Browse the corpus

Introduction Lipoprotein(a) (Lp(a)) consists of apolipoprotein(a) and apolipoprotein B-100 and has a similar structure both to LDL-cholesterol and to plasminogen.1,2 Therefore, Lp(a) has a proatherogenic and a prothrombotic component, and is associated with the pathogenesis of cardiovascular (CV) disease.3,4 Elevated Lp(a) levels were proven as a marker of increased cardiovascular risk in numerous epidemiological and genetic studies during the past decades.5–8 Due to the lack of selective Lp(a)-lowering therapies the use of Lp(a) in clinical practice remains scarce. Recently, new compounds with the potential to lower Lp(a) like PCSK9-antibodies or the phase 2 study-proved IONIS-Apo(a) Rx, an antisense oligonucleotide targeting hepatic apo(a) mRNA,9,10 are presently evaluated in clinical trials. Therefore, the need for a precise characterization of the Lp(a)- associated CV risk is of increasing importance. For this purpose comparability of Lp(a) among different studies is essential.11,12 Although the largest meta-analysis revealed no significant differences among various methods of Lp(a) determination, it remains unclear whether the observed variability between studies was due to regional differences or due to the use of different assays.6 Further, underlying datasets should offer sufficient power to perform a diversity of subgroup analyses, in order to identify those individuals at highest Lp(a)-associated risk.

on, it remains unclear whether the observed variability between studies was due to regional differences or due to the use of different assays.6 Further, underlying datasets should offer sufficient power to perform a diversity of subgroup analyses, in order to identify those individuals at highest Lp(a)-associated risk. To achieve a more comparable as well as precise and timely characterization of the Lp(a)-associated CV risk in Europe we aimed to analyse the harmonized data of centrally measured Lp(a) of 7 cohorts from 5 European countries with 56 804 individuals and a maximum follow-up time of 24 years within the Biomarker for Cardiovascular Risk Assessment in Europe (BiomarCaRE) project.13 Methods Study overview The design and rationale of the BiomarCaRE project have been described previously in detail.13 Briefly, BiomarCaRE is based on the MORGAM (MONICA Risk Genetics Archiving and Monograph) Project.14 The MORGAM/BiomarCaRE Data Center in Helsinki harmonized individual data from 21 population-based cohort studies with central storage of selected biomaterial of more than 300 000 participants in the central BiomarCaRE laboratory in Hamburg. All presented Lp(a) concentrations were measured centrally at this laboratory site using the same Lp(a) assay. Local ethics review boards approved all participating studies.

sed cohort studies with central storage of selected biomaterial of more than 300 000 participants in the central BiomarCaRE laboratory in Hamburg. All presented Lp(a) concentrations were measured centrally at this laboratory site using the same Lp(a) assay. Local ethics review boards approved all participating studies. Study cohorts The present analysis included data of 7 cohorts from 5 European countries comprising 56 804 individuals with available Lp(a) levels from 52 131 individuals. Cohorts involved were the FINRISK Study and the DanMONICA Study as Northern European cohorts, the Caerphilly Prospective Study and the Kooperative Gesundheitsforschung in der Region Augsburg (KORA) Study as Central European cohorts, and the MONICA Brianza Study, the MATISS cohort (Rome) and the Moli-Sani Project as Southern European cohorts. Each cohort is based on a well-defined population (see Supplementary material online, Table S1). Cohort descriptions are provided in Supplementary material online, Box S1.

A) Study as Central European cohorts, and the MONICA Brianza Study, the MATISS cohort (Rome) and the Moli-Sani Project as Southern European cohorts. Each cohort is based on a well-defined population (see Supplementary material online, Table S1). Cohort descriptions are provided in Supplementary material online, Box S1. For each cohort, the following harmonized variables were available at baseline: age, sex, body-mass-index (BMI), systolic blood pressure, lipid-lowering medication, anti-hypertensive medication, smoking status, history of diabetes, total cholesterol, low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL). LDL and HDL were not available for the Caerphilly Prospective Study. History of diabetes was defined as documented or self-reported, or diagnosed diabetes. Anti-hypertensive medication and smoking status were self reported. BMI, systolic blood pressure, and lipid parameters have been measured values. LDL cholesterol was calculated using the Friedewald formula.

rphilly Prospective Study. History of diabetes was defined as documented or self-reported, or diagnosed diabetes. Anti-hypertensive medication and smoking status were self reported. BMI, systolic blood pressure, and lipid parameters have been measured values. LDL cholesterol was calculated using the Friedewald formula. Study outcome We defined the following outcome measures: (i) first fatal or non-fatal major coronary event (MCE) including the definite, possible, definite or possible (if not specifiable) acute myocardial infarction, coronary death, unstable angina pectoris, and cardiac revascularization, (ii) first major cardiovascular disease (CVD) event including the first fatal or non-fatal coronary heart disease event or likely cerebral infarction, coronary death, unstable angina pectoris, cardiac revascularization, ischaemic stroke, and unclassifiable death, and (iii) total mortality defined as mortality due to any cause during follow-up. Detailed endpoint definitions are in Supplementary material online.

ronary heart disease event or likely cerebral infarction, coronary death, unstable angina pectoris, cardiac revascularization, ischaemic stroke, and unclassifiable death, and (iii) total mortality defined as mortality due to any cause during follow-up. Detailed endpoint definitions are in Supplementary material online. Laboratory procedures All Lp(a) measurements were performed in the central BiomarCaRE laboratory between 2011 and 2015 using a fully automated, particle-enhanced turbidimetric immunoassay (Biokit Quantia Lp(a)-Test; Abbott Diagnostics, USA). This assay is not affected by apo(a) size heterogeneity.15 The limit of detection (LOD) is 0.38 mg/dL. Measurements are linear in the range of 1.3–90.0 mg/dL. Intra-assay coefficients of variation were <4% and inter-assay coefficients of variation were <9% for each cohort (see Supplementary material online, Table S3). The further included lipid parameters were measured locally at each participating centre. Statistical methods Associations between baseline variables like total cholesterol, smoking status, hypertension, BMI, diabetes, sex (male), age at baseline examination, storage time, and Lp(a) concentrations were assessed using Spearman correlations. These were computed using mixed effects models combining individual participant data as described by Pigott and colleagues.16 This allowed us to consider cohort heterogeneity. Lp(a) concentrations >90 mg/dL were defined as 90 mg/dL.

tion, storage time, and Lp(a) concentrations were assessed using Spearman correlations. These were computed using mixed effects models combining individual participant data as described by Pigott and colleagues.16 This allowed us to consider cohort heterogeneity. Lp(a) concentrations >90 mg/dL were defined as 90 mg/dL. For prediction of a first ever MCE, CVD event, and total mortality, only participants without a prior history of major CVD such as myocardial infarction (MI), hospitalized unstable angina, coronary artery bypass grafting, percutaneous transluminal coronary angioplasty, or ischaemic or haemorrhagic stroke were included. Survival curves for MCE, CVD events, and total mortality were computed according to Lp(a) categories, using thirds and the 90th percentile, adapted from Kamstrup and colleagues.11 The last cut-off allowed to differentiate the risk for high and very high Lp(a) values. Furthermore, we performed additional analyses for Lp(a) levels ≥50 mg/dl vs.<50 mg/dL which is the suggested cut-off according to the ESC guidelines.7,17 These predefined categories were applied in the overall BiomarCaRE cohort.

.11 The last cut-off allowed to differentiate the risk for high and very high Lp(a) values. Furthermore, we performed additional analyses for Lp(a) levels ≥50 mg/dl vs.<50 mg/dL which is the suggested cut-off according to the ESC guidelines.7,17 These predefined categories were applied in the overall BiomarCaRE cohort. Sex- and cohort-stratified Cox proportional hazards models for MCE, CVD events, and total mortality were computed using the individual level data from the available cohorts. For these analyses, Lp(a) was used after applying the cubic root transformation as a continuous variable and using pre-defined categories mentioned above. The Cox models for all three endpoints were adjusted in Model 1 for age as time scale, sex as strata and cohort as strata, and in Model 2 additionally for classical cardiovascular risk factors (smoking status, arterial hypertension, BMI, history of diabetes, total cholesterol, and age as time scale). We corrected LDL cholesterol and total cholesterol for the Lp(a) contribution according to the compositional data of Lp(a) with a cholesterol proportion of 30%.18 Therefore, Lp(a) was multiplied by 0.3 and this term was subtracted from LDL cholesterol and total cholesterol for each individual as done by previous studies.11 Since there was a substantial number of missing values regarding intake of lipid-lowering medication in the KORA and Caerphilly cohorts we did not adjust for this variable in the main outcome analysis but performed separate analyses with additional adjustment for intake of lipid-lowering medication in all cohorts except KORA and Caerphilly.

