Performance of PREVENT and pooled cohort equations for predicting 10-Year ASCVD risk in the UK Biobank
Matthew Ambrosio, Pamela L. Alebna, Terence K. Lee, Alec Friedman, Nicholas Chew, Carolyn A. Burns, Phillip A. Duncan, W. Gregory Hundley, Le Kang, Anurag Mehta
Abstract
Background The Pooled Cohort Equations (PCE) were created in 2013 to assess ASCVD risk in primary prevention. In 2023 the American Heart Association published the PREVENT equations to assess the risk of cardiovascular disease in primary prevention. The comparative performance of PCE and PREVENT for predicting 10-year ASCVD risk has not been evaluated in an external large-scale epidemiologic cohort. Methods The study population includes participants of the UK Biobank who were free of clinical cardiovascular disease. 10-year ASCVD risk was calculated using the PCE and PREVENT equations. Harrel’s C-Statistics and delta C-Statistics were calculated for males and females to evaluate risk discrimination. Predicted 10-year risks were divided into deciles as well as risk groups for each equation and stratified by sex to compare predicted risk versus observed risk within each risk group, with calibration slopes calculated by decile. Sensitivity and specificity were also analyzed to assess statin eligibility. Results The final cohort was 368,125 individuals ages 40-73 (mean age 56.2, 54.7% female, 94.0% white). The C-statistics for PCE were 0.729 (0.722-0.736) for females and 0.688 (0.683-0.693) for males; C-Statistics for PREVENT were 0.728 (0.721-0.735) for females and 0.687 (0.682-0.692) for males, with delta C-Statistics being 0.001 (p=0.87) for females and 0.001 (p=0.82) for males. Predicted risks were closer to observed risks for PREVENT as compared to PCE, and PREVENT tended to estimate lower risk (mean risk of 4.6% compared to 8.3% for PCE). PREVENT demonstrated higher specificity but lower sensitivity than PCE using the current 7.5% risk threshold for statin eligibility, and a Youden index of 4.5% risk was found for PREVENT. Conclusions There is no significant difference in 10-year ASCVD risk discrimination between PCE and PREVENT equations. However, the PREVENT equations demonstrate better calibration in the UK Biobank.