Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Risk of Cardiovascular Disease from Antiretroviral Therapy for HIV: A Systematic Review

  • Clay Bavinger ,

    claybavinger@gmail.com

    Affiliation Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America

  • Eran Bendavid,

    Affiliations Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America, Division of General Internal Medicine, Stanford University Medical Center, Stanford, California, United States of America

  • Katherine Niehaus,

    Affiliation Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America

  • Richard A. Olshen,

    Affiliation Division of Biostatistics, Stanford University Medical Center, Stanford, California, United States of America

  • Ingram Olkin,

    Affiliation Department of Statistics, Stanford University, Stanford, California, United States of America

  • Vandana Sundaram,

    Affiliation Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America

  • Nicole Wein,

    Affiliation Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America

  • Mark Holodniy,

    Affiliation Veterans Affairs Palo Alto Health Care System, Palo Alto, California, United States of America

  • Nanjiang Hou,

    Affiliations Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America, Veterans Affairs Palo Alto Health Care System, Palo Alto, California, United States of America

  • Douglas K. Owens,

    Affiliations Veterans Affairs Palo Alto Health Care System, Palo Alto, California, United States of America, Center for Primary Care and Outcomes Research, and Center for Health Policy, Stanford University, Stanford, California, United States of America

  • Manisha Desai

    Affiliation Quantitative Sciences Unit, Department of Medicine, Stanford University Medical Center, Stanford, California, United States of America

Abstract

Background

Recent studies suggest certain antiretroviral therapy (ART) drugs are associated with increases in cardiovascular disease.

Purpose

We performed a systematic review and meta-analysis to summarize the available evidence, with the goal of elucidating whether specific ART drugs are associated with an increased risk of myocardial infarction (MI).

Data Sources

We searched Medline, Web of Science, the Cochrane Library, and abstract archives from the Conference on Retroviruses and Opportunistic Infections and International AIDS Society up to June 2011 to identify published articles and abstracts.

Study Selection

Eligible studies were comparative and included MI, strokes, or other cardiovascular events as outcomes.

Data Extraction

Eligibility screening, data extraction, and quality assessment were performed independently by two investigators.

Data Synthesis

Random effects methods and Fisher’s combined probability test were used to summarize evidence.

Findings

Twenty-seven studies met inclusion criteria, with 8 contributing to a formal meta-analysis. Findings based on two observational studies indicated an increase in risk of MI for patients recently exposed (usually defined as within last 6 months) to abacavir (RR 1.92, 95% CI 1.51–2.42) and protease inhibitors (PI) (RR 2.13, 95% CI 1.06–4.28). Our analysis also suggested an increased risk associated with each additional year of exposure to indinavir (RR 1.11, 95% CI 1.05–1.17) and lopinavir (RR 1.22, 95% CI 1.01–1.47). Our findings of increased cardiovascular risk from abacavir and PIs were in contrast to four published meta-analyses based on secondary analyses of randomized controlled trials, which found no increased risk from cardiovascular disease.

Conclusion

Although observational studies implicated specific drugs, the evidence is mixed. Further, meta-analyses of randomized trials did not find increased risk from abacavir and PIs. Our findings that implicate specific ARTs in the observational setting provide sufficient evidence to warrant further investigation of this relationship in studies designed for that purpose.

Introduction

Advances in HIV antiretroviral therapy (ART) have dramatically reduced mortality from HIV, such that a person receiving state-of-the-art ART may now expect to live 25 years and potentially longer [1]. Currently, about 50% of all patients with HIV die from causes considered unrelated to HIV [2]. Thus, management of HIV now involves the treatment of a chronic disease with the possibility of near normal life expectancy, but often with multiple comorbidities.

Recent studies suggest that some types of ART may be associated with increased risk of cardiovascular disease, a cause for concern given that people living with HIV may take ART for decades. The mechanisms causing an increased risk of cardiovascular disease are unclear, but according to a review by Grinspoon and Carr, “may relate to dyslipidemia, insulin resistance, diabetes mellitus, inflammation, impaired fibrinolysis, factors specific to antiretroviral medications, or combinations of these factors [3].” The authors further speculate that both HIV and ART might be associated with many of these risk factors [3].

Understanding the relationship between ART and cardiovascular risk is complex because the typical ART regimen contains at least three drugs from two drug classes, and many patients have had multiple regimens. Until recently, the three principal classes of ART have been: protease inhibitors (PIs), nucleoside reverse transcriptase inhibitors (NRTIs), and non-nucleoside reverse transcriptase inhibitors (NNRTIs). The evidence linking ART and cardiovascular disease has pointed specifically to PIs as a class, and specific agents (abacavir, didanosine) [4][7]. The evidence, however, has not been consistent. While some observational studies have found elevated risk with specific drugs or classes [4][7], another observational study has found contrasting evidence [8]. In addition, three recent meta-analyses of randomized trials evaluating abacavir, one of the implicated agents, did not find its exposure associated with an elevated risk of cardiovascular disease [9][11]. Our goal is to reconcile these inconsistencies. To that end, we performed a systematic review of studies that assess the risk of cardiovascular disease from ART. More specifically, we critically evaluated relevant studies to assess the strength of the evidence, to characterize the heterogeneity across studies, and when feasible to make use of information across studies in order to summarize statements regarding specific agents and classes.

