Objective Evaluate performance of the Predicting Risk of Cardiovascular Disease EVENTs (PREVENT) calculator in rheumatoid arthritis (RA). Methods Patients with RA were matched up to 10 controls on age, sex, and enrollment year using national Veterans Health Administration (VHA), Medicare, and National Death Index data (2006‐2020). Ten‐year estimated cardiovascular disease (CVD) risk was calculated using PREVENT. Calibration (standardized incidence ratio SIR; observed:predicted events) and discrimination (sensitivity, Harrel's C‐statistics) were compared between RA cases and controls. PREVENT performance was compared with the Pooled Cohorts Equations (PCE) for atherosclerotic CVD (ASCVD) prediction in RA, including net reclassification index (NRI) calculation. Results Among 30,687 RA and 231,752 non‐RA patients, 28,061 ASCVD and 13,851 heart failure (HF) outcomes were identified over >2 million person‐years. PREVENT underestimated overall CVD (SIR 1.83 1.79‐1.88), ASCVD (SIR 2.25 2.19‐2.32) and HF risk (SIR 1.41 1.36‐1.46) to a greater degree in RA compared to controls and exhibited poor sensitivity for ASCVD (61.9%) and HF (63.2%) development. PREVENT performance was poorer for ASCVD prediction compared to the PCE (SIR 1.38 1.34‐1.41; sensitivity 76.0%). NRI for PREVENT was modest (5.3%). Among 657 reclassified patients who experienced ASCVD, 626 were inappropriately reclassified as low/borderline risk. PREVENT performance significantly improved when including hemoglobin A1c (SIR Overall CVD 1.21 1.18‐1.24, ASCVD 1.45 1.41‐1.50, HF 0.79 0.76‐0.82; sensitivity ASCVD 80.3%). Conclusion PREVENT underestimates CVD risk in RA, consistent with suboptimal performance of existing risk calculators. Preferential use of PREVENT including hemoglobin A1c should be considered. Improving CVD risk stratification in RA remains a high priority.
Johnson et al. (Mon,) studied this question.