Browse the corpus

than 20 cigarettes per day having a HR of 1·63 (1·30–2·03; figure 4). By contrast with men, adiposity did not appear to modify the risk associated with smoking among women, but this might reflect inadequate power to examine such associations (appendix p 12).Figure 4 Adjusted HRs of incident diabetes by smoking in women Stratified by age at risk and study area and adjusted for education, alcohol consumption, physical activity, body-mass index, and waist circumference. Analyses examining smoking duration were additionally adjusted for age at baseline. Tests for trend include all smoking categories. Test for heterogeneity across smoking categories includes all smoking categories. Other tests for heterogeneity include only smokers. HR=hazard ratio. Discussion In this large, nationwide, prospective cohort study, smoking was associated in a dose-dependent manner with increased risk of incident diabetes among men and women. Among men, the relative risk appeared stronger among individuals with higher adiposity. Given the amount smoked, women tended to have somewhat greater relative risk than men. Furthermore, after allowing for adiposity, smoking cessation was associated with an elevated risk of diabetes only among men who stopped smoking because of illness, and not those who stopped by choice.

Introduction Globally, an estimated 425 million people have diabetes, causing a substantial burden of premature death and disability.1 In China, the prevalence of diabetes has increased dramatically over recent decades, from approximately 1% in 1980,2 to around 11% in 2013.3 However, the factors underlying this escalating prevalence are incompletely understood. Apart from increasing adiposity,4 other lifestyle factors, including smoking, might play a part.5 In China, cigarette consumption has increased markedly over a similar time period, first in urban areas, then in rural areas, although this increase has occurred almost exclusively in men.6 About two-thirds of Chinese men now smoke,6 consuming roughly 40% of the world's cigarettes.7 A few prospective studies8, 9 have examined the association between smoking and diabetes in China, but these were constrained by small sample size, short follow-up (usually <5 years), insufficiently detailed information on smoking patterns, and restriction to particular urban populations. As such, substantial uncertainty remains about the associations of smoking and smoking cessation with diabetes in China.

n China, but these were constrained by small sample size, short follow-up (usually <5 years), insufficiently detailed information on smoking patterns, and restriction to particular urban populations. As such, substantial uncertainty remains about the associations of smoking and smoking cessation with diabetes in China. Accumulating evidence, mainly from developed countries, suggests a positive dose-response association of smoking with risk of type 2 diabetes.10, 11 Questions persist, however, about the causal nature of the observed associations, about the effects of smoking cessation on diabetes risk, and about the potential role of adiposity—a major cause of diabetes4—in affecting the association. While smokers tend to have lower adiposity than non-smokers, heavy smokers tend to have higher abdominal fat accumulation than light or non-smokers.12 Moreover, smoking cessation can lead to weight gain, with previous studies10 reporting elevated risk of diabetes within a few years of stopping. While weight gain could possibly account for this association,12 the excess risk might also be due to reverse causality, since a proportion of ex-smokers might have stopped because of poor health. However, previous studies have lacked the detailed information (eg, specific reasons for stopping) to clarify this association. Reliable elucidation of the associations of smoking and smoking cessation with diabetes will inform tobacco control strategies. Research in context Evidence before this study

Accumulating evidence, mainly from developed countries, suggests a positive dose-response association of smoking with risk of type 2 diabetes.10, 11 Questions persist, however, about the causal nature of the observed associations, about the effects of smoking cessation on diabetes risk, and about the potential role of adiposity—a major cause of diabetes4—in affecting the association. While smokers tend to have lower adiposity than non-smokers, heavy smokers tend to have higher abdominal fat accumulation than light or non-smokers.12 Moreover, smoking cessation can lead to weight gain, with previous studies10 reporting elevated risk of diabetes within a few years of stopping. While weight gain could possibly account for this association,12 the excess risk might also be due to reverse causality, since a proportion of ex-smokers might have stopped because of poor health. However, previous studies have lacked the detailed information (eg, specific reasons for stopping) to clarify this association. Reliable elucidation of the associations of smoking and smoking cessation with diabetes will inform tobacco control strategies. Research in context Evidence before this study We searched PubMed for articles published in English up to Aug 31, 2017, using the search terms “smoking”, “diabetes”, “prospective”, and “cohort”. Four relevant meta-analyses were identified, which all found positive associations of current regular smoking with risk of incident diabetes (pooled hazard ratios ranging from 1·37 to 1·44), and higher risk with greater number of cigarettes smoked. One meta-analysis, including ten prospective studies, also examined the association with smoking cessation, showing that individuals who had stopped for less than 5 years had significantly elevated risk, which decreased as time since cessation increased. These meta-analyses included populations from predominantly developed countries, where widespread cigarette smoking among young adults has persisted for several decades. In 2014, about 40% of the world's cigarettes were consumed in China, almost exclusively by men, but the increase in cigarette usage, first urban then rural, has been relatively recent. Two prospective studies in China, which recruited participants between 2000 and 2006, have also examined the association between smoking and diabetes. However, these studies included relatively small sample sizes (3598 and 51 464) and few diabetes cases (160 and 1304), with inconsistent findings.

een relatively recent. Two prospective studies in China, which recruited participants between 2000 and 2006, have also examined the association between smoking and diabetes. However, these studies included relatively small sample sizes (3598 and 51 464) and few diabetes cases (160 and 1304), with inconsistent findings. Added value of this study Our aim was to assess the associations of smoking and smoking cessation with risk of incident diabetes among urban and rural Chinese adults. We showed that smoking is associated with a higher risk of diabetes among men and women. Dose–response effects were clear with age at first starting to smoke regularly and with amount smoked. Among men, the relative risks were more extreme in urban areas than in rural areas. Given the amount smoked, the relative risks appeared somewhat greater among women than men. We also found strong evidence that high body fat content might amplify the excess risk associated with smoking among men. Furthermore, stopping smoking by choice (ie, before development of serious illness) was not associated with excess risk of developing diabetes in the first 5 years of cessation, in contrast with the higher risk among those who stopped because of illness. Implications of all the available evidence

Our aim was to assess the associations of smoking and smoking cessation with risk of incident diabetes among urban and rural Chinese adults. We showed that smoking is associated with a higher risk of diabetes among men and women. Dose–response effects were clear with age at first starting to smoke regularly and with amount smoked. Among men, the relative risks were more extreme in urban areas than in rural areas. Given the amount smoked, the relative risks appeared somewhat greater among women than men. We also found strong evidence that high body fat content might amplify the excess risk associated with smoking among men. Furthermore, stopping smoking by choice (ie, before development of serious illness) was not associated with excess risk of developing diabetes in the first 5 years of cessation, in contrast with the higher risk among those who stopped because of illness. Implications of all the available evidence Smoking is an important modifiable risk factor for diabetes in diverse populations. Further studies, including Mendelian randomisation analyses, might enable investigation of the causality of the association. Irrespective of causality, smoking should be targeted as an important lifestyle factor in disease prevention strategies, including for diabetes, in China and elsewhere. Using data from the large nationwide China Kadoorie Biobank (CKB) prospective study, we evaluate prospective associations of smoking and smoking cessation with risk of incident diabetes, and explore potential effect modification by adiposity.

Smoking is an important modifiable risk factor for diabetes in diverse populations. Further studies, including Mendelian randomisation analyses, might enable investigation of the causality of the association. Irrespective of causality, smoking should be targeted as an important lifestyle factor in disease prevention strategies, including for diabetes, in China and elsewhere. Using data from the large nationwide China Kadoorie Biobank (CKB) prospective study, we evaluate prospective associations of smoking and smoking cessation with risk of incident diabetes, and explore potential effect modification by adiposity. Methods Study population Details of the CKB study design and methods have been described previously.13, 14 The CKB is a prospective cohort study of 512 891 (210 259 men and 302 632 women) adults recruited from ten areas (five urban and five rural) of China. These areas were chosen from China's nationally representative Disease Surveillance Points to ensure coverage of a range of socioeconomic levels, disease patterns, and risk factors. Between June 25, 2004, and July 15, 2008, all registered residents aged 35–74 years were invited to participate in the baseline survey; approximately 30% responded, including about 10 000 participants slightly outside the target age range (making the actual baseline age range from 30 years to 79 years).

d risk factors. Between June 25, 2004, and July 15, 2008, all registered residents aged 35–74 years were invited to participate in the baseline survey; approximately 30% responded, including about 10 000 participants slightly outside the target age range (making the actual baseline age range from 30 years to 79 years). At study assessment clinics, participants were interviewed by trained health workers, using laptop-based questionnaires, on sociodemographic status, lifestyle (eg, smoking, alcohol consumption, diet, and physical activity), and medical history. Anthropometric measurements (eg, height, weight, and waist and hip circumferences) were done using calibrated instruments with standard protocols, with participants wearing light clothing and no shoes. Weight was measured using a body composition analyser (TANITA-TBF-300GS; Tanita Corporation, Tokyo, Japan), which also measured body fat percentage with its inbuilt proprietary algorithm based on foot-to-foot bioelectrical impedance. Standing height was obtained to the nearest 0·1 cm with a stadiometer. Body-mass index (BMI) was calculated as weight in kilograms divided by the square of standing height in metres (kg/m2). Waist and hip circumferences were measured using a non-stretchable tape to the nearest 0·1 cm. Systolic blood pressure and diastolic blood pressure were measured in a seated position after at least 5 min rest using a digital sphygmomanometer (Omron UA-779; A&D Instruments, Abingdon, UK). A 10 mL non-fasting blood sample (with time since last meal recorded) was collected into an EDTA vacutainer for storage and onsite random plasma glucose testing using the SureStep Plus system (LifeScan, Milpitas, CA, USA). Two resurveys (from May 26 to Oct 10, 2008, and from Aug 4, 2013, to Sept 18, 2014) including a randomly selected sample of approximately 5% of participants were undertaken, collecting the same information as at baseline. Ethics approval was obtained from the Oxford University Tropical Research Ethics Committee and the Chinese Center for Disease Control and Prevention Ethical Review Committee.