Methods

Data Sources

We reviewed English-language articles on the association between antiretroviral drugs and cardiovascular outcomes published through 06/2011 in the Medline, Cochrane, and Web of Science databases, as well as abstracts from the two principal HIV-focused annual conferences: Conference on Retroviruses and Opportunistic Infections (CROI) and International AIDS Society (IAS). Our search terms included myocardial infarction, stroke, and antiretroviral therapy.

Study Selection

We included comparative studies that described the association between antiretroviral drugs and cardiovascular events, including myocardial infarction (MI) and stroke. We included abstracts only when they presented unique data not already included in our analysis from published studies. Studies were excluded from our analysis if they were not comparative, if they only researched intermediate cardiovascular outcomes such as blood pressure, and if subjects were not humans infected with HIV (see Figure 1). Also, non-English language studies were excluded. Two investigators (from JB, KN, VS, or NH) independently reviewed titles, abstracts, and full articles to determine whether studies met inclusion criteria. Conflicting assessment between reviewers were resolved through discussion and review by the two assigned reviewers.

Data Extraction

Two investigators (from JB, KN, VS, and NH) independently abstracted data on study design; eligibility and exclusion criteria; numbers of patients enrolled and lost to follow-up; method of treatment assessment; method of outcome assessment and results for each outcome.

Quality Assessment

We assessed the quality of the study based on features of study design. For observational studies, we designed a rating scheme based on a methodological guide published by the Agency for Healthcare Research and Quality [12]. The most important (major) criteria were ascertainment of exposure, ascertainment of outcome, patient selection criteria, and use of adjusted analyses (see Figure 2). Additional criteria included similarity of patients between treatment and control groups, clear definition of exposure to drugs and outcomes, and adequate description of patient characteristics. We rated studies as good, fair, or poor. Only studies that clearly defined exposure to drugs, outcomes, and patient selection criteria, used medical chart review or chart linkage to gather patient exposure and outcome data, adjusted for common cardiovascular confounders, and fulfilled all quality criteria were rated as good. Ratings of fair were given to studies that fulfilled criteria for ascertainment of exposure and outcomes, patient selection, and used adjusted analyses, but that did not meet all of the additional quality criteria. Any study that failed one or more of the major criteria was rated as poor. All included studies stated approval by an appropriate research ethics committee, or were exempt from such approval because the study was a chart review using no identifying information.

thumbnail
Figure 2. Quality of observational studies was judged according to 8 features of study design.

The 4 major and 4 minor features are shown in this figure. Studies were rated as being of good, fair, or poor quality. Rating scheme is described in the Methods Section.

https://doi.org/10.1371/journal.pone.0059551.g002

We rated randomized clinical trials (RCTs) using the Jadad scale [13]. According to this system, RCT quality is based on whether the study was randomized, whether the study was double-blind, and whether there was a description of withdrawals from the study. For meta-analyses of RCTs, we used the AMSTAR rating system to rate each article numerically [14]. This system rates quality based on whether explicit inclusion criteria were developed before the search, whether a list of included and excluded articles was provided, whether quality of included articles was assessed, and whether proper statistics were used to combine evidence. Articles receiving a score of less than 5 were considered to be of poor methodological quality; articles with a score of between 5 to 8 were considered fair; and articles with a score of 9 or greater were considered to be good quality, as has been done previously [15].

Data Synthesis and Analysis

We combined evidence from studies using two approaches. Our primary approach made use of random effects methods to combine point estimates of similar type [16] when a likelihood ratio test assessing heterogeneity was not rejected, implying the point estimates were not measuring inherently different quantities. Our secondary approach used Fisher’s method to combining p-values for summarizing evidence in cases where point estimates of different measures of risk were provided that could not be combined (for example, hazard ratios and odds ratios) [17], [18]. To ensure p-values across studies described associations in comparable directions we computed one-sided p-values for harm and one-sided p-values for protection and assessed significance of each at the 0.025 level. Finally, whereas our quantitative analyses addressed effects of regimens specifically on MI, we qualitatively compared the results from studies reporting on general cardiovascular events.

