) curve, 0–24 h. Data analyzed on a logarithmic scale; treatment effect reported as ratios per 10-mg increase in alteplase dose. Overall 317 (123) 4.54 ± 5.58 5.94 ± 7.53 5.84 ± 6.23 1.25 (1.07 to 1.46), 0.006 Ischemic time 0.191 <2 h 85 (22) 3.40 ± 5.62 3.63 ± 3.96 5.27 ± 5.53 1.41 (1.02 to 1.95), 0.036 ≥2 but <4 h 163 (72) 5.42 ± 5.96 6.23 ± 7.07 5.59 ± 5.50 1.10 (0.90 to 1.36), 0.353 ≥4 to 6 h 69 (29) 3.09 ± 3.44 8.43 ± 10.46 7.17 ± 8.53 1.54 (1.08 to 2.20), 0.016 Values are mean ± SD or n (%), unless otherwise indicated. All outcomes were pre-specified. Treatment effect estimates derived from linear or logistic regression models, modelling the treatment effect as a linear trend across alteplase dose groups (0 mg vs. 10 mg vs. 20 mg). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. Treatment effect estimates and tests of interaction are based on models assuming a linear trend with alteplase dose. The p values and 95% CI have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. CI = confidence interval; LV = left ventricular; MI = myocardial infarction; MVO = microvascular obstruction.

Overall 317 (123) 4.54 ± 5.58 5.94 ± 7.53 5.84 ± 6.23 1.25 (1.07 to 1.46), 0.006 Ischemic time 0.191 <2 h 85 (22) 3.40 ± 5.62 3.63 ± 3.96 5.27 ± 5.53 1.41 (1.02 to 1.95), 0.036 ≥2 but <4 h 163 (72) 5.42 ± 5.96 6.23 ± 7.07 5.59 ± 5.50 1.10 (0.90 to 1.36), 0.353 ≥4 to 6 h 69 (29) 3.09 ± 3.44 8.43 ± 10.46 7.17 ± 8.53 1.54 (1.08 to 2.20), 0.016 Values are mean ± SD or n (%), unless otherwise indicated. All outcomes were pre-specified. Treatment effect estimates derived from linear or logistic regression models, modelling the treatment effect as a linear trend across alteplase dose groups (0 mg vs. 10 mg vs. 20 mg). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. Treatment effect estimates and tests of interaction are based on models assuming a linear trend with alteplase dose. The p values and 95% CI have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. CI = confidence interval; LV = left ventricular; MI = myocardial infarction; MVO = microvascular obstruction. Central Illustration Efficacy of Intracoronary Alteplase and Mechanism of Increased Microvascular Injury in Patients With an Ischemic Time of ≥4 to 6 h

Overall 317 (123) 4.54 ± 5.58 5.94 ± 7.53 5.84 ± 6.23 1.25 (1.07 to 1.46), 0.006 Ischemic time 0.191 <2 h 85 (22) 3.40 ± 5.62 3.63 ± 3.96 5.27 ± 5.53 1.41 (1.02 to 1.95), 0.036 ≥2 but <4 h 163 (72) 5.42 ± 5.96 6.23 ± 7.07 5.59 ± 5.50 1.10 (0.90 to 1.36), 0.353 ≥4 to 6 h 69 (29) 3.09 ± 3.44 8.43 ± 10.46 7.17 ± 8.53 1.54 (1.08 to 2.20), 0.016 Values are mean ± SD or n (%), unless otherwise indicated. All outcomes were pre-specified. Treatment effect estimates derived from linear or logistic regression models, modelling the treatment effect as a linear trend across alteplase dose groups (0 mg vs. 10 mg vs. 20 mg). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. Treatment effect estimates and tests of interaction are based on models assuming a linear trend with alteplase dose. The p values and 95% CI have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. CI = confidence interval; LV = left ventricular; MI = myocardial infarction; MVO = microvascular obstruction. Central Illustration Efficacy of Intracoronary Alteplase and Mechanism of Increased Microvascular Injury in Patients With an Ischemic Time of ≥4 to 6 h The flow diagram groups participants by ischemic time into 3 categories (≥4 to 6 h, n = 98; ≥2 but <4 h, n = 235; <2 h, n = 107), those with an ischemic time of 4 h or more are subgrouped according to treatment group allocation (placebo, n = 29; 10 mg alteplase n = 38; 20 mg alteplase, n = 31). The effect of intracoronary alteplase on the extent of microvascular obstruction and myocardial hemorrhage is shown, including the effect estimates. The estimated mean difference on a square root scale is shown for the extent of microvascular obstruction and the estimated mean difference for myocardial hemorrhage. There was a statistically significant increase in microvascular obstruction and myocardial hemorrhage extent in those patients receiving alteplase. CI = confidence interval.

ifference on a square root scale is shown for the extent of microvascular obstruction and the estimated mean difference for myocardial hemorrhage. There was a statistically significant increase in microvascular obstruction and myocardial hemorrhage extent in those patients receiving alteplase. CI = confidence interval. Table 3 Pre-Specified Analyses of the Primary and Secondary Outcomes, Adjusting for Location of MI, by Subgroups of Ischemic Time and Interactions With Treatment Group, Effect Estimates, and Interactions Treatment Effect (Alteplase 20 mg vs. Alteplase 10 mg vs. Placebo) Treatment Effect (Alteplase vs. Placebo) 10 mg vs. Placebo Estimate (95% CI), p Value 20 mg vs. Placebo Estimate (95% CI), p Value Interaction p Value Estimate (95% CI), p Value Interaction p Value Primary Outcome Extent of MVO (% of LV mass) at 2–7 days. Treatment effects reported as mean differences in square root–transformed MVO between treatment groups (each dose vs. placebo separately, and both active treatment groups combined vs. placebo). Overall 0.11 (−0.21 to 0.43), 0.511 0.24 (−0.07 to 0.56), 0.128 0.18 (−0.10 to 0.45), 0.204 Ischemic time 0.090 0.061 <2 h 0.09 (−0.55 to 0.73), 0.783 0.25 (−0.43 to 0.92), 0.476 0.16 (−0.42 to 0.74), 0.592 ≥2 but <4 h −0.01 (−0.45 to 0.42), 0.947 −0.06 (−0.47 to 0.35), 0.790 −0.04 (−0.40 to 0.32), 0.837 ≥4 to 6 h 0.53 (−0.15 to 1.22), 0.128 1.12 (0.42 to 1.82), 0.002 0.81 (0.21 to 1.42), 0.009 Secondary Outcomes

Extent of MVO (% of LV mass) at 2–7 days. Treatment effects reported as mean differences in square root–transformed MVO between treatment groups (each dose vs. placebo separately, and both active treatment groups combined vs. placebo). Overall 0.11 (−0.21 to 0.43), 0.511 0.24 (−0.07 to 0.56), 0.128 0.18 (−0.10 to 0.45), 0.204 Ischemic time 0.090 0.061 <2 h 0.09 (−0.55 to 0.73), 0.783 0.25 (−0.43 to 0.92), 0.476 0.16 (−0.42 to 0.74), 0.592 ≥2 but <4 h −0.01 (−0.45 to 0.42), 0.947 −0.06 (−0.47 to 0.35), 0.790 −0.04 (−0.40 to 0.32), 0.837 ≥4 to 6 h 0.53 (−0.15 to 1.22), 0.128 1.12 (0.42 to 1.82), 0.002 0.81 (0.21 to 1.42), 0.009 Secondary Outcomes MVO present at 2–7 days. Treatment effects reported as odds ratios between groups. Overall 0.12 (0.68 to 1.84), 0.651 1.09 (0.67 to 1.77), 0.734 1.10 (0.72 to 1.69), 0.647 Ischemic time 0.240 0.147 <2 h 1.35 (0.49 to 3.78), 0.561 1.04 (0.35 to 3.11), 0.940 1.21 (0.47 to 3.09), 0.689 ≥2 but <4 h 0.89 (0.45 to 1.74), 0.726 0.77 (0.41 to 1.45), 0.410 0.82 (0.47 to 1.42), 0.476 ≥4 to 6 h 1.86 (0.61 to 5.61), 0.272 3.38 (1.08 to 10.55), 0.036 2.46 (0.92 to 6.60), 0.073 Myocardial hemorrhage (% of LV mass) at 2–7 days. Treatment effects reported as mean differences between groups. Overall 0.55 (−0.50 to 1.60), 0.304 0.93 (−0.09 to 1.94), 0.074 0.75 (−0.14 to 1.65), 0.100 Ischemic time 0.149 0.097 <2 h 0.94 (−1.16 to 3.05), 0.380 0.90 (−1.31 to 3.10), 0.425 0.93 (−0.99 to 2.85), 0.343 ≥2 but <4 h 0.00 (−1.43 to 1.44), 0.996 0.08 (−1.24 to 1.41), 0.903 0.05 (−1.12 to 1.22), 0.935 ≥4 to 6 h 1.63 (−0.64 to 3.91), 0.160 3.49 (1.22 to 5.75), 0.003 2.57 (0.59 to 4.54), 0.011 Myocardial hemorrhage present at 2–7 days. Treatment effects reported as odds ratios between groups. Overall 1.25 (0.75 to 2.08), 0.401 1.14 (0.69 to 1.89), 0.598 1.19 (0.77 to 1.85), 0.434 Ischemic time 0.150 0.059 <2 h 1.72 (0.58 to 5.09), 0.327 1.41 (0.45 to 4.43), 0.554 1.58 (0.58 to 4.28), 0.369 ≥2 but <4 h 0.89 (0.44 to 1.80), 0.748 0.72 (0.37 to 1.38), 0.322 0.79 (0.45 to 1.40), 0.418 ≥4 to 6 h 2.42 (0.76 to 7.64), 0.133 3.81 (1.19 to 12.25), 0.025 3.02 (1.08 to 8.42), 0.035 Infarct size (% of LV mass) at 2–7 days. Data analyzed on original scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Overall 1.23 (−1.66 to 4.11), 0.404 0.38 (−2.48 to 3.23), 0.795 0.79 (−1.68 to 3.27), 0.530 Ischemic time 0.600 0.475 <2 h 3.36 (−2.45 to 9.16), 0.258 −0.17 (−6.32 to 5.97), 0.956 1.82 (−3.44 to 7.08), 0.498 ≥2 but <4 h −0.47 (−4.44 to 3.50), 0.818 −0.46 (−4.20 to 3.28), 0.810 −0.46 (−3.73 to 2.80), 0.780 ≥4 to 6 h 2.85 (−3.41 to 9.11), 0.373 3.71 (−2.70 to 10.12), 0.257 3.26 (−2.25 to 8.77), 0.246 LV ejection fraction at 2–7 days. Treatment effects reported as mean differences between groups.

