Global Burden of Atherosclerotic Cardiovascular Disease in People Living With HIV: Systematic Review and Meta-Analysis

Background: With advances in antiretroviral therapy, most deaths in people with HIV are now attributable to noncommunicable illnesses, especially cardiovascular disease. We determine the association between HIV and cardiovascular disease, and estimate the national, regional, and global burden of cardiovascular disease attributable to HIV. Methods: We conducted a systematic review across 5 databases from inception to August 2016 for longitudinal studies of cardiovascular disease in HIV infection. A random-effects meta-analysis across 80 studies was used to derive the pooled rate and risk of cardiovascular disease in people living with HIV. We then estimated the temporal changes in the population-attributable fraction and disability-adjusted life-years (DALYs) from HIV-associated cardiovascular disease from 1990 to 2015 at a regional and global level. National cardiovascular DALYs associated with HIV for 2015 were derived for 154 of the 193 United Nations member states. The main outcome measure was the pooled estimate of the rate and risk of cardiovascular disease in people living with HIV and the national, regional, and global estimates of DALYs from cardiovascular disease associated with HIV. Results: In 793 635 people living with HIV and a total follow-up of 3.5 million person-years, the crude rate of cardiovascular disease was 61.8 (95% CI, 45.8–83.4) per 10 000 person-years. In comparison with individuals without HIV, the risk ratio for cardiovascular disease was 2.16 (95% CI, 1.68–2.77). Over the past 26 years, the global population–attributable fraction from cardiovascular disease attributable to HIV increased from 0.36% (95% CI, 0.21%–0.56%) to 0.92% (95% CI, 0.55%–1.41%), and DALYs increased from 0.74 (95% CI, 0.44–1.16) to 2.57 (95% CI, 1.53–3.92) million. There was marked regional variation with most DALYs lost in sub-Saharan Africa (0.87 million, 95% CI, 0.43–1.70) and the Asia Pacific (0.39 million, 95% CI, 0.23–0.62) regions. The highest population-attributable fraction and burden were observed in Swaziland, Botswana, and Lesotho. Conclusions: People living with HIV are twice as likely to develop cardiovascular disease. The global burden of HIV-associated cardiovascular disease has tripled over the past 2 decades and is now responsible for 2.6 million DALYs per annum with the greatest impact in sub-Saharan Africa and the Asia Pacific regions. Clinical Trial Registration: URL: https://www.crd.york.ac.uk/prospero. Unique identifier: CRD42016048257.

C urrently, >35 million people are infected with HIV, with two-thirds being resident in sub-Saharan Africa. 1 Although the global incidence for HIV has stabilized, the provision and widespread distribution of combined antiretroviral therapy 2 has dramatically improved survival with the prevalence of HIV steadily increasing over the past 2 decades. 3 This improvement in survival has been primarily attributed to a reduction in opportunistic infections, especially in the low-and middle-income nations. 1,4 Indeed, most deaths now arise from noncommunicable illnesses, especially cardiovascular disease. [5][6][7] Cardiovascular disease is the leading cause of morbidity and mortality worldwide. 8,9 The past 2 decades have seen a substantial increase in the morbidity attributable to cardiovascular disease, with a significant proportion of the burden borne by low-and middleincome nations. 10,11 The highest prevalence rates of HIV have been observed in sub-Saharan Africa. This region has also seen a steady increase in the burden of cardiovascular disease over the past 2 decades. 2,12 Recent studies have shown a link between the development of cardiovascular disease and HIV infection with multiple potential mechanisms, including direct vascular inflammation, 13,14 dyslipidemia, 15 and insulin resistance. 16,17 The aim of this systematic analysis was to review and to meta-analyze the rate of cardiovascular disease in people living with HIV, to determine the association between HIV infection and the risk of cardiovascular disease, and to estimate the national, regional, and global burden of HIV-associated cardiovascular disease.

