Introduction

Although cardiovascular disease (CVD) disease burden is declining in high-income countries, it is rising in low-income and middle-income countries (LMICs) and is the leading cause of death worldwide, accounting for an estimated 17.8 million deaths and an age-standardised death rate of 233 per 100,000 in 2017 [1]. CVD occurs at younger ages in LMICs than in high-income countries and exhibits strong socio-economic gradients, both in terms of disease burden and economic hardship in managing it [2]. Coordinated action with respect to CVD is especially important for several reasons: (1) the rising health and economic burden of CVD is placing considerable strain on the individuals affected by CVD, their families, and health systems more broadly; (2) the leading risk factors for CVD can be diagnosed and treated at relatively low cost compared with the cost of treating CVD events such as hospitalisation for myocardial infarction and stroke; and (3) better information on how CVD risk varies globally could inform health system planning and targeting of global and national CVD programmes and help accelerate progress in achieving Sustainable Development Goal 3.4 to reduce premature mortality from non-communicable diseases (NCDs) by one-third by 2030 [3–5].

The 2010 World Health Organization Package of Essential Noncommunicable Disease Interventions (WHO PEN) for primary healthcare in LMICs identifies a set of interventions for strengthening primary healthcare to tackle NCDs through the use of low-cost medicines, tools, and technologies [6]. WHO PEN is the most commonly used technical guidance for NCDs in LMICs and has been implemented with varying degrees of success in several countries [7–10]. Other more recent technical guidance includes the 2016 Global Hearts Initiative, launched by WHO and the US Centers for Disease Control and Prevention, which comprises 5 technical packages of evidence-based interventions for the prevention and management of CVDs in primary healthcare [11]. Despite the availability of guidance, however, in LMICs implementation of routine national surveillance and cost-effective interventions to address CVD and its risk factors remains challenging [12,13].

About 60% of all cardiovascular deaths will occur in asymptomatic people who have not had a previous cardiovascular event [14]. The challenges of identifying people at high risk are extensive, especially in settings with low-skilled health workforces and inadequate access to healthcare services. Over the past decade, there has been a shift in CVD prevention from assessing single risk factor abnormalities to management based on a person’s future risk of experiencing a cardiovascular event. This approach has been demonstrated to be superior to assessment of single risk factors when identifying who will benefit the most from treatment and is endorsed in WHO PEN and the Global Hearts Initiative [6,11,14].

In this paper, we analyse individual-level data from 45 nationally representative population-based surveys. Taking primarily a country-level health system perspective, we aim to (1) estimate CVD risk profiles, (2) determine management patterns in each country based on blood pressure (BP) medicine use as recommended by WHO PEN guidelines, and (3) determine whether wealth and educational status are associated with variation in CVD risk and management practices.

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist). A study outline, including proposed figures and tables, was developed prior to conducting the analyses. No changes were made to these analyses; however, an initial pre-planned analysis on BP control rates was excluded due to low numbers with available data.

Results

Sample characteristics

Complete case analysis was conducted for 600,484 people from 45 LMICs (Table 1). At least 1 of the variables needed to ascertain CVD risk was missing for 4.1% (n = 25,752) of the people in the overall sample, with this proportion varying by country (S1 Fig). These individuals were excluded from the analysis. Aside from female participants having a higher probability of being excluded due to missing data than male participants, the differences in participant characteristics between those included in the analysis and those excluded due to missing data were small (S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Weighted distribution of cardiovascular disease risk factors in participants aged 30–74 years from population-based surveys conducted in 45 low- and middle-income countries between 2005 and 2017.

https://doi.org/10.1371/journal.pmed.1003485.t001

The median (IQR) age for the total sample was 44.7 years (43.0–46.9), median (IQR) female proportion was 51.7% (49.5%–55.3%), median (IQR) proportion of current smokers was 18.9% (11.5%–25.3%), and median (IQR) BMI was 25.7 kg/m² (23.3–26.8). The proportion of individuals taking BP medication varied widely across countries—from 0.6% in Vanuatu to 25.5% in Belarus (median = 8.6%; IQR = 4.1%–14.2%). The median (IQR) SBP among people not taking and taking BP medication was 124.3 mm Hg (123.0–127.4) and 144.8 mm Hg (139.8–149.9), respectively. For the latter group, the pre-treatment median (IQR) SBP was 153.8 mm Hg (148.3–159.6), using the Wald and Law adjustments.