, 2014) including a randomly selected sample of approximately 5% of participants were undertaken, collecting the same information as at baseline. Ethics approval was obtained from the Oxford University Tropical Research Ethics Committee and the Chinese Center for Disease Control and Prevention Ethical Review Committee. Assessment of smoking Smoking information collected included frequency, amount, and type of tobacco smoked, currently and in the past, and the ages at which participants started smoking regularly and stopped smoking. Among current and former (or ex) regular smokers, additional information was collected on the types and amount of tobacco smoked when last smoking, with amount calculated in g/day, assuming 1 g of tobacco per factory cigarette, 2 g per cigar, and actual amount in pipes and hand-rolled cigarettes as reported. Smoking duration was derived from the age at starting smoking regularly to age at baseline (current smokers) or duration since quitting at baseline (ex-smokers). Among individuals who reported having stopped smoking, the main reason for smoking cessation (physical illness or by choice) was collected. Four categories of smoking status were defined: never-smokers (participants who reported not smoking at recruitment and who had smoked <100 cigarettes during their lifetime); ever-regular smokers (participants who reported ever smoking at least one cigarette or 1 g tobacco per day for ≥6 months, or who had stopped smoking ≥6 months before recruitment because of ill health [included in this group to avoid bias]); ex-smokers by choice (participants who stopped smoking ≥6 months before recruitment by choice [providing information on the effects of smoking cessation]); and occasional smokers (participants who did not meet the criteria for never-smoker or ever-regular smoker, and who had not stopped smoking completely for at least the 6 months before recruitment). Pack-years were not examined since smoking ten cigarettes per day for 30 years might have different effects than smoking 20 cigarettes per day for 15 years.15 Exhaled carbon monoxide was measured to validate smoking exposure (CareFusion MicroCO meter; CareFusion, Chatham, UK).

at least the 6 months before recruitment). Pack-years were not examined since smoking ten cigarettes per day for 30 years might have different effects than smoking 20 cigarettes per day for 15 years.15 Exhaled carbon monoxide was measured to validate smoking exposure (CareFusion MicroCO meter; CareFusion, Chatham, UK). Follow-up for incident diabetes Cause-specific morbidity and mortality were monitored through linkage, via unique national identification number, with disease (including diabetes) and death registries, and with the national health insurance system, which has almost universal coverage (>98%) across the ten study areas. Annual active follow-up was performed by checking against local residential records to confirm survival status and to minimise losses to follow-up. Causes of death were classified on the basis of official death certificates and were checked against available medical records when necessary. Linkage to the health insurance system enabled identification of diagnoses resulting in, or during, hospital admissions. Incident diabetes cases were identified through the diabetes disease surveillance system and through diabetes diagnoses (ICD-10 E10-E14 codes) recorded in health insurance databases or underlying or contributing to death on death certificates.

ed identification of diagnoses resulting in, or during, hospital admissions. Incident diabetes cases were identified through the diabetes disease surveillance system and through diabetes diagnoses (ICD-10 E10-E14 codes) recorded in health insurance databases or underlying or contributing to death on death certificates. Statistical analysis In this study, participants with self-reported diabetes that was clinically diagnosed, screen-detected diabetes (no self-reported diabetes and a plasma glucose concentration ≥7·0 mmol/L and a fasting time ≥8 h, a plasma glucose concentration ≥11·1 mmol/L and a fasting time <8 h, or a fasting plasma glucose concentration ≥7·0 mmol/L),16 or with missing BMI data at baseline were excluded. Sensitivity analyses further excluded participants with a history of cancer or cardiovascular disease at baseline, and, separately, participants who developed diabetes, died, or were lost to follow-up during the first 3 years of follow-up.

oncentration ≥7·0 mmol/L),16 or with missing BMI data at baseline were excluded. Sensitivity analyses further excluded participants with a history of cancer or cardiovascular disease at baseline, and, separately, participants who developed diabetes, died, or were lost to follow-up during the first 3 years of follow-up. All analyses were done separately in men and women. Prevalence and mean values of characteristics at baseline were calculated across smoking categories standardised by age (5-year age groups) and study area, when appropriate. Cox regression was used to estimate adjusted hazard ratios (HRs) of diabetes, with time since entry into the study as the underlying timescale. Participants contributed person-years until diagnosis of diabetes, death, loss to follow-up, or the censoring date (Dec 31, 2015). Models were stratified by age at risk (5-year groups) and study area, and adjusted for educational attainment (no formal education, primary school, middle or high school, or college or university and above), alcohol consumption (never, occasional, or ever-regular), and physical activity as metabolic equivalent of task hours per day (model 1), and further adjusted for BMI and waist circumference as continuous variables (model 2).

o formal education, primary school, middle or high school, or college or university and above), alcohol consumption (never, occasional, or ever-regular), and physical activity as metabolic equivalent of task hours per day (model 1), and further adjusted for BMI and waist circumference as continuous variables (model 2). The main analyses were done separately in urban areas and rural areas because of well documented differences in their past smoking patterns.6 χ2 tests for trend and heterogeneity were applied to the log HRs and their SEs.17 The floating absolute risk method was used; this method does not alter the value of the HRs but provides a 95% CI for each HR, derived from the variance of the log risk in that category, enabling comparisons between any two categories and not only with the reference group.18 Incidence rates per 1000 person-years (Ri) were calculated from the adjusted HRs using a weighted method with the number of events in each group as the weighting variable, Wi, using the formula: Ri=HRi×overall rate([∑Wi×HRi]total number of events) Analyses were done using SAS version 9.3 and R version 3.3.2. Role of the funding source The funders of the study had no role in the design of the study, collection, analysis, or interpretation of data, or in the writing of the report. FB and ZC had full access to the data and responsibility for the final decision to submit for publication.

Analyses were done using SAS version 9.3 and R version 3.3.2. Role of the funding source The funders of the study had no role in the design of the study, collection, analysis, or interpretation of data, or in the writing of the report. FB and ZC had full access to the data and responsibility for the final decision to submit for publication. Results 16 162 participants with self-reported diabetes that was clinically diagnosed, 14 138 participants with screen-detected diabetes, and two with missing BMI data at baseline were excluded, leaving 482 589 participants (198 574 men and 284 015 women) for the main analyses. Sensitivity analyses further excluded 21 376 participants with a history of cancer or cardiovascular disease at baseline, and, separately, 2760 participants who developed diabetes, died, or were lost to follow-up during the first 3 years of follow-up were excluded. By Jan 1, 2016, 37 289 (7·3%) participants had died, with only 4875 (<1%) lost to follow-up.

21 376 participants with a history of cancer or cardiovascular disease at baseline, and, separately, 2760 participants who developed diabetes, died, or were lost to follow-up during the first 3 years of follow-up were excluded. By Jan 1, 2016, 37 289 (7·3%) participants had died, with only 4875 (<1%) lost to follow-up. Overall 68% (n=134 975) of men (62% [n=51 990] urban and 72% [n=82 985] rural) and 3% (n=7811) of women (2% [n=2772] urban and 3% [n=5039] rural) were ever-regular smokers. Among men, ever-regular smokers were younger (p<0·0001), and ex-smokers older (p<0·0001), than never-smokers (table 1). Ever-regular smokers seemed more likely than never-smokers to have lower educational attainment, be regular alcohol or tea drinkers, and consume more meat, but less fresh fruit (table 1). Among ever-regular smokers, there was a positive dose-dependent association of amount smoked with waist circumference—a marker of central adiposity—independent of general adiposity, but a less clear association with general adiposity (BMI or body fat percentage; appendix p 7). Blood pressure was lower among ever-regular smokers and higher among ex-smokers, compared with never-smokers (table 1). Mean exhaled carbon monoxide was much higher among current regular smokers (15·4 ppm [SD 10·6]; ie, ever-regular smokers excluding ex-smokers who stopped because of illness) than among participants in other smoking categories, while random plasma glucose concentrations did not differ across smoking categories (table 1).Table 1 Baseline characteristics for men by smoking status

ividuals with higher adiposity. Given the amount smoked, women tended to have somewhat greater relative risk than men. Furthermore, after allowing for adiposity, smoking cessation was associated with an elevated risk of diabetes only among men who stopped smoking because of illness, and not those who stopped by choice. A 2015 meta-analysis10 of published data from 84 prospective studies, including almost 300 000 diabetes cases, reported an overall HR of diabetes of 1·37 (95% prediction interval 1·11–1·70), associated with active smoking, with a non-significantly stronger association among men (1·42) than women (1·33). Most previous studies10, 11, 19, 20, 21, 22 have been in high-income countries, where, by contrast with China, persistent use of cigarettes among young adults has lasted for a longer period of time. Before the 1970s, total cigarette consumption in China was low and stable. Since the early 1980s, however, cigarette consumption has increased five-fold, almost exclusively among men (as shown by the National Bureau of Statistics of China). In China, although more than two-thirds of men smoke6 and more than 10% of adults are affected by diabetes,3 evidence on the association of smoking with diabetes is sparse and inconsistent. In the Shanghai Men's Health study,9 with 1304 incident diabetes cases (identified through self-report) after 5 years of follow-up, the increased risk of diabetes was evident among men who smoked 20 cigarettes or more per day (HR 1·25, 95% CI 1·00–1·56), but not among men who smoked less. By contrast, a small prospective study8 in China, including only 160 diabetes cases reported an overall HR of 4·16 (2·77–6·24) among current smokers. Our study included almost ten times more diabetes cases than previous studies in China combined, and demonstrated a modest, but clear, positive dose-dependent association of smoking with risk of diabetes among men and women.

regular smokers (15·4 ppm [SD 10·6]; ie, ever-regular smokers excluding ex-smokers who stopped because of illness) than among participants in other smoking categories, while random plasma glucose concentrations did not differ across smoking categories (table 1).Table 1 Baseline characteristics for men by smoking status Never-smokers (n=28 214) Ex-smokers (n=12 950) Occasional smokers (n=22 435) Ever-regular smokers (n=134 975) Age (years) 53·4 (11·9) 55·3 (10·9) 50·0 (11·1) 51·7 (10·5) Urban residence 15 278 (53·7%) 6715 (49·9%) 9735 (43·4%) 51 990 (38·5%) Education >6 years 18 622 (60·5%) 7629 (59·3%) 14 834 (60·6%) 73 035 (55·6%) Exhaled carbon monoxide (ppm) 5·0 (5·5) 5·3 (5·5) 5·6 (5·0) 14·1 (9·5) Random plasma glucose (mmol/L)* 5·6 (1·3) 5·6 (1·3) 5·6 (1·2) 5·6 (1·2) Systolic blood pressure (mm Hg) 133·1 (20·3) 134·2 (20·0) 132·5 (20·5) 132·0 (19·5) Diastolic blood pressure (mm Hg) 79·6 (12·1) 80·4 (12·2) 79·4 (11·9) 78·8 (11·3) Weight (kg) 64·6 (10·7) 66·4 (10·8) 64·8 (10·2) 63·4 (9·8) Height (cm) 164·8 (6·6) 165·5 (6·5) 164·9 (6·1) 165·3 (5·8) BMI (kg/m2) 23·7 (3·4) 24·2 (3·3) 23·7 (3·2) 23·1 (3·1) Waist circumference (cm) 82·1 (10·1) 84·0 (10·1) 82·4 (9·5) 81·3 (9·3) Waist-hip ratio 0·9 (0·1) 0·9 (0·1) 0·9 (0·1) 0·9 (0·1) Body fat percentage (%)† 22·3 (6·5) 22·3 (6·2) 23·3 (6·5) 21·6 (5·9) Physical activity (MET h/day) 22·1 (15·5) 22·2 (14·9) 22·9 (13·8) 22·3 (13·2) Regular alcohol drinker 6627 (24·3%) 6213 (45·8%) 6719 (31·9%) 63 371 (46·8%) Regular tea drinker 10 166 (35·3%) 6184 (50·4%) 8170 (40·3%) 76 319 (55·3%) Regular meat consumption 15 356 (48·5%) 7013 (52·3%) 11 022 (49·2%) 68 106 (51·9%) Regular fresh fruit consumption 9310 (28·3%) 3944 (26·9%) 5609 (26·1%) 26 368 (20·5%) Family history of diabetes 1826 (6·2%) 789 (5·7%) 1472 (6·2%) 7898 (5·9%) Values are mean (SD) or number of participants (%), standardised to age and study area structure of study population. Ex-smokers=ex-smokers who stopped by choice. Ever-regular smokers=current smokers and ex-smokers who stopped because of illness. BMI=body-mass index. MET h=metabolic equivalent of task hours. Regular alcohol drinker=current or previous consumption at least once weekly. Regular tea drinker=consumption at least once weekly. Regular meat consumption=consumption on at least 4–6 days per week. Regular fruit consumption=consumption on at least 4–6 days per week.