6 1.82 (−3.44 to 7.08), 0.498 ≥2 but <4 h −0.47 (−4.44 to 3.50), 0.818 −0.46 (−4.20 to 3.28), 0.810 −0.46 (−3.73 to 2.80), 0.780 ≥4 to 6 h 2.85 (−3.41 to 9.11), 0.373 3.71 (−2.70 to 10.12), 0.257 3.26 (−2.25 to 8.77), 0.246 LV ejection fraction at 2–7 days. Treatment effects reported as mean differences between groups. Overall −0.9 (−2.8 to 1.0), 0.367 −0.3 (−2.2 to 1.6), 0.752 −0.6 (−2.2 to 1.1), 0.483 Ischemic time 0.117 0.027 <2 h −0.2 (−4.0 to 3.6), 0.915 0.9 (−3.2 to 4.9), 0.679 0.3 (−3.2 to 3.7), 0.884 ≥2 but <4 h 0.7 (−2.0 to 3.3), 0.616 0.6 (−1.8 to 3.1), 0.621 0.6 (−1.5 to 2.8), 0.557 ≥4 to 6 h −5.5 (−9.5 to −1.4), 0.008 −4.5 (−8.7 to −0.2), 0.039 −5.0 (−8.6 to −1.4), 0.007 LV end-systolic volume at 2–7 days. Data analyzed on a logarithmic scale; treatment effects reported as relative differences between groups. Overall 1.08 (1.01 to 1.16), 0.027 1.00 (0.94 to 1.08), 0.907 1.04 (0.98 to 1.11), 0.184 Ischemic time 0.222 0.053 <2 h 1.17 (1.01 to 1.34), 0.032 1.06 (0.92 to 1.23), 0.424 1.12 (0.99 to 1.27), 0.082 ≥2 but <4 h 1.00 (0.91 to 1.11), 0.943 0.96 (0.87 to 1.05), 0.346 0.98 (0.90 to 1.06), 0.576 ≥4 to 6 h 1.21 (1.04 to 1.40), 0.016 1.10 (0.94 to 1.28), 0.255 1.15 (1.01 to 1.32), 0.038 LV end-diastolic volume at 2–7 days. Data analyzed on a logarithmic scale; treatment effects reported as relative differences between groups. Overall 1.06 (1.01 to 1.12), 0.947 1.00 (0.95 to 1.05), 0.947 1.03 (0.98 to 1.08), 0.210 Ischemic time 0.314 0.081 <2 h 1.16 (1.04 to 1.29), 0.007 1.08 (0.96 to 1.21), 0.211 1.12 (1.02 to 1.24), 0.021 ≥2 but <4 h 1.01 (0.94 to 1.09), 0.743 0.97 (0.90 to 1.04), 0.349 0.99 (0.93 to 1.05), 0.676 ≥4 to 6 h 1.10 (0.98 to 1.23), 0.117 1.01 (0.90 to 1.14), 0.814 1.06 (0.95 to 1.17), 0.282 Myocardial salvage (% LV) at 2–7 days. Data analyzed on original scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Overall 0.04 (−2.26 to 2.34), 0.975 0.08 (−2.20 to 2.35), 0.948 0.06 (−1.92 to 2.03), 0.955 Ischemic time 0.071 0.337 <2 h −2.44 (−7.03 to 2.16), 0.299 1.99 (−2.87 to 6.86), 0.422 −0.51 (−4.70 to 3.68), 0.811 ≥2 but <4 h 2.00 (−1.14 to 5.14), 0.212 0.81 (−2.15 to 3.77), 0.592 1.35 (−1.25 to 3.94), 0.309 ≥4 to 6 h −1.60 (−6.55 to 3.35), 0.527 −4.54 (−9.61 to 0.53), 0.080 −3.00 (−7.39 to 1.38), 0.179 Area under the troponin T (mg/l) curve, 0–24 h. Data analyzed on a logarithmic scale; treatment effects reported as relative differences between groups.

o 5.14), 0.212 0.81 (−2.15 to 3.77), 0.592 1.35 (−1.25 to 3.94), 0.309 ≥4 to 6 h −1.60 (−6.55 to 3.35), 0.527 −4.54 (−9.61 to 0.53), 0.080 −3.00 (−7.39 to 1.38), 0.179 Area under the troponin T (mg/l) curve, 0–24 h. Data analyzed on a logarithmic scale; treatment effects reported as relative differences between groups. Overall 1.61 (1.17 to 2.22), 0.003 1.56 (1.14 to 2.13), 0.006 1.58 (1.21 to 2.08), 0.001 Ischemic time 0.257 0.081 <2 h 1.76 (0.96 to 3.21), 0.067 2.00 (1.05 to 3.79), 0.034 1.86 (1.08 to 3.19), 0.025 ≥2 but <4 h 1.24 (0.78 to 1.96), 0.362 1.21 (0.80 to 1.83), 0.363 1.22 (0.85 to 1.76), 0.278 ≥4 to 6 h 2.82 (1.44 to 5.54), 0.003 2.44 (1.20 to 4.94), 0.014 2.64 (1.44 to 4.83), 0.002 All outcomes were pre-specified. Treatment effect estimates derived from linear or logistic regression models, modelling the treatment effect as a 3-level categorical variable or as a 2-level categorical variable (active vs. placebo). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. P values and 95% CI presented in this table have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. Abbreviations as in Table 2.

Overall 1.61 (1.17 to 2.22), 0.003 1.56 (1.14 to 2.13), 0.006 1.58 (1.21 to 2.08), 0.001 Ischemic time 0.257 0.081 <2 h 1.76 (0.96 to 3.21), 0.067 2.00 (1.05 to 3.79), 0.034 1.86 (1.08 to 3.19), 0.025 ≥2 but <4 h 1.24 (0.78 to 1.96), 0.362 1.21 (0.80 to 1.83), 0.363 1.22 (0.85 to 1.76), 0.278 ≥4 to 6 h 2.82 (1.44 to 5.54), 0.003 2.44 (1.20 to 4.94), 0.014 2.64 (1.44 to 4.83), 0.002 All outcomes were pre-specified. Treatment effect estimates derived from linear or logistic regression models, modelling the treatment effect as a 3-level categorical variable or as a 2-level categorical variable (active vs. placebo). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. P values and 95% CI presented in this table have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. Abbreviations as in Table 2. Secondary outcomes CMR parameters at 2 to 7 days Although the interaction between ischemic time and treatment in relation to the binary outcome of the presence of any MVO did not reach statistical significance (odds ratio [OR]: 1.84; 95% CI: 1.04 to 3.24; p = 0.036; interaction p = 0.076), there was a trend toward a higher prevalence with increasing dose in patients presenting ≥4 to 6 h (Table 2), but no evidence of a treatment effect with shorter ischemic times. A similar pattern was observed regarding myocardial hemorrhage, with an increasing prevalence in those with ischemic times ≥4 to 6 h (OR per 10-mg increase in alteplase dose: 1.93; 95% CI: 1.09 to 3.45; p = 0.025), but no significant trend with shorter ischemic times (p value for interaction = 0.044), as well as an increasing extent of myocardial hemorrhage (percentage of left ventricular mass) in those with longer ischemic times (1.74% increase per 10-mg increase in alteplase dose; 95% CI: 0.61 to 2.87; p = 0.003) (Central Illustration), but no evidence of treatment effects with shorter ischemic times (p for interaction = 0.038). The statistical evidence for interactions was weaker when considering treatment effects categorically (Table 3), but the general pattern of associations was very similar, with poorer outcomes in those treated with alteplase (particularly the 20-mg dose) when the ischemic time was ≥4 to 6 h. Similar trends were observed when patients were categorized by the location of MI, anterior and nonanterior (Table 4).Table 4 Pre-Specified Analyses of the Primary and Selected Secondary Outcomes, Adjusting for Location of MI, by Subgroups of Ischemic Time and MI Location (Anterior/Nonanterior)

