Medications for blood pressure, blood glucose, lipids, and anti-thrombotic medications: relationship with cardiovascular disease and death in adults from 21 high-, middle-, and low-income countries with an elevated body mass index 

Abstract

Introduction

Elevated body mass index (BMI) is an important cause of morbidity and mortality¹ and is likely to grow in significance globally as obesity increases in low- and middle-income countries.² An elevated BMI leads to increased blood pressure, dysglycaemia, and dyslipidaemia which are key determinants of atherosclerotic vascular disease. For this reason, current guidelines on the management of obesity recommend weight loss for individuals with a BMI ≥30 kg/m², or with a BMI 25–29.9 kg/m² and additional cardiovascular risk factors.³ Intensive management of cardiovascular risk factors is also recommended for individuals with an elevated BMI.³ However, there is limited direct evidence to inform the population-level benefits of pharmacologic approaches to cardiovascular risk factors to reduce adverse cardiovascular outcomes in these individuals. Blood pressure-lowering, lipid-lowering, anti-thrombotic, and some oral hypoglycemic medications have been proven in clinical trials of highly selected adults to reduce cardiovascular morbidity and mortality.^4–7 Whether these benefits translate to reduced cardiovascular disease (CVD) and death among adults with an elevated BMI globally has not been described. Therefore, the objective of this study was to evaluate whether the use of medications to treat high blood pressure, diabetes and dyslipidaemia, and anti-thrombotic drugs is associated with improved clinical outcomes in individuals with an elevated BMI.

Methods

Study participants

The present analysis includes 145 985 participants from 21 high-, middle-, and low-income countries in the Prospective Urban and Rural Epidemiology (PURE) study who had no history of cancer, ischaemic heart disease, heart failure, or stroke. The design of the PURE study has been described previously.⁸ In brief, the PURE study is a large, international, prospective cohort study of adults aged 35–70 years at enrolment. The choice and number of countries included in the PURE study were planned so that many communities in countries at different economic levels, with substantial heterogeneity in social and economic circumstances and policies, could participate. The feasibility of achieving long-term follow-up was a key factor in the choice of participating centres. Thus, PURE included sites in which investigators were committed to collecting good-quality data for a modest budget over the planned 10-year follow-up period without strict proportionate sampling. PURE includes countries in three income strata based on World Bank classification in 2006: the high-income countries were Canada, Saudi Arabia, Sweden, and United Arabic Emirates. The middle-income countries were Argentina, Brazil, Chile, China, Colombia, Iran, Malaysia, Palestine, Philippines, Poland, Turkey, and South Africa. The low-income countries were Bangladesh, India, Pakistan, Tanzania, and Zimbabwe.

Within each country, urban and rural communities were selected based on broad guidelines. In PURE, a community was defined as a group of people who have common characteristics and reside in a defined geographic area. Communities from low-, middle-, and high-income areas were selected from sections of a city defined according to a geographical measure (e.g. a set of contiguous postal code areas or a group of streets or a village). The primary sampling unit for rural areas in many countries was the village. Within each community, sampling was designed to achieve a sample of adults aged between 35 and 70 years that was generally representative of their community. The choice of sampling frame within each centre was based on both ‘representativeness’ and feasibility of long-term follow-up, following broad study guidelines. Once a community was identified, where possible, common and standardized approaches were applied to the enumeration of households, identification of individuals, recruitment procedures, and data collection. Households were eligible if at least one member of the household was between the ages of 35 and 70 years and the household members intended to continue living in their current home for a further 4 years. For each approach, at least three attempts at contact were made. All participants provided written informed consent. The proportion of people approached to participate in the study that provided consent was 71%.⁹

To ensure standardization and high data quality, we used a comprehensive operation manual, training workshops, videos, and regular communication with study personnel and standardized report forms. We entered all data in a customized database programmed with range and consistency checks which were transmitted electronically to the Population Health Research Institute in Hamilton (Ontario, Canada) where further quality checks were implemented. The research ethics committees in each of the participating countries approved the PURE study.