ss index. MET h=metabolic equivalent of task hours. Regular alcohol drinker=current or previous consumption at least once weekly. Regular tea drinker=consumption at least once weekly. Regular meat consumption=consumption on at least 4–6 days per week. Regular fruit consumption=consumption on at least 4–6 days per week. * Data missing for 301 never-smokers, 229 ex-smokers, 338 occasional smokers, and 2426 ever-regular smokers. † Data missing for 15 never-smokers, six ex-smokers, ten occasional smokers, and 85 ever-regular smokers.

ss index. MET h=metabolic equivalent of task hours. Regular alcohol drinker=current or previous consumption at least once weekly. Regular tea drinker=consumption at least once weekly. Regular meat consumption=consumption on at least 4–6 days per week. Regular fruit consumption=consumption on at least 4–6 days per week. * Data missing for 301 never-smokers, 229 ex-smokers, 338 occasional smokers, and 2426 ever-regular smokers. † Data missing for 15 never-smokers, six ex-smokers, ten occasional smokers, and 85 ever-regular smokers. During roughly 9 years' follow-up (4·3 million person-years), 5194 men and 8458 women developed new-onset diabetes among 482 589 participants without previous diabetes, with incidence rates of 3·0 per 1000 person-years for men and 3·3 per 1000 person-years for women. Smoking was associated with higher risk of diabetes, with adjusted HRs of 1·15 (95% CI 1·05–1·27) for ex-smokers (stopped by choice), 1·11 (1·02–1·21) for occasional smokers, and 1·04 (1·00–1·08) for ever-regular smokers, among men (table 2). The HRs were greater in urban areas than in rural areas (table 2). After further adjustment for BMI and waist circumference, the risks were substantially attenuated in ex-smokers (1·05, 0·96–1·16), but increased in ever-regular smokers (1·14, 1·10–1·18; table 2). The associations were not materially changed by further excluding individuals with a history of cancer or cardiovascular disease at baseline, or individuals who developed incident diabetes during the first 3 years of follow-up, or by further adjusting for dietary factors (appendix p 2) or age at baseline as a continuous variable.Table 2 Adjusted HR of incident diabetes in men according to smoking status

istory of cancer or cardiovascular disease at baseline, or individuals who developed incident diabetes during the first 3 years of follow-up, or by further adjusting for dietary factors (appendix p 2) or age at baseline as a continuous variable.Table 2 Adjusted HR of incident diabetes in men according to smoking status Urban Rural Overall Cases (n) Rate* HR (95% CI) Cases (n) Rate* HR (95% CI) Cases (n) Rate* HR (95% CI) Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Never-smokers 400 2·80 1·00 (0·90–1·11) 1·00 (0·90–1·11) 319 2·60 1·00 (0·89–1·12) 1·00 (0·89–1·12) 719 2·68 1·00 (0·93–1·08) 1·00 (0·93–1·08) Ex-smokers 229 3·11 1·21 (1·06–1·38) 1·11 (0·98–1·27) 201 2·59 1·09 (0·95–1·25) 1·00 (0·87–1·14) 430 2·82 1·15 (1·05–1·27) 1·05 (0·96–1·16) Occasional smokers 268 3·07 1·10 (0·98–1·25) 1·09 (0·97–1·24) 291 2·96 1·11 (0·99–1·24) 1·14 (1·01–1·28) 559 3·00 1·11 (1·02–1·21) 1·12 (1·03–1·21) Ever-regular smokers 1426 3·31 1·11 (1·05–1·18) 1·18 (1·12–1·25) 2060 2·87 0·97 (0·93–1·02) 1·10 (1·05–1·15) 3486 3·05 1·04 (1·00–1·08) 1·14 (1·10–1·18) p for heterogeneity .. .. 0·14 0·045 .. .. 0·12 0·20 .. .. 0·061 0·019 Model 1 was stratified by age at risk and study area and adjusted for education, alcohol consumption, and physical activity. Model 2 was additionally adjusted for body-mass index and waist circumference. HR=hazard ratio. Ex-smokers=ex-smokers who stopped by choice. Ever-regular smokers=current smokers and ex-smokers who stopped because of illness.

at risk and study area and adjusted for education, alcohol consumption, and physical activity. Model 2 was additionally adjusted for body-mass index and waist circumference. HR=hazard ratio. Ex-smokers=ex-smokers who stopped by choice. Ever-regular smokers=current smokers and ex-smokers who stopped because of illness. * Incidence rate per 1000 person-years was calculated from the HRs (model 2) using a weighted method with the number of events in each group as the weighting variable. Among urban men, the HRs differed little by tobacco type, but were higher among those who started smoking at an earlier age (1·12, 1·20, and 1·27 at ≥25 years, 20–24 years, and <20 years, respectively; p for trend=0·00073; figure 1). Similarly, the HRs increased with increasing duration of smoking (p for trend=0·00073; figure 1), and amount smoked (p for trend <0·0001; figure 1), with a HR of 1·52 (95% CI 1·35–1·72) among those who smoked 30 cigarettes or more per day. The patterns of association were similar among current regular smokers (appendix p 8). Similar associations, albeit more modest, were observed among rural men (figure 1), with no significant heterogeneity across individual rural (p for heterogeneity=0·28) or urban (p for heterogeneity=0·57) study areas (appendix p 9). Similarly, the HRs for ever-regular smoking did not vary markedly across different population subgroups (eg, by age, education, physical activity, alcohol intake, and systolic blood pressure; appendix p 10).Figure 1 Adjusted HRs of incident diabetes by smoking in men

for heterogeneity=0·57) study areas (appendix p 9). Similarly, the HRs for ever-regular smoking did not vary markedly across different population subgroups (eg, by age, education, physical activity, alcohol intake, and systolic blood pressure; appendix p 10).Figure 1 Adjusted HRs of incident diabetes by smoking in men Stratified by age at risk and study area and adjusted for education, alcohol consumption, physical activity, body-mass index, and waist circumference. Analyses examining smoking duration were additionally adjusted for age at baseline. Ever-regular smokers excludes occasional smokers (n=22 435) and ex-smokers who stopped by choice (n=12 950). Tests for trend include all smoking categories. Tests for heterogeneity include only smokers. HR=hazard ratio. When stratified by levels of adiposity, the HRs associated with smoking amount appeared stronger among men with higher levels of adiposity (figure 2, appendix p 3). Compared with never-smokers, the HRs for those who smoked 30 cigarettes or more per day were 1·18 (95% CI 1·04–1·33), 1·22 (1·08–1·39), and 1·60 (1·39–1·85) among those with waist circumferences of less than 85 cm, 85 cm to less than 95 cm, and 95 cm or more, respectively (figure 2; p for trend=0·0016). Similar patterns were observed across strata of body fat percentage (p for trend <0·0001) and BMI (figure 2), although with a less extreme trend (p for trend=0·048).Figure 2 Adjusted HRs of incident diabetes by amount smoked, stratified by levels of adiposity, in men

r more, respectively (figure 2; p for trend=0·0016). Similar patterns were observed across strata of body fat percentage (p for trend <0·0001) and BMI (figure 2), although with a less extreme trend (p for trend=0·048).Figure 2 Adjusted HRs of incident diabetes by amount smoked, stratified by levels of adiposity, in men Stratified by age at risk and study area and adjusted for education, alcohol consumption, and physical activity. Ever-regular smokers excludes occasional smokers and ex-smokers who stopped by choice. HR=hazard ratio. BMI=body-mass index. WC=waist circumference. BF%=body fat percentage.

r more, respectively (figure 2; p for trend=0·0016). Similar patterns were observed across strata of body fat percentage (p for trend <0·0001) and BMI (figure 2), although with a less extreme trend (p for trend=0·048).Figure 2 Adjusted HRs of incident diabetes by amount smoked, stratified by levels of adiposity, in men Stratified by age at risk and study area and adjusted for education, alcohol consumption, and physical activity. Ever-regular smokers excludes occasional smokers and ex-smokers who stopped by choice. HR=hazard ratio. BMI=body-mass index. WC=waist circumference. BF%=body fat percentage. Among men who stopped smoking because of illness, the excess risk of diabetes was significant, which attenuated with increasing time since quitting: HRs were 1·44 (95% CI 1·25–1·66), 1·37 (1·20–1·56), and 1·19 (0·96–1·47) for those who had stopped for less than 5 years, 5–14 years, and 15 years or more before baseline, respectively (p for trend=0·16; figure 3). Consideration of time since smoking cessation as a time-updating variable showed similar results (appendix p 11). Additional adjustment for BMI and waist circumference slightly attenuated these associations (figure 3). By contrast, among ex-smokers who stopped by choice, the HRs tended to increase with longer time since quitting, with HRs of 0·96 (0·79–1·17), 1·20 (1·05–1·38) and 1·28 (1·07–1·52) among those who had stopped for less than 5 years, 5–14 years, and 15 years or more before baseline (p for trend=0·041; figure 3). However, these associations were attenuated after further adjustment for BMI and waist circumference (p for trend=0·077).Figure 3 Adjusted HRs of incident diabetes in male ex-smokers by reasons for, and years after, smoking cessation

years, 5–14 years, and 15 years or more before baseline (p for trend=0·041; figure 3). However, these associations were attenuated after further adjustment for BMI and waist circumference (p for trend=0·077).Figure 3 Adjusted HRs of incident diabetes in male ex-smokers by reasons for, and years after, smoking cessation Model 1 was stratified by age at risk and study area and adjusted for education, alcohol consumption, and physical activity. Model 2 was additionally adjusted for body-mass index and waist circumference. Tests for trend include only ex-smokers. HR=hazard ratio. Among women, ever-regular smokers were, on average, older, and had higher adiposity, than never-smokers (appendix p 4). Compared with never-smokers, the HR of diabetes was 1·33 (95% CI 1·20–1·47; figure 4), which was somewhat more extreme than that among men (p for heterogeneity=0·0049). The adjusted HRs by tobacco category, age of starting smoking, duration of smoking, and daily amount smoked were broadly similar to those in men, with women who smoked more than 20 cigarettes per day having a HR of 1·63 (1·30–2·03; figure 4). By contrast with men, adiposity did not appear to modify the risk associated with smoking among women, but this might reflect inadequate power to examine such associations (appendix p 12).Figure 4 Adjusted HRs of incident diabetes by smoking in women

uding only 160 diabetes cases reported an overall HR of 4·16 (2·77–6·24) among current smokers. Our study included almost ten times more diabetes cases than previous studies in China combined, and demonstrated a modest, but clear, positive dose-dependent association of smoking with risk of diabetes among men and women. Unlike in developed populations, where tobacco use is almost exclusively in the form of cigarettes, a high proportion of smokers in China (especially those born before the 1970s in rural areas) smoke non-cigarette tobacco (eg, pipes and water pipes).6 However, the present study showed that the risk of diabetes did not differ significantly by tobacco type. Total amount of tobacco smoked did not differ between cigarette (18·3 cigarettes or equivalents per day) and non-cigarette or mixed smokers (18·1 cigarettes or equivalents per day), nor in the proportions reporting inhaling deeply into the lungs (76% vs 73%), suggesting that tobacco dose, rather than type, might be more relevant for risk of diabetes. Overall, our relative risk estimates were more modest than those reported previously in developed populations and were less extreme among men from rural areas than from urban areas, consistent with previous findings showing that the tobacco epidemic is still at an early stage in China, particularly in rural areas.6 Cigarette consumption became widespread earlier in urban, than in rural, areas in China,6 but with rapid economic development, cigarettes have become more readily available and affordable in rural areas.23, 24 An upward trend in diabetes risk attributable to smoking would, therefore, be expected in rural areas, unless there is widespread smoking cessation.