emic time was ≥4 to 6 h. Similar trends were observed when patients were categorized by the location of MI, anterior and nonanterior (Table 4).Table 4 Pre-Specified Analyses of the Primary and Selected Secondary Outcomes, Adjusting for Location of MI, by Subgroups of Ischemic Time and MI Location (Anterior/Nonanterior) Randomized Treatment Group Treatment Effect (Trend per 10-mg Dose Increase) n (Missing) Placebo (n = 151) Alteplase 10 mg (n = 144) Alteplase 20 mg (n = 145) Estimate (95% CI), p Value Interaction p Value Primary Outcome Extent of MVO (% of LV mass) at 2–7 days. Summaries of data on square root–transformed scale, with treatment effect estimates (change in √MVO per 10-mg increase in alteplase dose); estimates reported for all patients, and by subgroups of ischemic time, with interaction test p value. Anterior MI Overall 178 (15) 1.16 ± 1.44 1.08 ± 1.28 1.42 ± 1.64 0.15 (−0.11 to 0.42), 0.249 Ischemic time 0.264 <2 h 50 (4) 0.81 ± 1.16 0.77 ± 0.98 1.16 ± 1.51 0.19 (−0.32 to 0.70), 0.458 ≥2 but <4 h 96 (9) 1.37 ± 1.55 1.28 ± 1.40 1.41 ± 1.61 0.02 (−0.33 to 0.36), 0.929 ≥4 to 6 h 32 (2) 0.77 ± 1.26 1.11 ± 1.37 2.10 ± 2.06 0.65 (−0.04 to 1.34), 0.064 Nonanterior MI Overall 218 (29) 0.72 ± 0.99 0.91 ± 1.29 0.92 ± 1.29 0.10 (−0.09 to 0.29), 0.316 Ischemic time 0.049 <2 h 48 (5) 0.47 ± 0.81 0.66 ± 1.04 0.56 ± 1.12 0.05 (−0.40 to 0.51), 0.823 ≥2 but <4 h 119 (11) 0.91 ± 1.09 0.96 ± 1.39 0.78 ± 1.24 -0.06 (−0.31 to 0.19), 0.630 ≥4 to 6 h 51 (13) 0.42 ± 0.74 1.16 ± 1.42 1.45 ± 1.41 0.51 (0.13 to 0.90), 0.009 Secondary Outcomes

Extent of MVO (% of LV mass) at 2–7 days. Summaries of data on square root–transformed scale, with treatment effect estimates (change in √MVO per 10-mg increase in alteplase dose); estimates reported for all patients, and by subgroups of ischemic time, with interaction test p value. Anterior MI Overall 178 (15) 1.16 ± 1.44 1.08 ± 1.28 1.42 ± 1.64 0.15 (−0.11 to 0.42), 0.249 Ischemic time 0.264 <2 h 50 (4) 0.81 ± 1.16 0.77 ± 0.98 1.16 ± 1.51 0.19 (−0.32 to 0.70), 0.458 ≥2 but <4 h 96 (9) 1.37 ± 1.55 1.28 ± 1.40 1.41 ± 1.61 0.02 (−0.33 to 0.36), 0.929 ≥4 to 6 h 32 (2) 0.77 ± 1.26 1.11 ± 1.37 2.10 ± 2.06 0.65 (−0.04 to 1.34), 0.064 Nonanterior MI Overall 218 (29) 0.72 ± 0.99 0.91 ± 1.29 0.92 ± 1.29 0.10 (−0.09 to 0.29), 0.316 Ischemic time 0.049 <2 h 48 (5) 0.47 ± 0.81 0.66 ± 1.04 0.56 ± 1.12 0.05 (−0.40 to 0.51), 0.823 ≥2 but <4 h 119 (11) 0.91 ± 1.09 0.96 ± 1.39 0.78 ± 1.24 -0.06 (−0.31 to 0.19), 0.630 ≥4 to 6 h 51 (13) 0.42 ± 0.74 1.16 ± 1.42 1.45 ± 1.41 0.51 (0.13 to 0.90), 0.009 Secondary Outcomes MVO present at 2–7 days. Treatment effect reported as odds ratio per 10-mg increase in alteplase dose. Anterior MI Overall 178 (15) 28 (46.7) 29 (49.2) 31 (52.5) 1.16 (0.80 to 1.66), 0.437 Ischemic time 0.613 <2 h 50 (4) 5 ± 38.5 8 ± 44.4 8 ± 42.1 1.07 (0.52 to 2.17), 0.862 ≥2 but <4 h 96 (9) 20 ± 52.6 14 ± 53.8 18 ± 56.2 1.08 (0.67 to 1.72), 0.764 ≥4 to 6 h 32 (2) 3 ± 33.3 7 ± 46.7 5 ± 62.5 1.82 (0.67 to 4.94), 0.237 Nonanterior MI Overall 218 (29) 31 (40.8) 29 (41.4) 28 (38.9) 0.96 (0.69 to 1.33), 0.788 Ischemic time 0.089 <2 h 48 (5) 4 ± 28.6 8 ± 36.4 3 ± 25.0 0.93 (0.41 to 2.15), 0.874 ≥2 but <4 h 119 (11) 22 ± 48.9 14 ± 42.4 14 ± 34.1 0.74 (0.48 to 1.14), 0.169 ≥4 to 6 h 51 (13) 5 ± 29.4 7 ± 46.7 11 ± 57.9 1.81 (0.91 to 3.59), 0.092 Myocardial hemorrhage (% of LV mass) at 2–7 days. Treatment effect reported as mean change per 10-mg increase in alteplase dose. Anterior MI Overall 148 (35) 2.29 ± 5.13 2.22 ± 3.54 3.41 ± 5.81 0.69 (−0.25 to 1.63), 0.148 Ischemic time 0.079 <2 h 44 (10) 0.38 ± 0.85 1.43 ± 2.94 1.86 ± 2.72 0.71 (−1.13 to 2.54), 0.450 ≥2 but <4 h 87 (18) 3.30 ± 6.07 2.89 ± 4.07 3.52 ± 6.01 0.10 (−1.09 to 1.29), 0.867 ≥4 to 6 h 27 (7) 0.26 ± 0.50 1.94 ± 3.14 6.89 ± 9.33 3.32 (0.79 to 5.84), 0.010 Nonanterior MI Overall 202 (45) 1.05 ± 2.34 1.78 ± 3.82 1.70 ± 3.70 0.32 (−0.23 to 0.86), 0.255 Ischemic time 0.250 <2 h 46 (7) 0.18 ± 0.61 1.04 ± 2.38 0.63 ± 1.95 0.24 (−1.06 to 1.53), 0.721 ≥2 but <4 h 109 (21) 1.52 ± 2.84 1.87 ± 4.14 1.52 ± 3.76 0.00 (−0.72 to 0.72), 1.000 ≥4 to 6 h 47 (17) 0.57 ± 1.49 2.88 ± 5.01 2.80 ± 4.29 1.11 (0.01 to 2.20), 0.047 Myocardial hemorrhage present at 2–7 days. Treatment effect reported as odds ratio per 10-mg increase in alteplase dose.