CVD risk distribution

The distribution of CVD risk, age-standardised using the GBD project’s 2017 global population, among people aged 30 to 64 years for the 45 countries is shown in Fig 1. The median (IQR) 10-year CVD risk overall was 2.7% (2.3%–4.2%) for males and 1.6% (1.3%–2.1%) for females, with wide variation across countries and regions. In a sensitivity analysis, risk was calculated excluding those under 40 years of age, and median (IQR) risk increased to 4.0% (3.4%–5.9%) for males and to 2.8% (2.2%–3.4%) for females (S2 Fig). Risk estimates tended to be lower in sub-Saharan Africa compared with countries in other world regions and were highest in Europe and the Eastern Mediterranean. The median proportion (IQR) of individuals overall at very low (<5%), low (5% to <10%), medium (10% to <20%), high (20% to <30%), and very high risk (≥30%) was 79.0% (59.1%–83.4%), 16.6% (12.4%–25.1%), 5.3% (2.6%–11.2%), 0.6% (0.2%–1.6%), and 0.1% (0.0%–0.6%), respectively, for males, and 88.5% (82.1%–92.5%), 9.7% (6.5%–14.2%), 1.8% (0.8%–4.6%), 0.1% (0.0%–0.6%), and 0.0% (0.0%–0.1%), respectively, for females.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CVD risk profile by country for men and women aged 30–64 years.The numbers in the green section of the stacked bars represent the median 10-year CVD risk along with the interquartile range for the country.

All estimates are age-standardised using the Global Burden of Disease project’s 2017 global population. *The age range included in the sample varied in these 5 countries: India—30–49 years for women and 30–54 years for men; Ecuador—30–59 years for both men and women; Namibia—35–64 years for both men and women; Lesotho—30–49 years for women and 30–59 years for men; Albania—30–49 years for both men and women. CVD, cardiovascular disease; EEM, Europe and the Eastern Mediterranean; StVG, Saint Vincent and the Grenadines.

https://doi.org/10.1371/journal.pmed.1003485.g001

Fig 2 examines the variation in 10-year CVD risk by education, wealth, marital status, and employment status from the multivariable regression. Having higher levels of education, working in the previous 12 months, and being married were consistently associated with lower CVD risk in most countries. There was a more mixed picture for household wealth, particularly with some sub-Saharan African countries showing higher median CVD risk for the wealthier quintiles. The magnitude of socio-demographic associations was broadly similar for females (S3 Fig) and males (S4 Fig), and there was a mixed contribution of individual risk factors driving these gradients (S3 Table) across all regions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Relative change (%) in 10-year CVD risk by educational level, household wealth, marital status, and employment status, from multivariable regression.

These are estimates from linear mixed models with the primary sampling units as the clusters. The outcome is logarithm of CVD risk, and the predictors are sex, educational level (primary school or higher education versus no schooling) household wealth quintile (middle/richer/richest versus poorer/poorest), marital status (married/cohabiting versus never married/separated/divorced/widowed), and employment status in the last 12 months (working versus not working). The estimates for Grenada and Morocco are based on linear regression as there are no primary sampling units and a single participant was sampled from each household. The countries with estimates not plotted either had a predictor missing or there were fewer than 5 participants in a category for 1 or more predictors. CVD, cardiovascular disease; EEM, Europe and the Eastern Mediterranean; StVG, Saint Vincent and the Grenadines.

https://doi.org/10.1371/journal.pmed.1003485.g002

Use of BP medication by CVD risk category

The median (IQR) percentage of people overall indicated for BP medication who were taking a BP medication was 24.2% (15.4%–37.2%) for males and 41.6% (23.9%–53.8%) for females. There was, however, a large variation in this pattern by country (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Percentage of people indicated for BP medication who were taking medication, by sex.