tantial number of missing values regarding intake of lipid-lowering medication in the KORA and Caerphilly cohorts we did not adjust for this variable in the main outcome analysis but performed separate analyses with additional adjustment for intake of lipid-lowering medication in all cohorts except KORA and Caerphilly. We classified subjects taking antihypertensive medication as being hypertensive. We examined the association between Lp(a) concentrations and time to event in different subgroups (age < 65 years vs. age ≥ 65 years, men vs. women, daily smoker vs. non-daily smoker, diabetes vs. no diabetes, hypertension vs. no hypertension, BMI < 30 vs. BMI ≥ 30, Southern, central Europe vs. Northern Europe, and LDL < 160 mg/dL vs. LDL ≥ 160 mg/dL). The previous models were extended by adding an interaction term between Lp(a) and subgroup indicator. This allowed to estimate subgroup specific Lp(a) hazard ratios and to test if the hazard ratios in the two categories of the subgroup variable were different. No adjustment for multiple testing was performed due to the exploratory nature of the analyses.19

dding an interaction term between Lp(a) and subgroup indicator. This allowed to estimate subgroup specific Lp(a) hazard ratios and to test if the hazard ratios in the two categories of the subgroup variable were different. No adjustment for multiple testing was performed due to the exploratory nature of the analyses.19 The C-index20 and the net reclassification improvement (NRI)21,22 were used to quantify the added predictive value of Lp(a) beyond that from a model including classical risk factors. For the computation of NRI follow-up times were censored at ten years. Ten-fold cross validation was used to control for the over-optimism of calculating performance measures on the same dataset from which the models were computed. The risk categories used for the NRI analysis were <1%, 1 to <5%, 5 to <10%, and ≥10%23 for MCE, CVD, and total mortality. A version of NRI appropriate for survival analyses was computed using the Kaplan–Meier method.22 The overall NRI does not represent a proportion and is therefore reported as a decimal number between −2 and 2 rather than a percentage, as recommended by Leening and colleagues.21 Differences in C-statistics (with 95% CIs) after the addition of Lp(a) to the model consisting of cardiovascular risk factors were computed using the method described by Antolini and colleagues.24 A two-sided P-value of <0.05 was considered statistically significant. All statistical methods were implemented in R statistical software version 3.2.2 (www.R-project.org).

The C-index20 and the net reclassification improvement (NRI)21,22 were used to quantify the added predictive value of Lp(a) beyond that from a model including classical risk factors. For the computation of NRI follow-up times were censored at ten years. Ten-fold cross validation was used to control for the over-optimism of calculating performance measures on the same dataset from which the models were computed. The risk categories used for the NRI analysis were <1%, 1 to <5%, 5 to <10%, and ≥10%23 for MCE, CVD, and total mortality. A version of NRI appropriate for survival analyses was computed using the Kaplan–Meier method.22 The overall NRI does not represent a proportion and is therefore reported as a decimal number between −2 and 2 rather than a percentage, as recommended by Leening and colleagues.21 Differences in C-statistics (with 95% CIs) after the addition of Lp(a) to the model consisting of cardiovascular risk factors were computed using the method described by Antolini and colleagues.24 A two-sided P-value of <0.05 was considered statistically significant. All statistical methods were implemented in R statistical software version 3.2.2 (www.R-project.org). Results Baseline characteristics Baseline characteristics for the overall study population are shown in Table 1, of each individual cohort in Supplementary material online, Table S1, and according to predefined categories of Lp(a) levels in Supplementary material online, Table S2.Table 1 Baseline characteristics of the study population

characteristics Baseline characteristics for the overall study population are shown in Table 1, of each individual cohort in Supplementary material online, Table S1, and according to predefined categories of Lp(a) levels in Supplementary material online, Table S2.Table 1 Baseline characteristics of the study population Characteristics Number of cohorts, N 7 Number of individuals, N 56 804 Years of baseline examinations, range in years 1986 − 2008 Men, N (%) 28 498 (50.2) Women, N (%) 28 306 (49.8) Age at baseline examination, y 52.4 (42.1, 62.0) Cardiovascular risk factors Daily smokers, N (%) 13 304 (23.8) Diabetes, N (%) 3063 (5.4) Hypertension, N (%) 27 233 (48.1) Body mass index, kg/m2 26.7 (24.0, 29.8) Systolic blood pressure, mmHg 134.0 (121.0, 150.0) Total cholesterol, mg/dL 215.0 (189.0, 245.0) HDL cholesterol, mg/dL 54.0 (45.0, 65.0) LDL cholesterol, mg/dL 122.0 (98.0, 147.0) Medication Antihypertensive, N (%) 11 340 (20.3) Cholesterol lowering, N (%) 2357 (5.2) Lipoprotein (a) Information on lipoprotein (a), N (%) 52 131 (91.8) Lipoprotein (a), mg/dL 8.7 (3.9, 19.1) Endpoints during follow-up Major coronary event, N (%) 2452 (4.5) Cardiovascular disease, N (%) 2966 (5.5) Total mortality, N (%) 4877 (9.0) Baseline characteristics are presented as absolute and relative frequencies for categorical variables, and quartiles for continuous variables as well as range in years for years of baseline examinations. Numbers of endpoints during follow-up are reported for individuals without CVD at baseline. HDL, high density lipoprotein; LDL, low density lipoprotein.

Characteristics Number of cohorts, N 7 Number of individuals, N 56 804 Years of baseline examinations, range in years 1986 − 2008 Men, N (%) 28 498 (50.2) Women, N (%) 28 306 (49.8) Age at baseline examination, y 52.4 (42.1, 62.0) Cardiovascular risk factors Daily smokers, N (%) 13 304 (23.8) Diabetes, N (%) 3063 (5.4) Hypertension, N (%) 27 233 (48.1) Body mass index, kg/m2 26.7 (24.0, 29.8) Systolic blood pressure, mmHg 134.0 (121.0, 150.0) Total cholesterol, mg/dL 215.0 (189.0, 245.0) HDL cholesterol, mg/dL 54.0 (45.0, 65.0) LDL cholesterol, mg/dL 122.0 (98.0, 147.0) Medication Antihypertensive, N (%) 11 340 (20.3) Cholesterol lowering, N (%) 2357 (5.2) Lipoprotein (a) Information on lipoprotein (a), N (%) 52 131 (91.8) Lipoprotein (a), mg/dL 8.7 (3.9, 19.1) Endpoints during follow-up Major coronary event, N (%) 2452 (4.5) Cardiovascular disease, N (%) 2966 (5.5) Total mortality, N (%) 4877 (9.0) Baseline characteristics are presented as absolute and relative frequencies for categorical variables, and quartiles for continuous variables as well as range in years for years of baseline examinations. Numbers of endpoints during follow-up are reported for individuals without CVD at baseline. HDL, high density lipoprotein; LDL, low density lipoprotein. Men and women were equally distributed (49.8%, n = 28 306 female subjects). The median age was 52.4 years. Study participants were slightly overweight (median BMI 26.7 kg/m2) and median systolic blood pressure was 134 mmHg. At baseline, 23.8% of the study cohort were daily smokers, 48.1% were diagnosed with hypertension, and 5.4% had diabetes. The median values for LDL-cholesterol, total cholesterol, and HDL-cholesterol at baseline were 122 mg/dL, 215 mg/dL, and 54 mg/dL, respectively.

/m2) and median systolic blood pressure was 134 mmHg. At baseline, 23.8% of the study cohort were daily smokers, 48.1% were diagnosed with hypertension, and 5.4% had diabetes. The median values for LDL-cholesterol, total cholesterol, and HDL-cholesterol at baseline were 122 mg/dL, 215 mg/dL, and 54 mg/dL, respectively. Distribution and correlations of Lp(a) Lp(a) was measured in 91.8% (52 131 of 56 804) of study participants. The distribution of Lp(a) was skewed to the right in the overall cohort with a median of 8.7 mg/dL (IQR 3.9 to 19.1 mg/dL), a mean of 15.8 mg/dL [standard deviation (SD) 18 mg/dL] and the 66th, and 90th percentiles at 14.1 mg/dL and 43.5 mg/dL, respectively (Figure 1). We found a trend towards lower Lp(a) levels in the Northern European cohorts with an Lp(a) median of 4.9 mg/dL compared to an Lp(a) median of 10.9 mg/dL in Southern Europe (see Supplementary material online, Figure S2). To test if the gradient in median Lp(a) concentrations among North, Central, and South European populations is statistically significant, we performed Jonckheere–Terpstra test which resulted in P-value <0.001.25 Detailed illustrations of the distribution of Lp(a) of Northern, Central, and Southern European cohorts and for each particular cohort are shown in Supplementary material online, Figure S1.

South European populations is statistically significant, we performed Jonckheere–Terpstra test which resulted in P-value <0.001.25 Detailed illustrations of the distribution of Lp(a) of Northern, Central, and Southern European cohorts and for each particular cohort are shown in Supplementary material online, Figure S1. Figure 1 Density of Lp(a) levels in the entire study population. Density (y-axis) of Lp(a) levels (x-axis) in the entire study population including 52 131 measurements. Each column indicates the density of an Lp(a) range of 1 mg/dL. The median, 33th, 66th, 80th, and 90th percentiles are marked separately. In the overall cohort total cholesterol and age at baseline correlated positively with Lp(a) levels whereas male sex, diabetes, and BMI correlated negatively with Lp(a) levels. All correlations were only modest in nature (Table 2). Hypertension, smoking status, and storage duration as a technical variable did not correlate significantly with Lp(a) levels. Table 2 Spearman correlations of Lp(a) and CV risk factors Total cholesterol Daily smokers Hyper- tension BMI Diabetes Sex (male) Age at baseline Storage time Correlation coefficient for Lp(a), 0.04 −0.02 0.01 −0.02 −0.02 −0.07 0.06 −0.01 P-value P < 0.001 P = 0.061 P = 0.090 P = 0.0033 P < 0.001 P < 0.001 P < 0.001 P = 0.29 Spearman correlations of lipoprotein(a) with total cholesterol (Lp(a) corrected levels), smoker status, hypertension, body mass index, diabetes, age at baseline, and sex (male), and storage time. A linear mixed model was used to consider cohort heterogeneity. BMI, body mass index.