1.95 0.24 (−1.06 to 1.53), 0.721 ≥2 but <4 h 109 (21) 1.52 ± 2.84 1.87 ± 4.14 1.52 ± 3.76 0.00 (−0.72 to 0.72), 1.000 ≥4 to 6 h 47 (17) 0.57 ± 1.49 2.88 ± 5.01 2.80 ± 4.29 1.11 (0.01 to 2.20), 0.047 Myocardial hemorrhage present at 2–7 days. Treatment effect reported as odds ratio per 10-mg increase in alteplase dose. Anterior MI Overall 168 (25) 25 (44.6) 27 (49.1) 28 (49.1) 1.12 (0.77 to 1.62), 0.558 Ischemic time 0.661 <2 h 48 (6) 4 ± 33.3 8 ± 47.1 8 ± 42.1 1.16 (0.56 to 2.41), 0.689 ≥2 but <4 h 90 (15) 18 ± 50.0 12 ± 50.0 15 ± 50.0 1.00 (0.62 to 1.62), 1.000 ≥4 to 6 h 30 (4) 3 ± 37.5 7 ± 50.0 5 ± 62.5 1.67 (0.61 to 4.59), 0.323 Nonanterior MI Overall 210 (37) 27 (37.5) 27 (40.9) 28 (38.9) 1.02 (0.73 to 1.43), 0.910 Ischemic time 0.045 <2 h 48 (5) 3 ± 21.4 7 (31.8) 3 ± 25.0 1.11 (0.47 to 2.64), 0.811 ≥2 but <4 h 112 (18) 20 ± 48.8 13 (43.3) 14 ± 34.1 0.74 (0.47 to 1.15), 0.181 ≥4 to 6 h 50 (14) 4 ± 23.5 7 (50.0) 14 ± 34.1 2.06 (1.02 to 4.18), 0.044 Infarct size (% of LV mass) at 2–7 days. Data analyzed on original scale; treatment effect reported as relative increase per 10-mg increase in alteplase dose. Anterior MI Overall 178 (15) 33.1 ± 14.3 33.8 ± 11.9 31.9 ± 15.3 −0.40 (−2.90 to 2.10), 0.756 Ischemic time 0.453 <2 h 50 (8) 27.5 ± 17.0 34.5 ± 13.8 28.1 ± 16.4 −0.16 (−5.01 to 4.70), 0.950 ≥2 but <4 h 96 (19) 36.1 ± 13.6 32.6 ± 12.1 33.4 ± 14.7 −1.41 (−4.66 to 1.85), 0.3974 ≥4 to 6 h 32 (13) 28.7 ± 10.3 35.1 ± 9.5 35.1 ± 14.9 3.30 (−3.30 to 9.89), 0.327 Nonanterior MI Overall 218 (29) 20.9 ± 10.6 21.8 ± 10.0 22.3 ± 9.9 0.65 (−0.97 to 2.27), 0.431 Ischemic time 0.874 <2 h 48 (5) 18.7 ± 12.8 18.9 ± 8.2 18.2 ± 10.5 −0.22 (−4.11 to 3.67), 0.912 ≥2 but <4 h 119 (11) 21.1 ± 10.1 23.2 ± 10.1 22.6 ± 10.3 0.73 (−1.41 to 2.86), 0.505 ≥4 to 6 h 51 (13) 22.3 ± 10.3 23.2 ± 11.7 24.5 ± 8.2 1.10 (−2.20 to 4.40), 0.513 Values are mean ± SD or n (%), unless otherwise indicated. All outcomes were pre-specified. Treatment effect estimates derived from linear or logistic regression models, modelling the treatment effect as a linear trend across alteplase dose groups (0 mg vs. 10 mg vs. 20 mg). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. Treatment effect estimates and tests of interaction are based on models assuming a linear trend with alteplase dose.

lase dose groups (0 mg vs. 10 mg vs. 20 mg). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. Treatment effect estimates and tests of interaction are based on models assuming a linear trend with alteplase dose. The p values and 95% CI have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. Abbreviations as in Table 2.

lase dose groups (0 mg vs. 10 mg vs. 20 mg). Interaction test p values reported from regression models with ischemic time included as a 3-level categorical variable and interaction with treatment effect. Treatment effect estimates and tests of interaction are based on models assuming a linear trend with alteplase dose. The p values and 95% CI have not been adjusted for multiplicity, therefore these analyses should be interpreted as exploratory and not definitive. Abbreviations as in Table 2. Left ventricular ejection fraction 2 to 7 days post-STEMI was lower in patients presenting ≥4 to 6 h who were treated with alteplase (10 mg or 20 mg) compared with in those who received placebo (mean difference: −5.0%; 95% CI: −8.6% to −1.4%; p = 0.007) (Table 3), with no evidence of treatment effects (active vs. placebo) with shorter ischemic times (interaction p value = 0.027). The interaction with ischemic time was not statistically significant when treatment was assessed as a 3-level categorical variable (Table 3) or as a trend across treatment groups (Table 2), though the treatment effect estimates demonstrated a similar pattern. No significant interactions were observed between ischemic time and treatment for left ventricular end-systolic or end-diastolic volumes, regardless of how the treatment effect was modelled. Patterns of treatment effects in relation to left ventricular measures at 3 months were similar, though with fewer statistically significant associations (Supplemental Table 2). There was no evidence of any treatment effects in relation to infarct size, or myocardial salvage index at 2 to 7 days or 3 months.

modelled. Patterns of treatment effects in relation to left ventricular measures at 3 months were similar, though with fewer statistically significant associations (Supplemental Table 2). There was no evidence of any treatment effects in relation to infarct size, or myocardial salvage index at 2 to 7 days or 3 months. Blood chemistry The AUC for troponin T (ng/l) measured at baseline and 2 and 24 h post-reperfusion in 317 subjects was increased in both treatment groups compared with the placebo group, for those treated with alteplase, the relative difference was 1.53 (95% CI: 1.12 to 2.11; p = 0.008) (16). Troponin T AUC was 35% higher in patients treated with 20 mg of alteplase versus placebo. There was no interaction among troponin T AUC, ischemic time, and treatment with alteplase compared with placebo (Table 2). Hematology and coagulation, 2-h time point By 2 h after study drug administration, circulating concentrations of fibrin D-dimers were increased in the alteplase groups compared with in the placebo group (Supplemental Table 3). There were no statistically significant interactions observed for fibrin D-dimers, prothrombin F1 + 2 (a measure of thrombin activation), tissue plasminogen activator (a measure of endogenous tissue plasminogen activator and any circulating alteplase), plasminogen, or fibrinogen (Supplemental Table 4).

mental Table 3). There were no statistically significant interactions observed for fibrin D-dimers, prothrombin F1 + 2 (a measure of thrombin activation), tissue plasminogen activator (a measure of endogenous tissue plasminogen activator and any circulating alteplase), plasminogen, or fibrinogen (Supplemental Table 4). Discussion The principal findings from the T-TIME trial were that the intervention was feasible but not effective (16). Adjunctive, low-dose intracoronary alteplase administered after coronary reperfusion and before stent implantation did not reduce the amount of MVO revealed by cardiac CMR 2 to 7 days post-STEMI.

mental Table 3). There were no statistically significant interactions observed for fibrin D-dimers, prothrombin F1 + 2 (a measure of thrombin activation), tissue plasminogen activator (a measure of endogenous tissue plasminogen activator and any circulating alteplase), plasminogen, or fibrinogen (Supplemental Table 4). Discussion The principal findings from the T-TIME trial were that the intervention was feasible but not effective (16). Adjunctive, low-dose intracoronary alteplase administered after coronary reperfusion and before stent implantation did not reduce the amount of MVO revealed by cardiac CMR 2 to 7 days post-STEMI. In this pre-specified analysis, low-dose intracoronary alteplase administered during PPCI was associated with an increase in the amount of MVO in patients with an ischemic time of 4 h or more. When the interaction test between ischemic time and treatment was performed as a trend across treatment groups, we observed a statistically significant interaction, indicating a dose-dependent increase in MVO with alteplase in association with the duration of ischemia. An increase in the proportion of patients with myocardial hemorrhage as well as an increase in the amount of hemorrhage by ischemic time and treatment with alteplase (10 mg, 20 mg) was observed. These dose effects were driven by those patients receiving 20 mg of alteplase. In the subgroup of patients with the longest ischemic time (≥4 to 6 h), treatment with 20 mg alteplase was also associated with a lower left ventricular ejection fraction at 2 to 7 days. The results do not support this therapeutic approach, especially in those STEMI patients presenting with an ischemic time of 4 h or more, in whom MVO and myocardial hemorrhage may be increased. Clinical case examples are shown in Figure 2. Whether giving low-dose fibrinolysis at the end of PPCI in patients presenting with an ischemic time <4 h might be beneficial merits prospective assessment.Figure 2 Clinical Case Examples

ischemic time of 4 h or more, in whom MVO and myocardial hemorrhage may be increased. Clinical case examples are shown in Figure 2. Whether giving low-dose fibrinolysis at the end of PPCI in patients presenting with an ischemic time <4 h might be beneficial merits prospective assessment.Figure 2 Clinical Case Examples Two patients, both with acute lateral ST-segment elevation myocardial infarction treated successfully with primary percutaneous coronary intervention. Each patient had TIMI (Thrombolysis In Myocardial Infarction) flow grade 0 at initial angiography and TIMI flow grade 3 (normal flow grade) at the end of percutaneous coronary intervention. The first with an ischemic time of 5 h and the second 3 h. Cardiac magnetic resonance (CMR) was performed at 3 days post-reperfusion. (A) Patient with hemorrhagic infarction on CMR. Diagnostic coronary angiogram demonstrated an occluded circumflex artery (yellow arrow). T2*-CMR (far right) revealed myocardial hemorrhage (white arrow) within the infarct core. Late gadolinium-enhanced CMR revealed microvascular obstruction (middle, red arrow) within the bright area of infarction. The microvascular obstruction within the infarct core spatially corresponded with the myocardial hemorrhage. This represents a case of failed microvascular reperfusion despite successful percutaneous coronary intervention. (B) Patient with a lateral infarct but no CMR evidence of reperfusion injury. Diagnostic coronary angiogram demonstrated an occluded circumflex artery (yellow arrow). Late gadolinium-enhanced CMR revealed a lateral infarct with no evidence of microvascular obstruction and no evidence of hemorrhagic transformation on T2*-CMR. This represents a case of successful microvascular reperfusion.