Study procedures

At baseline, trained study personnel administered a standardized set of questions to all consenting eligible individuals. The following were recorded: demographics, self-reported baseline diseases,¹⁰ education, physical activity levels, tobacco, alcohol, and medication use. Physical activity was measured using the International Physical Activity Questionnaire and classified as previously described¹¹ as low (<600 MET min/week), medium (600–3000 MET min/week), or high (>3000 MET min/week). Physical measurements were performed using standardized equipment. Blood pressure was measured with the participant sitting for 5 min using a standardized automated sphygmomanometer (OMRON Healthcare Inc, Toronto, Canada). Weight was measured using a digital scale (Tanita Corporation) with the participant lightly clothed with no shoes. Height was measured using a tape measure (H.A. KIDD) with the participant standing without shoes. Waist and hip circumferences were measured unclothed using a tape measure (H.A. KIDD). The waist circumference was considered the smallest circumference between the costal margin and the iliac crest. Blood was collected at baseline and lipids were measured. Cardiovascular medications included medications taken to lower blood pressure or cholesterol, medications taken for diabetes, angiotensin-converting enzyme-inhibitors, beta-blockers, calcium antagonists, diuretics, statins, aspirin, clopidogrel, warfarin, oral hypoglycaemic drugs, and insulin.

Participants were followed every 3 years, for a median of 10.2 years to document the clinical outcomes of myocardial infarction, stroke, heart failure, and death which were adjudicated using standardized definitions. CVD included the occurrence of myocardial infarction, stroke, heart failure, or cardiovascular death.

Statistical analysis

For CVD and death, shared frailty Cox proportional hazards models were constructed and hazard ratios were presented. In all models, the community within which the participant lived was included as a random effect to account for the correlation of characteristics among individuals from the same community. Models were adjusted for age, sex, region (South and East Asia, n = 45 915; China, n = 43 460; Middle East, n = 10 214; Africa, n = 5764; Latin American, n = 22 336; Europe/North America, n = 18 296), education, baseline physical activity, alcohol, and tobacco use. BMI was modelled in two ways: (i) as restricted cubic splines with four knots chosen according to the percentiles recommended by Harrell¹² and (ii) as categorical variables, with BMI classified as <20 kg/m², 20 to <25 kg/m², 25 to <30 kg/m², 30 to <35 kg/m², or ≥35 kg/m². To examine whether the relationship between BMI and clinical outcomes varied according to whether cardiovascular medications were used, cardiovascular medications were included in an interaction term with BMI.

We conducted three sensitivity analyses: (i) with BMI classified as <18.5 kg/m², 18.5 to <25 kg/m², 25 to <30 kg/m², 30 to <35 kg/m², 35 to <40 kg/m², or ≥40 kg/m²; (ii) in addition to excluding participants with known stroke, coronary heart disease, heart failure, or cancer, we also excluded those with prior tobacco use, chronic obstructive pulmonary disease, tuberculosis, malaria, Chagas’ disease, human immunodeficiency virus or dying within 2 years of enrolment; and (iii) with waist circumference (stratified by sex-specific quartile) as the measure of adiposity, rather than BMI.

To describe the population-level benefit of cardiovascular medication use among individuals with BMI ≥30 kg/m², we calculated its population attributable benefit by using logistic regression to estimate the differences in the odds of CVD and death in the observed cohort with the cohort under the assumption that no one was taking a cardiovascular medication. This approach is analogous to estimating the population attributable fraction according to the methods described by Greenland and Drescher¹³. Thus, the population attributable benefit = $ProbabilityCVD−ProbabilityCVD|noCVmedicationsusedProbabilityCVD$⁠. These models were adjusted for geographic region, age, sex, education, physical activity, tobacco, and alcohol use.

Results

Participants’ baseline characteristics stratified by cardiovascular medication use are presented in Table 1. Cardiovascular medications used included: aspirin (2.8%), anticoagulant (0.3%), oral hypoglycemic (3.3%), insulin (0.5%), a cholesterol-lowering drug (3.4%), and blood pressure-lowering drugs (14%). Individuals taking cardiovascular medications were older with a slight preponderance of women. Cardiovascular medication use was lowest in Tanzania, Pakistan, Bangladesh, Zimbabwe, India, and China; and highest in the United Arab Emirates, Brazil, Poland, and Saudi Arabia. There was a positive association between country income level and cardiovascular medication use. Former smokers and alcohol drinkers took cardiovascular medications more often than current or never smokers and alcohol drinkers. Individuals with higher BMI used cardiovascular medications more often than those with lower BMI. Despite taking cardiovascular medications, these individuals still had higher blood pressure, total and low-density lipoprotein cholesterol levels, and triglycerides than participants not taking cardiovascular medications.