Total cholesterol Daily smokers Hyper- tension BMI Diabetes Sex (male) Age at baseline Storage time Correlation coefficient for Lp(a), 0.04 −0.02 0.01 −0.02 −0.02 −0.07 0.06 −0.01 P-value P < 0.001 P = 0.061 P = 0.090 P = 0.0033 P < 0.001 P < 0.001 P < 0.001 P = 0.29 Spearman correlations of lipoprotein(a) with total cholesterol (Lp(a) corrected levels), smoker status, hypertension, body mass index, diabetes, age at baseline, and sex (male), and storage time. A linear mixed model was used to consider cohort heterogeneity. BMI, body mass index. Outcome analysis Two thousand four hundred and fifty-two incident MCE were observed during a median follow-up time of 8.8 years, 2966 incident CVD events after a median of 8.7 years, and 4877 deaths after a median of 9.2 years. As illustrated in the Kaplan–Meier survival analyses across the predefined categories of Lp(a) levels MCE and CVD event rates increased with increasing Lp(a) levels with the highest event rates in the upper third of the distribution. No significant association between Lp(a) categories and all-cause mortality was observed (Figure 2). Figure 2 Kaplan–Meier curves according to predefined Lp(a) categories for the endpoints MCE, CVD events, and total mortality. The 33rd percentile of LP(a) corresponds to the value of 5.3 mg/dL, the 66th percentile corresponds to the value of 14.1 mg/dL, and the 90th percentile corresponds to the value of 43.5 mg/dL. P, P-value of log-rank test.

es according to predefined Lp(a) categories for the endpoints MCE, CVD events, and total mortality. The 33rd percentile of LP(a) corresponds to the value of 5.3 mg/dL, the 66th percentile corresponds to the value of 14.1 mg/dL, and the 90th percentile corresponds to the value of 43.5 mg/dL. P, P-value of log-rank test. Cox regression analyses revealed a significant association of Lp(a) with incident MCE such as Lp(a) levels between the 66th and 90th percentile were associated with a HR of 1.30 compared to the lowest third (95% CI 1.15–1.46) P < 0.01) after adjustment for a broad spectrum of risk factors (Figure 3, Models 1 and 2). Individuals with Lp(a) levels above the 90th percentile had the highest risk for future MCE with a HR of 1.49 (95% CI 1.29–1.73). Similar results were obtained when addressing a broader definition of CVD endpoints (Figure 3), whereas no significant association was found for total mortality. Further adjustment for lipid-lowering medication did not change the observed associations appreciably (data not shown).

with a HR of 1.49 (95% CI 1.29–1.73). Similar results were obtained when addressing a broader definition of CVD endpoints (Figure 3), whereas no significant association was found for total mortality. Further adjustment for lipid-lowering medication did not change the observed associations appreciably (data not shown). Figure 3 Cox regression analysis according to predefined Lp(a) categories (below 33rd percentile, 33rd-66th percentile, 67–89th percentile, above the 90th percentile) for the endpoints MCE, CVD events, and total mortality for two models of adjustment Model 1 (red rhombus) —adjusted for age, sex and cohort. Model 2 (blue rhombus)—adjusted for age, sex, cohort, smoking status, total cholesterol, diabetes, hypertension, and BMI. N events for MCE = 2038, N events for CVD events = 2478, and N events for total mortality = 3978. HR (95% CI), hazard ratio with 95% confidence interval. Corresponding analyses for the clinically recommended threshold of Lp(a) ≥50 mg/dL vs. <50 mg/dL revealed comparable results for all investigated endpoints (see Supplementary material online, Figures S3 and S4). Cox regression analyses for Lp(a) treated as a continuous variable and the three outcome measures as well as regional stratified analysis by cohort and sorted by European region are displayed in Supplementary material online, Figures S5 and S6.

estigated endpoints (see Supplementary material online, Figures S3 and S4). Cox regression analyses for Lp(a) treated as a continuous variable and the three outcome measures as well as regional stratified analysis by cohort and sorted by European region are displayed in Supplementary material online, Figures S5 and S6. Subgroup analysis of the Lp(a)-associated risk Results of subgroup analyses of the Lp(a)-associated risk are shown in Figure 4. Hazard ratios for cube root transformed Lp(a) were comparable across various predefined subgroups except for individuals with diabetes. For these individuals, the Lp(a)-associated risk was higher (HR for MCE 1.31, for CVD 1.22, and for total mortality 1.15) compared to individuals without diabetes (HR for MCE 1.15, for CVD 1.13, and for total mortality 0.96). In individuals in other cardiovascular high risk states, e.g. smoking, hypertension, LDL ≥160 mg/dL or obesity (BMI ≥30 kg/m2) there was no relevant difference of the Lp(a)-associated risk compared to individuals without these cardiovascular high-risk states (Figure 4). Figure 4 Subgroups analysis for a continuous version of cube root transformed Lp(a) for the endpoints MCE, CVD events, and total mortality. Subgroups used: age (<65 vs. ≥65 years), sex (men vs. women), smoking status, diabetes, hypertension, BMI (<30 vs. ≥30), European region (Southern, Central vs. Northern), and LDL (<160 mg/dL vs. ≥160mg/dL). HR (95%CI), hazard ratio (95% confidence interval), P, P-value for HRs.

nts MCE, CVD events, and total mortality. Subgroups used: age (<65 vs. ≥65 years), sex (men vs. women), smoking status, diabetes, hypertension, BMI (<30 vs. ≥30), European region (Southern, Central vs. Northern), and LDL (<160 mg/dL vs. ≥160mg/dL). HR (95%CI), hazard ratio (95% confidence interval), P, P-value for HRs. Importantly, despite regional differences of Lp(a) levels across Europe with lower Lp(a) levels in Northern European cohorts, the Lp(a)-associated risk in Northern European cohorts was comparable to the Lp(a)-associated risk in Central and Southern European cohorts. Lp(a) and prediction of major coronary events, cardiovascular disease events, and total mortality Assessing C-statistics for prediction of MCE, CVD events, and total mortality we observed marginal but significant changes for MCE and CVD events and no changes for total mortality after the addition of Lp(a) to the base model (see Supplementary material online, Table S4).

iovascular disease events, and total mortality Assessing C-statistics for prediction of MCE, CVD events, and total mortality we observed marginal but significant changes for MCE and CVD events and no changes for total mortality after the addition of Lp(a) to the base model (see Supplementary material online, Table S4). Reclassification analyses after the addition of Lp(a) to a model consisting of cardiovascular risk factor (CVRF) variables are presented in Supplementary material online, Table S5. The addition of Lp(a) to the CVRF variables for MCE led to an NRI of 0.010 (95% CI −0.008 to 0.028), 0.006 (95% CI −0.011 to 0.023) for cases and 0.004 (95% CI 0.002 to 0.006) for non-cases. The addition of Lp(a) to the CVRFs algorithm for CVD events produced an NRI of 0.011 (95% CI from −0.006 to 0.028), 0.008 (95% CI −0.008 to 0.024) for cases and 0.003 (95% CI 0.001 to 0.005) for non-cases and almost no improvement for total mortality. Reclassification tables showing estimates of the expected number of reclassifications per risk category for cases and non-cases are provided in the Supplementary material online, Table S5.

8 (95% CI −0.008 to 0.024) for cases and 0.003 (95% CI 0.001 to 0.005) for non-cases and almost no improvement for total mortality. Reclassification tables showing estimates of the expected number of reclassifications per risk category for cases and non-cases are provided in the Supplementary material online, Table S5. Discussion Based on a harmonized large scale assessment of Lp(a) and cardiovascular outcome the main study findings are: (i) Lp(a) distribution has a north–south gradient with lower Lp(a) levels in Northern European populations compared to Central or Southern European populations. (ii) We confirm Lp(a) as a marker of cardiovascular risk in the European population with an particular increase of the Lp(a)-associated risk for MCE and CVD events above the 66th and the 90th percentile. (iii) The Lp(a)-associated risk was particularly observed in individuals with diabetes compared to those without diabetes.

e confirm Lp(a) as a marker of cardiovascular risk in the European population with an particular increase of the Lp(a)-associated risk for MCE and CVD events above the 66th and the 90th percentile. (iii) The Lp(a)-associated risk was particularly observed in individuals with diabetes compared to those without diabetes. Regional distribution of Lp(a) levels across Europe It is well known that Lp(a) levels vary strongly between ethnicities.26 However, because different Lp(a) assays lack precise comparability, differences of Lp(a) levels across individual participants from large-scale datasets have not yet been described in-depth. Our results showing lower Lp(a) levels in Northern European countries are consistent with a small cohort study measuring Lp(a) in 2164 participants older than 70 years from different regions of Europe.27 Furthermore, we observed a 0.582 fold lower Lp(a) median in the Finnish FINRISK cohort compared to the Lp(a) median in the German KORA cohort. Similar differences were seen between a German and another Finnish cohort by the group of Kronenberg et al (Florian Kronenberg, personal communication).