Our meta-analyses consisted of combining evidence for studies that addressed comparable questions. Because studies classified exposure differently and used different outcomes, we stratified analyses by two features: drug exposure (e.g., recent, usually defined as exposure within the last 6 months, or cumulative exposure measuring the number of years exposed to a drug or class) and the outcome of interest (MI, stroke, and any cardiovascular event). Even within a specific question (e.g., whether recent abacavir exposure affects MI relative to past abacavir exposure), study designs varied yielding variable types of estimates of risk including the relative risk (RR), the odds ratio (OR), or the hazard ratio (HR). Combining point estimates that measure different parameters is not recommended; more specifically, a combination of measures with different interpretations will not provide a summary statistic with a meaningful interpretation. Because MI is rare, however, the OR in this case may be viewed as a reasonable approximation of the RR, enabling us to formally combine evidence for studies that yield such estimates. All analyses were performed using the R statistical package (http://cran.r-project.org/) [19].

Results

We identified 1,458 articles; 27 met our inclusion criteria yielding 125 separate analyses (see Figure 1, Figure 3a–d, and Table 1). There was 1 RCT. All other studies were observational: 6 were case-control studies, and 20 were cohort studies. Of the observational studies, 5 studies were rated as good quality; 12 were rated as fair; and 9 were rated as poor (see Figure 2). The RCT was rated as good quality. We identified four meta-analyses of RCTs [9][11], [20]. Three of these [9][11] focused on abacavir, with significant overlap of studies analyzed. We therefore chose the meta-analysis that was most comprehensive [9], as well as the one meta-analysis of PI RCTs [20], and used the evidence from these as comparisons to our findings. Neither of these meta-analyses was rated as of good quality, because they did not provide a list of all included and excluded studies, did not provide an assessment of study quality, and did not assess likelihood of publication bias.

thumbnail
Figure 3. (a–d) Reported risk ratio and 95% confidence interval for each study group, organized by drug exposure, cardiovascular event, exposure definition, and risk ratio.

Note that risk of recent exposure represents the effect of exposure to the agent within the past 6 months relative to non-exposure in the past 6 months, and risk per year represents the effect of one additional year of exposure to the agent.

https://doi.org/10.1371/journal.pone.0059551.g003

thumbnail
Table 1. Description of All Included Studies, ART, NNRTI, and NRTI.

https://doi.org/10.1371/journal.pone.0059551.t001

We were able to combine results from 8 observational studies [4], [6], [8], [21][25] that described associations using odds ratios in a formal meta-analysis. The remaining observational studies reported associations using hazard ratios [26][30], did not report MI-specific cardiovascular outcomes [29], [31][38], were the only study reporting on a specific drug exposure [7], [39], or did not provide a quantified measure of their findings (only whether the findings were significant or not) [40][42]. As some studies investigated multiple drugs, they may have contributed to separate analyses of more than one drug class in our meta-analysis (see Tables 2, 3). Table 2 shows the exposures for which we were able to provide summary estimates of effect size, and Table 3 indicates the exposures for which we were able to combine p-values using Fisher’s method.

Yearly Exposure to NRTIs

Two studies reported cardiovascular risk assessed based on yearly exposure to NRTIs [4], [8]. We were unable to combine evidence on yearly use of abacavir due to heterogeneity of the results from the two reporting studies. Specifically, while Lang et al. [8] showed no association between yearly use of abacavir and risk of MI (OR 0.97, 95% CI: 0.86, 1.1), DAD reported a relative risk of 1.14 (95% CI: 1.08, 1.21) [4].

These two studies showed similar heterogeneity when assessing risk from yearly exposure to didanosine; Lang et al. reported no association and the DAD reported an increased risk (0.91, 95% CI: 0.82, 1.01 and 1.06, 95% CI: 1.01, 1.12).

No significant findings were observed in our meta-analysis for cumulative exposure to the other NRTI agents investigated (lamivudine, stavudine, tenofovir and zidovudine exposure).

Recent Exposure to NRTIs

Our pooled analysis of 2 studies [4], [21] demonstrated an association between recent exposure (usually defined as within last 6 months) of abacavir and risk of MI, with a summary RR of 1.91 (95% CI: 1.50, 2.42, see Figure 4a). In addition, combining p-values across all studies that evaluate comparable definitions of recent exposure to abacavir with comparable reference groups (See Table 4) using Fisher’s method suggested a harmful association between recent abacavir use and MI (p<0.001).

thumbnail
Figure 4. (a, b) Reported risk ratio and 95% confidence interval for exposure groups with sufficient evidence to summarize in meta-analysis.

Results of meta-analysis are shown in bottom row of each exposure group, denoted pictorially by the red diamond. Each study is given a weight based on its number of subjects and length of follow-up, denoted pictorially by size of its box in the plot. Note that risk of recent exposure represents the effect of exposure to the agent within the past 6 months relative to non-exposure in the past 6 months, and risk per year represents the effect of one additional year of exposure to the agent.

https://doi.org/10.1371/journal.pone.0059551.g004

Both DAD [4] and Lang et al. [8] assessed the association between recent exposure to didanosine and MI, but because a test of heterogeneity indicated the parameters were incompatible, we did not combine the evidence. However, using Fisher’s method of combining p-values, we were able to include these as well as the results of Lundgren [28] (see Table 4). The results indicated a harmful association (p = 0.001). One additional study found no association between didanosine use and MI, but could not contribute to quantitative analyses, as they did not report numerical results [39].