Cardiovascular medications and the risk of CVD

During a median (25th–75th percentile) 10.2 (8.4–12.1) years, 2.8% of participants were lost to follow-up (6.3% in South and East Asia, 0.3% in China, 2.0% in the Middle East, 6.2% in Africa, 1.3% in Latin America, and 0.2% in Europe and North America). During follow-up, 7928 (5.4%) participants developed CVD. There was a positive association between BMI and the risk of developing CVD (Figure 1). We found a significant interaction between cardiovascular medication use and BMI for the risk of developing CVD (interaction P < 0.0001), indicating that at different levels of BMI, the risk of CVD among those taking cardiovascular medications differed from those not taking cardiovascular medications (Figure 2). Among those not taking a cardiovascular medication, the risk of CVD rose continuously as BMI increased. As compared with those with BMI 20 to <25 kg/m², the hazard ratio (HR) 95% confidence interval (95% CI) for CVD were, respectively, 1.14 (1.06–1.23); 1.45 (1.30–1.61); and 1.53 (1.28–1.82) among those with BMI 25 to <30 kg/m²; 30 to <35 kg/m²; and ≥35 kg/m². However, among those taking a cardiovascular medication, there was a U-shaped relationship between BMI and the risk of CVD, with the lowest risk observed at a BMI of 25 to <30 kg/m² [for whom the HR (95% CI) was 0.79 (0.72–0.87) as compared with those with a BMI 20 to <25 kg/m²]. Among those with BMI 30 to <35 kg/m² and ≥35 kg/m² taking a cardiovascular medication, respective HR (95% CI) were 0.90 (0.79–1.01) and 1.14 (0.98–1.33). Findings were similar when BMI was categorized as <18.5 kg/m², 18.5 to <25 kg/m², 25 to <30 kg/m², 30 to <35 kg/m², 35 to <40 kg/m², and ≥40 kg/m² (see Supplementary material online, Appendix Table S1). When analyses were performed on the 120 625 participants remaining after excluding those with current or prior tobacco use, known stroke, coronary heart disease, heart failure, cancer, chronic obstructive pulmonary disease, tuberculosis, malaria, Chagas’ disease, human immunodeficiency virus, or dying within 2 years of enrolment, findings were similar to the primary analysis (see Supplementary material online, Appendix Table S2 and Appendix Figure S1). We also repeated the analyses with the cardiovascular outcomes of myocardial infarction, stroke, and heart failure considered separately. The relationships between cardiovascular medication use, BMI, and both myocardial infarction and stroke were similar to CVD overall (interaction P < 0.0001; Supplementary material online, Appendix Table S3). However, there was no interaction between BMI and cardiovascular medication use for the outcome of heart failure (interaction P = 0.66).

Figure 1

Spline curve (solid line) and 95% CIs (dashed lines) representing the relationship between BMI and the HR for CVD (myocardial infarction, stroke, heart failure, or cardiovascular death). Estimates are adjusted for age, sex, education, physical activity levels, tobacco, and alcohol use. BMI 25 kg/m² is the reference. A histogram illustrating the distribution of BMI values in the study population is overlaid with frequency represented in the leftward vertical axis.

Open in new tabDownload slide

Figure 2

Spline curves (solid lines) and 95% CIs (dashed curves) representing the relationship between BMI and the HR for CVD (myocardial infarction, stroke, heart failure, or cardiovascular death) among participants on cardiovascular (CV) medications (in black) vs. those not taking CV medications (in red). Estimates are adjusted for age, sex, education, physical activity levels, tobacco, and alcohol use. BMI 25 kg/m² is the reference.