injury. Diagnostic coronary angiogram demonstrated an occluded circumflex artery (yellow arrow). Late gadolinium-enhanced CMR revealed a lateral infarct with no evidence of microvascular obstruction and no evidence of hemorrhagic transformation on T2*-CMR. This represents a case of successful microvascular reperfusion. The mechanism for an increase in microvascular injury in patients with an ischemic time ≥4 to 6 h treated with alteplase likely involves hemorrhagic transformation within the infarct core. Prolonged ischemia leads to capillary degradation (26) and myocyte necrosis, and in these circumstances, alteplase appears to promote tissue hemorrhage. Myocardial hemorrhage underpins adverse left ventricular remodeling (27,28) and is independently predictive of an adverse cardiac prognosis in the longer term (27,29). An increase in the extravasation of blood into the interstitial space at the infarct core results in the external compression of the capillary bed with an associated exponential increase in microvascular resistance. This external compressive mass potentiates progression of microvascular damage. Myocardial hemorrhage is a pathological subset of MVO, as revealed by CMR imaging (27), in addition to the effect on microvascular injury, the increase in interstitial mass increases the extent of MVO as measured by CMR due to the associated mass effect.

compressive mass potentiates progression of microvascular damage. Myocardial hemorrhage is a pathological subset of MVO, as revealed by CMR imaging (27), in addition to the effect on microvascular injury, the increase in interstitial mass increases the extent of MVO as measured by CMR due to the associated mass effect. We observed a close relationship between MVO and myocardial hemorrhage. Myocardial hemorrhage did not occur in the absence of MVO, although hemorrhage was present in the majority of patients with MVO, it was not universal. We found that myocardial hemorrhage occurred in all patients with MVO who presented with a prolonged ischemic time (4 h or more) who then went on to receive alteplase (10-mg or 20-mg dose). This was not the case in those patients receiving placebo or in patients with an ischemic time of <4 h. This increase in the proportion of patients with myocardial hemorrhage versus MVO without myocardial hemorrhage by ischemic time and treatment group highlights the potential deleterious effects of adjunctive alteplase in patients with established microvascular injury. The increase in the extent of myocardial hemorrhage may be explained by the observation that MVO and myocardial hemorrhage are the same phenomenon in the majority of cases. A multicenter cohort study (30) previously reported increases in myocardial hemorrhage in patients receiving periprocedural glycoprotein IIb/IIIa inhibitor and an animal study (31) demonstrated an increased incidence of myocardial hemorrhage with the use of additional glycoprotein IIb/IIIa inhibitors. More aggressive antithrombotic treatment may promote tissue hemorrhage especially in the context of established microvascular injury.

cedural glycoprotein IIb/IIIa inhibitor and an animal study (31) demonstrated an increased incidence of myocardial hemorrhage with the use of additional glycoprotein IIb/IIIa inhibitors. More aggressive antithrombotic treatment may promote tissue hemorrhage especially in the context of established microvascular injury. The relationship between vascular permeability post-MI and tissue hemorrhage was highlighted in a study investigating the role of angiopoietin-like protein 2, which has been linked to endothelial cell junction stability and vascular permeability in mice. The investigators demonstrated that angiopoietin-like protein 2 mediates protection against post-ischemic tissue damage through preservation of the endothelial cell tissue barrier with associated reductions in myocardial hemorrhage and infarct size (26).

helial cell junction stability and vascular permeability in mice. The investigators demonstrated that angiopoietin-like protein 2 mediates protection against post-ischemic tissue damage through preservation of the endothelial cell tissue barrier with associated reductions in myocardial hemorrhage and infarct size (26). The detection of myocardial hemorrhage in vivo is limited by difficulty in obtaining reliable diagnostic quality images in a proportion of patients, which for T2* imaging typically requires long breath holds with minimal respiratory movement. This is highlighted by the observation that in our study, an assessment of the extent of MVO was possible in 396 of 440 participants compared with 360 of 440 for myocardial hemorrhage. This difference, reflecting a limitation in the diagnostic performance of T2* imaging, is comparable to previous reports (32). These results help explain why detection of myocardial hemorrhage may prove challenging, especially in those patients with a limited amount of myocardial hemorrhage. The result provides insights into why some patients may have detectable MVO but no myocardial hemorrhage. The overall clinical relevance of our findings is highlighted by a trend on ischemic time toward reduced ejection fraction in patients receiving alteplase versus placebo. We provide evidence that increased myocardial hemorrhage is causally related to a reduction in left ventricular function and adverse left ventricular remodeling.

clinical relevance of our findings is highlighted by a trend on ischemic time toward reduced ejection fraction in patients receiving alteplase versus placebo. We provide evidence that increased myocardial hemorrhage is causally related to a reduction in left ventricular function and adverse left ventricular remodeling. The T-TIME study has several strengths. The primary and secondary outcomes were analyzed using core laboratory methods. The study intervention and source data analyses were conducted in a double-blind manner, minimizing the risk of bias. The design specified multimodality testing including a time-course AUC analysis of the circulating concentrations of troponin T. The coagulation results have been useful to inform the safety of intracoronary alteplase as an adjunct during PPCI.

lyses were conducted in a double-blind manner, minimizing the risk of bias. The design specified multimodality testing including a time-course AUC analysis of the circulating concentrations of troponin T. The coagulation results have been useful to inform the safety of intracoronary alteplase as an adjunct during PPCI. MVO presents an unmet therapeutic need and there is widespread interest in the potential efficacy of intra-coronary fibrinolytic therapy during PPCI. Two multicenter, international trials are scheduled to investigate the efficacy of reduced doses of either alteplase (STRIVE [Adjunctive, Low-dose tPA in Primary PCI for STEMI]; NCT03335839) or tenecteplase (RESTORE-MI [Restoring Microcirculatory Perfusion in ST-Elevation Myocardial Infarction (STEMI)]; ACTRN12618000778280) (Supplemental Appendix). Considering eligibility criteria in these trials, the ischemic time limit is 12 h. Furthermore, RESTORE-MI selects patients with evidence of microvascular dysfunction (index of microcirculatory resistance >32) in the infarct-related artery at the end of PCI. Our results suggest this risk-based selection strategy may enroll patients at risk of myocardial hemorrhage that, based on our findings, may be exacerbated by intracoronary lytic therapy. The new knowledge from the T-TIME study seems relevant to the design of these trials and to clinicians in practice when considering the use of intracoronary alteplase as a bail-out option in patients with massive thrombosis. Finally, PPCI is not available for many patients due to both geographical and socioeconomic factors (33). As a result, intravenous thrombolysis is the primary reperfusion strategy for many STEMI patients worldwide. Our findings are potentially relevant for these patients. The GUSTO-1 (Global Utilization of t-Pa and Streptokinase for Occluded Coronary Arteries) trial evaluated the effects of intravenous thrombolysis in over 40,000 STEMI patients and those with a symptom onset to treatment time of 4 to 6 h had a >40% relative increase in mortality at 30 days when compared with patients with shorter treatment times (18). Increases in MVO and myocardial hemorrhage in patients with a prolonged ischemic time treated with thrombolysis may be a contributing factor for this increase in mortality. This could be considered by clinicians when a choice is available between prompt thrombolysis and delayed PCI beyond the guideline-directed 120-min target in patients with prolonged ischemic times.

ents with a prolonged ischemic time treated with thrombolysis may be a contributing factor for this increase in mortality. This could be considered by clinicians when a choice is available between prompt thrombolysis and delayed PCI beyond the guideline-directed 120-min target in patients with prolonged ischemic times. Study limitations First, the study was discontinued when pre-specified futility criteria were met. The objectives of this phase 2 trial included evidence synthesis for mechanisms evaluation as well as efficacy. To an extent, premature discontinuation limits mechanism evaluation. Second, although ischemic time was a pre-specified subgroup, no adjustment for multiplicity was made in this subgroup analysis. Finally, the decision to explore treatment effects as trends across treatment groups was made post hoc, this provided stronger evidence of the interaction based on ischemic time and treatment with alteplase. The results of this analysis should therefore be interpreted as exploratory and not definitive.

bgroup analysis. Finally, the decision to explore treatment effects as trends across treatment groups was made post hoc, this provided stronger evidence of the interaction based on ischemic time and treatment with alteplase. The results of this analysis should therefore be interpreted as exploratory and not definitive. Conclusions In patients presenting with acute STEMI and an ischemic time ≥4 to 6 h, adjunctive, low-dose, intracoronary alteplase given during PPCI may increase MVO and myocardial hemorrhage and reduce left ventricular ejection fraction. The mechanisms may involve hemorrhagic transformation within the infarct core. The results do not support administering intracoronary alteplase in patients with STEMI presenting with an ischemic time ≥4 to 6 h.Perspectives COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with acute STEMI and an ischemic time ≥4 to 6 h undergoing PPCI, low-dose intracoronary alteplase increases MVO and myocardial hemorrhage and worsens left ventricular function. TRANSLATIONAL OUTLOOK: Future studies of intracoronary thrombolysis should focus on patients presenting within 4 h of symptom onset. Appendix Online Data

Conclusions In patients presenting with acute STEMI and an ischemic time ≥4 to 6 h, adjunctive, low-dose, intracoronary alteplase given during PPCI may increase MVO and myocardial hemorrhage and reduce left ventricular ejection fraction. The mechanisms may involve hemorrhagic transformation within the infarct core. The results do not support administering intracoronary alteplase in patients with STEMI presenting with an ischemic time ≥4 to 6 h.Perspectives COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with acute STEMI and an ischemic time ≥4 to 6 h undergoing PPCI, low-dose intracoronary alteplase increases MVO and myocardial hemorrhage and worsens left ventricular function. TRANSLATIONAL OUTLOOK: Future studies of intracoronary thrombolysis should focus on patients presenting within 4 h of symptom onset. Appendix Online Data Acknowledgments The authors are grateful to the Efficacy and Mechanism Evaluation program of the National Institute for Health Research for its support. The University of Glasgow and Greater Glasgow and Clyde Health Board were independent co-sponsors of the trial. The authors acknowledge the support of the principal investigators, the trial steering committee, and the independent data monitoring committee. To the research nurses, the radiographers, clinical staff, and in particular the patients, the authors extend their gratitude.

