Baseline ECG and Cardiovascular Outcomes in People With HIV: Insights From REPRIEVE

With antiretroviral therapy, people with HIV (PWH) have an increased burden of cardiovascular disease. The REPRIEVE (Randomized Trial to Prevent Vascular Events in HIV) trial demonstrated that pitavastatin reduces major adverse cardiovascular events (MACEs) among PWH at low to moderate traditional atherosclerotic risk. Electrocardiographic abnormalities are common in PWH, but little is known about their association with MACEs. We sought to examine whether baseline electrocardiographic abnormalities are associated with increased MACE risk among a global primary cardiovascular disease prevention cohort of PWH in REPRIEVE.

In this observational analysis, entry electrocardiographic abnormalities were adjudicated and classified as major or minor abnormalities. Multivariable cause‐specific Cox proportional hazards models assessed the association of electrocardiographic abnormalities with MACEs while stratifying for treatment effect. The model improvement with the addition of the ECG to a model with the pooled cohort equations risk score was examined.

Among 7719 participants (median age, 50 years; 69% men), 49% had ≥1 electrocardiographic abnormality, with 3% classified as major. Over a median of 5.6 years, a major electrocardiographic abnormality was associated with a 2.42‐fold (95% CI, 1.49–3.91) higher hazard of incident MACEs, whereas minor abnormalities were not. Specific abnormalities associated with MACEs were chamber enlargement and infarct/ischemia pattern. No significant subgroup‐ or treatment‐related interaction was observed. Adding electrocardiographic findings to traditional risk factors increased the C‐statistic modestly (+0.01).

Among PWH in REPRIEVE, electrocardiographic abnormalities were common, but major electrocardiographic abnormalities were rare. Though major abnormalities were associated with increased hazard of MACEs, routine electrocardiographic screening is unlikely to improve the prediction of future cardiovascular events in this primary prevention population with low to moderate cardiovascular risk.

A total of 159 participants had a first hard MACE (2%) over the period of follow‐up. Having a major baseline electrocardiographic abnormality was again found to be associated with hard MACEs compared with not having any abnormalities (aHR, 2.28 [95% CI, 1.22, 4.24]), whereas 1 minor abnormality and >1 minor abnormality were not found to be significantly associated with the outcome. Chamber enlargement/hypertrophy was found to be associated with the outcome in both the unadjusted and adjusted models (aHR, 1.73 [95% CI, 1.07–2.79]; Figure S9).

Including electrocardiography (minor/major classification) improved the fit of the original model, increasing the model χ2 from 98.9 to 114.5 (likelihood ratio statistic, 15.6; df=3; P=0.001). The discriminative ability at full follow‐up increased marginally (Harrel C‐statistic increased from 0.67 [0.64–0.70] to 0.68 [0.65, 0.71], Uno C increased from 0.67 [0.63–0.70] to 0.68 [0.64–0.71]) (Table S7). Overall net reclassification showed a modest increase at 4.12% over a period of 4 years of follow‐up, the period over which most participants had follow‐up data. Similar modest increases were seen over a 5‐year period of follow‐up and the full length of follow‐up (Table S8). Overall performance showed a small increase in the scaled Brier score from 2.3% to 2.9% with no change in Brier score (2.8%).

In our global cohort of ART‐treated, low to moderate ASCVD risk, PWH without known CVD, we identified a significant association between major baseline electrocardiographic abnormalities and first primary MACE and hard MACE. The association persisted across different subgroups and was not modified by the presence or absence of traditional cardiovascular risk factors. We further identified a significant association between specific groups of electrocardiographic abnormalities and primary MACE, such as chamber enlargement/hypertrophy, bundle branch block/IVCD, and infarct/ischemia patterns. The finding of ≥1 major electrocardiographic abnormalities was uncommon. Therefore, the addition of major/minor electrocardiographic abnormalities to a model with traditional CVD risk factors resulted in only a modest increase in the discrimination, net reclassification, and overall performance of the model.

Cardiovascular risk prevention has become an important component of care for PWH, as HIV infection has evolved for most PWH to be a chronic condition causally linked to increased CVD risk.29 We have previously shown that both cardiovascular and HIV‐specific risk factors are associated with MACE and hard MACE outcomes in REPRIEVE. However, the role of baseline electrocardiographic abnormalities as risk factors has not been investigated in the contemporary era of highly effective ART. Our findings reveal that while minor electrocardiographic abnormalities are common, major abnormalities are rare, occurring in only 3% of the REPRIEVE population, without known CVD and of low to moderate predicted risk. Furthermore, our data demonstrate that only major electrocardiographic abnormalities were associated with an increased hazard of both MACEs and hard MACEs, whereas minor abnormalities showed no significant association. Specific abnormalities associated with the first primary MACE included chamber enlargement/hypertrophy, bundle branch block/IVCD, and infarct/ischemia patterns, which may indicate electrical or subclinical myocardial disease.

