Quantifying and understanding the higher risk of atherosclerotic cardiovascular disease among South Asians — results from the UK Biobank prospective cohort study

Abstract

Background:

Individuals of South Asian ancestry represent 23% of the global population – corresponding to 1.8 billion people – and suffer from substantially higher risk of atherosclerotic cardiovascular disease compared with most other ethnicities. U.S. practice guidelines now recognize South Asian ancestry as an important ‘risk-enhancing’ factor. The magnitude of enhanced risk within the context of contemporary clinical care, extent to which it is captured by existing risk estimators, and its potential mechanisms warrant additional study.

Methods:

Within the UK Biobank prospective cohort study, 8,124 middle-aged participants of South Asian ancestry and 449,349 participants of European ancestry who were free of atherosclerotic cardiovascular disease at time of enrollment were examined. The relationship of ancestry to risk of incident atherosclerotic cardiovascular disease – defined as myocardial infarction, coronary revascularization, or ischemic stroke – was assessed using Cox proportional hazards regression, along with examination of a broad range of clinical, anthropometric, and lifestyle mediators.

Results:

Mean age at study enrollment was 57 years and 202,405 (44%) were male. Over a median follow-up of 11 years, 554 of 8,124 (6.8%) individuals of South Asian ancestry experienced an atherosclerotic cardiovascular disease event, compared with 19,756 of 449,349 (4.4%) individuals of European ancestry, corresponding to an adjusted hazard ratio of 2.03 (95%CI 1.86–2.22; P<0.001). This higher relative risk was largely consistent across a range of age, sex, and clinical subgroups. Despite the >2-fold higher observed risk, the predicted 10-year risk of cardiovascular disease according to the AHA/ACC Pooled Cohort Equations and QRISK3 equations was nearly identical for South Asian and European ancestry individuals. Adjustment for a broad range of clinical, anthropometric, and lifestyle risk factors led to only modest attenuation of the observed hazard ratio to 1.45 (95%CI 1.28–1.65, P<0.001). Analysis of variance explained by eighteen candidate risk factors suggested higher importance of hypertension, diabetes, and central adiposity in South Asians.

Conclusions:

Within a large prospective study, South Asian individuals had substantially higher risk of atherosclerotic cardiovascular disease as compared with individuals of European ancestry, and this risk was not captured by the Pooled Cohort Equations.

INTRODUCTION:

Although atherosclerotic cardiovascular disease (ASCVD) is a leading cause of death across much of the world and all major ancestral groups, previous reports have suggested South Asians are at particularly high risk.^1–3 Given South Asian ancestry accounts for 1.8 billion individuals – 23% of the global population – and South Asian Americans comprise one of the fastest growing minority groups in the United States, additional study of this phenomenon may have important public health implications.^4,5 British South Asians make up the largest minority population in the United Kingdom and have provided insights into higher cardiovascular disease patterns in immigrant South Asian communities.⁶

Practice guidelines within North America and Europe have identified South Asians as a higher-risk ancestral group in recent years. In the United States, the American Heart Association/American College of Cardiology guidelines describe South Asian ancestry as a ‘risk-enhancing factor,’ and Canadian guidelines similarly recommend screening South Asians for dyslipidemias starting at a younger age.^7,8 In Europe, the European Society of Cardiology suggests practitioners multiply the calculated Systematic COronary Risk Evaluation (SCORE) risk estimate by 1.4 for South Asian individuals, and risk calculators such as QRISK3 – developed using large-scale electronic healthcare data – similarly incorporate variable multiplication factors ranging from 1.3 to 1.7 to increase the predicted risk for individuals of Indian, Pakistani, or Bangladeshi origin.^9,10

Prior studies in multiple countries have estimated a 1.7- to 4-fold higher risk of ASCVD among South Asian individuals compared with other ancestries but have important potential limitations.^11–15 First, case-control studies such as the seminal INTERHEART study that recruited cases with first myocardial infarction and matched controls may be subject to ascertainment biases and – based on recruitment between 1999 and 2003 – may not reflect recent advances in cardiovascular disease prevention.¹⁶ Second, most prior studies have been small, retrospective, or reliant on death registry data, known to imprecisely identify the cause of mortality.^17–23 Third, previous studies have largely pointed to differences in traditional risk factors – such as dyslipidemia or central adiposity – as the primary drivers of the higher prevalence of disease in South Asians, but lack of uniform risk factor assessment may have limited ability to fully account for the proportion of risk explained.^24–31

Based on these prior results, four key areas of uncertainty remain. First, the magnitude of the higher risk experienced by individuals of South Asian ancestry is uncertain, particularly in the context of contemporary clinical care that includes frequent use of preventive cholesterol- or blood pressure-lowering therapies. Second, whether any observed higher risk varies according to sex, ethnic subgroup, or immigration pattern has not been fully explored. Third, the extent to which any observed higher risk is attenuated by adjustment for biological factors, comorbid conditions, lifestyle habits, and psychosocial risk factors has not been fully explored within prospective studies. Fourth, the risk attributable to traditional and emerging risk factors – along with assessment of whether this varies in South Asian versus European ancestry individuals – warrants further study.

