Abstract 47: Calibration and Discrimination of PREVENT vs PCE in Hispanics/Latinos across Disaggregated Background Groups, Self-Reported Race, and Genetic Ancestry: The Hispanic Community Health Study/Study of Latinos (HCHS/SOL)

Abstract

Background: Unlike the race-specific pooled cohort equations (PCE) for atherosclerotic cardiovascular disease (ASCVD) risk prediction, PREVENT is a race-neutral tool but Hispanics/Latinos were underrepresented in its derivation. We evaluated calibration/discrimination of PREVENT vs PCE in HCHS/SOL, a large and diverse population-based cohort of US Hispanics/Latinos. Methods: A total of 10,927 HCHS/SOL participants met PREVENT-ASCVD and 5,416 met PCE criteria, Figure 1 . Ten-year ASCVD risk (2008–2019) was estimated with base PREVENT-ASCVD and non-Hispanic Black (NHB), non-Hispanic White (NHW) PCEs. ASCVD (incident myocardial infarction and stroke) were adjudicated from medical records. Mean time to first ASCVD event was 9.6 years (177 events). Hispanic/Latino background groups and race were self-reported. Genetic ancestry proportions (European, African, Amerindian) were estimated using ADMIXTURE for 6,802 PREVENT and 3,329 PCE eligible participants, Figure 1 , and dichotomized by median. Observed ASCVD risk was estimated using Kaplan–Meier analysis. Calibration was evaluated by predicted-to-observed (P/O) ratios, and discrimination by Harrell’s C-statistics. All analyses accounted for complex survey design. Results: Among baseline ASCVD-free adults (mean age 47.4, 52.7% female), the observed 10-year ASCVD event rate was 1.6%. PREVENT predicted event rate was 3.4% (P/O 2.1). PCE had higher predicted event rate than PREVENT, regardless of race-specific equation used ( Figure 2, Table 1 ) . The PREVENT calibrated most closely among adults of Cuban and Dominican descent, while PCE-NHW showed slightly better calibration among Puerto Ricans. PREVENT outperformed PCE across all self-reported race categories except among self-reported Black participants where calibration was paradoxically closer with NHW PCE. Across genetic ancestry, PREVENT performed best with lower African ancestry, whereas NHB and NHW PCEs performed best with lower European ancestry. PREVENT and PCE had higher overestimation among Hispanics/Latino adults with greater Amerindian ancestry. PREVENT discrimination was moderate-to-high overall (C-statistic 0.78–0.89), Table 1. Conclusions: PREVENT improved ASCVD risk prediction vs PCE across Hispanic/Latino backgrounds and most race categories, except self-reported Black participants. Both equations overestimated risk among Hispanics/Latinos with higher Amerindian ancestry. ASCVD risk prediction equations may not apply uniformly across Hispanic/Latino populations.