HIV infection has a known independent association with QTc prolongation,30, 31, 32 yet the clinical significance of this prolongation remains underexplored in this population. In our analysis, each 5‐ms increase in QTc was associated with a 4% increase in the hazard of MACEs, notably, the prevalence of major QTc prolongation (defined as QTc ≥500 ms regardless of sex) was low in REPRIEVE (0.14%). Although baseline QTc prolongation was higher among men, Asians, and participants with detectable HIV‐1 RNA,11 we did not observe a different associated hazard in these subgroups, potentially due to the limited number of events per subgroup.

Early repolarization was the most prevalent minor ECG abnormality in our cohort. Although it is generally viewed as benign, a few reports in the general population have linked early repolarization to ventricular fibrillation and sudden death.33, 34 However, in our analysis, we did not observe an association with either MACE or hard MACE outcomes. While some abnormalities did not achieve statistical significance after adjustment, the CIs and point estimates for bundle branch block/IVCD, ST‐T wave abnormalities, and sex‐specific QTc prolongation suggest a clinically relevant increased risk.

Our results echo findings from the general population, where resting baseline electrocardiographic abnormalities are also found to be associated with increased hazard for MACE and other cardiovascular outcomes as reported by multiple studies.13, 14, 15, 16, 17, 18 The question of whether ECG screening can be informative for CVD risk stratification beyond traditional CVD risk factors is a matter of discussion both in the general population and in PWH.35 The last revision of the current evidence by the US Preventive Taskforce in 2018 concluded that there is not enough evidence to recommend resting or exercise electrocardiography as a preventive measure for CVD in asymptomatic adults.36, 37 Unlike the general population, the data in PWH are much more limited. Given the higher risk profile of asymptomatic PWH with a greater prevalence of subclinical ASCVD and structural heart disease in the form of cardiac fibrosis,29 along with the high prevalence of baseline electrocardiographic abnormalities, our findings provide new information to the field from a large contemporary primary prevention cohort of PWH. These data address the question of whether incorporating electrocardiographic screening for CVD risk assessment is useful, especially considering that electrocardiography is widely available and low‐cost.

Our findings are consistent with and complementary to the prior report from the SMART (Strategies for Management of Antiretroviral Therapy) study,12 which enrolled participants between 2002 and 2006 and demonstrated a high prevalence of baseline ECG abnormalities among PWH compared with the general population,38, 39, 40 as well as an association between the presence of major baseline electrocardiographic abnormalities and incident CVD (aHR, 1.83 [95% CI, 1.12–2.97]). Our findings from REPRIEVE, observed in an era of effective ART, represent an important complement to the earlier SMART findings. Notably, SMART was also a higher CVD risk population than REPRIEVE. For example, 40% of the SMART population were smokers (versus 25% in REPRIEVE), and 8% of the SMART study cohort had a major ECG abnormality (versus 3% in REPRIEVE). Moreover, ASCVD events were prospectively adjudicated in REPRIEVE, and baseline cardiovascular risk phenotypes were explored in detail through prospective ASCVD risk profiling.

With respect to use of electrocardiography for risk prediction, our analysis showed only a modest improvement for MACE prediction when adding electrocardiographic findings. A possible explanation could be related to the nature of our cohort: a younger, healthier, ART‐treated population with no known CVD, and low to moderate ASCVD risk as assessed by the PCE (median ASCVD risk score, 4.5%). This is supported by prior findings where the addition of electrocardiography was mostly useful for intermediate‐higher risk general populations.15, 41 Second, we assessed our models over a median follow‐up time of 5.6 years given our data availability. A longer follow‐up time might provide additional insights for the clinical utility of electrocardiography.

Based on our findings, widespread screening of this population, with only rare major electrocardiographic abnormalities, is not likely to have clinical utility. In contrast, finding a major abnormality in an individual patient as an incidental finding may provide important information, given the higher risk of an event, and warrant a clinical decision on further workup. In REPRIEVE, a small number of individuals (N=41) exhibited potential Q waves and may have experienced silent myocardial infarctions without prior symptoms or clinical history. The clinical significance of this finding with respect to cardiovascular prevention strategies remains unclear.