To address these areas of uncertainty within the context of contemporary clinical care, we analyzed a prospective cohort study inclusive of 8,124 South Asian ancestry and 449,349 European ancestry individuals free of baseline ASCVD with median follow-up of 11 years.

METHODS:

All data are made available from the UK Biobank to researchers from universities and other institutions with genuine research inquiries following institutional review board and UK Biobank approval.

Study population

The UK Biobank is a prospective cohort study that enrolled over 500,000 individuals between the ages of 40 and 69 years between 2006 and 2010.^32,33 Within this cohort, a subset of 481,542 participants self-reported South Asian or European ancestry. South Asian ancestry was defined as self-identification as being of Indian, Pakistani, Bangladeshi origin or reporting other South Asian origin with country of birth as Bhutan, Maldives, Nepal, or Sri Lanka. European ancestry was based on self-identification as being white British, white Irish, or any other white European background.

After additional exclusion of 24,069 individuals with ASCVD diagnosed prior to enrollment, 457,473 individuals were included in subsequent analyses. Exclusion of prevalent ASCVD at enrollment was based on self-report of myocardial infarction or ischemic stroke, hospitalization records confirming a diagnosis of acute myocardial infarction, ischemic stroke, or their acute complications, or a coronary revascularization procedure.

This research was conducted using the UK Biobank resource under Application Number 7089 and approved by the Mass General Brigham institutional review board.

Assessment of atherosclerotic cardiovascular disease risk factors

A detailed questionnaire completed by UK Biobank participants at enrollment assessed family history of heart disease, female reproductive history, smoking history, medication list, immigration history, dietary patterns, activity patterns, psychosocial stressors, and household income. Anthropometric measurements including waist and hip circumference were measured at the initial enrollment visit. Bioelectrical impedance measurements made by trained staff using hand and foot electrodes on the Tanita BC-418MA body composition analyzer (Tanita, Tokyo, Japan) were used to estimate body and trunk fat mass and percentage, as described elsewhere.^34,35 Biomarkers including serum lipid, glycated hemoglobin, cystatin-C, and C-reactive protein concentrations were assessed at time of enrollment as part of the study protocol. Further details are available in the Supplemental Methods.

Additional clinical and risk-enhancing factors were defined based on recent primary cardiovascular disease prevention guidelines.³⁶ Diagnosis of hypertension was determined based on self-report, primary care records, or hospitalization records confirming a clinical diagnosis, self-reported consumption of antihypertensive medications, average systolic blood pressure measurement above 140 mm Hg, or average diastolic blood pressure measurement above 90 mm Hg at enrollment. Diagnosis of diabetes was determined based on self-report, primary care records or hospitalization records confirming a clinical diagnosis, self-reported consumption of medications to treat diabetes, or glycated hemoglobin above 6.5% at enrollment. Individuals with a diagnosis of diabetes with glycated hemoglobin greater than 7% were classified as having uncontrolled diabetes. Chronic kidney disease was defined as glomerular filtration rate less than 60 mL/min/1.73m², as estimated by the CKD-EPI cystatin-C equation.³⁷ Presence of chronic inflammatory diseases was defined as self-report, primary care records or hospitalization records confirming a clinical diagnosis of rheumatoid arthritis, psoriasis, lupus erythematosus, and human immunodeficiency virus infection. Female reproductive factors enhancing cardiovascular disease risk were defined as a history of menopause prior to age 40, preterm delivery, fetus with intrauterine growth retardation, gestational hypertension, pre-eclampsia, eclampsia, gestational diabetes, or polycystic ovary syndrome. Obesity was defined per World Health Organization recommendations as a body mass index ≥ 27.5 kg/m² in South Asians and ≥ 30 kg/m² in European ancestry individuals.³⁸ Central adiposity was defined as waist-hip ratio greater than 0.9 in men and greater than 0.85 in women.³⁹

High LDL-cholesterol was defined as directly measured concentration above 160 mg/dL (~4.1 mmol/L).³⁶ Low HDL-cholesterol was defined as concentration below 40 mg/dL (~1.0 mmol/L) in men and below 50 mg/dL (~1.3 mmol/L) in women. High triglycerides were defined as concentration above 150 mg/dL (~1.7 mmol/L) for individuals fasting more than 8 hours or above 200 mg/dL (~2.3 mmol/L) for non-fasting individuals. High lipoprotein(a) was defined as concentration above 50 mg/dL (~107.5 nmol/L). High C-reactive protein was defined as concentration above 3 mg/L.