d earlier in urban, than in rural, areas in China,6 but with rapid economic development, cigarettes have become more readily available and affordable in rural areas.23, 24 An upward trend in diabetes risk attributable to smoking would, therefore, be expected in rural areas, unless there is widespread smoking cessation. Obesity is a well recognised modifiable risk factor for diabetes.4 Smokers, on average, tend to be leaner than non-smokers, possibly reflecting appetite suppression and elevated resting metabolic rate associated with smoking.12 However, heavy smokers are more likely to have higher abdominal adiposity than light or non-smokers, as shown in previous studies12 and in the CKB. These adiposity associations could help to explain the excess diabetes risk among smokers, despite their relatively low average BMI. Previous evidence in developed populations has been conflicting, however, with regards to the interaction of adiposity and smoking on diabetes risk.8, 19, 20, 21, 22 Although two studies,19, 20 each with approximately 3000 diabetes cases, showed no such effect modification by BMI, a larger study,22 with approximately 12 000 incident cases, reported a stronger association among individuals with lower BMI, contrary to the present and another Chinese study's findings.8 Only two small studies20, 21 have examined the interaction of smoking amount with adiposity, showing inconsistent findings. In the present study, we observed a stronger association of amount smoked with risk of diabetes among individuals with higher adiposity, especially as assessed by body fat percentage. This observation might help to explain the greater HRs among women than men for a given amount smoked, because women have a greater proportion of body fat compared with men. The precise mechanism underlying this interaction is not clear, but animal studies have shown release of fatty acids from adipose tissue in response to nicotine administration, which might, in turn, promote insulin resistance25 and development of diabetes.

women have a greater proportion of body fat compared with men. The precise mechanism underlying this interaction is not clear, but animal studies have shown release of fatty acids from adipose tissue in response to nicotine administration, which might, in turn, promote insulin resistance25 and development of diabetes. A meta-analysis10 pooling published data from ten prospective studies reported that individuals who stopped smoking more recently (<5 years) had a higher risk of diabetes than those who had stopped for a longer duration. Some smokers are likely to have quit because of poor health (including development of diabetes), resulting in a spurious association, but information about reasons for quitting smoking was not available in these previous studies. This assertion is supported by data in the CKB, in which individuals who stopped smoking because of illness tended to have higher baseline prevalence of cardiovascular diseases, hypertension, and diabetes than those who stopped by choice and than never-smokers (appendix p 5). Moreover, the present study showed that the excess risk among recent quitters was evident among individuals who had stopped smoking because of illness but not among those who stopped by choice, suggesting that the previously reported findings were due chiefly to reverse causality. Among individuals who stopped smoking by choice, excess risk was small among those who had stopped for longer periods of time (≥5 years), which could be attributable (at least partly) to weight gain following smoking cessation (appendix p 13), in keeping with attenuation of this excess risk after adjustment for adiposity. The interaction between adiposity and smoking is also likely to explain the contrasting effects of additional adjustment for adiposity on the risks of diabetes among ex-smokers (more adipose) and ever-regular smokers (less adipose).

p 13), in keeping with attenuation of this excess risk after adjustment for adiposity. The interaction between adiposity and smoking is also likely to explain the contrasting effects of additional adjustment for adiposity on the risks of diabetes among ex-smokers (more adipose) and ever-regular smokers (less adipose). In addition to the large study population, the availability of uniquely detailed information on smoking and smoking cessation enabled more comprehensive assessment of the associations of smoking with diabetes in the current study than has previously been possible. Detailed review of medical records of about 1000 randomly selected diabetes cases in CKB showed a high positive predictive value of diabetes diagnosis (97% based on American Diabetes Association diagnostic criteria26 and medication use). Furthermore, the diverse study population and extremely low loss to follow-up ensure generalisability of the findings and limit potential for biased risk estimates. There are, however, limitations of our study. First, without data on weight change after smoking cessation, it was not possible to assess the direct effect of this factor on the association of smoking cessation with diabetes. Second, potential confounding by total energy or specific nutrient consumption could not be assessed adequately. Third, a proportion of new-onset diabetes cases might not have been captured during follow-up. However, the two resurveys of a random subset of participants enabled estimation of the extent of undiagnosed diabetes, demonstrating that diabetes prevalence in the CKB was moderately lower than in contemporaneous nationally representative surveys, which used a combination of self-report, fasting blood glucose and glycated haemoglobin measurement, and oral glucose tolerance testing to detect diabetes (appendix p 6).3, 27 However, sensitivity analyses comparing the association of regular smoking with diagnosed and undiagnosed diabetes among men included in both resurveys showed no significant difference in the magnitude of association (p for heterogeneity 0·86). Fourth, type 1 diabetes cases might have been included in the analyses. However, given the age of the study population, the number would be expected to be extremely small and unlikely to affect the risk estimates. Finally, reliable investigation of the effects of smoking cessation on risk of diabetes and of differential association by adiposity among women was limited by the low prevalence of smoking among women in China.

dy population, the number would be expected to be extremely small and unlikely to affect the risk estimates. Finally, reliable investigation of the effects of smoking cessation on risk of diabetes and of differential association by adiposity among women was limited by the low prevalence of smoking among women in China. Among men in China, the excess diabetes risk associated with smoking is likely to increase substantially in future generations because the tobacco epidemic is maturing and population mean adiposity is increasing.28 Encouragingly, despite increased risk of diabetes among female smokers, very few Chinese women now smoke, and stopping smoking before the onset of major illness could prevent the elevated risk of diabetes (and other diseases) associated with smoking. Irrespective of whether these associations are causal or not, smoking should be targeted as an important lifestyle factor in future disease prevention strategies, including for diabetes, in China and elsewhere. For more on the CKB see http://www.ckbiobank.org/ For more on the National Bureau of Statistics of China see http://data.stats.gov.cn/ Supplementary Material Supplementary appendix

Among men in China, the excess diabetes risk associated with smoking is likely to increase substantially in future generations because the tobacco epidemic is maturing and population mean adiposity is increasing.28 Encouragingly, despite increased risk of diabetes among female smokers, very few Chinese women now smoke, and stopping smoking before the onset of major illness could prevent the elevated risk of diabetes (and other diseases) associated with smoking. Irrespective of whether these associations are causal or not, smoking should be targeted as an important lifestyle factor in future disease prevention strategies, including for diabetes, in China and elsewhere. For more on the CKB see http://www.ckbiobank.org/ For more on the National Bureau of Statistics of China see http://data.stats.gov.cn/ Supplementary Material Supplementary appendix Acknowledgments The baseline survey and the first resurvey were supported by a research grant from the Kadoorie Charitable Foundation in Hong Kong. The long-term continuation of the project is supported by programme grants from the UK Wellcome Trust (088158/Z/09/Z, 104085/Z/14/Z), the Chinese Ministry of Science and Technology (2011BAI09B01, 2012–14), the Chinese National Natural Science Foundation (81390540, 81390541, 81390544), and the National Key Research and Development Program of China (2016YFC0900500, 2016YFC0900501, 2016YFC0900504, 2016YFC1303904). The British Heart Foundation (BHF), Medical Research Council, and Cancer Research UK provide core funding to the Oxford Clinical Trial Service Unit. XL acknowledges support from the China Scholarship Council (201606285145). FB acknowledges support from the BHF Centre of Research Excellence, Oxford (RE/13/1/30181). The chief acknowledgment is to the participants, the project staff, and the China National Centre for Disease Control and Prevention and its regional offices for access to death and disease registries. The Chinese National Health Insurance scheme provides electronic linkage to all hospital admission data.

/13/1/30181). The chief acknowledgment is to the participants, the project staff, and the China National Centre for Disease Control and Prevention and its regional offices for access to death and disease registries. The Chinese National Health Insurance scheme provides electronic linkage to all hospital admission data. Contributors XL and FB contributed to the concept and design of the study, statistically analysed data, drafted the manuscript, and provided critical revision of the manuscript for important intellectual content. LY was involved in the acquisition of data, contributed to the concept and design of the study, and provided critical revision of the manuscript for important intellectual content. CK contributed to the concept and design of the study, statistically analysed data, and provided critical revision of the manuscript for important intellectual content. YG, HD, ZB, CY, JL, KW, and HZ were involved in the acquisition of data and provided critical revision of the manuscript for important intellectual content. YC, JC, and LL were involved in the acquisition of data, obtained funding, and provided critical revision of the manuscript for important intellectual content. RCl, RCo, and RP obtained funding and provided critical revision of the manuscript for important intellectual content. ZC obtained funding, contributed to the concept and design of the study, was involved in the acquisition of data, drafted the manuscript, and provided critical revision of the manuscript for important intellectual content. FB and ZC are the guarantors of this work and, as such, had full access to all the data and take responsibility for the integrity of the data and the accuracy of the data analysis.