METHODS
Data and the corresponding R analysis code will be available at https://github.com/anoopsshah/hiv_cvd.

Databases, Sources, and Searches
We searched MEDLINE, EMBASE, Global Health, Cumulative Index to Nursing and Allied Health Literature, and Web of Science by using the following key words: myocardial infarction, stroke, cerebrovascular disease, cardiovascular disease, and HIV (Text I in the online-only Data Supplement). Bibliographic reference lists of studies selected for inclusion in our meta-analysis and relevant review articles were manually searched ( Figure 1). We limited our search to studies published between 1948 and August 30, 2016.

Selection of Articles, Extraction of Data, and Data Synthesis
All longitudinal studies, including case-control, cohort, and randomized controlled trials, were included. There were no language restrictions, and only peer-reviewed original articles were included. Many studies provided data on the same cohort at extended follow-up time points. In such cases, we selected the published study with the longest follow-up period. Data were extracted independently, and any discrepancies were adjudicated by 4 investigators (A.S.V.S., D.S., K.K.L., and S.A.). We contacted authors for additional data or clarification where required. The study methodology, results, and presentation were conducted as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines ( Table I in

Crude Rate
The pooled incident crude rates of cardiovascular disease studies were only included if they provided the number of cardiovascular events (fatal or nonfatal myocardial infarction or stroke) and the follow-up period. Crude incidence rates were pooled per 10 000 person-years and presented for all incident cardiovascular events and deaths. The analysis was substratified by etiology classified as either myocardial infarction or cerebrovascular disease, where applicable.

Risk Ratio
For estimating the pooled risk ratios of cardiovascular disease, cases were defined as any hospitalization with,

Clinical Perspective
What Is New?
• Recent studies have identified plausible biological mechanisms, including endothelial dysfunction and arterial inflammation, to explain the association between HIV infection and atherosclerotic disease. • This article represents a systematic analysis to evaluate the association between HIV and cardiovascular disease and estimate the burden of HIV-associated cardiovascular disease at a national, regional, and global level. • We report that the risk of cardiovascular disease is increased 2-fold in people living with HIV, and the global burden of HIV-associated cardiovascular disease has tripled over the past 2 decades with the majority of the burden experienced in sub-Saharan Africa and the Asia Pacific region.
What Are the Clinical Implications?
• The combined burden of HIV and cardiovascular disease, especially in high-prevalence regions, has important implications with respect to regional health policies, guidelines, and resource allocation. • Risk stratification and identification of patients at risk of future cardiovascular disease are already challenging in these regions. • Whether patients living with HIV should be considered high-risk and appropriate primary prevention pharmacotherapy such as statin therapy should be implemented remains unclear. • Our estimates have important policy implications for implementing appropriate cardiovascular risk stratification and treatment strategies across healthcare systems, especially in low-and middleincome nations where both HIV and cardiovascular disease remain highly prevalent.

ORIGINAL RESEARCH ARTICLE
or mortality from, cardiovascular events (fatal or nonfatal myocardial infarction or stroke) in individuals with and without HIV. Studies were also included if the outcome was defined as a cardiovascular event and included but was not exclusive to myocardial infarction or stroke. Risk ratio estimates for disease incidence and case fatality for cardiovascular events in HIV-infected populations in comparison with populations not infected by HIV were pooled and presented with 95% CIs. Pooled risk ratio estimates were assessed for publication bias by visually assessing funnel plots and for asymmetry by using the Egger regression test. 19 The trim-and-fill method was used to adjust for selection bias attributable to potential unpublished studies or bias attributable to small-study effects. 20 Sensitivity analysis was performed by removing studies where adjustment for important confounders was omitted ( Table II in the online-only Data Supplement). Across both the pooled rate and risk ratios, subgroup analysis was only performed where ≥3 estimates were available. Studies providing estimates for the risk ratio were further stratified by type of event and publication year.