An unhealthy dietary pattern was ascertained based on a diet score of three or less, computed by assigning one point for adherence to each of the following seven categories: eating at least three pieces of fruit per day, at least nine heaping table-spoons of vegetables per day, at least three servings of whole grains per day, at least two servings of fish per week, no more than 1.5 servings of refined grains and starches a day, no more than one serving of processed meat per week, and no more than 2.5 servings of red meat per week, as previously described.^40,41 Sedentary lifestyle was defined as more than a combined total eight hours a day of: watching television, sitting in front of a computer, or driving, as previously described.⁴²

Family history referred to heart disease of any type reported in a first-degree relative at any age. Psychosocial stressors were defined as experiencing any of the following in the two years prior to enrollment: serious illness, injury or assault to self or a relative, death of a close relative or partner, marital separation or divorce, or financial difficulties. Low socioeconomic status was defined as average total pre-tax household income of less than £18,000 or Townsend Deprivation Index—a composite measure of unemployment, household crowding and lack of car or home ownership—in the top decile of the study population (≥3.41).⁴³ Additional phenotype variable descriptions are available in the Supplemental Methods. Missing data totals for the main risk factor variables are included in Table I in the Supplement.

Clinical endpoint

The primary endpoint was atherosclerotic cardiovascular disease (ASCVD). Incidence of ASCVD was defined based on hospitalization records indicating a diagnosis of acute myocardial infarction, ischemic stroke, or their acute complications, coronary revascularization procedures (coronary artery bypass graft surgery or percutaneous angioplasty/stent placement), or death register indicating myocardial infarction or ischemic stroke as a cause of death.⁴⁴ Nonfatal outcomes were ascertained based on hospitalization and procedural records while fatal outcomes were ascertained from death registry data. Additional details are provided in the Supplemental Methods.

Statistical analysis

All statistical analyses were performed with the use of R software, version 3.5 (R Project for Statistical Computing). Comparison of baseline characteristics between individuals of South Asian versus European ancestry was performed with the chi-squared test for categorical variables, analysis of variance (ANOVA) for continuous variables, and Mann-Whitney U test for continuous variables with nonparametric distributions. Standardized mean differences were estimated using the ‘tableone’ package in R, and a difference of less than 0.1 was taken to indicate a negligible difference in the mean or prevalence of a variable between ancestry groups.^45,46 Unadjusted cumulative incidence of ASCVD was determined for each ancestry through summation of events over follow-up time. The ACC/AHA Pooled Cohort Equations and the QRISK3 Equations were used to predict the unadjusted 10-year risk estimates for subgroup of individuals of South Asian and European ancestry without prior ASCVD noted in primary care, hospitalization, or procedural records or statin use.^9,47,48 Risk for incident ASCVD for South Asian relative to European ancestry was calculated using Cox proportional hazards regression models, including covariates of age, sex, enrollment center, and self-reported ancestry. ASCVD incidence rates were estimated as events per 1000 person-years of follow up time and adjusted for age and sex using Poisson regression and the ‘epiR’ package in R.^49,50

Risk for development of incident ASCVD after enrollment associated with a given risk factor category was computed using Cox proportional hazards regression models for each ancestry, including covariates of age, sex, enrollment center, and risk factor of interest. Population attributable fractions for each risk factor were also calculated and compared among South Asians and Europeans using categorized risk factors using the ‘epiR’ package in R.⁵⁰ Proportion of disease variance explained in each ancestry was calculated for each risk factor group using the Nagelkerke’s pseudo-R² metric, as previously described.⁵¹ The change in R² was calculated for the model inclusive of the risk factor group of interest plus the covariates of age, sex, and enrollment center minus R² for model with these covariates alone, thus yielding an estimate of the explained variance attributed to the risk factor group. Bootstrapping was performed 200 times for the estimation of the 95% confidence intervals for each of these estimates. Prior diagnosis of hypertension, systolic blood pressures, and diastolic blood pressures were grouped as hypertension measures. Prior diagnosis of diabetes and glycated hemoglobin concentrations were grouped as diabetes measures. Body mass index and body fat percentage values derived from bioelectrical impedance measurements were grouped as obesity measures. Waist-hip ratio and trunk fat percentage values derived from bioelectrical impedance measurements were grouped as central adiposity measures. Townsend deprivation index values and average household income were grouped as socioeconomic measures. Triglyceride, lipoprotein(a), and C-reactive protein concentrations were log-transformed prior to inclusion in statistical models. Additional details are provided in the Supplemental Methods.

The extent to which the higher risk for incident ASCVD in South Asian versus European ancestry individuals was associated with traditional and emerging risk factors was assessed using a baseline Cox proportional hazard models with age, sex, and enrollment center as covariates, followed by sequential addition of additional risk factor groups of interest into the model. For these Cox proportional hazards regressions models, individuals missing data for any covariate in the model of interest were excluded from that analysis. Additional mediation analysis was performed to evaluate the proportional association of each risk factor within the association between ancestry and ASCVD risk using the ‘mediation’ package in R.⁵² Further details are provided in the Supplemental Methods.

RESULTS:

In the UK Biobank, 457,473 participants with median follow-up of 11.1 years (IQR 10.4–11.8) were studied. This study population included 449,349 participants of European ancestry and 8,124 participants of South Asian ancestry. Mean age at enrollment was 57 years, and 44% were male.