No significant findings were observed in our meta-analysis for recent exposure to the other NRTI agent investigated (stavudine).

Yearly Exposure to PIs

Cumulative exposure to individual PIs was investigated by the DAD and by Lang et al. [6], [8] (see Table 2). Combining evidence from these two studies demonstrated significantly increased risks of MI with cumulative indinavir use (1.11, 95% CI: 1.05, 1.17) and cumulative use of lopinavir with or without ritonavir (1.22, 95% CI: 1.01, 1.47). Lang et al. also found an increased risk associated with amprenavir, although no other studies identified such an association. Neither nelfinavir nor saquinavir were found to be significantly harmful in any study. Only one study by Friis-Moller et al. examined the effect of cumulative exposure to PIs as an entire class on MI, where a significantly increased risk of MI was observed [7].

Recent Exposure to PIs

Recent PI exposure was examined by nine studies, with five finding significantly increased cardiovascular risk (see Figure 3d). Summarizing the three studies that reported odds ratios for MI [23][25] yielded an OR of 2.13 (95% CI: 1.06, 4.28) (see Figure 4b). Using Fisher’s method of summarizing p-values for the six studies reporting on recent PI use and MI [23][27], [41] we found an overall significant risk for MI associated with recent exposure to PIs as a class (summary p-value = 0.003). One additional study investigated PI drugs individually, finding a 75% and 93% increased risk of MI (95% CIs: 1.02, 3.01 and 1.04, 3.57) for nelfinavir and indinavir, respectively [39].

Discussion

Our analysis combined evidence across studies investigating the association between cumulative and recent exposure to specific ART drugs as well as to classes of ART drugs and the risk of MI. Our findings implicated recent exposure to abacavir, recent exposure to PIs in general, and cumulative exposure to PIs indinavir and lopinavir. There are several issues, however, that need to be considered when interpreting our findings.

Exposure to Abacavir

While our meta-analysis suggested an association between recent abacavir use and risk of MI, we note there were inconsistencies across studies in both findings and study quality. Six studies reported the association between recent abacavir use and risk of MI: three reported significant increases in risk [4], [21], [28]. One of these six studies was of good quality [30], four were of fair quality [4], [8], [28], [43], and one other was of poor quality; its patients were not similar across control and treatment groups [21]. Thus, although the observational studies point towards an increase in MI risk (Figure 3a), the evidence is not fully consistent. In contrast, three meta-analyses of randomized clinical trials reported no evidence of an association between abacavir use and MI [9], [10], [11]. Of the three studies, the meta-analysis by Ding et al. included a greater number of studies, including those used in the other two analyses, so we chose to compare our results to those from the study by Ding et al. [9], [44]. The RCTs included in this meta-analysis were designed for the purpose of establishing drug efficacy, and they are thus of short duration; the average length of patient follow-up was 1.62 person-years per subject. MIs are rare enough events that studies with short follow-up time are unlikely to have the power to detect differential risk. Indeed, the meta-analysis itself reports only 62% power to detect an odds ratio of 1.8 for MI [9], [44]. As each individual trial is relatively small, the number of events for many of the trials is 0 for both exposure groups, again reflecting the scarcity of information.

The evidence for risk from cumulative exposure to abacavir is also mixed. Only three studies reported on this relationship, with conflicting point estimates. Lang reported no association with cumulative abacavir use (OR = 0.97, 95% CI: 0.86, 1.1) [8], whereas DAD reported a relative risk of 1.14 per year (95% CI: 1.08, 1.21) [4]. Bedimo et al., who reported a HR, found no association (HR 1.18, 95% CI: 0.92, 1.50) [43]. The heterogeneity between these fair-quality observational studies suggests that there is still uncertainty about the cumulative risk of MI from abacavir. An additional question is how to reconcile uncertainty about cumulative risk with the finding that recent exposure to abacavir is associated with increased risk. A possible explanation might be that those who remain on abacavir have cardiovascular risk profiles that continue to be favorable while on the regimen and that those whose profiles become unfavorable while on abacavir are removed from the regimen; if those with unfavorable profiles were more likely to experience an MI, it would thereby implicate recent use of abacavir but allow cumulative use to not appear harmful. In conclusion, the available evidence on the association between abacavir use and MI is not definitive.