Regional and Global Estimates for HIV Prevalence and Cardiovascular Disability-Adjusted Life-Years
Global and regional estimates of cardiovascular disease attributable to HIV were derived annually from 1990 to 2015.

Cardiovascular Disability-Adjusted Life-Years
National disability-adjusted life-year (DALY) estimates for cardiovascular disease (consisting of the sum of DALYs attributable to ischemic heart disease and stroke) were available from the Institute of Health Metrics and Evaluation. Countries were grouped into Joint United Nations Programme on HIV/AIDS (UNAIDS) regions (Table III in the online-only Data Supplement). The DALYs for cardiovascular disease in each country were summed to derive the regional cardiovascular burden. These data were available at 5-year intervals, from 1990 to 2015, with intervening years obtained via linear interpolation.

Regional Prevalence of HIV
Prevalence estimates for HIV were available from 1990 to 2015, for the population from 15 to 49 years of age, from UNAIDS at a global and regional level.

Population-Attributable Fraction and HIV-Attributable Cardiovascular DALYs
At the global and regional levels, we applied the HIV prevalence estimates for the 15-to 49-year age group across the entire adult population. As such, the calculated populationattributable fraction was applied to the entire adult population when calculating the cardiovascular DALYs attributable to HIV at global and regional levels.

National Estimates for HIV Prevalence and Cardiovascular DALYs
National estimates of HIV prevalence and cardiovascular burden were available from 160 and 179 countries, respectively, of the 193 United Nations member states, and both were available from 154 countries. National HIV prevalence estimates were available for the >15-years age group for 2016, and national DALY estimates for cardiovascular disease were available for 2015.
We also calculated estimates of burden in the nations with a high HIV prevalence by combining data for the UNAIDSdefined 21 Global Plan priority countries (Table III in

Statistical Analysis
Using the pooled risk ratio for cardiovascular disease in people living with HIV and the prevalence of HIV, we estimated the population-attributable risk fraction at national, regional, and global levels. At regional and global levels, we further estimated the yearly changes in attributable risk to take into account temporal changes in the population prevalence of HIV. The population-attributable fraction for HIV for cardiovascular disease was calculated as described previously: 23,24 Population Attributable Fraction We anticipated heterogeneity between studies when estimating both the crude incidence rates and the risk ratio because of different study designs, methods of analysis and varying adjustment, and geographical and population differences. We therefore used a random-effects model with the maximum likelihood estimator to account for both withinand between-study heterogeneity. 25 Heterogeneity, when estimating the pooled estimate of the risk ratio, was examined using the standard I 2 test. Publication or small-study bias was assessed using the regression test, and the trimand-fill method was used for correcting funnel plot asymmetry. Risk of bias was assessed at a study level according to the level of adjustment undertaken (Table II in the online-only Data Supplement). Studies at low risk of bias were defined if adjustment of age, sex, and at least 1 other covariate was undertaken. Moderate risk was defined as adjustment of at least age or sex, and studies were classified as high-risk if no adjustment was undertaken. Full statistical methods are explained in the online-only Data Supplement (Text II in the online-only Data Supplement, Figure IA and IB in the onlineonly Data Supplement). All analyses were performed in R Version 3.2.3 with the estimates derived using the metafor package. Statistical significance was taken as a 2-sided P<0.05.

RESULTS
A total of 80 studies were identified to estimate the rate and risk ratio of cardiovascular disease in people living with HIV. One hundred twenty-two estimates from 73 studies were used to calculate the pooled crude incident rates of fatal and nonfatal cardiovascular disease in people living with HIV (Table IV in  A further 17 estimates from 16 studies were identified to estimate the pooled risk ratio of incident cardiovascular disease in individuals with HIV infection (Table, Figure 2B). Studies originated mainly from Europe, North America, and the Asia Pacific region with few studies from low-and middle-income nations (Table) and primarily involving black and white participants ( for stroke. Risk ratios for older studies, those with moderate/high risk of bias, and those with longer follow-up were larger (Table VII in the online-only Data Supplement). Selection bias attributable to potential unpublished studies or to small-study effects was noticed for the overall risk ratio. Imputing for asymmetry using the trim-and-fill method did not alter the effect direction, but, as expected, did attenuate the effect size (Table VII in the online-only Data Supplement and Figure III in the online-only Data Supplement). We observed substantial heterogeneity for the overall estimate ( Figure 2).
There was marked regional variation in the temporal change in the DALYs because of cardiovascular disease attributable to HIV ( Figure 3C). In 2015, East and Southern Africa, Asia and the Pacific, and West and Central Africa accounted for over two-thirds of all DALYs (Figure 3C). The largest annual increase across the 26-