At the time of enrollment, several differences in prevalence of cardiovascular risk factors between individuals of South Asian and European ancestry were present (TABLE 1). Individuals of South Asian ancestry were younger than their European counterparts, with mean age 53.5 years versus 57.0 years, respectively. A larger proportion of South Asians relative to Europeans were male (51.6% versus 44.1%), had history of diabetes (19.5% versus 5.3%), chronic kidney disease (7.7% vs. 4.2%), sedentary lifestyle (16.8% vs. 12.2%), psychosocial stressors (51.7% vs. 43.6%), and income less than £18,000 (31.4% versus 21.4%). Relative to individuals of European ancestry, on average South Asians had higher waist-hip ratios (a marker of central adiposity, 0.90 versus 0.87, Figure I in the Supplement), lower low-density lipoprotein cholesterol (130.9 mg/dL versus 139.3 mg/dL), lower high-density lipoprotein cholesterol (49.0 mg/dL versus 56.5 mg/dL), higher triglycerides (median 148.7 mg/dL versus 131.2 mg/dL), and higher Townsend Deprivation Indices (0.2 vs. −1.5). At enrollment, a significantly higher proportion of South Asians reported prescription of blood pressure medications (24.0% vs. 19.4%), diabetes medications (12.8% vs. 3.2%), and statin medications (20.9% vs. 13.1%) when compared with European ancestry individuals. Among South Asian ethnic subgroups, individuals of Bangladeshi origin had the highest prevalence of diabetes (29.9%) while individuals of Indian origin (17.6%) had the lowest prevalence. Bangladeshi participants also had the lowest HDL cholesterol concentrations (43.7 mg/dL) while Indian participants had the highest (50.2 mg/dL). Similarly, Bangladeshi origin individuals had on average the highest Townsend indices (+2. 9) while Indian origin individuals had the lowest (−0.2). (Table II in the Supplement)

TABLE 1:

Baseline characteristics of individuals of European versus South Asian ancestry

	European (N=449,349)	South Asian (N=8,124)	p-value	SMD
Age at Enrollment (years, mean (SD))	57.0 (8.0)	53.5 (8.4)	<0.001	0.42
Male Sex (%)	198,210 (44.1)	4,195 (51.6)	<0.001	0.15
PCE 10yr Risk (%, median [IQR])	6.0 [2.6, 12.0]	4.8 [2.1, 10.6]	<0.001	0.07
QRISK3 10 yr Risk (%, median [IQR])	8.3 [4.0, 14.7]	9.7 [4.5, 18.9]	<0.001	0.27
Hypertension (%)	174,139 (38.8)	3,397 (41.8)	<0.001	0.06
Taking Blood Pressure Medication (%)	87208 (19.4)	1948 (24.0)	<0.001	0.11
Diabetes (%)	23,733 (5.3)	1,586 (19.5)	<0.001	0.44
Taking Diabetes Medication (%)	14312 (3.2)	1037 (12.8)	<0.001	0.36
Chronic Kidney Disease (%)	17,689 (4.2)	579 (7.7)	<0.001	0.15
Chronic Inflammatory Diseases (%)	16,420 (3.7)	313 (3.9)	0.357	0.01
Female Reproductive Factors (%)	25,096 (5.6)	428 (5.3)	0.227	0.01
Body Mass Index (kg/m2, mean (SD))	27.3 (4.8)	27.2 (4.4)	0.003	0.04
Waist-hip Ratio (mean (SD))	8.68 (0.89)	9.00 (0.85)	<0.001	0.37
LDL Cholesterol (mg/dL, mean (SD))	139.3 (33.0)	130.9 (31.9)	<0.001	0.26
HDL Cholesterol (mg/dL, mean (SD))	56.5 (14.7)	49.0 (12.4)	<0.001	0.55
Triglycerides (mg/dL, median [IQR])	131.2 [92.7, 189.6]	148.7 [104.9, 214.2]	<0.001	0.21
Lipoprotein(a) (nmol/L, median [IQR])	18.5 [7.4, 70.8]	31.2 [12.0, 69.6]	<0.001	0.004
C-reactive Protein (mg/L, median [IQR])	1.3 [0.7, 2.7]	1.6 [0.8, 3.3]	<0.001	0.07
Taking Statin (%)	58842 (13.1)	1694 (20.9)	<0.001	0.21
Family History of Heart Disease (%)	196,462 (43.7)	3,357 (41.3)	<0.001	0.05
Current Smoking (%)	46,373 (10.4)	706 (8.8)	<0.001	0.05
Unhealthy Diet (%)	136,506 (30.4)	2,260 (27.8)	<0.001	0.06
Sedentary Lifestyle (%)	53,307 (12.2)	1,209 (16.8)	<0.001	0.13
Psychosocial Stressors (%)	193,873 (43.6)	4,013 (51.7)	<0.001	0.16
Townsend Deprivation Index (mean (SD))	−1.5 (3.0)	0.2 (3.1)	<0.001	0.56
Household Income (%)		<0.001	0.24
Less than £18,000	82,984 (21.4)	1,909 (31.4)
£18,000 to £30,999	98,379 (25.4)	1,458 (24.0)
£31,000 to £51,999	103,210 (26.6)	1,251 (20.6)
£52,000 to £100,000	81,368 (21.0)	1,097 (18.0)
Greater than £100,000	21,628 (5.6)	366 (6.0)

