PREVENT vs pooled cohort equations for 10-year mortality prediction in symptomatic patients undergoing evaluation for suspected coronary artery disease
J Brendel, T Mayrhofer, N M Kerkovits, S Ersoezlue, M Kolossvary, B Alhanti, N J Pagitipati, S H Shah, M Ferencik, M T Lu, U Hoffmann, P S Douglas, B FoldynaAbstract
Background
The 2013 Pooled Cohort Equations (PCE) are widely used for 10-year atherosclerotic cardiovascular disease (ASCVD) risk estimation.1 The recently introduced PREVENT equations were designed to improve performance in contemporary populations.2 Comparative performance in symptomatic patients undergoing evaluation for suspected coronary artery disease (CAD) remains uncertain.
Objective
To compare discrimination, calibration, and risk stratification of PCE and PREVENT for long-term outcomes in symptomatic patients evaluated for suspected CAD.
Methods
We studied 9,355 participants from the PROMISE randomized clinical trial conducted at 193 North American sites. Median (IQR) follow-up was 10.6 (9.9-11.3) years (maximum 12.4 years). Outcomes were all-cause death and cardiovascular (CV) death. Model discrimination was assessed using Harrell’s c-statistics. Calibration was evaluated by comparing predicted and observed risks. Cox regression was used to estimate hazard ratios (HRs) across predefined risk categories (<5%, 5–<7.5%, 7.5–<20%, ≥20%); cumulative incidence by risk category is shown in Figure 2.
Results
PREVENT produced a substantially lower predicted 10-year risk than PCE (7.4±4.8% vs. 14.9±11.9%; p<0.001 for difference; Figure 1). During follow-up, 11.5% died, and 3.2% experienced CV death. PREVENT showed higher discrimination than PCE for all-cause death (C-statistic 0.702 vs 0.693, p=0.005) and CV death (0.722 vs 0.709, p=0.025). Calibration plots showed systematic overestimation of risk by PCE, particularly at higher predicted risk, whereas PREVENT showed closer agreement between predicted and observed CV mortality. Cumulative incidence curves (Figure 2) demonstrated clearer separation across increasing risk categories with PREVENT. For CV mortality, compared with <5% risk, PCE HRs (95% CI) were 2.17 (1.14–4.13), 2.90 (1.69–4.99), and 8.55 (5.04–14.51) across increasing categories, compared with 2.64 (1.72–4.05), 5.86 (4.08–8.41), and 20.92 (12.89–33.93) for PREVENT.
Conclusions
In symptomatic patients undergoing CAD evaluation, PREVENT demonstrated modestly improved discrimination, better calibration, and stronger risk differentiation compared with PCE for 10-year mortality.Figure-1_HistogramFigure-2_Incidence-Curves