Exposure to PIs

Some of the studies investigating PIs found an association between cumulative use of PIs and cardiovascular disease, and others found an association with recent exposure. Our meta-analysis based on three observational studies indicated that recent PI use was associated with an odds ratio of 2.13 for MI. We caution, however, that this combined estimate is based upon studies that did not meet important criteria for quality [23][25]. In contrast to the findings from observational studies, Coplan et al. conducted a meta-analysis of RCTs [20] and found no association between nelfinavir and risk of MI (point estimate not reported) or between indinavir exposure and risk of MI (0.7, 95% CI: 0.1, 7.75). Similar issues present here as with the meta-analysis of RCTs presented by Ding et al. that evaluated the association between abacavir and MI [9]. These include drawing inference from studies of short duration that are not designed to evaluate endpoints such as MI and that consequently are underpowered to detect such associations.

Finally, we found a significant increase in risk associated with cumulative lopinavir and indinavir use. These results are based upon only two studies, and their quality was fair [6], [8]. We therefore caution against interpreting these findings as conclusive.

Methodological Challenges

When possible, we combined estimates from studies to assess the risk of cardiovascular disease associated with ART. There were significant challenges, however, to achieving this in our study. Barriers included heterogeneity across the studies with regard to definitions of drug exposure (e.g., time-varying or fixed; cumulative exposure or recent exposure), populations investigated, designs employed (e.g., longitudinal or cross-sectional), and finally, specification of the statistical models (e.g., assessing cumulative or recent exposure separately or jointly). While some of the studies allowed the association of drug exposure to vary over time (e.g., DAD, Lang, Bedimo, and Choi) [4], [8], [30], [43], some considered drug exposure to be a fixed effect (e.g., Daftary and Holmberg) [23], [24]. Incorporating information about how exposure changes over time for an individual (e.g., whether the subject is currently exposed to the drug at a specific time point versus whether the subject was ever exposed over their observation period) will impact the estimates of the coefficients, their interpretation, and therefore their comparability across studies. In addition, some of the studies were based on models that specified risk as a function of cumulative exposure to a drug whereas others were based on models that specified risk as a function of both cumulative exposure as well as indicators for past and recent exposure. Cumulative exposure in the former model represents the change in risk corresponding to a 1-year increase in exposure and includes the risk for someone exposed for 1 year relative to someone never exposed. Cumulative exposure in the latter model, on the other hand, represents the change in risk corresponding to a 1-year increase in exposure only among those exposed.

An illustration of the impact of these analytic choices on findings may be provided by the discrepancy in estimates obtained by Bedimo et al. and Choi et al. [30], [43]. Both studies were conducted using the same data on the same population – the national sample of HIV-infected US veterans (with some differences in inclusion/exclusion criteria as well as follow-up time). Bedimo et al. estimated the effect of recent exposure to abacavir on the hazard of MI relative to exposure to neither abacavir nor tenofovir and demonstrated no association with a point estimate indicating protection among abacavir users against acute MI (HR = 0.67; 95% CI: 0.43, 1.03). In contrast, Choi et al. estimated the effect of recent abacavir exposure on the hazard of MI to be 1.64 (95% CI = 0.88–3.08). It is difficult to say whether these differences can be attributed to how the patients were selected from the larger cohort, the chosen reference group for comparisons, how the models were specified including which covariates were used for adjustment, how exposure was defined, or a combination of all of these factors. However, this discrepancy exemplifies well some of the challenges faced with combining evidence across studies to evaluate the effect of ART on MI.

An additional issue with these analyses involved the potential for confounding by indication. In the observational setting, patients with HIV start and stop regimens and initiate new ones over time. Factors governing such decisions likely relate to a patient’s response to treatment. When decisions for initiation or termination of regimens are related to risk of MI, however, drawing inference about risk of MI may be confounded by one’s history of treatment and risk factor profile. Tools such as marginal structural models and propensity score methods may be useful in addressing this issue [45][52]. For example, Obel et al. [21] incorporated the use of propensity scores in their analysis, and we advocate the use of such methods to attempt to handle this complex confounding. Of course, application of these methods designed to handle challenging statistical issues is no guarantee of valid findings and also rely on their own sets of assumptions. We therefore recommend sensitivity analyses that vary methods, which rely on different assumptions, and that vary definitions of drug exposure. This can provide insight into the robustness of findings to different assumptions. Some of the studies described here made use of such approaches to provide insight into the interpretation of their results [4], [8].

Finally, we should note that there is some overlap between our study and that by Islam et al. who address topics similar to those discussed in our paper [53]. While our work highlights comparisons of the risks of MI among HIV-positive people exposed to different regimens of ART, the study by Islam et al. has a different scope that also includes comparisons between HIV-positive and HIV-negative subjects. Where our goals overlap, however, our findings are largely in agreement. Islam et al. observed an increased risk of CV events with exposure to PIs as a class, lopinavir, and abacavir, and our analysis additionally revealed an increased risk in MI from exposure to indinavir. However, our findings are based on methodological choices regarding pooling or summarizing estimates that differ from those of Islam and colleagues. We chose not to combine results from studies that reported hazard ratios with studies that reported odds ratios or relative risks, due to concerns about the interpretability of a summary statistic. We also chose not to use a study more than once in any specific summary estimate, even if it had relevant data (for example, two NRTI drugs), because of concerns that the lack of independence of observations could lead to underestimation of the standard errors for the summary ORs. In sum, although we made different methodological choices, our findings are largely in agreement with those of Islam and colleagues.

