Differential Performance of Polygenic Score for Coronary Artery Disease Based on Coronary Artery Calcium Between Men Living With and Without HIV.

BACKGROUND

Individuals with HIV have increased risk for coronary artery disease (CAD). However, the utility of the polygenic risk score (PRS) for estimating CAD risk (PRSCAD) is unknown in people living with HIV (PLWH). We evaluated the performance of PRSCAD on coronary artery calcium (CAC), a measure of subclinical atherosclerosis, by HIV serostatus.

METHODS

A PRSCAD was applied to 1071 men (PLWH, n=659; people living without HIV [PWOH], n=412) within MACS (Multicenter AIDS Cohort Study), then stratified by non-Hispanic White or Black race and genetic similarity to references (non-Hispanic White/genetically similar to European reference group, n=731; non-Hispanic Black/genetically similar to African reference group, n=340). PRSCAD performance for CAC, estimated by Agatston score (AS) from cardiac computed tomography scans, was evaluated independently and jointly with the American College of Cardiology/American Heart Association Pooled Cohort Equations by HIV serostatus using adjusted R2 for log-transformed CAC and area under the curve (AUC) for binary outcomes (absence/presence of CAC, absence of CAC [AS=0] versus low CAC [AS=1-100] or high CAC [AS >100]).

RESULTS

Predicting the presence/absence of CAC in non-Hispanic White/genetically similar to European reference group adding PRSCAD significantly improved the performance compared with the base model (AUCPLWH=0.738, AUCPWOH=0.759; P=0.015) but not the Pooled Cohort Equations (AUCPLWH=0.722, AUCPWOH=0.736; P=0.26). Performance was lower in PLWH than PWOH among outcomes, especially comparing high with absent AS (AUCPLWH=0.785, AUCPWOH=0.865; P=0.025). Joint modeling of PRSCAD and Pooled Cohort Equations decreased these gaps by HIV serostatus. PRSCAD performed worse in non-Hispanic Black/genetically similar to African reference groups compared with non-Hispanic White/genetically similar to European reference groups, with lower performance in PLWH (AUCPLWH=0.671, AUCPWOH=0.769; P=0.22).

CONCLUSIONS

The application of a PRSCAD to CAC showed lower performance for PLWH than PWOH men. Further research is needed to develop accurate risk prediction models with clinical utility in all patients.