Abstract 4136554: Comparison of short- and long-term atherosclerotic cardiovascular disease risk assessment tools in US young adults

Abstract

Background: In 2023, the AHA published the PREVENT equations for estimating atherosclerotic cardiovascular disease (ASCVD) risk in adults aged 30-79 years. Research Questions: In young adults aged 20-39 years, does PREVENT improve risk prediction for 10- and 30-year ASCVD compared with existing risk assessment tools recommended in the current US guidelines (i.e., Pooled Cohort Equations [PCEs] and Pencina et al. equations)? Aims: To compare the performance of PREVENT vs. PCEs in predicting 10-year ASCVD, and PREVENT vs. Pencina equations in predicting 30-year ASCVD in young adults. Methods: We analyzed data from two complementary sources: (1) pooled data from two large cohorts: Coronary Artery Risk Development in Young Adults (CARDIA) and Framingham Heart Study (FHS; including the Offspring, Third Generation, Omni 1, and Omni 2 cohorts), and (2) electronic health records from Kaiser Permanente Southern California (KPSC). We included adults aged 20-39 years without a history of ASCVD at baseline. The outcome was incident ASCVD (defined as myocardial infarction, fatal coronary heart disease, fatal and nonfatal stroke) at 10 or 30 years. Model discrimination (Harrell’s C) and mean calibration (estimated as the ratio of predicted to observed event rates) were calculated for the overall population and stratified by sex and race/ethnicity. Results: We included 7,606 young adults (mean age 29 years, 53% female, 30% Black) from the pooled cohorts, and 284,667 (mean age 32 years, 61% female, 8% Black, 46% Hispanic) from KPSC. When predicting 10-year risk, PREVENT improved discrimination in both the pooled cohort (ΔHarrell’s C=0.052; 95% CI: 0.014, 0.095) and KPSC (ΔHarrell’s C=0.039; 95% CI: 0.028, 0.049) compared with the PCEs. PREVENT had good calibration (mean calibration ranged from 0.77 to 1.54), whereas the PCEs overestimated 10-year risk (mean calibration ranged from 1.99 to 4.82). When predicting 30-year risk, discrimination was similar for PREVENT and Pencina equations, but both algorithms underestimated 30-year risk with PREVENT showing worse calibration (mean calibration 0.61). Conclusion: PREVENT improved 10-year ASCVD risk prediction in young adults compared to the PCEs but underestimated 30-year risk.