Conclusion

Our study found evidence based on observational studies to suggest a harmful association between abacavir and risk of MI. In addition, there is evidence from observational studies that use of PIs increases risk of MI. Evidence from the observational and randomized trials are at odds, however. While the randomized clinical trial setting would provide the least biased approach to assessing cardiovascular risk, the clinical trials included in our investigation were not designed for that purpose. Consequently they were short-term, and were underpowered for assessing cardiovascular risk. Compared to these meta-analyses of RCTs, the observational studies include much longer follow up and a more representative sample, but they are subject to confounding by indication, in which risk of cardiovascular disease may influence both ART choice and MI, which may lead to spurious associations. In addition, combining evidence across studies proves challenging in the presence of heterogeneity of study designs and analytic plans. Based on the overall evidence, we believe there is still uncertainty whether ART leads to increased cardiovascular risk, and if so, the magnitude of that risk. The current evidence provided by the observational studies is sufficient, however, to warrant further study in prospective studies designed to assess cardiovascular risk from ART.

Acknowledgments

Disclaimer

This serves as a disclaimer that the views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.

Author Contributions

Conceived and designed the experiments: JCB EB KN VS NW DKO MD. Performed the experiments: JCB EB KN VS NW NH. Analyzed the data: JCB EB KN RAO IO VS NW MH DKO MD. Wrote the paper: JCB EB KN RAO IO VS NW MH DKO MD.