The numbers on the right for each country and sex present univariate risk ratio of taking BP medication for an individual indicated for medication per World Health Organization/International Society of Hypertension guidelines, compared to an individual not indicated for medication. Indication for use of BP medication was defined as the presence of any of the following: an extreme blood pressure elevation (systolic BP ≥ 160 mm Hg or diastolic BP ≥ 100 mm Hg), a 10-year CVD risk ≥ 30%, or a 10-year CVD risk of 20%–29% and elevated blood pressure (systolic BP ≥ 140 mm Hg or diastolic BP ≥ 90 mm Hg). The risk ratio for men in Vanuatu was not calculated as among men indicated for medication, fewer than 5 men were on medication. BP, blood pressure; CVD, cardiovascular disease; EEM, Europe and the Eastern Mediterranean; StVG, Saint Vincent and the Grenadines.

https://doi.org/10.1371/journal.pmed.1003485.g003

When compared with females at low CVD risk who were not indicated for BP medication, the univariable RR (95% CI) of receiving BP medication ranged from 2.1 (1.6–2.9) in Costa Rica to 15.5 (9.8–24.6) in Albania among females. For males, the corresponding RR (95% CI) ranged from 2.8 (1.8–4.4) in Costa Rica to 16.5 (6.6–41.0) in Burkina Faso. Few countries had an association between BP medication use and socio-demographic characteristics regardless of whether BP medication use was indicated (S5 Fig) or not (S6 Fig).

Although a greater proportion of people at elevated CVD risk are taking BP medication compared with those at low CVD risk, almost half of survey participants who were taking BP medication overall were individuals at low CVD risk in most countries (Fig 4). Of all people taking BP medication, the median (IQR) proportion who were at low CVD risk and not indicated for BP medication was 47.1% (36.1%–58.6%), with minimal difference between males (45.1% [32.9%–57.3%]) and females (48.3% [37.2%–59.4%]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Proportion of people taking blood pressure medication who were indicated/not indicated for medication based on guideline recommendations.

EEM, Europe and the Eastern Mediterranean; StVG, Saint Vincent and the Grenadines.

https://doi.org/10.1371/journal.pmed.1003485.g004

Discussion

In this study of 600,484 adults from 45 LMICs, we estimate the CVD risk profile using the recently published WHO region-specific, recalibrated risk prediction equations. Overall, we observed large variations in risk profile; an inverse association between CVD risk and higher education and employment in most countries; an overuse of medicines in people at lower levels of CVD risk; and an underuse of medicines in people at elevated CVD risk across all countries.

The large variation in risk profiles across and within countries is due to considerable variation in the presence of CVD risk factors. Countries with the highest median CVD risk tended to be middle-income countries. Females tended to have lower median CVD risk than males in almost all countries, which, in part, reflects the different coefficients used in sex-specific equations. These findings are consistent with those from the WHO CVD Risk Chart Working Group using the same equations for 79 countries (the majority of which are LMICs) [16] and those from Ueda et al. who used a different risk prediction model (Globorisk) in 10 high-, middle- and, low-income countries [21].

An important new finding from this study is the inverse socio-demographic gradients in CVD risk profiles, which appear to be driven by moderate elevations of multiple risk factors in all world regions. Although CVD risk levels tended to be higher overall in middle-income countries among the 45 countries studied, this study highlights the need to depart from narrowly conceptualising CVD as a disease of affluence and to apply an equity lens to implementing CVD risk programmes both within and across countries. Given that males tend to underutilise health services and people with lower education, lower household wealth, and unemployment may experience access barriers to high-quality care, intensified efforts to increase coverage and quality of care for these populations is warranted.

The other key study finding was that around one-half of all BP medication (47.1%) is being used by people at lower levels of CVD risk, and only 24.2% of males and 41.6% of females at high risk were taking guideline-recommended medication. Given that a BP treatment strategy based on predicted CVD risk is more effective than one based on BP levels alone [14,22–24], the pattern of relative overuse in people at low CVD risk and undertreatment of people at high CVD risk (especially for men) represents inefficient use of scarce resources for many countries. An important explanation for this pattern is the likely persistence of single-risk-factor-based treatment over risk-based treatment. Many countries have local or regional hypertension guidelines, and these guidelines, many of which have been in operation for decades, generally recommend that if patients with hypertension do not respond to lifestyle interventions (diet, weight loss, exercise) within 3–6 months, they should be treated with an antihypertensive agent. Given the superiority of risk-based management, there needs to be harmonisation of conflicting guidelines and a shift toward targeting treatment to individuals who will gain the greatest benefit in terms of CVD events avoided. Factors influencing appropriate use of medicines are complex and go beyond improved risk stratification. Health systems barriers related to unreliable supply chains, healthcare providers (e.g., an insufficiently trained and supported workforce), and patients (e.g., financial barriers and other factors influencing non-adherence) all need to be considered [25,26].