Open in new tabDownload slide

Our findings suggest that intervening in metabolic pathways that mediate the relationship between obesity and CVD may reduce the risk of future CVD events. We sought to examine this hypothesis further by generating time-to-CVD models that were sequentially adjusted for baseline cholesterol levels, systolic blood pressure, and blood glucose levels among participants not receiving cardiovascular medications. These models demonstrate attenuation of the relationship between BMI and CVD with the inclusion of these metabolic mediators of CVD, indicating that the association between an elevated BMI and CVD is in part accounted for by cholesterol levels, blood pressure, and blood glucose levels (Supplementary material online, Appendix Table S4).

Similar patterns were observed for waist circumference, whereby there was a continuous positive association between waist circumference and the risk of CVD but with significant differences in this association between those taking a cardiovascular medication and those not taking a cardiovascular medication (interaction P < 0.0001; Supplementary material online, Appendix Table S5). Among those not taking a cardiovascular medication, as compared with those whose waist circumference was less than the lowest quartile, the adjusted HR (95% CI) for CVD for each successive quartile of waist circumference was 1.25 (1.15–1.35); 1.39 (1.28–1.52); and 1.83 (1.66–2.02). The corresponding adjusted HR (95% CI) among those taking a cardiovascular medication were 0.96 (0.83–1.10); 0.90 (0.78–1.03); and 1.12 (0.97–1.28).

The use of aspirin for the primary prevention of CVD has led to mixed results in clinical trials, so the role of aspirin in this setting remains debated, especially among adults 40–70 years with cardiovascular risk factors¹⁴. Since obesity is an important cardiovascular risk factor, we evaluated whether the relationship between BMI and CVD varied according to aspirin use. We found a significant interaction between aspirin and BMI (P = 0.0038). Among the 97.2% of participants not taking aspirin, compared with BMI 20 to <25 kg/m², the adjusted HR (95% CI) for CVD in those with BMI <20 kg/m², 25 to <30 kg/m², 30 to <35 kg/m², and ≥35 kg/m² were, respectively, 0.81 (0.75–0.87), 1.09 (1.03–1.15), 1.40 (1.30–1.51), and 1.64 (1.47–1.83). Among those taking aspirin, the respective HR (95% CI) were 1.40 (0.73–2.70), 0.80 (0.61–1.04), 0.76 (0.55–1.05), and 0.97 (0.68–1.40).

Cardiovascular medications and the risk of death

Overall, 9863 (6.8%) participants died. The associations between BMI and mortality differed according to whether cardiovascular medications were used (interaction P = 0.0020). Compared with those with a BMI of 25 kg/m², there was a continuous increase in the risk of death among those with progressively higher BMI values if they were not taking cardiovascular medications (Figure 3). As compared with BMI 20 to <25 kg/m², the respective HR (95% CI) for death were 0.93 (0.87–1.00); 1.03 (0.93–1.15); and 1.44 (1.24–1.67) among those with BMI 25 to <30 kg/m²; 30 to <35 kg/m²; and ≥35 kg/m². However, among those taking a cardiovascular medication, as compared with a BMI of 20 to <25 kg/m², the respective HRs (95% CI) for death were 0.77 (0.69–0.84); 0.88 (0.78–0.99); and 1.12 (0.96–1.30) at BMI 25 to <30 kg/m²; 30 to <35 kg/m²; and ≥35 kg/m². When we excluded participants with current or prior tobacco use, known stroke, coronary heart disease, heart failure, cancer, chronic obstructive pulmonary disease, tuberculosis, malaria, Chagas’ disease, human immunodeficiency virus, or dying within 2 years of enrolment, findings were similar: among those not taking a cardiovascular medication, the adjusted HR (95% CI) in participants with BMI ≥35 kg/m² was 1.45 (1.18–1.78) when compared with those with BMI 20 to <25 kg/m² (see Supplementary material online, Appendix Table S2). In contrast, among those taking a cardiovascular medication, the adjusted HR (95% CI) in participants with BMI ≥35 kg/m² was 1.04 (0.85–1.28) when compared with those with BMI 20 to <25 kg/m².