ed in the acquisition of data, drafted the manuscript, and provided critical revision of the manuscript for important intellectual content. FB and ZC are the guarantors of this work and, as such, had full access to all the data and take responsibility for the integrity of the data and the accuracy of the data analysis. China Kadoorie Biobank collaborative group International Steering Committee: Junshi Chen, Zhengming Chen (principal investigator [PI]), Robert Clarke, Rory Collins, Yu Guo, Liming Li (PI), Jun Lv, Richard Peto, Robin Walters. International Co-ordinating Centre, Oxford: Daniel Avery, Derrick Bennett, Ruth Boxall, Fiona Bragg, Yumei Chang, Yiping Chen, Zhengming Chen, Robert Clarke, Huaidong Du, Simon Gilbert, Alex Hacker, Michael Holmes, Christiana Kartsonaki, Rene Kerosi, Garry Lancaster, Kuang Lin, John McDonnell, Iona Millwood, Qunhua Nie, Jayakrishnan Radhakrishnan, Paul Ryder, Sam Sansome, Dan Schmidt, Rajani Sohoni, Becky Stevens, Iain Turnbull, Robin Walters, Jenny Wang, Lin Wang, Neil Wright, Ling Yang, Xiaoming Yang. National Co-ordinating Centre, Beijing: Zheng Bian, Ge Chen, Yu Guo, Xiao Han, Can Hou, Jun Lv, Pei Pei, Shuzhen Qu, Yunlong Tan, Canqing Yu. Ten Regional Co-ordinating Centres. (Qingdao) Qingdao CDC: Zengchang Pang, Ruqin Gao, Shaojie Wang, Yongmei Liu, Ranran Du, Yajing Zang, Liang Cheng, Xiaocao Tian, Hua Zhang. Licang CDC: Silu Lv, Junzheng Wang, Wei Hou. (Heilongjiang) Provincial CDC: Jiyuan Yin, Ge Jiang, Xue Zhou. Nangang CDC: Liqiu Yang, Hui He, Bo Yu, Yanjie Li, Huaiyi Mu, Qinai Xu, Meiling Dou, Jiaojiao Ren. (Hainan) Provincial CDC: Shanqing Wang, Ximin Hu, Hongmei Wang, Jinyan Chen, Yan Fu, Zhenwang Fu, Xiaohuan Wang. Meilan CDC: Min Weng, Xiangyang Zheng, Yilei Li, Huimei Li, Yanjun Wang. (Jiangsu) Provincial CDC: Ming Wu, Jinyi Zhou, Ran Tao, Jie Yang. Suzhou CDC: Chuanming Ni, Jun Zhang, Yihe Hu, Yan Lu, Liangcai Ma, Aiyu Tang, Shuo Zhang, Jianrong Jin, Jingchao Liu. (Guangxi) Provincial CDC: Zhenzhu Tang, Naying Chen, Ying Huang. Liuzhou CDC: Mingqiang Li, Jinhuai Meng, Rong Pan, Qilian Jiang, Weiyuan Zhang, Yun Liu, Liuping Wei, Liyuan Zhou, Ningyu Chen, Hairong Guan. (Sichuan) Provincial CDC: Xianping Wu, Ningmei Zhang, Xiaofang Chen, Xuefeng Tang. Pengzhou CDC: Guojin Luo, Jianguo Li, Xiaofang Chen, Xunfu Zhong, Jiaqiu Liu, Qiang Sun. (Gansu) Provincial CDC: Pengfei Ge, Xiaolan Ren, Caixia Dong. Maiji CDC: Hui Zhang, Enke Mao, Xiaoping Wang, Tao Wang, Xi Zhang.

Chen, Hairong Guan. (Sichuan) Provincial CDC: Xianping Wu, Ningmei Zhang, Xiaofang Chen, Xuefeng Tang. Pengzhou CDC: Guojin Luo, Jianguo Li, Xiaofang Chen, Xunfu Zhong, Jiaqiu Liu, Qiang Sun. (Gansu) Provincial CDC: Pengfei Ge, Xiaolan Ren, Caixia Dong. Maiji CDC: Hui Zhang, Enke Mao, Xiaoping Wang, Tao Wang, Xi Zhang. (Henan) Provincial CDC: Ding Zhang, Gang Zhou, Shixian Feng, Liang Chang, Lei Fan. Huixian CDC: Yulian Gao, Tianyou He, Huarong Sun, Pan He, Chen Hu, Qiannan Lv, Xukui Zhang. (Zhejiang) Provincial CDC: Min Yu, Ruying Hu, Hao Wang. Tongxiang CDC: Yijian Qian, Chunmei Wang, Kaixue Xie, Lingli Chen, Yidan Zhang, Dongxia Pan. (Hunan) Provincial CDC: Yuelong Huang, Biyun Chen, Li Yin, Donghui Jin, Huilin Liu, Zhongxi Fu, Qiaohua Xu. Liuyang CDC: Xin Xu, Hao Zhang, Youping Xiong, Huajun Long, Xianzhi Li, Libo Zhang, Zhe Qiu. Declaration of interests RCo reports support from Nuffield Department of Population Health, during the conduct of the study; grants and personal fees from British Heart Foundation, grants from Cancer Research UK, Medical Research Council, Merck & Co, National Institute for Health Research, and Wellcome Trust, personal fees from UK Biobank, and other from Pfizer, outside the submitted work. Additionally, RCo has a patent for a statin-related myopathy genetic test licensed to University of Oxford from Boston Heart Diagnostics (he has waived any personal reward). All other authors declare no competing interests.

Introduction Cuba is a middle-income country with universal health coverage.1 The health system focuses particularly on primary care and preventive medicine. This approach has delivered substantial reductions in infant and child mortality over the past few decades,2 but premature mortality in middle age (35–69 years) remains high. At 2015 mortality rates, 25% of men and 17% of women would die in middle age, mostly from non-communicable diseases, including about a third from cardiovascular disease.3 In 2013, WHO set global targets for the control of non-communicable diseases, including a 25% relative reduction in the prevalence of elevated blood pressure by 2025.4 Meta-analyses of prospective studies have shown that moderate differences in blood pressure have important implications for cardiovascular risk.5 Trials of blood-pressure-lowering medication have confirmed that much of the excess risk can be reversed within a few years of starting treatment.6, 7 In Cuba, hypertension control in primary care has been prioritised as a cost-effective means of addressing premature cardiovascular mortality.8 However, there are few large-scale studies on the prevalence and management of hypertension in Cuba, and there is no direct evidence from large prospective studies showing the expected benefit on cardiovascular mortality of efforts to improve hypertension control. Research in context Evidence before this study

In Cuba, hypertension control in primary care has been prioritised as a cost-effective means of addressing premature cardiovascular mortality.8 However, there are few large-scale studies on the prevalence and management of hypertension in Cuba, and there is no direct evidence from large prospective studies showing the expected benefit on cardiovascular mortality of efforts to improve hypertension control. Research in context Evidence before this study We did a literature search to identify cross-sectional or prospective studies done in Cuba reporting on the prevalence of hypertension or its associated risks for cardiovascular disease. We searched PubMed for articles published between Jan 1, 1960, and Feb 2, 2018, in any language, using the search string “Cuba (OR Cuban) AND hypertension (OR blood pressure) AND cross-sectional study (OR prospective study OR cohort study)”. We also searched the reference lists of retrieved articles to identify further relevant publications. We identified several small studies (<5000 adults) and two larger studies (a national survey in 2001 of 23 000 adults living in urban areas and a regional survey of 55 000 adults in 2004–06). Neither of the larger studies reported the proportions of people who had diagnosed hypertension, were receiving treatment, or in whom blood pressure was controlled or assessed the excess cardiovascular risks of uncontrolled hypertension. Added value of this study

We did a literature search to identify cross-sectional or prospective studies done in Cuba reporting on the prevalence of hypertension or its associated risks for cardiovascular disease. We searched PubMed for articles published between Jan 1, 1960, and Feb 2, 2018, in any language, using the search string “Cuba (OR Cuban) AND hypertension (OR blood pressure) AND cross-sectional study (OR prospective study OR cohort study)”. We also searched the reference lists of retrieved articles to identify further relevant publications. We identified several small studies (<5000 adults) and two larger studies (a national survey in 2001 of 23 000 adults living in urban areas and a regional survey of 55 000 adults in 2004–06). Neither of the larger studies reported the proportions of people who had diagnosed hypertension, were receiving treatment, or in whom blood pressure was controlled or assessed the excess cardiovascular risks of uncontrolled hypertension. Added value of this study This study is to our knowledge the largest prospective analysis of hypertension-related mortality done in Cuba, and one of the largest in Latin America. The prevalence, diagnosis, and management of hypertension are assessed by age, sex, education, area, and previous diagnosis of cardiovascular disease. We have also assessed the age-specific and sex-specific effects of uncontrolled hypertension on cardiovascular mortality. At ages 35–79 years, about a third of participants had hypertension. The proportions of people in whom hypertension was diagnosed and being treated were commensurate with those in some high-income countries, but control of blood pressure was low. Overall, about 80% of all people with hypertension had uncontrolled blood pressure, which was associated with roughly a doubling of risk of premature cardiovascular death.

whom hypertension was diagnosed and being treated were commensurate with those in some high-income countries, but control of blood pressure was low. Overall, about 80% of all people with hypertension had uncontrolled blood pressure, which was associated with roughly a doubling of risk of premature cardiovascular death. Implications of all the available evidence Addressing the burden of hypertension in Cuba will require not only improved detection of hypertension in primary care but also greater control of blood pressure among people with hypertension. Several initiatives using community-wide awareness campaigns and simplified hypertension treatment regimens are ongoing in Cuba, the findings of which will be relevant nationally and to other low-income and middle-income countries. Cuba's health system, which focuses on primary care and preventative medicine, has delivered substantial reductions in infant and child mortality over the past few decades, but death from the chronic diseases of middle age remains high. Hypertension control in primary care has been prioritised as a cost-effective means of addressing premature death in Cuba, and this study helps to inform the delivery of such programmes and estimate their potential impact on cardiovascular mortality.

ades, but death from the chronic diseases of middle age remains high. Hypertension control in primary care has been prioritised as a cost-effective means of addressing premature death in Cuba, and this study helps to inform the delivery of such programmes and estimate their potential impact on cardiovascular mortality. We report a large prospective study of the burden of hypertension and its associated risks of premature cardiovascular mortality in Cuba. In reporting, we had several aims: to describe the prevalence, diagnosis, and management of hypertension by age, sex, education, area, and previous diagnosis of cardiovascular disease; assess the age-specific and sex-specific effects of uncontrolled hypertension on cardiovascular mortality; and estimate the proportion of cardiovascular deaths attributable to hypertension in Cuba. Methods Study design and participants This was a prospective cohort study that, from Jan 1, 1996, to Nov 24, 2002, recruited members of the general population in five areas of Cuba, including the capital city (Havana) and four geographically dispersed provinces (Pinar del Río, Matanzas, La Habana, and Camagüey). The populations of these areas collectively accounted for just under half of Cuba's population in 2000. Within each area, family medical practices were randomly selected with use of a computer-generated random allocation sequence, and local health-care staff (mostly physicians) were asked to recruit all residents aged 30 years or older within their catchment area. On average, each practice provided care for around 150 families.2

ach area, family medical practices were randomly selected with use of a computer-generated random allocation sequence, and local health-care staff (mostly physicians) were asked to recruit all residents aged 30 years or older within their catchment area. On average, each practice provided care for around 150 families.2 Ethics approval was provided by the National Institute of Cardiology and Cardiac Surgery in Havana. All participants provided written informed consent. Study procedures Trained health-care staff from the family practices made household visits and invited eligible household members to participate in the study. They recorded age, sex, education, occupation, information on health-related behaviours (including smoking and alcohol intake), and medical history. Blood pressure was measured twice while the participant was seated, using standard techniques and a calibrated manual sphygmomanometer. The mean of the two measurements was used in our analyses. Height and weight were measured separately at the medical practice because the mechanical scales commonly used at the time of survey in Cuba were not easily portable. The original data collection form was in Spanish, but an English translation is available in the appendix. In 2006–08, participants in two municipalities (Colón and Jagüey Grande) in the province of Matanzas who had participated in the first survey were invited to be resurveyed. The same procedures were used as for the baseline survey.

