A critical epidemiological review of cardiovascular disease risk in HIV‐infected adults: the importance of the HIV‐uninfected comparison group, confounding, and competing risks – authors' reply

Abstract

HIV infection is no longer the death sentence it once was. Attention in some resource-rich settings is turning to the field of ageing because of the realization that people living with HIV (PLHIV) age at greater rates than the non-HIVinfected population. It is suggested that PLHIV are more at risk for a host of comorbidities including cancers, renal failure, mental health disorders and also cardiovascular disease (CVD) [1]. We decided to collate the existing evidence and determine what the current body of knowledge indicates about the relative extent of risk of CVD for PLHIV. In doing so, we found that there was no consistency in the way that studies were designed, their data were analysed and/or their results were reported, which makes external comparison and drawing overarching conclusions difficult. The studies were also not ideally designed to address the broader research question of our review and meta-analysis. This could have been a condequence of feasibility issues within the studies or because the studies were addressing related but different questions. Despite this, we believed it was timely and important to conduct a synthesis and meta-analyses but there are factors to be acknowledged in interpreting our findings. We thank Althoff and Gange for addressing some of these factors.Firstly, we agree with Althoff and Gange that the comparison between HIV-infected and HIV-uninfected adults was not ideal. However, as they point out this was a consequence of the scarcity of data from the few studies designed in the best epidemiological manner for our purposes. The factors identified by Althoff and Gange are important in this context and it is unfortunate that underlying studies did not control for them completely. However, it is unreasonable to expect all studies to match all conceivable factors for a true control and each study to match the same factors. As a result of heterogeneities, we were unable to quantify estimates from all the studies as a whole. Rather, we had to conduct a series of meta-analyses. That does not alleviate the problem of incomparability with the full demographics and other characteristics of the setting-specific populations. As that was out of our control, we reported on pooled estimates of meta-analyses from collated relevant studies in subgroups. The outcomes from our analyses should consider these caveats, but it is not clear whether, and if so to what extent, they may bias results.Secondly, as we previously stated in discussion of our analysis, the included studies could not adjust for all confounders and our synthesis could only report on studies as they were conducted and reported. As in all epidemiological and clinical studies, exogenous factors that are believed to be associated with the dependent and independent variables should be assessed and accounted for in study analyses. We chose to include studies that met selection criteria and, although there were some inconsistencies in the chosen confounders or ignored potential confounders, each study had some merit in contributing...