ORIGINAL RESEARCH ARTICLE
year period was observed in East and Southern Africa

National Estimates
National estimates of prevalence and cardiovascular burden were available for 154 countries. The highest population-attributable fraction was observed in countries within sub-Saharan Africa, with HIV accounting for >15% of the cardiovascular burden in Swaziland,  Figure 4A, Table  X in the online-only Data Supplement). Similarly, the largest burden was observed in sub-Saharan Africa (Figure 4B). In the UNAIDS Global Plan priority countries, the population-attributable fraction was comparable to other traditional cardiovascular risk factors (Table XI in

DISCUSSION
In this systematic review, meta-analysis, and burden assessment, we evaluated the association between HIV infection and cardiovascular disease, and estimated the national, regional, and global burden of cardiovascular disease attributable to HIV infection. We make a number of important and novel observations. First, the crude rate for incident cardiovascular disease was 60 per 10 000 person years and is comparable to other high-risk cardiovascular groups, such as diabetes mellitus. 42 Second, the risk of incident cardiovascular disease was 2-fold higher in people living with HIV. Third, the number of DALYs attributable to HIV-associated cardiovascular disease has increased 3-fold over the past 2 decades, but has now plateaued. Finally, there were major regional variations in both the attributable fraction and the rates of cardiovascular disease attributable to HIV, with much of the burden seen in sub-Saharan Africa, followed by Asia and the Pacific.
Many factors may have affected the estimates that we have derived and are based on several assumptions that merit discussion. First, the pooled risk ratios used to calculate the population-attributable fraction and the subsequent cardiovascular burden were primarily obtained from developed nations but were applied to all regions. This approach is ubiquitous in these types of analyses 43,44 and highlights the paucity of data from these regions. In a recent analysis evaluating the global burden of cardiovascular disease attributable to hypertension and obesity, <10% of cohorts originated from low-and middle-income nations. 44,45 Second, the incidence rate does not consider competing risk from noncardiovascular mortality. This would further underestimate the rate of cardiovascular disease in people living with HIV, especially in earlier studies when antiretroviral therapy was not widely available. Third, although many of the individual studies evaluating the risk ratio of cardiovascular disease adjusted for important traditional risk factors, there remains the risk of residual confounding. Previous studies have already shown higher fre- population-attributable fraction for DALY for ischemic heart and cerebrovascular disease was obtained solely from studies including acute myocardial infarction or stroke and so did not specifically include estimates for angina pectoris or other chronic manifestations of atherosclerotic disease. However, relative to the overall cardiovascular DALY attributable to ischemic heart disease, the burden from angina is minimal. 10 Therefore, the impact of this limitation on the overall burden estimate is likely to be small. Fifth, the majority of studies evaluating the risk of cardiovascular disease in people living with HIV have recruited participants before 2010, with a large number conducted in the previous century. The epidemiology of HIV diagnosis and care has changed significantly over the past 2 decades, with better provision of antiretroviral therapy and improved survival, resulting in an increased prevalence of HIV. 46 Sixth, when calculating the burden at a global and regional level, we have made the assumption that the prevalence of HIV infection in the younger age group (15-49 years old) is consistent across the entire age range. There is a paucity of data in the prevalence of HIV in the older population, especially in high-prevalence regions such as sub-Saharan Africa. 47 However, analysis of the populations in these regions shows that the prevalence in the older population remains similar to that of the 15-to 49-year group. 47 Finally, we noticed substantial heterogeneity for our overall pooled risk ratios. The source for this degree of heterogeneity is likely to be multifactorial and reflect differences in population demographics, sample size and small-study effect, patient characteristics, selection or publication bias, and case ascertainment bias because the majority of data were based on national statistics. Many studies have evaluated the association between HIV infection and the risk of atherosclerotic disease including the potential role of antiretroviral therapy. 17,[48][49][50] This is the first study to review and metaanalyze systematically the association between HIV infection and cardiovascular disease, and to estimate the burden of cardiovascular disease attributable to HIV. The mechanisms underlying this association remain poorly understood. 17 Possible mechanisms include endothelial dysfunction 51 and increased systemic 13 and coronary arterial inflammation 14 associated with elevated inflammatory markers. 13 Furthermore, patients with HIV have more traditional metabolic risk factors for cardiovascular disease 17 including dyslipidemia, 15 insulin resistance and abnormal glucose homeostasis, 16 and abnormalities in body fat composition. [52][53][54] The increased risk of cardiovascular disease in people living with HIV is thus a consequence of both accelerated atherosclerosis attributable to chronic infection and the increased prevalence of traditional risk factors.
The global burden of cardiovascular disease attributable to HIV infection has tripled over the past 2 decades,

