Abstract Body (Do not enter title and authors here): Background
The AHA recently developed the Predicting Risk of Cardiovascular Disease EVENTS (PREVENT) equations. PREVENT estimates risk for total cardiovascular disease (CVD: a composite of atherosclerotic cardiovascular disease [ASCVD] and heart failure [HF]), includes kidney function (eGFR) as a predictor, and did not include race as a predictor.
Aim
We evaluated model performance of the PREVENT equations and the pooled cohort equations (PCEs) in a large multi-ethnic sample of US Veterans and compared accuracy of PREVENT and PCEs among self-identified race and ethnic groups.
Methods
The Loma Linda VA IRB provided ethical approvals. Data were accessed through the Veterans Integrated Networking and Computing Infrastructure, with death data provided through the VA Information Resource Center. Individuals aged 30-79 with available risk factor data between 1/1/2007 and 12/31/2009 and without baseline CVD were eligible for inclusion. For PREVENT, we evaluated the base models predicting total CVD and ASCVD. For PCEs, we evaluated model performance predicting ASCVD. We assessed model discrimination with the C-statistic and calibration with observed vs. predicted curves. All analyses were stratified by self-reported race and ethnicity and utilized R statistical software.
Results
Among 2,500,291 adults, mean age was 60.3 years and 1.2% were Asian/Native Hawaiian/Pacific Islander (AANHPI), 5.2% Hispanic, 15% non-Hispanic Black (NHB), and 70% non-Hispanic White (NHW). Over a median follow-up of 8.7 years, the incident CVD rate was 16.5%. The C-statistic (95% CI) for PREVENT for total CVD was 0.70 (0.69-0.71) in AANHPI, 0.70 (0.69-0.70) in Hispanic, 0.68 (0.68-0.69) in NHB, and 0.65 (0.64-0.65) in NHW individuals. Similar C-statistics were observed for PREVENT-ASCVD and PCEs. Calibration curves demonstrated accurate prediction with PREVENT for CVD and ASCVD across the range of predicted risk. In contrast, PCEs overpredicted ASCVD at high predicted risk for all race and ethnic groups (Figure).
Conclusion
The AHA PREVENT equations more accurately and precisely estimates risk across race and ethnic groups. This demonstrates that removing race as a variable from outcomes prediction does not diminish predictive performance.