In our analysis, taking a statin did not appear to modify the hazard associated with most baseline electrocardiographic findings, with the caveat that REPRIEVE was not powered to detect this treatment effect based on electrocardiographic subgroups. For major/minor abnormalities, the interaction effect was borderline, and effect modification cannot be ruled out. There was some indication of possible modification of the association between electrocardiographic abnormalities and MACEs by treatment group, although this was of a borderline significance and potentially due to chance.

In this analysis, we used electrocardiographic information that was analyzed using standard criteria. However, this traditional interpretation may have limitations. Recent advances in artificial intelligence and machine learning that leverage raw ECG waveforms hold promise for detecting subtle and interrelated nonlinear patterns that are overlooked by human experts.42 A few studies have already explored predicting long‐term cardiovascular death from resting ECGs using these techniques, showing a Harrel C‐statistic of an average of 0.8 for prediction of cardiovascular death when combined with the PCE risk score in the general population.43, 44 However, as we continue to address the complexities of clinically deploying such models, such as ensuring the consistency of ground‐truth labels and adopting appropriate evaluation metrics,45, 46 it remains important to recognize the value of readily available and interpretable electrocardiographic data, especially given the high prevalence of PWH in resource limited settings.

This study has a number of strengths, including assessment of ECGs from a global cohort with diverse racial and sociodemographic characteristics who were confirmed to be asymptomatic at the time of electrocardiographic screening and enrollment, addressing a key limitation noted in several prior studies. Moreover, all findings were interpreted by a core laboratory with expertise in electrocardiographic interpretation, ensuring accuracy. Our study also has several limitations: First, we obtained only a baseline ECG, before treatment initiation, and electrocardiographic abnormalities may have occurred after enrollment, which we did not capture serially. Second, there has been no universally accepted consensus on classification of minor and major abnormalities, determined from prior studies.14, 15, 16, 17 Due to the relatively low frequency of individual abnormalities, we were underpowered to detect associations with each specific abnormality. To address this, we used classification approaches that grouped abnormalities on the basis of shared clinical and pathophysiological relevance. Third, we used a composite MACE end point assuming a common pathophysiology among its components, and small event numbers limited subgroup analyses. Finally, only participants with a baseline ECG were included, but selection bias was minimal, as 99% of the cohort had a baseline ECG. Finally, we had high missingness for HIV‐1 viral load, but we used multiple sensitivity analysis that showed consistency with the described associations.

Among an ART‐treated population with no known CVD and low to moderate 10‐year ASCVD risk as assessed by the PCE, our data show that minor electrocardiographic abnormalities are frequent but not associated with future MACEs. In contrast, major abnormalities are rare but associated with MACEs. Nonetheless, given the low prevalence, the presence of major electrocardiographic abnormalities did not add significantly to overall discrimination or net reclassification of events in our population of PWH who were participating in a primary cardiovascular prevention trial. Taken together, these data argue against widespread electrocardiographic screening for event prediction in asymptomatic PWH with low to moderate predicted 10‐year ASCVD risk.

This study is supported through National Institutes of Health grants U01HL123336 and 1UG3HL164285 to the Clinical Coordinating Center, and U01HL123339 and 1U24HL164284 to the Data Coordinating Center, as well as funding from Kowa Pharmaceuticals America, Inc., Gilead Sciences, and ViiV Healthcare. The National Institute of Allergy and Infectious Diseases supported this study through grants UM1 AI068636, which supports the AIDS Clinical Trials Group Leadership and Operations Center; and UM1 AI106701, which supports the AIDS Clinical Trials Group Laboratory Center. This work was also supported by the Nutrition Obesity Research Center at Harvard (P30DK040561 to S.K.G.). The views expressed in this article are those of the authors and do not necessarily represent the views of the National Heart, Lung, and Blood Institute or the National Institute of Allergy and Infectious Diseases; the National Institutes of Health; or the US Department of Health and Human Services.

The study investigators thank the study participants, site staff, and study‐associated personnel for their participation in the trial. In addition, the authors thank the following: the AIDS Clinical Trials Group for clinical site support; AIDS Clinical Trials Group Clinical Trials Specialists (Laura Moran, MPH, and Jhoanna Roa, MD) for protocol development and implementation support; the data management center, Frontier Science Foundation, for data support; the Center for Biostatistics in AIDS Research for statistical support; and the Community Advisory Board for input from the community.