Summary of baseline characteristics of individuals studied in the UK Biobank. PCE: American Heart Association/American College of Cardiology Pooled Cohort Equations. LDL: Low-density lipoprotein. HDL: High-density lipoprotein. Chronic kidney disease: glomerular filtration rate less than 60 mL/min/1.73m². Female reproductive factors: history of menopause before age 40, preterm delivery, fetus with intrauterine growth retardation, gestational hypertension, pre-eclampsia, eclampsia, gestational diabetes, or polycystic ovary syndrome. Chronic inflammatory diseases: rheumatoid arthritis, psoriasis, lupus erythematosus, or human immunodeficiency virus infection.

Over a median follow-up of 11.1 years, corresponding to 5.0 million person-years of follow-up time, 554 of 8,124 (6.8%) individuals of South Asian ancestry experienced an ASCVD event, compared with 19,756 of 449,349 (4.4%) individuals of European ancestry, corresponding to an adjusted hazard ratio of 2.03 (95%CI 1.86–2.22, P<0.001) as illustrated in FIGURE 1A. Despite this higher observed risk, there was no significant difference in predicted 10-year risk of incident cardiovascular disease for individuals of South Asian and European ancestry, with median risks predicted by AHA/ACC Pooled Cohort Equations (PCE) of 4.8% versus 6.0% respectively, respectively, and 9.7% and 8.3% for the QRISK3 score respectively (FIGURE 1B, 1C). To adjust for baseline differences in age and sex between ancestries, risk estimates were standardized for age and sex to the overall study means, yielding 10-year risk estimates for individuals of South Asian and European ancestry of 8.3% and 7.5% using the PCE (1.1-fold higher for South Asians) and 13.7% and 9.6% using QRISK3 (1.4-fold higher for South Asians), respectively.

FIGURE 1:

Observed and predicted incidence of atherosclerotic cardiovascular disease in individuals of South Asian versus European ancestry

A: Unadjusted cumulative incidence of initial atherosclerotic cardiovascular disease (ASCVD) events over length of follow-up stratified by South Asian or European ancestry. B: Distribution of unadjusted 10-year predicted risk of ASCVD by the AHA/ACC Pooled Cohort Equations, stratified by individuals of South Asian and European ancestry. C: Distribution of unadjusted 10-year predicted risk of ASCVD by the QRISK3 equations, stratified by individuals of South Asian and European ancestry.

Secondary analyses using alternate endpoints – restricted to coronary artery disease, myocardial infarction (fatal and nonfatal), coronary revascularization, and ischemic stroke (fatal and nonfatal) – demonstrated no significant difference in results. (FIGURE 2). Beyond ASCVD, a higher risk of heart failure (HR 1.83, 95%CI 1.63–2.05, P<0.001) and a lower risk of atrial fibrillation or atrial flutter (HR0.76, 95%CI 0.67–0.86, P<0.001) were observed in South Asians relative to European ancestry individuals. (Figure II in the Supplement)

FIGURE 2:

Adjusted hazard ratios of atherosclerotic cardiovascular disease and component endpoints for South Asians relative to individuals of European ancestry

Adjusted hazard ratios with corresponding 95% confidence intervals and p values for composite atherosclerotic cardiovascular disease endpoint (coronary artery disease and ischemic stroke), composite coronary artery disease endpoint (myocardial infarction and coronary revascularization procedure) and individual component endpoints stratified by nonfatal and fatal outcomes, comparing individuals of South Asian ancestry to individuals of European ancestry, calculated using Cox proportional hazards regression models with covariates of enrollment age, sex, and testing center. Adjusted incidence rates estimated as events per 1000 person-years (PY) of follow up time and adjusted for age and sex using Poisson regression.

The observed higher risk for ASCVD in South Asian relative to European ancestry individuals was largely consistent across subgroups divided by sex, age, and whether born in the United Kingdom (FIGURE 3). There was no evidence of statistical interaction of ancestry with sex in assessing ASCVD risk, and analyses stratified by sex yielded no significant difference in results for women (HR 1.92, 95% CI 1.62–2.28, P<0.001) and men (HR 2.06, 95% CI 1.87–2.28, P<0.001, p-value for sex interaction 0.12). Among South Asians, individuals of Bangladeshi (HR 3.66, 95%CI 2.38–5.62, P<0.001) origin experienced the highest proportional hazard versus individuals of European ancestry, followed by individuals of Pakistani origin (HR 2.45, 95%CI 2.06–2.91, P<0.001), other South Asian origin (HR 2.41, 95%CI 1.86–3.14, P<0.001), and Indian origin (HR 2.08, 95%CI 1.85–2.33, P<0.001) (FIGURE 4).