Board were independent co-sponsors of the trial. The authors acknowledge the support of the principal investigators, the trial steering committee, and the independent data monitoring committee. To the research nurses, the radiographers, clinical staff, and in particular the patients, the authors extend their gratitude. The trial was funded by Efficacy and Mechanism Evaluation program of the National Institute for Health Research. Boehringer Ingelheim U.K. Ltd. provided the study drugs including alteplase (10 mg, 20 mg), matched placebo, and sterile water for injection. These organizations had no other involvement in the conduct of the study or in any aspect of this manuscript. The chief investigator had full access to the study data and had final responsibility for the decision to submit for publication. Dr. Maznyczka has received grants from British Heart Foundation Clinical Research Training Fellowship (FS/16/74/32573). Dr. McEntegart has a proctoring agreement with Boston Scientific and Vascular Perspectives. Prof. Oldroyd has received consultant and speaker fees from Abbott Vascular and Boston Scientific. Prof. McCann has received research grants from the National Institute for Health Research and the British Heart Foundation. Prof. Tait has received personal fees from Bayer Healthcare, Pfizer, Shire, Novo Nordisk, Sanofi, Sobi, CSL Behring, and Daiichi-Sankyo. Dr. Welsh has received grants from Boehringer Ingelheim, Roche, and AstraZeneca. Dr. McConnchie has received grants from Medical Research Council and the National Institute for Health Research—Efficacy and Mechanism Evaluation Programme. Prof. Berry is employed by the University of Glasgow, which holds consultancy and research agreements with Abbott Vascular, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, HeartFlow, Menarini, Ospens, Philips, and Siemens Healthcare. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

iversity of Glasgow, which holds consultancy and research agreements with Abbott Vascular, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, HeartFlow, Menarini, Ospens, Philips, and Siemens Healthcare. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Listen to this manuscript's audio summary by Editor-in-Chief Dr. Valentin Fuster onJACC.org. Appendix For supplemental methods and tables, please see the online version of this paper.

Healthcare systems need to quantify the survival benefit of interventions such as biventricular pacing (cardiac resynchronization therapy [CRT]) (1) and often discuss which patients gain most (2). Clinicians commonly use metrics such as relative risk reduction and number-needed-to-treat (NNT), although there are more sophisticated variables available such as quality-adjusted life-years (QALYs). For patients, however, the most easily understood metric is additional lifespan gained. Lifespan gain can be evaluated from trial data. It is the area between survival curves for the device and nondevice arms. However, trials rarely run to the lifespan of the device and, even if they do, staggered enrollment means few patients with long follow-up, making the later parts of the area between survival curves noisy. Whether addressing lifespan gain over a shorter period is acceptable has not been determined for biventricular pacing. It is also unclear whether a single device should be considered, rather than a commitment to sequential devices. Competing risks from other causes of death can be expected to rise with aging, but how much does this attenuate the lifespan benefit estimated per device?

le has not been determined for biventricular pacing. It is also unclear whether a single device should be considered, rather than a commitment to sequential devices. Competing risks from other causes of death can be expected to rise with aging, but how much does this attenuate the lifespan benefit estimated per device? In the current study, we examined the impact of time window over which lifespan gain is quantified, on the size of that lifespan gain, and on who gains most. We conducted this analysis in 3 ways. First, within the duration of randomized controlled trials of biventricular pacing, in terms of lifespan gain per patient; second, over a typical battery life of a biventricular pacemaker; and third, over >1 battery life, quantifiable as lifespan gain per patient or per device. Methods Search strategy We searched MEDLINE from inception to March 2013 using a combination of key words, including cardiac resynchronization therapy, biventricular pacemaker, mortality, survival, and randomized controlled trial. We also searched the bibliographies of published systematic reviews (3,4). Trials that compared biventricular pacing against no biventricular pacing and reported Kaplan-Meier survival curves for at least 6 months were identified. Trials with implantable cardioverter-defibrillator therapy were not excluded as long as it was present in both study arms. Data analysis We quantified at 3-month intervals the segmental area between the 2 curves (Fig. 1) and the cumulative area representing lifespan gained per patient up to that time.

Trials that compared biventricular pacing against no biventricular pacing and reported Kaplan-Meier survival curves for at least 6 months were identified. Trials with implantable cardioverter-defibrillator therapy were not excluded as long as it was present in both study arms. Data analysis We quantified at 3-month intervals the segmental area between the 2 curves (Fig. 1) and the cumulative area representing lifespan gained per patient up to that time. Effect of analysis duration on lifespan gained The shortest duration of survival curves presented by all trials was 2 years. To examine how lifespan gained changed with follow-up duration, we calculated for each trial the lifespan gained at each time point as a proportion of lifespan gained in that trial at 2 years. The results for different trials were then scaled to permit easy comparison between trials with different mortality rates (5). Calculated survival gain for the duration of 1 device Hazard ratios reported by trials are the most precise estimate of the mortality effect of the decision to implant. Making the assumption that they remain similar during the life of 1 device, we then calculated survival curves and lifespan gain to 5 years.

Effect of analysis duration on lifespan gained The shortest duration of survival curves presented by all trials was 2 years. To examine how lifespan gained changed with follow-up duration, we calculated for each trial the lifespan gained at each time point as a proportion of lifespan gained in that trial at 2 years. The results for different trials were then scaled to permit easy comparison between trials with different mortality rates (5). Calculated survival gain for the duration of 1 device Hazard ratios reported by trials are the most precise estimate of the mortality effect of the decision to implant. Making the assumption that they remain similar during the life of 1 device, we then calculated survival curves and lifespan gain to 5 years. Calculated survival gain for >1 sequential device It is excessively pessimistic to halt analysis at the duration of 1 device (5) because it ignores the subsequent lifespan gain of patients who were enabled to survive until then. Conversely, during use of subsequent devices, patients are older and have more competing risks that a biventricular pacemaker may not reduce. Thus, we partitioned risk into that on which the pacemaker might have a physiological effect and that on which it would not. The pacemaker-relevant risk component was conservatively considered to remain constant with age, and the relative risk reduction for that component was also considered constant with age. In contrast, the nonpacemaker-relevant risk component was made to rise progressively with time in the standard Gompertz manner (6), and the risk reduction for that component was kept at zero.

onsidered to remain constant with age, and the relative risk reduction for that component was also considered constant with age. In contrast, the nonpacemaker-relevant risk component was made to rise progressively with time in the standard Gompertz manner (6), and the risk reduction for that component was kept at zero. Results Characteristics of included trials Seven trials met the criteria of comparing biventricular pacemaker implantation with no such implantation and publishing Kaplan-Meier survival curves. Two were excluded because follow-up was ≤6 months (7,8). The 5 eligible trials (9–13) totaled 6,561 patients (Table 1). All provided survival curves for at least 2 years. For the COMPANION (Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure) trial, we used only the CRT-pacemaker and no-device arms. Lifespan gained from device implantation averaged across all trials The average pattern of lifespan gained, weighted according to study size between 0 and 24 months across all trials, is shown in Figure 2. Because the lifespan gain curve for each trial was rescaled to run from 0 at 0 year to 100% at 24 months, the averaged curve also does the same. The shape was nonlinear, with a slow early development and later progressively faster development of life-years gained with time. The lifespan gain at 24 months was >4 times the lifespan gain at 12 months.

e for each trial was rescaled to run from 0 at 0 year to 100% at 24 months, the averaged curve also does the same. The shape was nonlinear, with a slow early development and later progressively faster development of life-years gained with time. The lifespan gain at 24 months was >4 times the lifespan gain at 12 months. Calculated lifespan gained for the duration of 1 device From the trial data, Table 2 shows the lifespan gained at 1, 2, 3, and 5 years after device implantation. In parallel is shown the number of devices needed to be implanted (NNT) to gain 1 life-year. Lifespan gained rises progressively, and NNT falls progressively, for progressively longer time windows. The nonlinear pattern evident across all studies in aggregate (Fig. 2) is also visible in all 5 individual trials (Table 2, Fig. 3). Figure 3 illustrates the similar nonlinear pattern of lifespan gained across all 5 trials: slow early and progressively faster later. Within the 2 years available in all trials, absolute life-years gained is much larger in trials in advanced heart failure (COMPANION and CARE-HF) than in those in milder heart failure (MADIT-CRT [Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy] and REVERSE [Resynchronization Reverses Remodeling in Systolic Left Ventricular Dysfunction]), as shown in the middle row of panels. However, when the curves showing development of life-years gained are re-scaled (bottom panels), a similar curvilinear progression is seen across all 5 trials.

