EVALUATING THE ACCURACY OF THE PREVENT CALCULATOR COMPARED TO THE PCE IN ASSESSING ATHEROSCLEROTIC CARDIOVASCULAR DISEASE RISK: THE DALLAS HEART STUDY

Therapeutic Area

ASCVD/CVD Risk Assessment

Background

The predicting risk of cardiovascular disease events (PREVENT) calculator was recently (2023) developed as an updated cardiovascular disease risk calculator from the prior Pooled Cohort Equations (PCE) calculator. Few studies are available comparing the accuracy and implications on risk categorization of using the PREVENT vs. PCE calculator.

Methods

Participants from the Dallas Heart Study first phase (DHS1) aged 40 to 65 without known cardiovascular disease at baseline and with complete follow up data for atherosclerotic cardiovascular disease (ASCVD) events (fatal or non-fatal myocardial infarction or stroke) were included. Discrimination was assessed using the Harrell C-statistic. Calibration was assessed evaluating observed vs. predicted 10-year ASCVD risk across risk deciles using the Nam-D'Agostino χ2 test. Categorical net reclassification was performed by cross-tabulating risk estimates from PREVENT and PCE in those with and without ASCVD events. The predicted risk categories based on clinically relevant treatment thresholds were: <5%, <7.5% and ³7.5% 10-year ASCVD risk. Replication was performed in the Dallas Heart Study second phase (DHS2) cohort which was slightly more contemporary (enrollment 2009 vs. 2001).

Results

The DHS1 cohort comprised 1346 individuals, mean age 49.6 (±6.6) years, 42% male, 48% Black individuals. Applying the PREVENT and PCE calculators resulted in similar c-statistics (0.7291 vs. 0.7253). Both calculator risk estimates diverged from the observed risk deciles (p<0.0001 each) with PREVENT generally underestimating risk and PCE overestimating risk, particularly in the middle and higher deciles (Fig 1,2). Among the 170 individuals who experienced an ASCVD event, PREVENT incorrectly down-classified ASCVD risk compared with PCE in 70 (41%), and only up-classified risk in 3 (Fig 3). Among the 1176 who did not experience an event, PREVENT appropriately down-classified risk in 324 (28%), and up-classified risk in 6 (Fig 4). When evaluating the DHS2 cohort (n=1742, mean age 52 years), similar results were found.

Conclusions

In a large, multiethnic, population-based cohort, the PREVENT calculator had comparable discrimination of ASCVD events as the PCE. Both miscalibrated observed risk, with PREVENT underestimating and PCE overestimating risk. The lower estimates by PREVENT could result in fewer individuals recommended preventive therapies, reducing therapy burden but also potentially increasing ASCVD events.