Study procedures Trained health-care staff from the family practices made household visits and invited eligible household members to participate in the study. They recorded age, sex, education, occupation, information on health-related behaviours (including smoking and alcohol intake), and medical history. Blood pressure was measured twice while the participant was seated, using standard techniques and a calibrated manual sphygmomanometer. The mean of the two measurements was used in our analyses. Height and weight were measured separately at the medical practice because the mechanical scales commonly used at the time of survey in Cuba were not easily portable. The original data collection form was in Spanish, but an English translation is available in the appendix. In 2006–08, participants in two municipalities (Colón and Jagüey Grande) in the province of Matanzas who had participated in the first survey were invited to be resurveyed. The same procedures were used as for the baseline survey. Deaths were identified through linkage by national identification numbers to the Cuban Public Health Ministry records until Dec 31, 2016. We were able to capture deaths for participants who moved out of the study areas to other parts of Cuba, but not for those who emigrated. The underlying causes of death were coded according to the International Classification of Diseases ninth and tenth editions (ICD-9 and ICD-10). We used ICD-9 codes for deaths between 1996 and 2000, and ICD-10 codes for deaths from 2001 onwards. Cardiovascular deaths were defined as deaths from myocardial infarction, stroke, or other vascular disease (ICD-9 codes 390–459, and 798, and ICD-10 codes I00–I99, and R96).

ninth and tenth editions (ICD-9 and ICD-10). We used ICD-9 codes for deaths between 1996 and 2000, and ICD-10 codes for deaths from 2001 onwards. Cardiovascular deaths were defined as deaths from myocardial infarction, stroke, or other vascular disease (ICD-9 codes 390–459, and 798, and ICD-10 codes I00–I99, and R96). Statistical analysis The analyses excluded participants with any missing data on the demographic characteristics of age, sex, education, and area, missing or implausible blood pressure values, and age outside the range of interest (35–79 years). Mean systolic blood pressure (SBP), mean diastolic blood pressure (DBP), and prevalence of hypertension were calculated for men and women separately in the age groups 35–39, 40–49, 50–59, 60–69, and 70–79 years. We classified participants as having hypertension if they had SBP 140 mm Hg or greater or DBP 90 mm Hg or greater on measurement at baseline or if they reported that they had a diagnosis of hypertension and were receiving blood-pressure-lowering medication (irrespective of measured blood pressure at baseline).9

years. We classified participants as having hypertension if they had SBP 140 mm Hg or greater or DBP 90 mm Hg or greater on measurement at baseline or if they reported that they had a diagnosis of hypertension and were receiving blood-pressure-lowering medication (irrespective of measured blood pressure at baseline).9 To assess the management of hypertension, we calculated the proportion of people at baseline who had hypertension diagnosed by a doctor, the proportion of those diagnosed who were being treated with blood-pressure-lowering medication, and the proportion of those receiving treatment who had controlled blood pressure (SBP <140 mm Hg and DBP <90 mm Hg). These proportions were estimated overall and by age, sex, area, education, ethnicity, history of cardiovascular disease, and season of baseline survey (because seasonal variation in blood pressure has been found in some populations10). Analyses were standardised, where appropriate, for age, sex, and area.

P <90 mm Hg). These proportions were estimated overall and by age, sex, area, education, ethnicity, history of cardiovascular disease, and season of baseline survey (because seasonal variation in blood pressure has been found in some populations10). Analyses were standardised, where appropriate, for age, sex, and area. Cox regression was used to calculate rate ratios (RRs) and 95% CIs for cardiovascular death among participants with and without uncontrolled hypertension. To limit reverse causality, these analyses excluded people with previously diagnosed cardiovascular disease. RRs were estimated by sex and age at risk (age groups 35–59, 60–69, and 70–79 years) and adjusted for age at risk in 5-year age groups, area, and level of education. The regression analysis used time in study as the underlying time variable and adjusted for current age by fitting a term that allows the RR in each decade of age to be estimated as the geometric mean of the RRs in the first and second half of that decade. Age-specific and sex-specific population-attributable fractions were calculated with the equation

Cox regression was used to calculate rate ratios (RRs) and 95% CIs for cardiovascular death among participants with and without uncontrolled hypertension. To limit reverse causality, these analyses excluded people with previously diagnosed cardiovascular disease. RRs were estimated by sex and age at risk (age groups 35–59, 60–69, and 70–79 years) and adjusted for age at risk in 5-year age groups, area, and level of education. The regression analysis used time in study as the underlying time variable and adjusted for current age by fitting a term that allows the RR in each decade of age to be estimated as the geometric mean of the RRs in the first and second half of that decade. Age-specific and sex-specific population-attributable fractions were calculated with the equation Pe(RR-1)/RR where Pe is the proportion of cardiovascular deaths occurring in participants with uncontrolled hypertension and RR is the adjusted cardiovascular death RR in participants with versus those without uncontrolled hypertension.11 Estimates of the number of deaths attributable to uncontrolled hypertension in Cuba were made by applying the population-attributable fractions to the age-specific and sex-specific number of deaths from cardiovascular disease in 2015, using data from the Global Burden of Disease study.3 All analyses were done with SAS version 9.3, and figures were plotted with R version 2.14.1.

ed hypertension in Cuba were made by applying the population-attributable fractions to the age-specific and sex-specific number of deaths from cardiovascular disease in 2015, using data from the Global Burden of Disease study.3 All analyses were done with SAS version 9.3, and figures were plotted with R version 2.14.1. Role of the funding source The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data and had final responsibility for the decision to submit for publication.

ed hypertension in Cuba were made by applying the population-attributable fractions to the age-specific and sex-specific number of deaths from cardiovascular disease in 2015, using data from the Global Burden of Disease study.3 All analyses were done with SAS version 9.3, and figures were plotted with R version 2.14.1. Role of the funding source The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data and had final responsibility for the decision to submit for publication. Results Of 215 practices approached, none refused to participate. Between Jan 1, 1996, and Nov 24, 2002, of 198 049 residents in the catchment areas, 146 556 (74%) men and women were interviewed in the first survey. We excluded 164 participants with missing demographic information, 68 with missing or implausible blood pressure values, and 10 111 outside the age range of interest, leaving 136 111 participants eligible for inclusion in the analyses (table 1). The mean age of participants was 54 (SD 12) years, just over half were women, and most were white. Almost all participants (127 355 [94%]) had received some formal education, and two-thirds (89 460 [66%]) were educated to secondary school level or higher. Most participants were recruited from Matanzas in Central Cuba and Camagüey in eastern Cuba. 35 703 (59%) men smoked (cigarettes or cigars) and 17 311 (29%) drank alcohol at least weekly. By contrast, 25 025 (33%) women smoked and only 3434 (5%) drank at least weekly. Overall, mean SBP, DBP, and body-mass index were similar in men and women. By Dec 31, 2016, 26 465 (19%) participants had died and 204 (<1%) were lost to follow-up through emigration or uncertainty about the national identification number provided at baseline.Table 1 Baseline characteristics

(5%) drank at least weekly. Overall, mean SBP, DBP, and body-mass index were similar in men and women. By Dec 31, 2016, 26 465 (19%) participants had died and 204 (<1%) were lost to follow-up through emigration or uncertainty about the national identification number provided at baseline.Table 1 Baseline characteristics Men (n=60 164) Women (n=75 947) All (n=136 111) Age at entry (years) 35–39 9291 (15%) 12 301 (16%) 21 592 (16%) 40–49 16 774 (28%) 21 651 (29%) 38 425 (28%) 50–59 15 347 (26%) 19 014 (25%) 34 361 (25%) 60–69 11 192 (19%) 13 920 (18%) 25 112 (18%) 70–79 7560 (13%) 9061 (12%) 16 621 (12%) Mean (SD) 54 (12) 53 (12) 54 (12) Blood pressure (mm Hg) Systolic 126 (14) 124 (16) 125 (15) Diastolic 81 (9) 80 (10) 80 (10) Area of Cuba Matanzas 26 778 (45%) 32 877 (43%) 59 655 (44%) Camagüey 24 160 (40%) 30 619 (40%) 54 779 (40%) Pinar del Río 4873 (8%) 5903 (8%) 10 776 (8%) Ciudad de La Habana 3088 (5%) 4946 (7%) 8034 (6%) La Habana 1265 (2%) 1602 (2%) 2867 (2%) Educational level Less than primary 3240 (5%) 5516 (7%) 8756 (6%) Primary 14 758 (25%) 23 137 (30%) 37 895 (28%) Lower secondary 16 142 (27%) 21 106 (28%) 37 248 (27%) High school/technical college 20 068 (33%) 19 637 (26%) 39 705 (29%) University 5956 (10%) 6551 (9%) 12 507 (9%) Ethnicity White 46 605 (77%) 58 307 (77%) 104 912 (77%) Black 8763 (15%) 10 791 (14%) 19 554 (14%) Mixed 4596 (8%) 6597 (9%) 11 193 (8%) Other 200 (<1%) 252 (<1%) 452 (<1%) Previous cardiovascular disease No 57 782 (96%) 73 664 (97%) 131 446 (97%) Yes 2382 (4%) 2283 (3%) 4665 (3%) Season of baseline survey Spring 12 649 (21%) 16 242 (21%) 28 891 (21%) Summer 19 612 (33%) 23 812 (31%) 43 424 (32%) Autumn 12 151 (20%) 16 055 (21%) 28 206 (21%) Winter 15 752 (26%) 19 838 (26%) 35 590 (26%) Data are n (%) or mean (SD). Data exclude participants with missing demographic information, missing or implausible blood pressure values, and age outside the range of interest (35–79 years).

612 (33%) 23 812 (31%) 43 424 (32%) Autumn 12 151 (20%) 16 055 (21%) 28 206 (21%) Winter 15 752 (26%) 19 838 (26%) 35 590 (26%) Data are n (%) or mean (SD). Data exclude participants with missing demographic information, missing or implausible blood pressure values, and age outside the range of interest (35–79 years). We found a positive and roughly linear association between age and mean SBP at baseline in both sexes (figure 1). Mean SBP was higher in men than women at younger ages but increased more steeply with age in women than men, and was higher in women from around age 65 years and older. By contrast, the association of mean DBP with age was non-linear. As with SBP, mean DBP was higher in men than women at younger ages and increased more steeply with age in women, but from age 65 years there was little association with age and no difference between the sexes.Figure 1 Mean blood pressure, by age and sex (A) SBP. (B) DBP. Data are mean (95% CI). Means are standardised for area. Analyses in 136 111 participants. DBP=diastolic blood pressure. SBP=systolic blood pressure.