ent regions of Europe.27 Furthermore, we observed a 0.582 fold lower Lp(a) median in the Finnish FINRISK cohort compared to the Lp(a) median in the German KORA cohort. Similar differences were seen between a German and another Finnish cohort by the group of Kronenberg et al (Florian Kronenberg, personal communication). Lp(a) levels are mainly genetically determined by the number of kringle IV type 2 repeats which correlate strongly and inversely with Lp(a) levels whereas the influence of nutrition or lifestyle is rather weak.28 Further, Lp(a) levels are much higher in people with African ancestry compared to Caucasians. It is conceivable that over time inhabitants of Southern Europe have mixed more genetic characteristics with people of African origin than inhabitants of Northern Europe. Therefore, regional differences of Lp(a) levels in European populations might be due to lower numbers of kringle IV type 2 repeats in Southern European populations compared to Northern European populations. Interestingly, the regional differences of Lp(a) did not lead to differences regarding the Lp(a)-associated cardiovascular risk in different European regions. However, this issue has to be subject of further studies.

f kringle IV type 2 repeats in Southern European populations compared to Northern European populations. Interestingly, the regional differences of Lp(a) did not lead to differences regarding the Lp(a)-associated cardiovascular risk in different European regions. However, this issue has to be subject of further studies. Lp(a) as a marker of cardiovascular risk Elevated Lp(a) levels have been demonstrated to be a marker of increased cardiovascular risk for a broad spectrum of subgroups. As the endpoint classification of CVD events was mainly driven by myocardial infarction, risk estimates for the endpoints CVD events and MCE are rather similar. The significant increase of the Lp(a)-associated risk for MCE and CVD events for Lp(a) levels above the 66th percentile is in line with previous studies. Also, the magnitude of the associations in our study is similar to others: The emerging risk factors collaboration meta-analysis found after adjustment for cardiovascular risk factors a HR of 1.27 for myocardial infarction or fatal coronary events for the top third vs. the lowest third of Lp(a) levels which is comparable to our HRs for MCE of 1.30 and 1.49 for Lp(a) levels in the 67–89th percentile and ≥90th percentile compared to the lowest third of Lp(a) levels.6 Using a lower-risk reference category of Lp(a)-levels <22nd percentile Kamstrup and colleagues found somewhat higher multivariable-adjusted HRs for myocardial infarction with 1.6 for Lp(a) levels in the 67–89th percentile, 1.9 for 90–95th percentile, and 2.6 for ≥95th percentile.5 In addition to the percentile-based analyses which are performed by the vast majority of Lp(a) studies we applied the clinically important threshold of 50 mg/dL and found a strong association for MCE and CVD events for Lp(a) levels ≥50 mg/dL compared to Lp(a) levels <50 mg/dL. This confirms the desirable Lp(a) level of <50 mg/dL recommended by the guidelines.17

s which are performed by the vast majority of Lp(a) studies we applied the clinically important threshold of 50 mg/dL and found a strong association for MCE and CVD events for Lp(a) levels ≥50 mg/dL compared to Lp(a) levels <50 mg/dL. This confirms the desirable Lp(a) level of <50 mg/dL recommended by the guidelines.17 The considerable size of the dataset allowed us to perform powerful subgroup analysis. The subgroup with the highest Lp(a)-associated cardiovascular risk were n = 2785 individuals with diabetes representing 5.4% of the total study population, while other cardiovascular high risk states did not influence the Lp(a)-associated cardiovascular risk. In previous studies, the results regarding the Lp(a)-associated risk in diabetic vs. non-diabetic individuals are heterogeneous or this issue has not be addressed.6,29–31 Since the presence of diabetes correlated negatively with Lp(a) levels in our analysis and in other studies, the increased risk in individuals with diabetes cannot be explained by higher Lp(a) levels in these individuals. Hence, there might be factors which induce a higher atherogenic or thrombogenic potency of Lp(a) e.g. due to glycosylation as known from LDL in a diabetic environment.32

in our analysis and in other studies, the increased risk in individuals with diabetes cannot be explained by higher Lp(a) levels in these individuals. Hence, there might be factors which induce a higher atherogenic or thrombogenic potency of Lp(a) e.g. due to glycosylation as known from LDL in a diabetic environment.32 In the present study, we found only slight NRI and C-Index change. The issue whether Lp(a) improves cardiovascular risk prediction has been addressed only by few studies yet. Comparability of reclassification analyses is often limited due to different base models and risk categories. Recently, Willeit and colleagues found a more relevant NRI and increase of the C-index for addition of Lp(a) to Framingham risk score and Reynolds risk score variables in 826 participants of the Bruneck study.8 Our data and results are rather comparable to the emerging risk factors collaboration analysis reporting only slight improvement of CVD prediction when adding Lp(a) to conventional CV risk factors in 165 544 participants from 37 cohorts.33 It is conceivable that in the emerging risk factors collaboration analysis and in our study residual confounding of large-scale data may have weakened the results compared to the more pronounced cardiovascular risk prediction in the small, very precisely characterized cohort of the Bruneck study. However, our results with only marginal risk prediction improvement confirm the guideline recommendation to determine Lp(a) not routinely in the setting of primary prevention.

e results compared to the more pronounced cardiovascular risk prediction in the small, very precisely characterized cohort of the Bruneck study. However, our results with only marginal risk prediction improvement confirm the guideline recommendation to determine Lp(a) not routinely in the setting of primary prevention. Comparison of Lp(a) levels to previous studies Due to non-standardization of Lp(a)-assays the comparability of absolute Lp(a) levels in general is limited across different studies. Therefore, the largest meta-analysis of 26 cohort studies using a wide variety of different assays showed a range of Lp(a) medians between 3.0 and 23.0 mg/dL across the included studies.6 The present study, using the same assay in a central laboratory, revealed a range of cohort specific median values of Lp(a) between 4.6 and 11.3 mg/dL. One factor for the relatively low Lp(a) medians in our study could have been the prolonged storage duration of blood samples. However, the correlation of storage time as a technical variable and Lp(a) levels was only modest and statistically not significant. We noted comparable Lp(a) levels for cohorts within similar regions despite large differences in the storage duration (Brianza MONICA 25 years, Moli-Sani 8 years, MATISS 21 years) (see Supplementary material online, Table S2). Furthermore, the same Lp(a) assay used for the present study was evaluated in fresh plasma samples of a Northern Spanish population with a median of 25.3 nmol/L.15 Although the validity of a conversion from nmol/L to mg/dL is generally limited due to large differences of the molecular weight of Lp(a), applying the mean conversion factor of 2.4 suggested by Marcovina and colleagues34 to our measurements of stored samples results in a median of 26.1 nmol/L in the Southern European cohorts of the present study, which is comparable to the median of 25.3 nmol/L of Simo and colleagues in fresh samples of a Northern Spanish population.15

an conversion factor of 2.4 suggested by Marcovina and colleagues34 to our measurements of stored samples results in a median of 26.1 nmol/L in the Southern European cohorts of the present study, which is comparable to the median of 25.3 nmol/L of Simo and colleagues in fresh samples of a Northern Spanish population.15 Strengths and limitations Some strengths and limitations of the present study merit consideration. Despite long-standing expertise in international standardization of data collection and data harmonization in the MORGAM Data Centre since 1984, resulting in excellent risk factor and endpoint validation, we cannot exclude a potential contribution of residual confounding for some of the observed effects among the more than 56 000 individuals investigated in 7 European population-based cohort studies. Of these, some variables are missing in some cohorts e.g. information on LDL and HDL values are not available for the Caerphilly cohort.

de a potential contribution of residual confounding for some of the observed effects among the more than 56 000 individuals investigated in 7 European population-based cohort studies. Of these, some variables are missing in some cohorts e.g. information on LDL and HDL values are not available for the Caerphilly cohort. On the one hand we present a large scale dataset of Lp(a) measured centrally with the same assay, but on the other hand differences in storage duration among the included cohorts may have contributed to differences in the Lp(a) levels across populations. However, the resulting effect appears rather minor without a significant correlation for Lp(a) and storage duration. The leading cause for the remarkable differences across European population with lower Lp(a) levels in Northern European cohorts might be differences in the prevalence of kringle IV type 2 repeats. However, as we cannot provide genetic data for our analyses, this issue remains hypothetical and has to be addressed in future studies. Further, Lp(a) measurements were not performed consecutively so that we cannot correct for regression dilution bias. Despite the remarkable stability of Lp(a) this could have led to an underestimation of risk estimates.35

data for our analyses, this issue remains hypothetical and has to be addressed in future studies. Further, Lp(a) measurements were not performed consecutively so that we cannot correct for regression dilution bias. Despite the remarkable stability of Lp(a) this could have led to an underestimation of risk estimates.35 Conclusion In this large Lp(a) dataset on harmonized Lp(a) determination, we observed a north–south gradient of Lp(a) levels across Europe with lower Lp(a) levels in Northern European cohorts. Further, we could confirm Lp(a) as a marker of cardiovascular risk in the European population with an increasing cardiovascular risk for Lp(a)-levels above the 66th percentile. Individuals with diabetes had a particularly high Lp(a)-associated risk. These results may lead to better identification of target populations who might benefit most from future Lp(a)-lowering therapies. Supplementary material Supplementary material is available at European Heart Journal online. Funding The BiomarCaRE Project is funded by the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement No. HEALTH-F2-2011-278913. The activities of the MORGAM Data Centre have been sustained by recent funding from European Union FP 7 project CHANCES (HEALTH-F3-2010-242244). A part of the biomarker determinations in the population cohorts was funded by the Medical Research Council London (G0601463, identification No. 80983: Biomarkers in the MORGAM Populations). The Open Access funding was provided by the BiomarCaRE Project.