Resynchronization Therapy] and REVERSE [Resynchronization Reverses Remodeling in Systolic Left Ventricular Dysfunction]), as shown in the middle row of panels. However, when the curves showing development of life-years gained are re-scaled (bottom panels), a similar curvilinear progression is seen across all 5 trials. Calculated survival gain for >1 sequential device In practice, patients who receive 1 device and survive its entire battery life will usually have it replaced. However, the hazard ratio with extended aging will depend on the extent to which deaths are heart failure related. If almost all mortality is heart failure related, then even substantial scaling up of the few non-heart-failure–related deaths with aging might not greatly influence the overall hazard ratio.

it replaced. However, the hazard ratio with extended aging will depend on the extent to which deaths are heart failure related. If almost all mortality is heart failure related, then even substantial scaling up of the few non-heart-failure–related deaths with aging might not greatly influence the overall hazard ratio. For example, in CARE-HF, one-quarter of deaths were noncardiac related. To achieve its overall 0.64 hazard ratio, the hazard ratio for cardiac death would have been ∼0.52 because (0.75 × 0.52) + (0.25 × 1) = 0.64. Had the population had a greater proportion of noncardiac deaths (e.g., 50%), then the overall hazard ratio might have been (0.50 × 0.52) + (0.50 × 1) = 0.76. Figure 4 shows the results for 3 different possibilities for hazard ratios after trial end. For Figure 4A, annual mortality is initially 12.6% (annual mortality at 1 year in CARE-HF [9]) and initially one-quarter of this is non-heart-failure related, and the absolute rate scales up by a factor of 1.1 with every year of aging, as is typical (6). For Figures 4B and 4C, initially one-half and three-quarters of mortality is non-heart-failure related and has the same age scaling.

y at 1 year in CARE-HF [9]) and initially one-quarter of this is non-heart-failure related, and the absolute rate scales up by a factor of 1.1 with every year of aging, as is typical (6). For Figures 4B and 4C, initially one-half and three-quarters of mortality is non-heart-failure related and has the same age scaling. Figure 5 shows a perhaps surprising phenomenon. For the same hazard ratio, the patient group with the highest baseline risk (Fig. 5F) exhibited more lifespan gain than the group at lower baseline risk (Fig. 5D), at the early 5-year time point. However, by 15 years, the situation has reversed. The reason for this finding is that a lower-risk group has more survivors at 5 years than a high-risk group and so continuing the calculation for longer reveals more additional lifespan gain. In our example, over a short window of 2 years, using the average hazard ratio of all 5 trials (0.71), those with a lower mortality risk (10%) would gain 1.4 months while those at higher mortality (20%) would gain much more (2.3 months). By 15 years, however, the gain in the lower-risk group increased to 16.0 months, whereas the gain in the higher-risk group reached only 13.7 months.

hazard ratio of all 5 trials (0.71), those with a lower mortality risk (10%) would gain 1.4 months while those at higher mortality (20%) would gain much more (2.3 months). By 15 years, however, the gain in the lower-risk group increased to 16.0 months, whereas the gain in the higher-risk group reached only 13.7 months. Discussion Which risk stratum of patients gains the most lifespan from biventricular pacemaker implantation depends on the time window over which gain is assessed. Over a short window, higher-risk patients may gain more lifespan than lower-risk patients, but over a longer window, this outcome is reversed. Thus, even well-designed, well-conducted, and well-reported trials of patients at lower annual risk such as MADIT-CRT (13) and REVERSE (10) ended after exposing only a tiny fraction of the full lifespan gain available from biventricular pacing.

an lower-risk patients, but over a longer window, this outcome is reversed. Thus, even well-designed, well-conducted, and well-reported trials of patients at lower annual risk such as MADIT-CRT (13) and REVERSE (10) ended after exposing only a tiny fraction of the full lifespan gain available from biventricular pacing. Appropriate index of benefit depends on disease and intervention. We focused on lifespan gained, because it is relevant to patients, and on lifespan gained per device implanted because of its health/economic importance. More common metrics such as NNT and absolute risk reduction have the disadvantage of implicitly assuming no benefit after trial end; disease-modifying interventions may give continued benefit (14), which they do not capture (15). The choice of statistical metric depends on disease and intervention. For example, for brief therapies with immediate consequences (e.g., antibiotics for acute infection), NNT may be the most appropriate metric (16). For lifelong treatment with progressively accumulating cost such as drug therapy for chronic disease, a more appropriate metric is years-needed-to-treat to gain 1 life-year (15). Calculation of years-needed-to-treat must address survival gain in the posttrial period to give a correct quantification. For on–off interventions that may have sustained effect on mortality (e.g., device implantation), lifespan gain per device may be the most appropriate metric. In addition to assessing benefit in the posttrial period, this method also begins to address cost-effectiveness.

sttrial period to give a correct quantification. For on–off interventions that may have sustained effect on mortality (e.g., device implantation), lifespan gain per device may be the most appropriate metric. In addition to assessing benefit in the posttrial period, this method also begins to address cost-effectiveness. Impact of time-window in assessing lifespan gain Trials rarely continue with randomization intact until survival is zero in both arms; therefore, the observed lifespan gain within trials is much less than the potential gain. The only practical way of assessing lifespan gain over a satisfactorily long period is with the use of modeling, but this method must take into account the progressive increase in noncardiac mortality. We used the Gompertz method for this. At 1 year, the lifespan gain was only (10) 0.1 month, but by 2 years this had grown to 0.5 month, by 5 years to 6.5 months, and by 15 years to 13.7 months. Such calculations are dependent on the hazard ratio being preserved in the longer term, supported by the CARE-HF (17) finding of no attenuation of hazard ratio in its extension period.

pan gain was only (10) 0.1 month, but by 2 years this had grown to 0.5 month, by 5 years to 6.5 months, and by 15 years to 13.7 months. Such calculations are dependent on the hazard ratio being preserved in the longer term, supported by the CARE-HF (17) finding of no attenuation of hazard ratio in its extension period. Impact of patient group studied in assessing lifespan gain Trials in patients at high mortality risk may have greater statistical power, but these patient groups may not be those who gain the greatest lifespan benefit over their entire lifetime, since their lifetime may be short (2). Conversely, trials of the same duration in patients at lower mortality risk are inherently less powered, but these patient groups may gain the greatest increment in lifespan if given a lifetime of biventricular pacing. Thus, even though CARE-HF and COMPANION were the trials reporting the most encouraging mortality effects from biventricular pacing, we should not assume that REVERSE and MADIT-CRT populations, if given a lifetime of pacing, would not exhibit a lifespan benefit. When faced with patients such as those of REVERSE or MADIT-CRT, we should consider carefully whether it is wise to delay device implantation until they reach the CARE-HF/COMPANION stage of disease (18). Delaying may cause the curves to diverge quicker, but our analysis indicates it may also miss the majority of the opportunity for lifespan gain.

ch as those of REVERSE or MADIT-CRT, we should consider carefully whether it is wise to delay device implantation until they reach the CARE-HF/COMPANION stage of disease (18). Delaying may cause the curves to diverge quicker, but our analysis indicates it may also miss the majority of the opportunity for lifespan gain. Although the etiology of heart failure affects lifespan gain, quantitative analysis is limited because Kaplan-Meier plots are not available for these subgroups. Gain is likely to be greater for patients with nonischemic heart failure than ischemic heart failure. Need for modeling to assess the cost-effectiveness of biventricular pacing Previous analyses (19–21) exploring the cost-effectiveness of biventricular pacing have used trial data to estimate the incremental cost-effective ratios. Device implantation has front-loaded costs, and using trial data only will therefore tend to underestimate cost-effectiveness by concentrating on the early period when cost is already fully exposed but lifespan gain is only partly revealed. This pattern has also been observed for implantable defibrillators (22,23).

os. Device implantation has front-loaded costs, and using trial data only will therefore tend to underestimate cost-effectiveness by concentrating on the early period when cost is already fully exposed but lifespan gain is only partly revealed. This pattern has also been observed for implantable defibrillators (22,23). Study limitations Our study analyzed survival to 2 years because this was the period for which all the trials showed Kaplan-Meier survival data. We used the Gompertz method to address posttrial survival although there are alternatives, including the Deale method (24). Modeling post-trial survival can only provide an estimate of the potential benefit of an intervention. However, prolonged clinical trials are expensive, and it may be ethically unacceptable to withhold devices from patients in the control arm long after the device is proven beneficial for mortality. Modeling may therefore be the best way of quantifying ultimate survival gain. Our analysis used only total mortality data rather than attempting to assess quality of life. For formal cost-effectiveness analysis, it is usual to assess QALYs. Because patients with milder heart failure often have higher quality of life, each incremental life-year gained for them would contribute more QALYs than a life-year gained in a patient with more severe disease. Therefore, this effect of patients with milder disease showing more gain over the long horizon may be even greater when analyzed with QALY data.