We found a positive and roughly linear association between age and mean SBP at baseline in both sexes (figure 1). Mean SBP was higher in men than women at younger ages but increased more steeply with age in women than men, and was higher in women from around age 65 years and older. By contrast, the association of mean DBP with age was non-linear. As with SBP, mean DBP was higher in men than women at younger ages and increased more steeply with age in women, but from age 65 years there was little association with age and no difference between the sexes.Figure 1 Mean blood pressure, by age and sex (A) SBP. (B) DBP. Data are mean (95% CI). Means are standardised for area. Analyses in 136 111 participants. DBP=diastolic blood pressure. SBP=systolic blood pressure. Overall, about a third of adults had hypertension (32% of men and 35% of women; table 2, figure 2). 6% had isolated systolic hypertension, 8% isolated diastolic hypertension, and 14% both systolic and diastolic hypertension, and 6% had controlled hypertension (appendix). 20% of all participants had blood pressure consistent with stage 1 hypertension (SBP 140–159 mm Hg, DBP 90–99 mm Hg, or both12) and 7% had higher blood pressure values (appendix). The prevalence of hypertension increased substantially with age in both sexes and was higher in men than women at younger ages (19% in men and 14% in women aged 35–39 years) but higher in women than men at older ages (50% in women and 43% in men at age 70–79 years; appendix).Table 2 Prevalence, diagnosis, treatment, and control of hypertension, by age and sex*

ubstantially with age in both sexes and was higher in men than women at younger ages (19% in men and 14% in women aged 35–39 years) but higher in women than men at older ages (50% in women and 43% in men at age 70–79 years; appendix).Table 2 Prevalence, diagnosis, treatment, and control of hypertension, by age and sex* Number of participants Prevalence of hypertension Proportion with diagnosed hypertension Proportion of diagnosed patients treated Controlled hypertension Among treated Among all hypertensives Men 35–39 years 9291 18·9% 48·2% 67·7% 36·0% 11·7% 40–49 years 16 774 26·4% 53·0% 71·2% 34·8% 13·1% 50–59 years 15 347 34·9% 60·9% 74·0% 34·2% 15·4% 60–69 years 11 192 41·2% 61·0% 76·4% 31·0% 14·5% 70–79 years 7560 42·5% 57·0% 77·1% 33·0% 14·5% All 60 164 32·0% 57·3% 74·1% 33·4% 14·2% Women 35–39 years 12 301 14·1% 62·0% 71·4% 44·4% 19·6% 40–49 years 21 651 24·7% 69·9% 76·1% 39·6% 21·0% 50–59 years 19 014 40·6% 76·2% 78·1% 36·6% 21·8% 60–69 years 13 920 49·4% 76·6% 77·2% 34·7% 20·5% 70–79 years 9061 50·2% 71·7% 75·5% 33·4% 18·1% All 75 947 34·7% 73·4% 76·7% 36·5% 20·6% Men and women 35–39 years 21 592 16·2% 54·9% 69·7% 40·7% 15·5% 40–49 years 38 425 25·5% 62·1% 74·2% 37·8% 17·4% 50–59 years 34 361 38·1% 70·0% 76·7% 35·8% 19·2% 60–69 years 25 112 45·8% 70·4% 76·9% 33·5% 18·1% 70–79 years 16 621 46·8% 65·8% 76·1% 33·2% 16·6% Overall 136 111 33·5% 66·5% 75·8% 35·5% 17·9% * Proportions are standardised for area and, where appropriate, age and sex. Figure 2 Prevalence of hypertension, by age and sex

Number of participants Prevalence of hypertension Proportion with diagnosed hypertension Proportion of diagnosed patients treated Controlled hypertension Among treated Among all hypertensives Men 35–39 years 9291 18·9% 48·2% 67·7% 36·0% 11·7% 40–49 years 16 774 26·4% 53·0% 71·2% 34·8% 13·1% 50–59 years 15 347 34·9% 60·9% 74·0% 34·2% 15·4% 60–69 years 11 192 41·2% 61·0% 76·4% 31·0% 14·5% 70–79 years 7560 42·5% 57·0% 77·1% 33·0% 14·5% All 60 164 32·0% 57·3% 74·1% 33·4% 14·2% Women 35–39 years 12 301 14·1% 62·0% 71·4% 44·4% 19·6% 40–49 years 21 651 24·7% 69·9% 76·1% 39·6% 21·0% 50–59 years 19 014 40·6% 76·2% 78·1% 36·6% 21·8% 60–69 years 13 920 49·4% 76·6% 77·2% 34·7% 20·5% 70–79 years 9061 50·2% 71·7% 75·5% 33·4% 18·1% All 75 947 34·7% 73·4% 76·7% 36·5% 20·6% Men and women 35–39 years 21 592 16·2% 54·9% 69·7% 40·7% 15·5% 40–49 years 38 425 25·5% 62·1% 74·2% 37·8% 17·4% 50–59 years 34 361 38·1% 70·0% 76·7% 35·8% 19·2% 60–69 years 25 112 45·8% 70·4% 76·9% 33·5% 18·1% 70–79 years 16 621 46·8% 65·8% 76·1% 33·2% 16·6% Overall 136 111 33·5% 66·5% 75·8% 35·5% 17·9% * Proportions are standardised for area and, where appropriate, age and sex. Figure 2 Prevalence of hypertension, by age and sex (A) Men. (B) Women. Controlled hypertension at baseline is defined as systolic blood pressure less than 140 mm Hg and diastolic blood pressure less than 90 mm Hg. Prevalence is standardised for area and, where appropriate, age The analysis included 136 111 participants. All=participants aged 35–79 years.

Figure 2 Prevalence of hypertension, by age and sex (A) Men. (B) Women. Controlled hypertension at baseline is defined as systolic blood pressure less than 140 mm Hg and diastolic blood pressure less than 90 mm Hg. Prevalence is standardised for area and, where appropriate, age The analysis included 136 111 participants. All=participants aged 35–79 years. Among participants with hypertension at baseline, 67% had a diagnosis of hypertension and among these 76% were receiving treatment. In 36% of participants receiving treatment, blood pressure was controlled (table 2). Overall, 18% of participants with hypertension (14% of men and 21% of women) had controlled blood pressure (table 2). The proportion of participants with diagnosed hypertension increased with increasing age, but at all ages was higher in women than men (table 2). The proportion of participants with diagnosed hypertension who were treated also increased with age but did not differ substantially between the sexes. By contrast, control of hypertension among people receiving treatment declined with age, but was low even at younger ages in both men and women (table 2). 69% of participants treated for hypertension were taking one blood-pressure-lowering medication, 28% were taking two medications, and 3% were taking three medications (appendix). 61% of treated participants were taking diuretics (27% in combination with other antihypertensive drugs and 34% alone) and only 5% of all patients being treated for hypertension were taking angiotensin-converting-enzyme inhibitors (appendix).

were taking two medications, and 3% were taking three medications (appendix). 61% of treated participants were taking diuretics (27% in combination with other antihypertensive drugs and 34% alone) and only 5% of all patients being treated for hypertension were taking angiotensin-converting-enzyme inhibitors (appendix). The prevalence of hypertension did not differ greatly by survey area, educational level, or season of survey (table 3), but was somewhat higher among black participants and those with mixed ethnicity than white participants, and was substantially higher among those with previous cardiovascular disease than those without. The proportion of all participants with hypertension in whom blood pressure was controlled varied by area, from 14% in Matanzas to 27% in Pinar del Río, and was slightly greater among participants with higher levels of education, those of white ethnicity, and those with previous cardiovascular disease (table 3). However, even among participants with previous cardiovascular disease, blood pressure was controlled in only 26%.Table 3 Prevalence, diagnosis, treatment, and control of hypertension* by selected characteristics†

er levels of education, those of white ethnicity, and those with previous cardiovascular disease (table 3). However, even among participants with previous cardiovascular disease, blood pressure was controlled in only 26%.Table 3 Prevalence, diagnosis, treatment, and control of hypertension* by selected characteristics† Number of participants Prevalence of hypertension Proportion with diagnosed hypertension Proportion of diagnosed patients treated Controlled hypertension Among treated Among all hypertensives Area Matanzas 59 655 34·5% 62·3% 69·8% 31·6% 13·8% Camagüey 54 779 32·5% 67·9% 77·5% 37·3% 19·6% Pinar del Río 10 776 32·2% 77·8% 89·0% 38·8% 26·7% Ciudad de la Habana 8034 35·7% 75·3% 83·6% 39·8% 25·0% La Habana 2867 31·3% 63·1% 76·4% 36·9% 17·9% Educational level Less than primary 8756 33·5% 61·8% 74·9% 34·8% 16·5% Primary 37 895 34·3% 64·9% 75·0% 34·9% 17·0% Lower secondary 37 248 33·0% 66·5% 75·8% 35·5% 18·0% High school/technical college 39 705 32·8% 67·6% 76·2% 38·2% 19·6% University 12 507 33·2% 70·0% 76·6% 39·8% 21·3% Ethnicity White 104 912 31·7% 65·6% 75·7% 37·3% 18·6% Black 19 554 41·8% 69·4% 75·9% 29·2% 15·4% Mixed 11 193 36·5% 67·4% 75·5% 32·2% 16·5% Other 452 32·8% 65·9% 65·6% 27·7% 12·8% Previous cardiovascular disease No 131 446 32·8% 65·5% 75·1% 35·2% 17·4% Yes 4665 58·1% 81·9% 83·7% 37·7% 26·0% Season of baseline survey Spring 28 891 35·1% 63·0% 77·2% 32·9% 16·1% Summer 43 424 32·1% 66·7% 74·9% 35·8% 18·0% Autumn 28 206 33·1% 69·7% 80·0% 38·7% 21·5% Winter 35 590 35·7% 66·4% 74·8% 33·6% 16·7% Overall 136 111 33·5% 66·5% 75·8% 35·5% 17·9% * Hypertension is categorised as undiagnosed, diagnosed but not treated, treated but not controlled, or controlled (ie, blood pressure at baseline <140 mm Hg systolic and < 90 mm Hg diastolic).