from European Union FP 7 project CHANCES (HEALTH-F3-2010-242244). A part of the biomarker determinations in the population cohorts was funded by the Medical Research Council London (G0601463, identification No. 80983: Biomarkers in the MORGAM Populations). The Open Access funding was provided by the BiomarCaRE Project. Ethics Our study complies with the Declaration of Helsinki. The local ethics committees approved the research protocol. Informed consent has been obtained from the subjects (or their legally authorized representative). Conflict of interest: Dr S.B. reports investigator-initiated grants from SIEMENS, Abbott Diagnostics and Thermofisher. Dr. W.K. reports receiving fees for serving on advisory boards from Novartis, Pfizer, The Medicines Company, Amgen, AstraZeneca, MSD, and Kowa, lecture fees from Amgen, AstraZeneca, Novartis, MSD, Sanofi, Actavis and berlin-Chemie as well as research grants from Roche, Beckmann, Singulex, and Abbott Diagnostics. Dr C.W. reports lecture fees from AstraZeneca. Dr U.L. received lecture or advisory fees from Amgen, Sanofi, MSD, Berlin-Chemie, Medicines Company, Abbott. Supplementary Material Supplementary Tables and Figures Click here for additional data file.

Introduction Acute and chronic alcohol intake result in desirable and undesirable health effects.1 Modest recreational alcohol consumption may lead to perceived wellbeing,2 beyond moderate intake confers the risk of adverse outcomes. Common symptoms of acute alcohol consumption are disorientation, disinhibition, nausea, and vomiting.3,4 Excessive acute abuse may lead to death.5 Chronic consumption has been ascribed cardio-protective effects.6 More commonly, chronic alcohol use dose-dependently is associated with detrimental effects including substance addiction and severe nutritional, hepatic, neurological and social consequences.7 From a cardiovascular perspective, acute excessive alcohol consumption has been associated with the so called ‘Holiday Heart Syndrome’.8 This syndrome affects individuals without a specific cardiac history resulting in both ventricular and supraventricular arrhythmias, predominantly atrial fibrillation.8,9 Small case series confirmed a relation between alcohol use and atrial fibrillation.10,11 Regarding chronic alcohol consumption, population-based studies reported a dose-response effect on the development of atrial fibrillation.12

n both ventricular and supraventricular arrhythmias, predominantly atrial fibrillation.8,9 Small case series confirmed a relation between alcohol use and atrial fibrillation.10,11 Regarding chronic alcohol consumption, population-based studies reported a dose-response effect on the development of atrial fibrillation.12 Many open questions remain. All reports on effects of acute alcohol intake were derived from small retrospective analyses with relations to arrhythmias as secondary findings. In our study, we thus intended to conduct a large, sufficiently powered, observational, cross-sectional analysis of cardiac arrhythmia prevalence in participants with quantitatively measured acute alcohol intake. Investigating visitors of the 2015 Munich Octoberfest, we hypothesized that increased breath alcohol concentration (BAC) is associated with a higher burden of cardiac arrhythmias. To parallel acute with chronic alcohol use, we studied participants of the community-based KORA Study.

tively measured acute alcohol intake. Investigating visitors of the 2015 Munich Octoberfest, we hypothesized that increased breath alcohol concentration (BAC) is associated with a higher burden of cardiac arrhythmias. To parallel acute with chronic alcohol use, we studied participants of the community-based KORA Study. Methods Study cohorts For the analysis of BAC and arrhythmia prevalence, we designed an observational, cross-sectional cohort study and recruited voluntary visitors at the annual Octoberfest in Munich, Germany between September and October 2015 (acute alcohol cohort). Participants had to be ≥18 years of age and provide written informed consent to study inclusion. We screened and consented 3042 individuals. After exclusion of individuals with a BAC ≥3.00 g/kg (n = 4) and individuals with uninterpretable electrocardiogram (ECG) recordings (n = 8), the final study cohort comprised 3028 participants. Individuals with a BAC ≥ 3.00 g/kg are considered disabled due to intoxication according to German law, and must not be consented. The ethics committee at the Ludwig Maximilians University of Munich, Germany approved the study, which was registered at clinicaltrials.org (NCT02550340).

ohort comprised 3028 participants. Individuals with a BAC ≥ 3.00 g/kg are considered disabled due to intoxication according to German law, and must not be consented. The ethics committee at the Ludwig Maximilians University of Munich, Germany approved the study, which was registered at clinicaltrials.org (NCT02550340). To assess chronic alcohol consumption on arrhythmia prevalence, we investigated participants of the Survey S4 of the community-based Co-operative Health Research in the region of Augsburg Study (KORA S4, chronic alcohol cohort).13 Study participants were identified through the registration office. Ten strata of equal size according to sex and age comprised 4261 individuals; after exclusion of those without (n = 95) or with uninterpretable (n = 17) ECGs, or without information on alcohol consumption (n = 18), the final cohort consisted of 4131 individuals. The ethics committee of the Bavarian Medical Association approved the study. Electrocardiogram recordings In the acute alcohol cohort, 30 s ECG recordings were obtained using the smart phone based AliveCor device (AliveCor, San Francisco, CA, USA). A two-electrode hardware extension wirelessly communicating with a software application was held with both hands by the participant, resembling a lead I ECG. In KORA S4, digital 10 s 12-lead ECGs were obtained using the Hannover ECG System version 3.22-12 (HES, Corscience, Erlangen, Germany). ECGs were recorded after 10 min rest in supine position.

Electrocardiogram recordings In the acute alcohol cohort, 30 s ECG recordings were obtained using the smart phone based AliveCor device (AliveCor, San Francisco, CA, USA). A two-electrode hardware extension wirelessly communicating with a software application was held with both hands by the participant, resembling a lead I ECG. In KORA S4, digital 10 s 12-lead ECGs were obtained using the Hannover ECG System version 3.22-12 (HES, Corscience, Erlangen, Germany). ECGs were recorded after 10 min rest in supine position. Electrocardiogram analysis was performed in parallel by two senior cardiologists blinded to BAC, chronic alcohol consumption levels, or clinical covariates. Discrepant findings were resolved by consent. Assessment of arrhythmias employed a standardized coding scheme. Arrhythmias were classified as sinus tachycardia (heart rate >100 b.p.m.), sinus arrhythmia, premature atrial complex, premature ventricular complex, and atrial fibrillation or flutter. We further assessed all ECGs for respiratory sinus arrhythmia as a measure of autonomic tone. For this, we modified the previously described respiratory sinus arrhythmia bedside test.14 Per 30 s ECG recording, we measured RR intervals with a scaled caliper, neglecting RR intervals before and after premature beats, and determined the shortest, longest, and mean RR interval of each participant. We considered respiratory sinus arrhythmia present, when the absolute difference of the shortest and longest RR interval was ≥20% of the mean RR interval duration.

a scaled caliper, neglecting RR intervals before and after premature beats, and determined the shortest, longest, and mean RR interval of each participant. We considered respiratory sinus arrhythmia present, when the absolute difference of the shortest and longest RR interval was ≥20% of the mean RR interval duration. Alcohol assessment In the acute alcohol cohort, BAC was determined using a Dräger Alcotest 7510 handheld device (Drägerwerk AG, Lübeck, Germany). The device accounts for remaining oral alcohol. To further reduce influence of the latter, BAC was obtained at the end of participant recruitment, when individuals had not ingested alcohol for several minutes. Alcohol was measured in gram per kilogram (g/kg). In KORA S4, chronic alcohol consumption was assessed in gram alcohol per day (g/d), surveyed using the validated 7-day-recall method.15 Clinical covariates In our acute alcohol cohort, assessment of clinical covariates was restricted by the study setting and the aim to maintain participant privacy. We were thus limited to survey self-reported age, sex, country of origin, history of heart disease, use of cardiovascular and antiarrhythmic drugs, and active smoking status. In KORA S4, participants underwent a standardized interview and examination.13 We used information on age, sex, hypertension, smoking status, history of angina, myocardial infarction, diabetes mellitus, and stroke, as well as the use of cardiovascular and antiarrhythmic medication.