FIGURE 3:

Adjusted hazard ratios of atherosclerotic cardiovascular disease for South Asians relative to individuals of European ancestry stratified by demographic subgroups

Adjusted hazard ratios with corresponding 95% confidence intervals and p values for composite cardiovascular disease endpoint, comparing individuals of South Asian ancestry to individuals of European ancestry, divided by subgroups of sex, and age (in years), and place of birth, calculated using Cox proportional hazards regression models with covariates of enrollment age, sex, and testing center. Adjusted incidence rates estimated as events per 1000 person-years (PY) of follow up time and adjusted for age and sex using Poisson regression.

FIGURE 4:

Adjusted hazard ratios of coronary artery disease for South Asian ethnic subgroups relative to individuals of European ancestry

Adjusted hazard ratios with corresponding 95% confidence intervals and p values for composite atherosclerotic cardiovascular disease endpoint, comparing individuals of different South Asian ancestry sub-groups to individuals of European ancestry, calculated using Cox proportional hazards regression models with covariates of enrollment age, sex, and testing center. Other South Asian denotes individuals identifying country of origin as Bhutan, Maldives, Nepal, or Sri Lanka. Adjusted incidence rates estimated as events per 1000 person-years (PY) of follow up time and adjusted for age and sex using Poisson regression.

Next, no significant differences were observed in relative risks associated with most clinical and lifestyle factors in South Asians versus European ancestry individuals (FIGURE 5, Figures III–VII in the Supplement). For example, low socioeconomic status was associated with a hazard ratio of 1.4 in both South Asian and European ancestry individuals, respectively. Although there was no difference in the association of the diagnosis of hypertension with ASCVD—hazard ratio of 1.95 versus 1.75 in South Asian and European individuals, respectively—every 20 mm Hg rise in systolic blood pressure measured at enrollment associated with higher ASCVD risk in individuals of South Asian ancestry (HR 1.33, 95%CI 1.21–1.45) relative to individuals of European ancestry (HR 1.12, 95%CI 1.11–1.13), p-heterogeneity <0.001 (Figure III in the Supplement). Potential heterogeneity in relative risk estimates was also noted for obesity, which was associated with a lower hazard ratio of 1.20 in South Asian individuals versus 1.45 in European individuals (p-value for heterogeneity 0.035) as well as smoking which was associated with a lower hazard ratio of 1.21 in South Asian individuals versus 1.85 in European individuals (p-value for heterogeneity 0.002). However, this heterogeneity was largely eliminated after additionally accounting for body mass index modeled as a continuous variable (p-value for heterogeneity 0.52) and lower numbers of cigarettes consumed per day in South Asian smokers as compared with European ancestry individuals (p-value for heterogeneity 0.53), respectively. (Figures IV and VI in the Supplement).

FIGURE 5:

Adjusted hazard ratios for atherosclerotic cardiovascular disease, per variable category of interest

Adjusted hazard ratios with corresponding 95% confidence intervals and p values for atherosclerotic cardiovascular disease risk imparted by variable of interest, stratified by ancestry, calculated using Cox proportional hazards regression models with covariates of enrollment age, sex, and testing center. LDL: Low-density lipoprotein. HDL: High-density lipoprotein. Chronic kidney disease: glomerular filtration rate less than 60 mL/min/1.73m². Female reproductive factors: history of menopause before age 40, preterm delivery, fetus with intrauterine growth retardation, gestational hypertension, pre-eclampsia, eclampsia, gestational diabetes, or polycystic ovary syndrome. Chronic inflammatory diseases: rheumatoid arthritis, psoriasis, lupus erythematosus, or human immunodeficiency virus infection. P-heterogeneity compares hazard ratios between ancestry groups and is deemed significant if <0.05.

Next, there were no significant differences observed in the proportions of variance explained by each risk factor group in between individuals of South Asian and European ancestry outside of hypertension, diabetes, and central adiposity – all of which explained a significantly larger proportion of ASCVD variance in South Asians (FIGURE 6, Table III in the Supplement). Relative to individuals of European ancestry, South Asians had significantly lower amount of ASCVD variance explained by current smoking. In line with findings from the analysis of variance explained, there were no significant differences in population attributable fractions (PAF) – determined by both the prevalence and associated relative risk – with most risk factors in between individuals of South Asian and European ancestry. However, diabetes (0.22, 95%CI 0.17–0.27) and central adiposity (0.49, 95%CI 0.40–0.58) accounted for a significantly higher fraction of ASCVD cases for South Asians (Figure VIII in the Supplement). Sex-stratified PAF estimates reflected these findings (Table IV in the Supplement).