28 206 33·1% 69·7% 80·0% 38·7% 21·5% Winter 35 590 35·7% 66·4% 74·8% 33·6% 16·7% Overall 136 111 33·5% 66·5% 75·8% 35·5% 17·9% * Hypertension is categorised as undiagnosed, diagnosed but not treated, treated but not controlled, or controlled (ie, blood pressure at baseline <140 mm Hg systolic and < 90 mm Hg diastolic). † Proportions are standardised for age, sex, and, where appropriate, area. The resurvey was done between July 14, 2006, and Oct 19, 2008. Of 27 983 participants invited to take part, 24 345 (87%) were interviewed. After exclusions, 23 114 were included in the analysis (appendix). The characteristics of people resurveyed were similar to those of the cohort as a whole (appendix). The prevalence of hypertension among participants aged 40–79 years (age-standardised to the baseline findings) was unchanged from baseline (39% at resurvey vs 38% at baseline), as were the proportions of those with hypertension that were diagnosed and being treated (appendix). By contrast, the proportion of participants with treated hypertension who had controlled blood pressure had increased from 36% in the first survey to 59% in the resurvey. This difference was seen despite only a slight increase in the proportion of participants receiving more than one blood-pressure-lowering medication (from 31% at baseline to 36%, appendix).

nts with treated hypertension who had controlled blood pressure had increased from 36% in the first survey to 59% in the resurvey. This difference was seen despite only a slight increase in the proportion of participants receiving more than one blood-pressure-lowering medication (from 31% at baseline to 36%, appendix). During 1·7 million person-years of follow-up (mean 17 [SD 4] years per person), 5707 cardiovascular deaths occurred in participants aged 35–79 years (figure 3). Uncontrolled hypertension at baseline was associated with RRs of 2·15 (95% CI 1·88–2·46), 1·86 (1·69–2·05) and 1·41 (1·32–1·52) at ages 35–59, 60–69, and 70–79 years, respectively. RRs were similar in men and women and were not changed substantially by further adjustment for education, smoking, alcohol intake, or body-mass index (appendix). The excess cardiovascular mortality associated with uncontrolled hypertension at baseline accounted for 20% (95% CI 18–22) of all cardiovascular deaths in the age group 35–69 years, and 13% (10–15) in the age group 70–79 years (figure 3). These population-attributable proportions would be equivalent to around 2000 excess cardiovascular deaths due to uncontrolled hypertension among people aged 35–69 years and around 1200 among those aged 70–79 years in Cuba in 2015.Figure 3 RRs for deaths in people with uncontrolled hypertension and cardiovascular mortality in Cuba for 2015

ributable proportions would be equivalent to around 2000 excess cardiovascular deaths due to uncontrolled hypertension among people aged 35–69 years and around 1200 among those aged 70–79 years in Cuba in 2015.Figure 3 RRs for deaths in people with uncontrolled hypertension and cardiovascular mortality in Cuba for 2015 (A) RRs for death are calculated for participants with versus those without uncontrolled hypertension, and are adjusted for area, level of education, and age within each age group. For each RR, the area of the square is inversely proportional to the variance of the log RR. (B) Cardiovascular mortality attributable to uncontrolled hypertension is calculated by applying PAFs to the estimated age-specific and sex-specific number of cardiovascular deaths in Cuba for 2015. PAF=population-attributable fraction. RR=rate ratio. Discussion In the Cuban population we assessed, a third of participants had hypertension. Two-thirds of those participants had a previous diagnosis, among whom three-quarters were receiving treatment. However, only about a third of treated patients had controlled blood pressure. The prevalence of hypertension was strongly related to age and was higher in men than women at younger ages but higher in women than men at older ages. Blood pressure was not well controlled in any subgroup, even in participants with previous cardiovascular disease. Uncontrolled hypertension was estimated to account for 20% of premature cardiovascular deaths in 2015 in this cohort.

as higher in men than women at younger ages but higher in women than men at older ages. Blood pressure was not well controlled in any subgroup, even in participants with previous cardiovascular disease. Uncontrolled hypertension was estimated to account for 20% of premature cardiovascular deaths in 2015 in this cohort. Although the study was not designed to be nationally representative, the sociodemographic characteristics of participants in the baseline survey (1996–2002) and the estimated prevalence of hypertension are consistent with a national survey of risk factors for chronic disease that was done in Cuba in 2001,13 which supports the generalisability of our findings. The national survey involved 23 000 participants (mean age 44 years) and found that 34% of adults had hypertension.13, 14 Another national survey done among adults in 2010 (mean age 46 years), found that prevalence of hypertension was largely unchanged (31%).13, 14 These data concur with our finding of similar hypertension prevalence at the time of the resurvey in 2006–08 (34%). The proportions of people with hypertension that was diagnosed, treated, and controlled were not reported in the national studies, and such results are only available from a few cross-sectional surveys in local areas,15, 16, 17 which cannot be generalised reliably to Cuba as a whole.

e of the resurvey in 2006–08 (34%). The proportions of people with hypertension that was diagnosed, treated, and controlled were not reported in the national studies, and such results are only available from a few cross-sectional surveys in local areas,15, 16, 17 which cannot be generalised reliably to Cuba as a whole. The age-specific prevalence values we report for hypertension are somewhat lower than those reported in surveys in many high-income countries, including those in western Europe and North America, even when more inclusive definitions of hypertension are used (appendix). Our values are, however, higher than in some other populations, such those in parts of east Asia.18 The proportion of participants in this study with hypertension that had been diagnosed and treated is broadly consistent with some high-income countries, despite the more limited resources of the Cuban health system.19 However, our finding that only 36% of treated participants with hypertension had controlled blood pressure was much lower than has generally been reported in high-income countries, such as 50% in the USA and 66% in Canada.19

th some high-income countries, despite the more limited resources of the Cuban health system.19 However, our finding that only 36% of treated participants with hypertension had controlled blood pressure was much lower than has generally been reported in high-income countries, such as 50% in the USA and 66% in Canada.19 Meta-analyses of randomised clinical trials of blood-pressure-lowering medication have shown that vascular risk is lowered by reducing SBP to 140 mm Hg and DBP to 90 mm Hg.6, 20 Some trials, however, that have involved selected populations, have found benefits with further reductions.21, 22, 23 Achieving controlled blood pressure in people with hypertension often requires several blood-pressure-lowering medications, but in our study most participants with treated hypertension at baseline and at the time of the resurvey were taking only one blood-pressure-lowering medication. The undertreatment of hypertension at baseline and resurvey was unexpected, given that Cuba's health system focuses on preventive medicine in primary care. This finding might in part reflect the availability of blood pressure-lowering medications. Shortages of common medications have occurred several times over the past few decades in Cuba, even in 2017,24 mainly during periods of economic hardship. Indeed, the baseline survey was done immediately following a time of economic crisis in Cuba that began in 1991.1

vailability of blood pressure-lowering medications. Shortages of common medications have occurred several times over the past few decades in Cuba, even in 2017,24 mainly during periods of economic hardship. Indeed, the baseline survey was done immediately following a time of economic crisis in Cuba that began in 1991.1 We found that uncontrolled hypertension was associated with roughly a doubling of the risk of premature death from cardiovascular disease. Meta-analyses of prospective studies, however, have shown direct and continuous associations between blood pressure and cardiovascular mortality, with no evidence of a threshold down to at least SBP 115 mm Hg.5 As such, the excess risk of cardiovascular death reflects the mean difference in the long-term average blood pressure between people with uncontrolled hypertension and those with controlled blood pressure (roughly 20 mm Hg systolic in the present study). Equivalent reductions in blood pressure within the hypertensive range would be expected to have the same effect on cardiovascular risk.

ference in the long-term average blood pressure between people with uncontrolled hypertension and those with controlled blood pressure (roughly 20 mm Hg systolic in the present study). Equivalent reductions in blood pressure within the hypertensive range would be expected to have the same effect on cardiovascular risk. The strengths of the present study include the very large sample size, the diverse areas surveyed, the range of population subgroups assessed, the stable population with low loss to follow-up, and the reliable linkage to cause-specific mortality. It is a limitation of the study, however, that our analyses could only describe associations with mortality because data were not available for non-fatal myocardial infarction and stroke events. Furthermore, hypertension was identified by measurement of blood pressure on one occasion at baseline. This is the standard approach in epidemiological surveys but differs from that for the diagnosis of hypertension in clinical practice, which is usually based on blood pressure recorded on at least two separate occasions. Additionally, any real improvements in blood pressure control during follow-up are likely to have attenuated the associations between uncontrolled hypertension at baseline and cardiovascular disease. Finally, the resurvey was done in only two municipalities. The initial plan had included more areas, but the number had to be reduced due to resource limitations.

od pressure control during follow-up are likely to have attenuated the associations between uncontrolled hypertension at baseline and cardiovascular disease. Finally, the resurvey was done in only two municipalities. The initial plan had included more areas, but the number had to be reduced due to resource limitations. Addressing the burden of hypertension in Cuba will require not only improved detection of hypertension in primary care (as in high-income countries), but also greater control of blood pressure among people known to have hypertension. The people with most to gain from lowering of blood pressure would be those at greatest absolute risk of hypertension (ie, those with previous cardiovascular disease or those with diabetes, who would also benefit from other treatments to address cardiovascular risk, such as statins, irrespective of blood pressure). The findings from our resurvey indicate that blood pressure control might have improved in some regions of Cuba, and an ongoing national survey will provide evidence on whether effects are similar at the national level, including for secondary prevention.

vascular risk, such as statins, irrespective of blood pressure). The findings from our resurvey indicate that blood pressure control might have improved in some regions of Cuba, and an ongoing national survey will provide evidence on whether effects are similar at the national level, including for secondary prevention. Several initiatives using community-wide awareness campaigns and simplified hypertension treatment regimens are ongoing in Cuba,25 and our findings should provide useful insights into addressing uncontrolled hypertension in primary care. Further research is also required to develop a cardiovascular risk score that is validated in the general population in Cuba to identify those at high absolute risk of cardiovascular events. Ideally, such efforts to improve the management of hypertension in primary care should be accompanied by public health programmes at the local level, national level, or both, to address the major determinants of hypertension, including harmful alcohol intake, excess adiposity, and high salt intake. The latter factor has been promoted by the Pan American Health Organization as an important public health measure to address raised blood pressure in the Americas.26

vel, national level, or both, to address the major determinants of hypertension, including harmful alcohol intake, excess adiposity, and high salt intake. The latter factor has been promoted by the Pan American Health Organization as an important public health measure to address raised blood pressure in the Americas.26 In this Cuban cohort study, a third of participants at baseline and resurvey had hypertension. Although the proportions of participants with diagnosed and treated hypertension were commensurate with those in some high-income countries, overall control of blood pressure was low. In addition to public health measures to reduce the prevalence of hypertension, improvement of blood-pressure control among people with diagnosed hypertension is required to prevent premature cardiovascular deaths in Cuba. This online publication has been corrected. The corrected version first appeared at thelancet.com/public-health on February 6, 2019 Supplementary Material Supplementary appendix Acknowledgments This study was funded by the UK Medical Research Council, British Heart Foundation and Cancer Research UK for the Clinical Trial Service Unit and Epidemiological Studies Unit (CTSU) and Population Health Research Unit (both now part of the Nuffield Department of Population Health, University of Oxford). BL is supported by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). We thank the study participants, fieldworkers, and physicians in Cuba who contributed to this study.

n Health Research Unit (both now part of the Nuffield Department of Population Health, University of Oxford). BL is supported by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). We thank the study participants, fieldworkers, and physicians in Cuba who contributed to this study. Contributors NAR, RP, and ADH designed the study. NAR and ADH directed the study and provided field supervision. NAR, PV-P, EL-V, MCM, SBC, JMMR, OJHL, MAMM, IAA, FAE, MDG, NRM, MCA, and ADH collected data. NAR, ED, BL, JAB, JE, SL, and ADH analysed, interpreted and reported the data. NAR, ED, BL, JAB, JE, SL, and ADH were responsible for writing and revising the manuscript. Declaration of interests JE received a grant from the Medical Research Council during the conduct of the study, and grants from the British Heart Foundation and Boehringer Ingelheim outside the study. The other authors declare no competing interests.