References

  1. 1. The Antiretroviral Therapy Cohort Collaboration (2008) Life expectancy of individuals on combination antiretroviral therapy in high-income countries: a collaborative analysis of 14 cohort studies. Lancet 372: 293–299.
  2. 2. The Antiretroviral Therapy Cohort Collaboration (2010) Causes of death in HIV-1-infected patients treated with antiretroviral therapy, 1996–2006: collaborative analysis of 13 HIV cohort studies. Clin Infect Dis 15: 1387–1396.
  3. 3. Grinspoon S, Carr A (2005) Cardiovascular Risk and Body-Fat Abnormalities in HIV-Infected Adults. N Engl J Med 352: 48–62.
  4. 4. Sabin CA, Worm SW, Weber R, Reiss P, El-Sadr W, et al. (2008) Use of nucleoside reverse transcriptase inhibitors and risk of myocardial infarction in HIV-infected patients enrolled in the D:A:D study: a multi-cohort collaboration. Lancet 371: 1417–1426.
  5. 5. Friis-Moller N, Sabin CA, Weber R, d’Arminio Monforte A, El-Sadr WM, et al. (2003) Combination antiretroviral therapy and the risk of myocardial infarction. N Engl J Med 349: 1993–2003.
  6. 6. Worm SW, Sabin C, Weber R, Reiss P, El-Sadr W, et al. (2010) Risk of myocardial infarction in patients with HIV infection exposed to specific individual antiretroviral drugs from the 3 major drug classes: the data collection on adverse events of anti-HIV drugs (D:A:D) study. J Infect Dis 201: 318–330.
  7. 7. Friis-Moller N, Reiss P, Sabin CA, Weber R, Monforte AD, et al. (2007) Class of antiretroviral drugs and the risk of myocardial infarction. N Engl J Med 356: 1723–1735.
  8. 8. Lang S, Mary-Krause M, Cotte L, Gilquin J, Partisani M, et al. (2010) Impact of individual antiretroviral drugs on the risk of myocardial infarction in human immunodeficiency virus-infected patients: a case-control study nested within the French Hospital Database on HIV ANRS cohort CO4. Arch Intern Med 170: 1228–1238.
  9. 9. Ding X, Andraca-Carrera E, Cooper C, Miele P, Kornegay C, et al.. (2012) No Association of Abacavir Use with Myocardial Infarction: Findings of an FDA Meta-analysis. JAIDS Ahead of Print.
  10. 10. Ribaudo JH, Constance A Benson, Yu Zheng, Susan L Koletar, et al. (2011) No Risk of Myocardial Infarction Associated With Initial Antiretroviral Treatment Containing Abacavir: Short and Long-Term Results from ACTG A5001/ALLRT. Clin Infect Dis 52: 929–940.
  11. 11. Brothers CH, Hernandez JE, Cutrell AG, Curtis L, Ait-Khaled M, et al. (2009) Risk of myocardial infarction and abacavir therapy: no increased risk across 52 GlaxoSmithKline-sponsored clinical trials in adult subjects. J Acquir Immune Defic Syndr 51: 20–28.
  12. 12. Agency for Healthcare Research and Quality (2007) Methods Reference Guide for Effectiveness and Comparative Effectiveness Reviews. Available: http://effectivehealthcare.ahrq.gov/repFiles/2007_2010DraftMethodsGuide.pdf.
  13. 13. Jadad A, Moore RA, Carroll D, Jenkinson C, Ryeynolds DJ, et al. (1996) Assessing the Quality of Reports of Randomized Clinical Trials: Is Blinding Necessary? Control Clin Trials 17: 1–12.
  14. 14. Shea BJ, Grimshaw JM, Wells GA, Boers M, Andersson N, et al. (2007) Development of AMSTAR: a measurement tool to assess the methodological quality of systematic reviews. BMC Med Res Methodol 7: 10.
  15. 15. Braga LH, Pemberton J, Demaria J, Lorenzo AJ (2011) Methodological concerns and quality appraisal of contemporary systematic reviews and meta-analyses in pediatric urology. J Urol 186: 266–271.
  16. 16. Viechtbauer W (2010) Conducting Meta-Analyses in R with the metafor Package. J Stat Softw 36.
  17. 17. Fisher RA (1948) Questions and Answers #14. The American Statistician 2: 30–31.
  18. 18. Fisher RA (1925) Statistical Methods for Research Workers. Edinburgh: Oliver and Boyd.
  19. 19. R Development Core Team (2011) R: A language and environment for statistical computing Vienna, Austria: R Foundation for Statistical Computing.
  20. 20. Coplan PM, Nikas A, Japour A, Cormier K, Maradit-Kremers H, et al. (2003) Incidence of myocardial infarction in randomized clinical trials of protease inhibitor-based antiretroviral therapy: an analysis of four different protease inhibitors. Aids Res Hum Retroviruses 19(6): 449–455.
  21. 21. Obel N, Farkas DK, Kronborg G, Larsen CS, Pedersen G, et al. (2010) Abacavir and risk of myocardial infarction in HIV-infected patients on highly active antiretroviral therapy: a population-based nationwide cohort study. HIV Med 11: 130–136.
  22. 22. Durand M, Sheehy O, Baril J-G, Lelorier J, Tremblay C, et al.. (2009) Relation between use of nucleoside reverse transcriptase inhibitors (NRTI) and risk of myocardial infarction (MI): a nested case control study using Quebec’s public health insurance database (QPHID). 5th IAS Conference on HIV Pathogenesis and Treatment.
  23. 23. Daftary M, Dutta AP, Xue Z, Wyl Vv, Young M, et al.. (2004) Women’s Intragency HIV Study (WIHS): cardiovascular outcomes in women on PI therapy. The XV International AIDS Conference.
  24. 24. Holmberg SD, Moorman AC, Tong TC, Ward DJ, Wood KC, et al.. (2002) Protease inhibitor drug use and adverse cardiovascular events in ambulatory HIV-infected patients. The XIV International AIDS Conference.
  25. 25. Rickerts V, Brodt H, Staszewski S, Stille W (2000) Incidence of myocardial infarctions in HIV-infected patients between 1983 and 1998: the Frankfurt HIV-cohort study. Eur J Med Res 5: 329–333.