FIGURE 6:

Proportion of variance in atherosclerotic cardiovascular disease risk explained by risk factor group

Proportion of variance explained was calculated for each disease using the Nagelkerke’s pseudo-R2 metric. The R2 was calculated for the full model inclusive of the risk factor group of interest plus the baseline covariates (age at enrollment, sex, and enrollment center) minus R2 for the baseline covariates alone, thus yielding an estimate of the explained proportion of variance attributable to each risk factor. Hypertension measures: Prior diagnosis of hypertension, systolic blood pressure, and diastolic blood pressure. Diabetes measures: Prior diagnosis of diabetes and glycated hemoglobin concentration. Obesity measures: body mass index and impedance-measured body fat percentage. Central adiposity measures: waist-hip ratio and impedance-measured trunk fat percentage. LDL: Low-density lipoprotein. HDL: High-density lipoprotein. Chronic kidney disease: glomerular filtration rate less than 60 mL/min/1.73m². Female reproductive factors: history of menopause before age 40, preterm delivery, fetus with intrauterine growth retardation, gestational hypertension, pre-eclampsia, eclampsia, gestational diabetes, or polycystic ovary syndrome. Chronic inflammatory diseases: rheumatoid arthritis, psoriasis, lupus erythematosus, or human immunodeficiency virus infection. Socioeconomic measures: Average household income and Townsend Deprivation Index.

Additionally, the observed higher risk of ASCVD in South Asians was mitigated by inclusion of various risk factors as covariates in the Cox proportional hazard models. Although each risk factor when added alone as a covariate to the Cox regression model did not significantly alter the risk estimate, when all of the risk factors were sequentially added to the model, the collective hazard ratio estimates decreased to a nadir of 1.45 (95%CI 1.28–1.65) (FIGURE 7). Similar trends were observed in a subset of 4,210 South Asians and 216,165 European ancestry individuals in whom longitudinal data on control of LDL cholesterol, glycated hemoglobin, and blood pressure were available and included in sensitivity analyses.(Table V in the Supplement) Additional mediation analyses suggest that a large proportion of the association of South Asian ancestry and ASCVD was associated with variation in glomerular filtration rate (10.0%, 95%CI 7.3%−14.7%), HDL cholesterol (10%, 95% CI 7.2%−15%), hemoglobin A1c (9.4%, 95% CI 6.7% - 14.3%), lipoprotein(a) (8.9%, 95%CI 6.5%−13.6%), and household income (4.9%, 95%CI 3.1%−8.2%). However, negative mediation associations were estimated for smoking (−8.6%, 95% −6.2% - −14.4%) and LDL cholesterol (−8.2%, 95% CI −5.6% - −13.6%).(Table VI in the Supplement)

FIGURE 7:

Adjusted hazard ratios for individuals of South Asian ancestry compared with European ancestry, with effects of additive covariates

Adjusted hazard ratios with corresponding 95% confidence intervals and p values for composite cardiovascular disease endpoint, comparing individuals of South Asian ancestry to individuals of European ancestry, calculated using Cox proportional hazards regression models with covariates of enrollment age, sex, and testing center, and iterative addition of additional covariates as ordered. Hypertension measures: Prior diagnosis of hypertension, systolic blood pressure, and diastolic blood pressure. Diabetes measures: Prior diagnosis of diabetes and glycated hemoglobin concentration. Obesity measures: body mass index and impedance-measured body fat percentage. Central adiposity measures: waist-hip ratio and impedance-measured trunk fat percentage. LDL: Low-density lipoprotein. HDL: High-density lipoprotein. Female reproductive factors: history of menopause before age 40, preterm delivery, fetus with intrauterine growth retardation, gestational hypertension, pre-eclampsia, eclampsia, gestational diabetes, or polycystic ovary syndrome. Chronic inflammatory diseases: rheumatoid arthritis, psoriasis, lupus erythematosus, or human immunodeficiency virus infection. Socioeconomic measures: Average household income and Townsend Deprivation Index.

DISCUSSION:

In a large prospective cohort study, individuals of South Asian ancestry were found to experience more than two-fold higher risk of incident atherosclerotic cardiovascular disease when compared with individuals of European ancestry. This risk was not captured by the Pooled Cohort Equations and was only modestly attenuated in mediation analyses that additionally adjusted for traditional and emerging risk factors.

These results build on previous efforts to understand the nature of ASCVD risk in South Asians in several key ways. First, the risk for incident ASCVD events was examined in a large number of individuals with a median follow-up of 11.1 years, enabling analyses of important demographic and clinical subgroups and well-powered estimations of risk. Second, the influence of a breadth of lifestyle and risk factors was uniformly assessed using detailed baseline survey, anthropometric, and biomarker data with events ascertained within national health care systems. Third, these findings confirm and extend prior reports that hypertension, diabetes, and central adiposity are the leading associations in this observed disparity.^53,54 Fourth, residual unexplained risk remained after accounting for a comprehensive list of risk factors, suggesting the potential for variable frequency of healthcare engagement, delayed diagnosis or undertreatment of comorbidities, genetic variation, or as-yet unidentified or unmeasured factors that may play a key role.

