Abstract 4360198: PREVENT Heart Failure Equations Underpredict Physician-Adjudicated Heart Failure Events Among People with HIV in the Multi-Center CNICS Cohort

Background: HIV infection is associated with increased risk of heart failure (HF). The American Heart Association recently developed the Predicting Risk of Cardiovascular Events (PREVENT) equations, which includes a calculator for 10-year risk of HF events. This calculator has not yet been validated among people with HIV (PWH) at risk for HF. Purpose: The purpose of this study was to evaluate the PREVENT 10-year HF risk prediction calibration and discrimination among a diverse, multicenter cohort study of PWH with physician-adjudicated heart failure events. Hypothesis: The PREVENT 10-year HF equations will have acceptable discrimination and strong calibration. Methods: We included adults ages 30-79 years old with HIV and without HF enrolled in the Center for AIDS Research Network of Integrated Clinical Systems (CNICS) Cohort at two sites (University of Washington and University of Alabama at Birmingham) that participated in physician adjudication of HF events from 2008 to 2021. We calculated predicted 10-year HF risk at baseline using the PREVENT 10-year base equation. HF events were adjudicated according to a standard protocol by two cardiologists independently, with discrepancies resolved by a third reviewer. Follow-up was until first HF event, death, loss to follow-up, or end of the study period. We used Harrell’s C-index to assess discrimination and the Kaplan-Meier approximation of the Greenwood-Nam-D’Agostino estimator (to account for variable follow-up time) and goodness-of-fit test to assess calibration. Results: We included 4,970 individuals with a mean age of 44 and 20% female ( Table ). The mean predicted 10-year risk of HF by PREVENT was 2.5% (SD: 4.9) [median=1.1% (IQR: 0.5-2.5)]. Over a median of 6.1 years of follow-up (IQR: 2.4-9.4), 125 individuals had an incident HF event. Discrimination by Harrell’s C-index was 0.760 (95% CI: 0.713-0.808). Using the Greenwood-Nam-D’Agostino estimator, the ratio of observed events to predicted events by the PREVENT HF equation over ten years was 1.47 (p<0.001). The calibration slope by decile of predicted risk was 1.035 for PREVENT ( Figure ). Conclusions: Among a cohort of people with HIV in the United States, the PREVENT 10-year HF equation has acceptable discrimination but is poorly calibrated, with the observed probability of events ~50% higher than the predicted risk across deciles of predicted risk.