There are a number of caveats to the study findings. First, the year of survey varied for the countries included in the study, and given rapid demographic transitions in many countries, the findings are most relevant to the year in which the country survey was conducted. Second, in some countries, most notably India, the upper age range limit of the sample was less than 65 years. Although we age-standardised each country’s sample to the world population, the lower age range in some surveys will underestimate risk, given increasing age is the strongest predictor of elevated risk. Third, the majority of country surveys did not ascertain if participants had a prior CVD event, and given that the risk of a subsequent CVD event is considerably higher in this group, this will have underestimated the true risk profile. Fourth, although the proportion of missing data overall was small (4.1%) and the country-level median (IQR) proportion with missing outcome data across the 45 countries was 3.6% (1.8%–6.1%), there is potential for bias due to missing data, with this potential being greater in countries with higher proportions of missing data. Fifth, use of the non-laboratory-based equations may mean that some people, particularly those with diabetes, who were assessed as being at low risk would be reclassified to be at higher risk if laboratory values were available. This could lead to an overestimate in the proportion found to be taking BP medication without a clear indication, and, conversely, for those not on BP medication, it could lead to an underestimate of the treatment gap. Despite this important limitation, in many countries laboratory cholesterol values are not routinely ordered, and assessment of risk is dependent on non-laboratory-based equations. Also, given that STEPS and other surveys generally do not routinely collect laboratory measures, non-laboratory-based estimates are generally the only means by which population-representative risk estimates can be obtained.

If current trends continue, most LMICs are unlikely to achieve Sustainable Development Goal 3.4, to reduce premature mortality from NCDs by one-third by 2030 [27]. The case for investing in CVD prevention is therefore greater than ever. The WHO estimated that a US$120 billion investment in 20 high-CVD-burden countries (amounting to $1.50 per capita over 15 years) would avert 15 million deaths and 21 million incident ischaemic heart disease or stroke incidents in these countries [4]. This investment includes funding population-level strategies to lower CVD risk and strategies to improve healthcare performance. Although undoubtedly increased investment is needed, this study has shown that considerable improvements in health system performance could be achieved by better harnessing existing resources to appropriately use guidelines for managing CVD. Systematic CVD risk screening programmes from around 35–45 years of age have been shown to be more cost-effective than universal screening [7,28]. Although most countries report the availability of the basic requirements for non-laboratory-based CVD risk assessments in primary healthcare facilities, there remain major barriers to implementing a risk-based approach [29]. Implementation programmes to improve identification of people at high CVD risk in the community, intensify efforts to target subgroups that may be at elevated CVD risk, and shift treatments away from people at lower levels of risk towards these higher CVD risk groups may have the greatest potential to generate benefit both at the individual and health system levels [30].

The large heterogeneity of the findings in this study reflects varying country contexts. Factors such as subnational wealth distribution, overall health system expenditure, health system capability, and the shifting dynamics of CVD risk population profiles over time vary substantially across countries. Hence a one-size-fits-all approach is most likely not appropriate, and it would be necessary for international guidelines such as WHO PEN to incorporate country-specific modifications to ensure context specificity [7].

Although most country and international guidelines recommend CVD risk estimation as a critical first step in determining treatment decisions, this approach appears to be loosely adhered to in most countries in this study. Improved adoption of risk-based guidelines requires a substantial change in management approach at multiple levels in the health system. This includes targeted policies that are responsive to each country’s context, engagement with professional bodies for workforce training and education, improved data systems and decision support that can be readily used by varying cadres of health workers, quality improvement processes in health facilities to enhance care effectiveness, and strategies to increase access and adherence to recommended medicines.

Supporting information

Acknowledgments

We would like to thank each of the country-level survey teams and study participants who made this analysis possible.