Figure 3

(A) Spline curve (dashed line) demonstrating the relationship between BMI and the HR for death relative to a BMI 25 kg/m². The solid lines represent 95% confidence limits. Estimates are adjusted for age, sex, education, physical activity levels, tobacco, and alcohol use. A histogram illustrating the distribution of BMI values in the study population is overlaid with frequency represented in the leftward vertical axis. (B) Spline curves representing the relationship between BMI and the HR for death relative to a BMI 25 kg/m² among participants taking cardiovascular (CV) medications (dashed curve) vs. not taking a CV medication (solid curve).

Open in new tabDownload slide

We found a nominally significant interaction between waist circumference and cardiovascular medication use with respect to the risk of death (interaction P = 0.045). Among those not taking a cardiovascular medication, when compared with individuals whose waist circumference was less than the lowest quartile, the respective HR (95% CI) for death in successive quartiles were 0.91 (0.84–0.97); 0.93 (0.86–1.01); and 1.15 (1.05–1.25). Among those taking a cardiovascular medication, when compared with individuals whose waist circumference was less than the lowest quartile, the respective HR (95% CI) for death in successive quartiles were 0.89 (0.77–1.01); 0.78 (0.69–0.90); and 0.99 (0.87–1.13).

Subclasses of cardiovascular medication

To evaluate whether the association between cardiovascular medication use and favourable outcomes among those with an elevated BMI was driven by a specific subclass of medication, we repeated our analyses by considering blood pressure-lowering (n = 22 204), cholesterol-lowering (n = 5511), blood glucose-lowering (n = 8061), and anti-thrombotic (n = 4648) medications separately.

There were significant interactions between blood pressure-lowering medication use and BMI with respect to the risk of both CVD (interaction P < 0.0001) and death (interaction P = 0.040). When stratified by blood pressure-lowering medication use, the association between BMI and both CVD and death mirrored what we observed for cardiovascular medication as a whole, with lower relative hazards of these outcomes among those with an elevated BMI taking a blood pressure-lowering medication as compared with those not taking a blood pressure-lowering medication (Table 2).

The association between BMI and CVD differed according to whether blood glucose-lowering medications were used (interaction P = 0.0001). The relative hazard of CVD was lower among those with an elevated BMI taking a blood glucose-lowering medication as compared with those not taking a blood glucose-lowering medication. The association between BMI and CVD also differed according to whether anti-thrombotic medications were used (interaction P = 0.011). The relative hazard of CVD was lower among those with an elevated BMI taking an anti-thrombotic medication as compared with those not taking an anti-thrombotic medication. The association between BMI and death was not affected by the use of blood glucose-lowering medications (interaction P = 0.22) or anti-thrombotic medications (interaction P = 0.40). There was no interaction between cholesterol-lowering medication and BMI with respect to the risk of either CVD (interaction P = 0.39) or death (interaction P = 0.11), indicating that the relationships between BMI and both CVD and mortality were consistent irrespective of whether cholesterol-lowering medications were used.

Population attributable benefit of cardiovascular medications in those with BMI ≥30 kg/m²

For participants with BMI ≥30 kg/m², we estimated the population attributable benefit of those cardiovascular medications that had significant interactions with BMI for CVD or death. These estimates are presented in Figure 4 and indicate that among those with BMI ≥30 kg/m², taking a cardiovascular medication might account for a 23% (95% CI 20–25%) reduction in the odds of CVD and a 16% (95% CI 13–19%) reduction in the odds of death. Blood pressure-lowering medications accounted for most of the population attributable benefit of cardiovascular medications.

Figure 4

Population attributable benefit (PAB) from the use of any CV medication (med), blood pressure (BP)-lowering medication, blood glucose-lowering medication, and anti-thrombotic medications in the prevention of CVD and mortality in participants with BMI ≥30 kg/m². CI, confidence interval.

Open in new tabDownload slide

Discussion

Our major findings are that among adults without a history of CVD or cancer from countries at all income levels, there is a positive relationship between an elevated BMI and the risk of CVD and death among those not taking a cardiovascular medication; however, in those taking a cardiovascular medication, this association is attenuated, so that as compared with individuals with a BMI 20–25 kg/m², the risk of CVD and death was only elevated for those with BMI >35 kg/m². We found that this blunting of the risk of CVD and death observed among those taking a cardiovascular medication was mostly driven by the use of blood pressure-lowering medications.