Clinical covariates In our acute alcohol cohort, assessment of clinical covariates was restricted by the study setting and the aim to maintain participant privacy. We were thus limited to survey self-reported age, sex, country of origin, history of heart disease, use of cardiovascular and antiarrhythmic drugs, and active smoking status. In KORA S4, participants underwent a standardized interview and examination.13 We used information on age, sex, hypertension, smoking status, history of angina, myocardial infarction, diabetes mellitus, and stroke, as well as the use of cardiovascular and antiarrhythmic medication. Statistical considerations We express categorical variables as frequency (percentage). Continuous data are presented as mean ± standard deviation or median (25th;75th percentile) as appropriate. We applied logistic regression models to assess the relation of alcohol as predictor on the primary and secondary outcomes of arrhythmia prevalence, adjusted for age and sex or in addition for the remaining available covariates. Odds ratios are provided per unit (i.e. 1 g/kg) increase of BAC. Arrhythmia prevalence across quartiles of BAC was compared by χ2 tests for trend. Primary outcome was the occurrence of any arrhythmia including sinus arrhythmia, sinus tachycardia, premature atrial complex, premature ventricular complex, and atrial fibrillation or flutter. Secondary outcomes were arrhythmia subtype separately or combinations thereof. Another secondary outcome in our acute alcohol cohort was the presence of respiratory sinus arrhythmia. For sensitivity analyses to adjudicate the influence of alcohol consumption on the baseline prevalence of sinus tachycardia, we selected a subgroup of participants from our acute alcohol cohort with no exposure to acute alcohol consumption (BAC 0 g/kg). Similarly, in KORA S4 we selected participants without chronic alcohol use. In these sensitivity analyses, we determined the prevalence of cardiac arrhythmias as detailed above. Finally, we assessed interaction with sex by adding multiplicative interaction terms (sex*BAC) to our models.

lcohol consumption (BAC 0 g/kg). Similarly, in KORA S4 we selected participants without chronic alcohol use. In these sensitivity analyses, we determined the prevalence of cardiac arrhythmias as detailed above. Finally, we assessed interaction with sex by adding multiplicative interaction terms (sex*BAC) to our models. For sample size calculations in the acute alcohol cohort, we assumed a 1% prevalence of arrhythmias in the general population assessed by a 10 s ECG. In our 30 s recordings, we assumed a 1.5% prevalence in those under no or low influence of alcohol (<0.5 g/kg). For the group under intermediate (≥0.5–<1.5 g/kg) and high (≥1.5 g/kg) influence, we assumed an odds of 2 and 4 for the prevalence of arrhythmias, respectively. For 85% power with a two-sided α-level set at 5%, we required 2754 participants. To account for a 10% drop-out rate, we aimed at 3029 enrolled individuals. Results During all 16 days of the 2015 Munich Octoberfest we included 3028 participants for analysis. Their mean age was 34.7 ± 13.3 years, 905 (29.9%) were women (Table 1). Reflecting the international attendance, individuals originated from 60 different countries, whereas the majority of 69% was from Germany. Table 1 Characteristics of the study cohorts

5 Munich Octoberfest we included 3028 participants for analysis. Their mean age was 34.7 ± 13.3 years, 905 (29.9%) were women (Table 1). Reflecting the international attendance, individuals originated from 60 different countries, whereas the majority of 69% was from Germany. Table 1 Characteristics of the study cohorts Acute alcohol cohort Chronic alcohol cohort Men Women Men Women n 3028 2123 905 4131 2021 2110 Sex (women), n(%) 905 (29.9%) - - 2110 (51.1%) - - Age, years 34.7 ± 13.3 35.5 ± 13.4 33.1 ± 12.8 49.1 ± 13.9 49.5 ± 14.0 48.7 ± 13.8 History of Heart disease, n(%) 175 (5.8%) 140 (6.6%) 35 (3.9%) - - - Angina, n(%) - - - 249 (6.0%) 116 (5.7%) 133 (6.3%) Myocardial infarction, n(%) - - - 82 (2.0%) 71 (3.5%) 11 (0.5%) Diabetes mellitus, n(%) - - - 159 (3.8%) 82 (4.1%) 77 (3.6%) Stroke, n(%) - - - 49 (1.2%) 33 (1.6%) 16 (0.8%) Arrhythmias, n(%) 80 (2.6%) 57 (2.7%) 23 (2.5%) - - - Medication use, n(%) 185 (6.1%) 149 (7.0%) 36 (4.0%) 859 (20.8%) 409 (20.2%) 450 (21.3%) Current smoking, n(%) 858 (28.3%) 605 (28.5%) 253 (28.0%) 1066 (25.6%) 616 (30.5%) 450 (21.3%) The KORA S4 cohort consisted of 4131 individuals with a mean age of 49.1 ± 13.9 years and 2110 (51.1%) females (Table 1). Owing to the study composition, >99.5% were of German descent.

Acute alcohol cohort Chronic alcohol cohort Men Women Men Women n 3028 2123 905 4131 2021 2110 Sex (women), n(%) 905 (29.9%) - - 2110 (51.1%) - - Age, years 34.7 ± 13.3 35.5 ± 13.4 33.1 ± 12.8 49.1 ± 13.9 49.5 ± 14.0 48.7 ± 13.8 History of Heart disease, n(%) 175 (5.8%) 140 (6.6%) 35 (3.9%) - - - Angina, n(%) - - - 249 (6.0%) 116 (5.7%) 133 (6.3%) Myocardial infarction, n(%) - - - 82 (2.0%) 71 (3.5%) 11 (0.5%) Diabetes mellitus, n(%) - - - 159 (3.8%) 82 (4.1%) 77 (3.6%) Stroke, n(%) - - - 49 (1.2%) 33 (1.6%) 16 (0.8%) Arrhythmias, n(%) 80 (2.6%) 57 (2.7%) 23 (2.5%) - - - Medication use, n(%) 185 (6.1%) 149 (7.0%) 36 (4.0%) 859 (20.8%) 409 (20.2%) 450 (21.3%) Current smoking, n(%) 858 (28.3%) 605 (28.5%) 253 (28.0%) 1066 (25.6%) 616 (30.5%) 450 (21.3%) The KORA S4 cohort consisted of 4131 individuals with a mean age of 49.1 ± 13.9 years and 2110 (51.1%) females (Table 1). Owing to the study composition, >99.5% were of German descent. In the acute alcohol cohort, the mean BAC was 0.85 ± 0.54 g/kg (range 0–2.94 g/kg). Generally, men had a higher BAC compared to women. Across quartiles of age, the highest BAC was found in the 2nd quartile (25–30 years of age, Figure 1A). We noted limited day-by-day variation, tending to higher BAC measurements towards the weekends (Figure 1C). The circadian distribution confirmed a constant rise in BAC toward the closing hour of each day. A slight decline in the afternoon likely reflects a partial exchange of visitors due to seat reservation time slots (Figure 1D).

ted day-by-day variation, tending to higher BAC measurements towards the weekends (Figure 1C). The circadian distribution confirmed a constant rise in BAC toward the closing hour of each day. A slight decline in the afternoon likely reflects a partial exchange of visitors due to seat reservation time slots (Figure 1D). Figure 1 Alcohol consumption. A. Distribution of breath alcohol concentration (BAC) across the acute alcohol cohort in g/kg. Results presented for the entire cohort, and stratified by sex and quartiles of age. B. Distribution of chronic alcohol consumption in KORA S4 in g/d. Results presented for the entire cohort, and stratified by sex and quartiles of age. Outliers truncated at 80 g/d. C. Day-by-day variability of mean BAC for each of the 16 days of the Octoberfest. D. Circadian variability of mean BAC across recruitment days at the Octoberfest. The mean chronic consumption of alcohol in KORA S4 was 15.8 ± 21.4 g/d. On average, men consumed nearly three time as much alcohol as women. The highest chronic alcohol consumption was observed in the 3rd age quartile (37–48 years of age, Figure 1B).

Figure 1 Alcohol consumption. A. Distribution of breath alcohol concentration (BAC) across the acute alcohol cohort in g/kg. Results presented for the entire cohort, and stratified by sex and quartiles of age. B. Distribution of chronic alcohol consumption in KORA S4 in g/d. Results presented for the entire cohort, and stratified by sex and quartiles of age. Outliers truncated at 80 g/d. C. Day-by-day variability of mean BAC for each of the 16 days of the Octoberfest. D. Circadian variability of mean BAC across recruitment days at the Octoberfest. The mean chronic consumption of alcohol in KORA S4 was 15.8 ± 21.4 g/d. On average, men consumed nearly three time as much alcohol as women. The highest chronic alcohol consumption was observed in the 3rd age quartile (37–48 years of age, Figure 1B). The quality of ECG recordings in our acute alcohol cohort was high, despite disadvantageous recording conditions. Only eight ECGs were uninterpretable. Examples of encountered arrhythmias are presented in Figure 2A–E. In the acute alcohol cohort, the primary outcome of any arrhythmia occurred in 30.5%. Thereby, sinus tachycardia was most common and affected 25.9% of participants. The combined prevalence of other arrhythmias was 5.4%, including sinus arrhythmia, premature atrial and ventricular complexes, and atrial fibrillation or flutter. The secondary outcome of respiratory sinus arrhythmia as a qualitative measure of autonomic tone was noted in 22.2% of participants (Table 2).