ailure often have higher quality of life, each incremental life-year gained for them would contribute more QALYs than a life-year gained in a patient with more severe disease. Therefore, this effect of patients with milder disease showing more gain over the long horizon may be even greater when analyzed with QALY data. Conclusions Lifespan gain from biventricular pacemaker implantation rises rapidly with time and much more than linearly. It continues to grow as long as patients continue to survive and are free of competing mortality risks. For this reason, although higher-risk patients may show clear gain at early time points when lower-risk patients (e.g., the MADIT-CRT [13] and REVERSE [10] cohorts) show no significant gain, this situation may reverse with time. Such analytical approaches to quantifying lifespan gain may be useful because designing trials to directly observe lifespan gain in its entirety would need maintenance of randomization for decades. Dr. Levy has received research support from the National Heart, Lung, and Blood Institute, Thoratec, HeartWare, and GE Healthcare; licensing support from Epocrates; is on the speakers’ bureau of Amarin, GlaxoSmithKline, and Boehringer-Ingelheim; is a consultant for Cardiac Dimensions; and is on the steering committee of Amgen. Dr. Francis is supported by the British Heart Foundation (FS/10/038) and is a consultant to Medtronic and Sorin. Dr. Whinnett is supported by the British Heart Foundation (FS/13/44/30291). All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

ittee of Amgen. Dr. Francis is supported by the British Heart Foundation (FS/10/038) and is a consultant to Medtronic and Sorin. Dr. Whinnett is supported by the British Heart Foundation (FS/13/44/30291). All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Figure 1 Graphic Showing How Lifespan Gained Is Related to Survival Curves Horizontal distance between the curves is lifespan gained. The distance is often greater farther down the graph (less sick patients, who survived longer). The total area between the curves is the average lifespan gained per patient. Figure 2 Pattern of Growth of Lifespan Gain From Device Implantation Weighted According to Study Size and Averaged Across All Trials Lifespan gained calculated at each time point as proportion of lifespan gained at 24 months. Bars show SEM. Figure 3 Survival, Life-Years Gained and Life-Years Gained Re-Scaled, for All 5 Trials Kaplan-Meier curves (upper panels), life-years gained (middle panels) and life-years gained as a proportion of gain at 2 years (lower panels). All 5 trials show a gradual, nonlinear increase in lifespan gain with time. CARE-HF = CArdiac REsynchronization–Heart Failure (9); COMPANION = Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure (11); MADIT-CRT = Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy (13); RAFT = Resynchronization–Defibrillation for Ambulatory Heart Failure (12); REVERSE = Resynchronization Reverses Remodeling in Systolic Left Ventricular Dysfunction.

, and Defibrillation in Heart Failure (11); MADIT-CRT = Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy (13); RAFT = Resynchronization–Defibrillation for Ambulatory Heart Failure (12); REVERSE = Resynchronization Reverses Remodeling in Systolic Left Ventricular Dysfunction. Figure 4 Impact of Extent of Competing Mortality Risks on Lifespan Gain From Biventricular Pacing As the proportion of non-heart-failure–related mortality is increased from one-quarter (A, D) to one-half (B, E) to three-quarters (C, F), the Kaplan-Meier curves (A to C) become progressively closer together. Correspondingly, lifespan gain becomes progressively smaller (D to F). This is true at every timepoint and regardless of whether it is calculated per patient (red curve) or per device (black curve). Figure 5 Which Risk Group Gains the Most? The answer to this question depends on the time window over which the evaluation is made. The panels show progressively increasing mortality: 10% (A, D), 15% (B, E), and 20% (C, F); in all cases, one-quarter of the mortality is non-heart-failure related and the same initial hazard ratio from CARE-HF (0.64) is used. At the 5-year time point, the highest-risk group (A, D) gained the most lifespan, but at the 20-year time point, the lowest-risk group (C, F) gained the most. Table 1 Characteristics of Included Studies Study, Year (Ref.

The answer to this question depends on the time window over which the evaluation is made. The panels show progressively increasing mortality: 10% (A, D), 15% (B, E), and 20% (C, F); in all cases, one-quarter of the mortality is non-heart-failure related and the same initial hazard ratio from CARE-HF (0.64) is used. At the 5-year time point, the highest-risk group (A, D) gained the most lifespan, but at the 20-year time point, the lowest-risk group (C, F) gained the most. Table 1 Characteristics of Included Studies Study, Year (Ref. #) Trial Participants n Mean Follow-Up (months) % Male Mean Age (yrs) Ischemic Etiology (%) NYHA Functional Class LVEF (%) Mean QRS Duration (ms) Hazard Ratio: Mortality Hazard Ratio: Hospitalization CRT versus medical therapy CARE-HF, 2005 (9) Patients with NYHA class III or IV heart failure and with LVEF ≤35%, a left ventricular end-diastolic dimension of at least 30 mm (indexed to height), and a QRS interval >120 ms were randomly assigned to optimal medical therapy or CRT-P 813 29.4 73 67 38 III to IV 25 160 0.64 (0.48–0.85) 0.61 (0.49–0.77) COMPANION, 2004 (11) Patients with NYHA class III or IV ischemic or dilated cardiomyopathy and QRS duration >120 ms were randomly assigned in a 1:2:2 ratio to optimal medical therapy, CRT-P, or CRT-D 1,520 16.5 (median) 68 67 55 III to IV 22 160 0.76 (0.58–1.01) CRT + ICD versus ICD alone REVERSE, 2008 (10) Patients with NYHA class I or II heart failure with CRT-P or CRT-D and QRS ≥120 ms and LVEF ≤40% were randomly assigned to CRT-on versus CRT-off 610 12 82 61 43 I to II 28 156 0.40 0.39 MADIT-CRT, 2009 (13) Patients with NYHA class I or II ischemic or nonischemic cardiomyopathy, LVEF ≤30%, a QRS duration of ≥130 ms Patients were randomly assigned in a 3:2 ratio to receive CRT plus an ICD or an ICD alone 1,820 29 75 65 55 I to II 24 1.00 (0.69–1.44) 0.59 (0.47–0.74) RAFT, 2010 (12) Patients with NYHA class II or III heart failure with LVEF ≤30% and intrinsic QRS duration ≥120 ms or a paced QRS duration ≥200 ms were randomly assigned to either an ICD alone or an ICD plus CRT 1,798 40 83 66 67 II to III 23 158 0.75 (0.62–0.91) 0.68 (0.56–0.83) CARE-HF = CArdiac REsynchronization–Heart Failure; COMPANION = Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure; CRT = cardiac resynchronization therapy; CRT-D = CRT-defibrillator; CRT-P = CRT-pacemaker; ICD = implantable cardioverter-defibrillator; LVEF = left ventricular ejection fraction; MADIT-CRT = Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy; NYHA = New York Heart Association; RAFT = Resynchronization–Defibrillation for Ambulatory Heart Failure; REVERSE = Resynchronization Reverses Remodeling in Systoli

lator; LVEF = left ventricular ejection fraction; MADIT-CRT = Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy; NYHA = New York Heart Association; RAFT = Resynchronization–Defibrillation for Ambulatory Heart Failure; REVERSE = Resynchronization Reverses Remodeling in Systoli c Left Ventricular Dysfunction. Table 2 Lifespan Gained and NNT to Gain 1 Life-Year Compared With Duration of Follow-Up After Device Implantation Lifespan Gained/Device Implanted (months) Size of NNT to Gain 1 Life-Year 1 Year 2 Years 3 Years 5 Years 1 Year 2 Years 3 Years 5 Years REVERSE 0.00 0.40 NA 30.1 CARE-HF 0.16 0.82 2.01 75.8 14.6 6.0 COMPANION 0.12 0.84 1.62 101.8 14.4 7.4 MADIT-CRT 0.00 0.09 0.10 NA 137.1 119.8 RAFT 0.13 0.37 0.77 2.52 91.3 32.7 15.5 4.8 NA = not assessable; NNT = number-needed-to-treat; other abbreviations as in Table 1.

Hypertrophic cardiomyopathy (HCM) is the most common genetic heart disease, generally caused by mutations in cardiac sarcomere genes (1). Most genotyped HCM patients harbor defects in the thick-filament genes, myosin heavy chain (MYH7) and myosin binding protein C (MYBPC3) (2). However, in a distinct patient subgroup, the disease is caused by mutations in thin-filament genes, including cardiac troponin T (TNNT2) and I (TNNI3), α-tropomyosin (TPM1), and cardiac actin (ACTC) (3,4). Each accounts for a small proportion of HCM cohorts, with TNNT2, the most common, accounting for only 2% to 5% (5).