  26. 26. Kwong GP, Ghani AC, Rode RA, Bartley LM, Cowling BJ, et al. (2006) Comparison of the risks of atherosclerotic events versus death from other causes associated with antiretroviral use. AIDS 20: 1941–1950.
  27. 27. Iloeje UH, Yuan Y, L’Italien G, Mauskopf J, Holmberg SD, et al. (2005) Protease inhibitor exposure and increased risk of cardiovascular disease in HIV-infected patients. HIV Med 6: 37–44.
  28. 28. Lundgren JD, Neuhaus J, Babiker A, Cooper D, Duprez D, et al. (2008) Use of nucleoside reverse transcriptase inhibitors and risk of myocardial infarction in HIV-infected patients. AIDS 22: F17–F24.
  29. 29. Bedimo R, A Westfall, H Drechsler, Tebas P (2009) Abacavir use and risk of acute myocardial infarction and cerebrovascular disease in the HAART era. 5th IAS Conference on HIV Pathogenesis and Treatment.
  30. 30. Choi A, Vittinghoff E, Deeks SG, Weekleya C, Liaand Y (2011) Cardiovascular risks associated with abacavir and tenofovir exposure in HIV-infected persons AIDS. 25: 1289–1298.
  31. 31. Belloso WH, Orellana LC, Grinsztejn B, Madero JS, La Rosa A, et al. (2010) Analysis of serious non-AIDS events among HIV-infected adults at Latin American sites. HIV Med 11: 554–564.
  32. 32. Martin A, Bloch M, Amin J, Baker D, Cooper DA, et al. (2009) Simplification of Antiretroviral Therapy with Tenofovir-Emtricitabine or Abacavir-Lamivudine: A Randomized, 96-Week Trial. Clin Infect Dis 49: 1591–1601.
  33. 33. Corral I, Quereda C, Moreno A, Perez-Elias MJ, Dronda F, et al. (2009) Cerebrovascular ischemic events in HIV-1-infected patients receiving highly active antiretroviral therapy: incidence and risk factors. Cerebrovasc Dis 27: 559–563.
  34. 34. Levy AR, Sobolev B, Hogg RS, Iloeje U, Mukherjee J, et al.. (2003) Risks of cardiovascular disease associated with highly active anti-retroviral therapy among persons treated for HIV/AIDS. The 2nd IAS Conference on HIV Pathogenesis and Treatment.
  35. 35. Bozzette SA, Ake CF, Tam HK, Chang SW, Louis TA (2003) Cardiovascular and cerebrovascular events in patients treated for human immunodeficiency virus infection. N Engl J Med 348: 702–710.
  36. 36. David MH, Hornung R, Fichtenbaum CJ (2002) Ischemic cardiovascular disease in persons with human immunodeficiency virus infection. Clin Infect Dis 34: 98–102.
  37. 37. Vaughn G, Detels R (2007) Protease inhibitors and cardiovascular disease: analysis of the Los Angeles County adult spectrum of disease cohort. AIDS Care 19: 492–499.
  38. 38. The Writing Committee of the D:A:D Study Group (2004) Cardio- and cerebrovascular events in HIV-infected persons. AIDS 18: 1811–1817.
  39. 39. Triant V, Regan S, Lee H, Sax P, Meigs J, Grinspoon S (2010) Association of antiretroviral therapy and HIV-related factors with acute myocardial infarction rates. International AIDS Conference. Vienna.
  40. 40. Quiros-Roldan E, Torti C, Tinelli C, Moretti F, Zanini B, et al. (2005) Risk factors for myocardial infarction in HIV-positive patients. Int J STD AIDS 16: 14–18.
  41. 41. Barbaro G, Di Lorenzo G, Cirelli A, Grisorio B, Lucchini A, et al. (2003) An open-label, prospective, observational study of the incidence of coronary artery disease in patients with HIV infection receiving highly active antiretroviral therapy. Clin Ther 25: 2405–2418.
  42. 42. Klein D, Hurley LB, Quesenberry CP Jr, Sidney S (2002) Do protease inhibitors increase the risk for coronary heart disease in patients with HIV-1 infection? J Acquir Immune Defic Syndr 30: 471–477.
  43. 43. Bedimo RJ, Westfall AO, Drechsler H, Vidiella G, Tebas P (2011) Abacavir Use and Risk of Acute Myocardial Infarction and Cerebrovascular Events in the Highly Active Antiretroviral Therapy Era. Clin Infect Dis 53: 84–91.
  44. 44. Ding X, Andraca-Carrera E, Cooper C, Miele P, Kornegay C, et al.. (2011) No Association of Myocardial Infarction with Abacavir Use: Findings of an FDA Meta-analysis. Conference on Retroviruses and Opportunistic Infections. Boston.
  45. 45. Robins JM, Hernán MA, Brumback B (2000) Marginal structural models and causal inference in epidemiology. Epidemiology 11: 550–560.
  46. 46. Hernán MA, Brumback B, Robins J (2000) Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology 11: 561–570.
  47. 47. Hernán M, Brumback B, Robins JM (2001) Marginal structural models to estimate the joint causal effect of nonrandomized treatments. J Am Stat Assoc 96: 440–448.
  48. 48. Hernán MA, Brumback B, Robins JM (2002) Estimating the causal effect of zidovudine on CD4 count with a marginal structural model for repeated measures. Stat Med 21: 1689–1709.
  49. 49. Sturmer T, Schneeweiss S, Avorn J, Glynn RJ (2005) Adjusting effect estimates for unmeasured confounding with validation data using propensity score calibration. Am J Epidemiol 162(3): 279–289.
  50. 50. Joffe MM, Rosenbaum PR (1999) Invited commentary: propensity scores. Am J Epidemiol 150: 327–33.
  51. 51. Hernan MA, Robins JM (2006) Instruments for causal inference: an epidemiologist’s dream? Epidemiology 17: 360–372.
  52. 52. Stürmer T, Schneeweiss S, Rothman KJ, Avorn J, Glynn RJ (2007) Performance of propensity score calibration–a simulation study. Am J Epidemiol 165: 1110–1118.
  53. 53. Islam F, Wu J, Jansson J, Wilson D (2012) Relative risk of cardiovascular disease among people living with HIV: a systemic review and meta-analysis. HIV Med 13(8): 453–68.