participants. The combined prevalence of other arrhythmias was 5.4%, including sinus arrhythmia, premature atrial and ventricular complexes, and atrial fibrillation or flutter. The secondary outcome of respiratory sinus arrhythmia as a qualitative measure of autonomic tone was noted in 22.2% of participants (Table 2). Figure 2 Examples and Prevalence of Cardiac Arrhythmias. A–E. Representative ECG recordings obtained in our acute alcohol cohort. ECG recordings show sinus rhythm (A), sinus tachycardia (B), premature atrial complex (C), premature ventricular complex (D), atrial fibrillation (E). F–G. Clustered bars represent the prevalence of the primary outcome of any cardiac arrhythmia (F) and sinus tachycardia (G) in our acute alcohol cohort by quartiles of BAC. Within each cluster, bars represent the overall cohort (green), and sex-stratified results for men (blue) and women (red). Clusters compared by χ2 test for trend. Table 2 Prevalence of arrhythmias

Figure 2 Examples and Prevalence of Cardiac Arrhythmias. A–E. Representative ECG recordings obtained in our acute alcohol cohort. ECG recordings show sinus rhythm (A), sinus tachycardia (B), premature atrial complex (C), premature ventricular complex (D), atrial fibrillation (E). F–G. Clustered bars represent the prevalence of the primary outcome of any cardiac arrhythmia (F) and sinus tachycardia (G) in our acute alcohol cohort by quartiles of BAC. Within each cluster, bars represent the overall cohort (green), and sex-stratified results for men (blue) and women (red). Clusters compared by χ2 test for trend. Table 2 Prevalence of arrhythmias Acute alcohol cohort Chronic alcohol cohort Men Women Men Women 1 Sinus arrhythmia 51 (1.7%) 40 (1.9%) 11 (1.2%) 9 (0.2%) 4 (0.2%) 5 (0.2%) 2 Sinus tachycardia 785 (25.9%) 514 (24.2%) 271 (29.9%) 17 (0.4%) 9 (0.4%) 8 (0.4%) 3 Premature atrial complexes 39 (1.3%) 30 (1.4%) 9 (1.0%) 26 (0.6%) 11 (0.7%) 11 (0.5%) 4 Premature ventricular complexes 52 (1.7%) 39 (1.8%) 13 (1.4%) 46 (1.1%) 19 (0.9%) 27 (1.3%) 5 Atrial fibrillation/flutter 25 (0.8%) 12 (0.6%) 13 (1.4%) 22 (0.5%) 18 (0.9%) 4 (0.2%) Combination of 1, 2, 3, 4, 5 925 (30.5%) 614 (28.9%) 311 (34.4%) 112 (2.7%) 58 (2.9%) 54 (2.6%) Combination of 1, 3, 4, 5 164 (5.4%) 118 (5.6%) 46 (5.1%) 95 (2.3%) 50 (2.2%) 45 (2.1%) Combination of 3, 4, 5 113 (3.7%) 78 (3.7%) 35 (3.9%) 87 (2.1%) 47 (2.0%) 40 (1.9%) Respiratory sinus arrhythmia 673 (22.2%) 509 (24.0%) 163 (18.0%) - - -

2, 3, 4, 5 925 (30.5%) 614 (28.9%) 311 (34.4%) 112 (2.7%) 58 (2.9%) 54 (2.6%) Combination of 1, 3, 4, 5 164 (5.4%) 118 (5.6%) 46 (5.1%) 95 (2.3%) 50 (2.2%) 45 (2.1%) Combination of 3, 4, 5 113 (3.7%) 78 (3.7%) 35 (3.9%) 87 (2.1%) 47 (2.0%) 40 (1.9%) Respiratory sinus arrhythmia 673 (22.2%) 509 (24.0%) 163 (18.0%) - - - In comparison, arrhythmias occurred less frequently in KORA S4. Only 2.7% of the participants met the primary outcome, at which 0.4% had sinus tachycardia and 2.3% presented with sinus arrhythmia, premature atrial and ventricular complexes, and atrial fibrillation or flutter (Table 2). In our sensitivity analysis of study participants with 0 g/kg BAC in the acute alcohol cohort, the primary outcome of any arrhythmia occurred in 23.9%, and sinus tachycardia occurred in 18.5%. In those without alcohol use in the chronic alcohol cohort, prevalence of any arrhythmia was 2.5% and of sinus tachycardia was 0.4%.

r sensitivity analysis of study participants with 0 g/kg BAC in the acute alcohol cohort, the primary outcome of any arrhythmia occurred in 23.9%, and sinus tachycardia occurred in 18.5%. In those without alcohol use in the chronic alcohol cohort, prevalence of any arrhythmia was 2.5% and of sinus tachycardia was 0.4%. For our primary outcome (sinus tachycardia; sinus arrhythmia; premature atrial and ventricular complexes; atrial fibrillation or flutter), we identified a robust association with higher BAC in our acute alcohol cohort. Analysis by arrhythmia subtype revealed that this association was driven by sinus tachycardia, both after age and sex, and after multivariable adjustment. (Table 3) Across quartiles of BAC in the overall and sex-stratified cohort, we found increasing prevalences of any arrhythmia and sinus tachycardia, respectively (Figure 2F and G). In sex-stratified analyses, the effect of BAC on sinus tachycardia was similar in males and females, both after adjustment for age (males: OR 2.12 (95% CI 1.75–2.58), P < 0.001; females: OR 2.02 (95% CI 1.51–2.69), P < 0.001) and after multivariable adjustment (males: OR 2.00 (95% CI 1.64–2.44), P < 0.001; females: OR 1.89 (95% CI 1.40–2.55), P < 0.001). We also noted a significant inverse association of respiratory sinus arrhythmia with BAC (Table 3). Interaction analyses did not suggest interaction between sex and BAC. Table 3 Association of arrhythmia prevalence with alcohol consumption

(95% CI 1.64–2.44), P < 0.001; females: OR 1.89 (95% CI 1.40–2.55), P < 0.001). We also noted a significant inverse association of respiratory sinus arrhythmia with BAC (Table 3). Interaction analyses did not suggest interaction between sex and BAC. Table 3 Association of arrhythmia prevalence with alcohol consumption Acute alcohol cohort Chronic alcohol cohort Adjusted for Multivariable Adjusted for Multivariable Age and Sex Adjustment Age and Sex Adjustment OR (95%CI) P OR (95%CI) P OR (95%CI) P OR (95%CI) P 1 Sinus arrhythmia 0.74 (0.43–1.30) 0.3 0.77 (0.44–1.35) 0.36 0.97 (0.92–1.03) 0.31 0.97 (0.91–1.02) 0.25 2 Sinus tachycardia 2.08 (1.77–2.45) <0.001 1.96 (1.66–2.31) <0.001 1.02 (1.01–1.04) 0.003 1.03 (1.01–1.05) 0.006 3 Premature atrial complexes 0.88 (0.46–1.70) 0.71 0.93 (0.48–1.81) 0.84 0.99 (0.96–1.01) 0.31 0.99 (0.96–1.01) 0.35 4 Premature ventricular complexes 1.11 (0.64–1.92) 0.71 1.07 (0.62–1.86) 0.81 0.99 (0.97–1.01) 0.43 0.99 (0.98–1.01) 0.45 5 Atrial fibrillation/flutter 1.45 (0.67–3.13) 0.35 1.39 (0.64–3.00) 0.83 1.00 (0.98–1.02) 0.80 1.00 (0.98–1.02) 0.84 Combination of 3, 4, 5 1.03 (0.70–1.51) 0.87 1.03 (0.70–1.51) 0.89 0.99 (0.98–1.00) 0.21 0.99 (0.98–1.01) 0.27 Combination of 1, 3, 4, 5 0.93 (0.68–1.27) 0.64 0.93 (0.68–1.28) 0.67 0.99 (0.98–1.00) 0.15 0.99 (0.98–1.00) 0.18 Combination of 1, 2, 3, 4, 5 1.87 (1.60–2.18) <0.001 1.75 (1.50–2.05) <0.001 1.00 (0.99–1.01) 0.94 1.00 (0.99–1.01) 0.90 Respiratory sinus arrhythmia 0.52 (0.44–0.63) <0.001 0.54 (0.45–0.65) <0.001 – – – – Odds ratios (OR) are presented per unit (i.e. 1 g/kg) increase of continuously measured breath alcohol concentration. Multivariable adjustment in the acute alcohol cohort included age, sex, history of heart disease, use of cardiovascular drugs, use of antiarrhythmic drugs, and active smoking status. Multivariable adjustment in the chronic alcohol cohort included age, sex, hypertension, smoking status, history of angina, myocardial infarction, diabetes mellitus, stroke, and use of cardiovascular and antiarrhythmic medication. Significant P-values are highlighted by bold print.

hmic drugs, and active smoking status. Multivariable adjustment in the chronic alcohol cohort included age, sex, hypertension, smoking status, history of angina, myocardial infarction, diabetes mellitus, stroke, and use of cardiovascular and antiarrhythmic medication. Significant P-values are highlighted by bold print. In KORA S4, we confirmed an association between sinus tachycardia and chronic alcohol consumption. Yet, the associated effect size was remarkably smaller than with acute alcohol consumption. No associations with chronic alcohol consumption were found for other arrhythmias or a combination of sinus tachycardia with other arrhythmia subtypes (Table 3). Discussion In our observational, cross-sectional acute alcohol cohort, we have recruited over 3000 participants with quantitatively measured acute alcohol consumption. Each participant received a 30 s ECG for arrhythmia analysis. Our main finding was a significant association of sinus tachycardia with acute alcohol intake, occurring in 25.9% of individuals. We also confirmed this association with chronic alcohol consumption in the community-based cohort KORA study. In the acute alcohol cohort, respiratory sinus arrhythmia as a marker of balanced autonomic tone was more common in those exposed to less alcohol.