ORIGINAL RESEARCH ARTICLE
especially in the low-and middle-income nations, and is likely to be a product of both temporal increases in the prevalence of HIV and the morbidity and mortality from cardiovascular disease. The prevalence of HIV varies by region with the greatest proportions seen in sub-Saharan Africa and the Asia Pacific region. Cardiovascular disease now accounts for >10% of all morbidity and mortality in sub-Saharan Africa with rates that are comparable to high-income regions. Consequently, the sub-Saharan region accounted for half of all DALYs from cardiovascular disease attributable to HIV. The population-attributable fraction of HIV-associated cardiovascular disease in the UNAIDS high-priority countries was up to 25% and similar to traditional lifestyle, metabolic, and environmental risk factors. 23,24 The combined burden of HIV and cardiovascular disease in the UNAIDS high-priority countries is of growing concern and has important implications with respect to regional health policies, guidelines, and resource allocation. Risk stratification and identification of patients at intermediate or high risk of future cardiovascular disease are already challenging in resource-limited nations. 55 Furthermore, traditional risk scores perform poorly because they consistently underestimate risk in HIV-infected populations. 35,56,57 Whether patients living with HIV should be considered high risk and started on primary prevention, such as statin therapy, remains unclear. A recent randomized controlled trial of rosuvastatin in patients with HIV demonstrated a reduction in carotid artery intimamedia thickness despite these individuals having low low-density lipoprotein cholesterol concentrations at baseline. 58 Although the latest international guidelines have expanded the use of lipid-lowering therapy in the general population, over two-thirds of people living with HIV with evidence of high-risk morphology coronary atherosclerotic plaque would not have been recommended for statin therapy. 59 The REPRIEVE study (Randomised Trial to Prevent Vascular Events in HIV) is now underway to evaluate the efficacy of statin therapy in people living with HIV who are deemed low-risk based on traditional risk scores. 60,61 CONCLUSIONS This analysis evaluates the association between HIV and cardiovascular disease, and estimates the global burden of HIV-associated cardiovascular disease. We report that the risk of cardiovascular disease was 2-fold higher in people living with HIV. Moreover, the global burden of HIV-associated cardiovascular disease has tripled over the past 2 decades and is now responsible for 2.6 million DALYs per annum, with the majority in sub-Saharan Africa and the Asia Pacific regions. Our estimates have important policy implications for implementing appropriate cardiovascular risk stratification and treatment strategies across healthcare systems, especially in those countries with the greatest burden where resources remain limited.