Obesity and the risk of adverse outcomes

The association between an elevated BMI and both CVD and reduced life-expectancy is well-established and is corroborated by Mendelian randomization studies.¹⁵ There are several ways in which obesity might lead to pre-mature CVD. An elevated BMI is associated with increased blood pressure, diabetes, hypertriglyceridaemia, and inflammation, which are mediators of CVD. Mendelian randomization studies suggest that these metabolic mediators might account for an important part of the CVD risk attributable to obesity.¹⁶ Furthermore, treating these CVD risk factors has been demonstrated to lower the risk of recurrent CVD events among those with manifest CVD. There is also evidence to support the treatment of hypertension for the primary prevention of CVD, although there are fewer data to suggest that reducing blood glucose, triglycerides, or inflammation reduces CVD risk among those without prior CVD. Therefore, our findings add information to address this knowledge gap. We found that among individuals with no known CVD but with an elevated BMI, the use of medications to lower blood pressure or blood glucose, or anti-thrombotic medications was associated with a reduced risk of CVD as compared with individuals not taking these medications. These results lead us to hypothesize that the use of these medications in overweight or obese adults might represent a means to reduce CVD risk in this population. Our findings are consistent with evidence from the Emerging Risk Factors Collaboration that indicates that after accounting for systolic blood pressure, diabetes, and lipids, BMI may not be independently associated with risk for future cardiovascular events,¹⁷ suggesting that an elevated blood pressure, diabetes, and lipids may account for the excess risk conferred by an elevated BMI. Our research corroborates these data by indicating that if blood pressure and an elevated blood glucose are treated, the risk of future cardiovascular events is attenuated.

We found that cholesterol-lowering medications did not influence the relationship between BMI and CVD. This observation might be explained by the fact that statins are the most commonly used lipid-lowering medication. Statins pre-dominantly reduce low-density lipoprotein cholesterol with lower effects on triglycerides. Thus, because the lipid abnormalities related to obesity are most often elevated triglycerides and reduced high-density lipoprotein cholesterol, it is plausible that statins have a limited effect on the CVD risk conferred by obesity because they act on lipid pathways less closely linked with obesity.

Clinical implications

Although reducing obesity among those with an elevated BMI is an important means to reduce the accompanying risk of CVD, there are few strategies that have proven effective in the durable reduction of body fat. Therefore, adjunctive strategies to reduce CVD risk among those with an elevated BMI are valuable. To the extent that CVD risk among those with an elevated BMI is related to hypertension, diabetes, and an elevated thrombotic milieu, targeting these pathways pharmacologically may represent an important complementary means of reducing the CVD burden caused by an elevated BMI.

Strengths and limitations

The strengths of our study are that it was conducted in a large, multi-national cohort, including participants from countries at all income levels. To our knowledge, this represents the only study in which detailed information on cardiovascular medication use, participant characteristics, and clinical outcomes have been collected from diverse countries. The major limitation of our study is that we cannot exclude the possibility that residual confounding from unmeasured factors may account for the reduction in adverse cardiovascular outcomes seen among those taking a cardiovascular medication. We adjusted for education as an important marker of socioeconomic status—we have previously shown that education is among the most powerful predictors of mortality⁹. However, we are still unable to draw definitive causal inferences from the present observations. Our findings suggest that a pragmatic trial of cardiovascular pharmacotherapies in individuals with an elevated BMI could be helpful to establish whether this may be a useful strategy to lower their risk of adverse cardiovascular outcomes. A further limitation is that the relationship between BMI and mortality may be influenced by unmeasured confounders so that the higher risk of death in those with lower BMI might be driven by factors that lead to weight loss, rather than the weight loss itself.

Author contributions

D.P.L. designed and performed this analysis and drafted the manuscript. S.Y. is the global principal investigator of the PURE study and designed the overall study and revised the manuscript. All other authors contributed to data collection, study design, and critically revised the manuscript. All approved of the manuscript and agreed to be accountable for all aspects, ensuring integrity, and accuracy.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology.

Funding

A full list of study funding sources is provided in the Appendix.

Data availability

The data analysed in this study are not publicly available because we have not obtained participants' consent to share their data.

References

Author notes