The higher ASCVD risk for South Asians was not predicted by the AHA/ACC Pooled Cohort Equations, likely because the populations used to derive these models lacked individuals of South Asian ancestry.⁴⁷ The current QRISK3 calculator accounts for South Asian subgroup ancestry using a multiplicative factor variable ranging from 1.3 to 1.7 based on country of origin.⁹ The findings of residual risk in this study after accounting for the list of risk factors lends further credence to such an approach, but suggests and that even these correction factors may not be adequate to fully capture the higher risk. The underperformance of the QRISK3 score may in part be due to a more limited cardiovascular disease endpoint being considered in this study, however further study is warranted.

Consistent with prior studies, diabetes appears to play an outsized role in driving risk among South Asians.^53–55 There was a significant difference in the population attributable fraction between South Asian and European ancestries for diabetes, explaining 23% of risk among South Asians versus 8% in Europeans. A recent study in the U.S. suggested high prevalence of diabetes in South Asians, and that this often occurs even in the absence of obesity.⁵⁶ This study’s findings confirm this observation, noting a lower average body-mass index of 28.7 kg/m² in South Asians with diabetes versus 31.5 kg/m² in European individuals. Moreover, South Asians with diabetes were less likely to have achieved adequate glycemic control. Further investigation into potential novel metabolite biomarkers and imaging phenotypes may prove useful.

In addition to South Asians having a higher overall burden of comorbidities, social determinants of health may be contributing to the observed excess risk of ASCVD, particularly among subgroups of South Asians. The findings in this study confirm and extend earlier findings of significantly higher risk factor burden and lower socioeconomic metrics among Bangladeshi and Pakistani individuals as well as higher ASCVD incidence in these groups.^30,57 Prior studies have suggested that numerous other factors may be contributing to this disparity, including education, systemic discrimination, perception of biomedicine, and other cultural factors in South Asian migrant populations.^8,58,59 Developing South Asian-specific patient education tools and interventions that bridge language and cultural gaps, as is being carried out in the South Asian Heart Lifestyle Intervention (SAHELI, NCT03336255) study, may shed light on methods of overcoming cultural barriers in addressing this disparity.^60,61

Despite accounting for these imbalances in prevalence and influence of traditional and emerging risk factors between individuals of South Asian and European ancestry, there remains a residual unexplained risk, raising the possibility of genetic etiologies. Studies to date have suggested that the genetic predictors of ASCVD are largely similar across South Asian versus European ancestry but have remained underpowered to systematically characterize.⁶² We and others have recently started to assess the relationship of a polygenic predictor of ASCVD in South Asian individuals and establish larger reference populations that may identify common or rare genetic factors distinct to South Asian individuals.^63–66

These results should be interpreted within the context of potential limitations. First, participant volunteers were recruited at sites across the United Kingdom at age 40–69 years, raising the possibility of selection bias that limits generalizability to younger patients or individuals in other countries. Second, individuals who volunteered for the UK Biobank are known to be more healthy when compared with the general population, which likely led to reduced event incidence in both ancestries.⁶⁷ Third, disease endpoints were ascertained through participant self-report, diagnosis codes from inpatient admissions, and national procedure and death registries rather than by manual review of medical records, and therefore may be subject to mis- or under-ascertainment, especially in the context of limited access to medical care that might prevent inclusion of a diagnosis in the medical record.

CONCLUSION:

In a large prospective study of South Asians with comprehensive examination of risk factors, substantially higher risk of atherosclerotic cardiovascular disease among South Asian individuals relative to European ancestry individuals was observed. These findings confirm and extend recent guideline recommendations to consider South Asian ancestry as a ‘risk-enhancer’ not captured by the Pooled Cohort Equations and support ongoing efforts to identify mechanisms underlying this higher risk.

CLINICAL PERSPECTIVE:

What is new?

• Data from a large prospective cohort study were used to study the relationship between South Asian ancestry and incident atherosclerotic cardiovascular disease (ASCVD) events within the context of contemporary medical care.

• Our findings confirm an approximate doubling of ASCVD risk among South Asians compared with Europeans that is not captured by the Pooled Cohort Equations.

• This higher risk of ASCVD persisted despite adjustment for a broad range of potential clinical, anthropometric, and lifestyle mediators.

• Hypertension, diabetes and central adiposity explained a greater proportion of risk for ASCVD in South Asians as compared with Europeans.

What are the clinical implications?

• These results confirm and extend current guidelines to consider South Asian ancestry a ‘risk-enhancing’ factor in assessing future risk for atherosclerotic cardiovascular disease.

• Residual risk that persisted after accounting for a range of potential mediators may relate to differences in social determinants of health, unmeasured risk factors, and genetics and warrants further investigation.

• Whether or not targeted interventions attenuate the outsized impact of diabetes or central adiposity among South Asian individuals warrants further attention.

NON-STANDARD ABBREVIATIONS AND ACRONYMS:

UK

United Kingdom

ASCVD

Atherosclerotic cardiovascular disease

SCORE

Systematic coronary risk evaluation

LDL

Low-density lipoprotein

HDL

High-density lipoprotein

ANOVA

Analysis of variance

PCE

Pooled cohort equations

PAF

Population attributable fraction

PY

Person-year