a with acute alcohol intake, occurring in 25.9% of individuals. We also confirmed this association with chronic alcohol consumption in the community-based cohort KORA study. In the acute alcohol cohort, respiratory sinus arrhythmia as a marker of balanced autonomic tone was more common in those exposed to less alcohol. The Octoberfest is a traditional public festival in the city of Munich, Germany that is celebrated annually since 1810. It has gained strong international recognition and is renowned for serving Munich-brewed beer. Numerous visitors attend the festival primarily for the purpose of beer consumption. In 2015, 5.9 million visitors frequented the Octoberfest and consumed 7.5 million liters of beer. We thus considered the setting most suitable for conducting our acute alcohol study. We anticipated that by recruiting participants during all 16 days of the festival, we would be able to enrol a sufficient number of individuals presenting with a continuously distributed range of BAC. With an average BAC of 0.85 g/kg (range: 0–2.94 g/kg) we clearly achieved this goal. The modest day-by-day and circadian variabilities of BAC support the robustness of our results. The lively atmosphere in a beer tent deviated from AliveCor manufacturer recommendations for ECG recording. Despite adverse recording conditions, with >99.5% interpretable readings the obtained ECG quality was very high.

al. The modest day-by-day and circadian variabilities of BAC support the robustness of our results. The lively atmosphere in a beer tent deviated from AliveCor manufacturer recommendations for ECG recording. Despite adverse recording conditions, with >99.5% interpretable readings the obtained ECG quality was very high. We conservatively based sample size considerations on an estimated 1.5% prevalence of any arrhythmia under no acute influence of alcohol and at rest. Even without considering sinus tachycardia, we detected arrhythmias in 2.3% of our chronic and in 5.4% of our acute alcohol cohorts. We thus submit that we present sufficiently powered results. In our acute alcohol cohort, we found a strong and robust association of increased BAC with sinus tachycardia in particular. We were not able to associate increased BAC with other arrhythmia subtypes and specifically atrial fibrillation. This is despite prior reports that repeatedly suggested such a relation coining the term ‘holiday heart syndrome’.8 It is notable that so far studies have considered patients presenting with atrial fibrillation and have then retrospectively identified recent acute alcohol consumption as a presumed cause.8–10 Ours is the first large and prospective investigation to systematically analyse the immediate occurrence of arrhythmias under the influence of acute alcohol intake. We thus conclude that acute alcohol consumption leads to increased arrhythmia prevalence overall, with sinus tachycardia as the immediate main effect.

8–10 Ours is the first large and prospective investigation to systematically analyse the immediate occurrence of arrhythmias under the influence of acute alcohol intake. We thus conclude that acute alcohol consumption leads to increased arrhythmia prevalence overall, with sinus tachycardia as the immediate main effect. For chronic alcohol consumption, prior results are more equivocal. Based on large community-based studies, modest use does not appear to be a major risk factor for atrial fibrillation.16 Only long-term use in the Framingham Heart Study and beyond moderate use (>35 standard drinks per week) in the Copenhagen City Heart Study resulted in an increased risk for atrial fibrillation.12,17 Along these findings, in our own community-based analysis of low to moderate chronic alcohol consumption with less than 20 g/d on average, we did not find an association with an increased arrhythmia prevalence.

inks per week) in the Copenhagen City Heart Study resulted in an increased risk for atrial fibrillation.12,17 Along these findings, in our own community-based analysis of low to moderate chronic alcohol consumption with less than 20 g/d on average, we did not find an association with an increased arrhythmia prevalence. A main result of our study is the profound association of acute alcohol consumption with sinus tachycardia. As demonstrated in our sensitivity analysis, it is intuitive to assume that participants of a festival present with significantly higher rates of sinus tachycardia than under resting conditions. However, sinus tachycardia occurred almost 30% more often in Octoberfest participants under the influence of alcohol compared with those without alcohol consumption. Furthermore, the association of sinus tachycardia with higher BAC remained even after accounting for confounders. Although less pronounced, also in KORA S4 increased chronic alcohol consumption was associated with sinus tachycardia.

cipants under the influence of alcohol compared with those without alcohol consumption. Furthermore, the association of sinus tachycardia with higher BAC remained even after accounting for confounders. Although less pronounced, also in KORA S4 increased chronic alcohol consumption was associated with sinus tachycardia. Whereas ours is the largest prospective analysis to describe the association of acute alcohol consumption and sinus tachycardia under real life conditions, the relation has previously been noted. Prior studies have reported an increase in heart rate following alcohol intake in experimental settings.18–21 Consistently, these studies have identified alcohol induced alterations of the autonomic nervous system to elevate heart rate. Thereby, both an increase in sympathetic activity18,21 and a decrease in vagal tone19,20 have been described. Respiratory sinus arrhythmia is a simple measure of cardiac vagal tone and autonomic balance.22 In our acute alcohol cohort, participants with increased BAC presented with a significantly reduced prevalence of respiratory sinus arrhythmia. This finding concurs with the interpretation that the observed alcohol induced increase in heart rate is the consequence of changes in autonomic balance. Autonomic imbalance is strongly predisposing to the development of atrial fibrillation.23,24 Alcohol induced autonomic imbalance could thus hypothetically be a link to the later development of atrial fibrillation. Possibly, the short duration of our ECG recordings immediately during acute alcohol exposition resulted in an underestimation of alcohol induced cardiac arrhythmias and atrial fibrillation in particular. These considerations ask for additional research including longer ECG recordings during follow-up of hours and days after acute alcohol consumption.

our ECG recordings immediately during acute alcohol exposition resulted in an underestimation of alcohol induced cardiac arrhythmias and atrial fibrillation in particular. These considerations ask for additional research including longer ECG recordings during follow-up of hours and days after acute alcohol consumption. A number of considerations are warranted when interpreting our study. Given the public environment of our study, we were restricted in assessing personal questions and conducting physical examinations. Hence, several possible confounding factors remained unaddressed, including but not limited to the amount of alcohol consumption prior to BAC measurement, the participants’ common alcohol consumption behaviour, their use of recreational drugs, or their usual physical activity. A single 30 s ECG recording prevented the investigation of temporal relations between alcohol consumption and arrhythmia occurrence. Importantly, we cannot comment on the relevance of baseline heart rate prior to alcohol intake. We were thus only able to assess the prevalence but not the incidence of arrhythmias during follow-up. Future studies with longer ECG recordings will need to fill this gap. Respiratory sinus arrhythmia only partially reflects the influence of autonomic tone on heart rate variability. Additional research is warranted to investigate more elaborate measures of autonomic tone in relation to alcohol consumption. Due to our exclusion criteria, we were not able to study severely intoxicated individuals (BAC ≥3.00 g/kg). In KORA S4, chronic alcohol consumption was rather low compared with other community-based cohorts. We might thus have underestimated arrhythmia prevalence secondary to long-term alcohol use. Information on chronic alcohol consumption in our acute alcohol cohort was unavailable.

ntoxicated individuals (BAC ≥3.00 g/kg). In KORA S4, chronic alcohol consumption was rather low compared with other community-based cohorts. We might thus have underestimated arrhythmia prevalence secondary to long-term alcohol use. Information on chronic alcohol consumption in our acute alcohol cohort was unavailable. In conclusion, we have conducted a large prospective analysis of acute alcohol consumption on ECG-assessed cardiac arrhythmias. We thereby report good technical feasibility of ECG screening even under lively conditions at the Munich Octoberfest. Acute and—to a lesser extent—chronic alcohol consumption were associated with sinus tachycardia. Analysis of respiratory sinus arrhythmia as a measure of autonomic tone suggested that acute alcohol intake confers autonomic imbalance. Additional research is warranted to investigate if autonomic imbalance constitutes the link between sinus tachycardia and the occurrence of arrhythmias like atrial fibrillation, as implicated by reports of the so-called ‘Holiday Heart Syndrome’ (Figure3). Figure 3 MunichBREW study conclusions. The figure summarizes the study procedures and results in the panel shaded in green. These findings influence the generated hypothesis on ‘Holiday Heart Syndrome’ pathophysiology illustrated in the panel shaded in red. Importantly, additional research is warranted to support this hypothesis.

nichBREW study conclusions. The figure summarizes the study procedures and results in the panel shaded in green. These findings influence the generated hypothesis on ‘Holiday Heart Syndrome’ pathophysiology illustrated in the panel shaded in red. Importantly, additional research is warranted to support this hypothesis. Acknowledgements We are grateful to Staatliches Hofbräuhaus in München (director Dr Michael Möller) for supporting the conduction of our research. This work is part of the doctoral theses of Rebecca Herbel and Cathrine Drobesch. Funding This study was funded by the Stiftung Biomedizinische Alkoholforschung, institutional funds of the Department of Medicine I, University Hospital Munich, and by the European Commission’s Horizon 2020 research and innovation programme [grant number 633196]: CATCH ME. The KORA study was initiated and financed by the Helmholtz Zentrum München, German Research Centre for Environmental Health, funded by the German Federal Ministry of Education and Research (BMBF) and by the State of Bavaria. Conflict of interest: none declared.