Background:
People with HIV (PWH) have a higher burden of cardiovascular (CV) disease than the general population; estimating this risk is an essential component of prevention. However, it is unknown how well the 2013 ACC/AHA guideline-recommended Pooled Cohort Equation (PCE) estimates risk among PWH globally. Leveraging the international REPRIEVE Trial, which prospectively adjudicated incident CV events, we compared observed vs. predicted event rates in PWH not taking statins.
Methods:
The REPRIEVE Trial used the PCE to determine eligibility of PWH at low-moderate CV risk for a statin primary prevention trial. We now assess discrimination and calibration of the PCE in those randomized to placebo (n=3869) as well as those randomized to statin but never starting treatment (n=24). To align with the median 5-year follow up in REPRIEVE, a 5-year risk score was recalculated for this analysis per established method, and follow-up beyond 5 years was censored. We limited outcomes to the specific CV events predicted by the PCE (hard MACE): CV death, myocardial infarction (MI), and stroke. We calculated the C-statistic, the observed: expected (O:E) event ratio and the Nam-D’Agostino goodness-of-fit (GND) statistic overall and in subgroups by race, natal sex, and Global Burden of Disease region. Small GND p-value indicates poor calibration.
Results:
Participants were mean age 50 years, 31% female, 65% nonwhite, with a median (Q1, Q3) 10-yr PCE risk of 4.5% (2.2, 7.1). Overall, discrimination was moderate (C-statistic 0.72) and calibration was good (O:E events, 84:81, ratio 1.03, GND P=0.81). However, calibration demonstrated over-prediction of risk outside High Income regions. When restricted to High Income regions, under-prediction (O:E ratio>1.0) was suggested among females (2.56) and Black or African American participants (1.66). See Figure, points above and below the line y=x for overall cohort calibration plot indicate relative under and overestimation, respectively.
Conclusions:
Among a global cohort of PWH with low-to-moderate traditional CV risk, the PCE was moderately effective to predict CV death, MI or stroke over 5 years but under-predicted events in females, Blacks or African Americans and participants from high-income regions. Performance in selected subgroups should be considered when using the PCE to guide prescribing statin therapy for CV prevention among PWH.
