Comparative performance of the two pooled cohort equations for predicting atherosclerotic cardiovascular disease

Area of Science:

• Cardiology
• Epidemiology
• Health Informatics

Background:

• Multivariable algorithms, such as the AHA/ACC Pooled Cohort Equations (PCE), are used to predict atherosclerotic cardiovascular disease (ASCVD) risk.
• A systematic overestimation of risk was identified with the original PCE, leading to the proposal of a revised PCE.
• Accurate ASCVD risk prediction is crucial for identifying high-risk patients and guiding preventative strategies.

Purpose of the Study:

• To compare the accuracy of the original and revised PCE in predicting ASCVD risk.
• To evaluate the performance of both PCE models in a large, real-world patient sample.
• To examine potential effect modification by race and sex on PCE accuracy.

Main Methods:

• Retrospective cohort study of 20,843 patients aged 40-75 without prior ASCVD.
• Comparison of model fit, calibration, and discrimination using metrics like Bayesian Information Criterion (BIC), Hosmer-Lemeshow test, AUC, Brier score, and precision-recall analysis.
• Analysis of race and sex subgroups to assess effect modification.

Main Results:

• Both PCE models demonstrated poor calibration (Hosmer-Lemeshow χ² > 20; p < 0.05) and discrimination (AUC < 0.7).
• The revised PCE showed no significant improvement in discrimination compared to the original PCE (AUC: 0.677 vs 0.679; p = 0.357).
• The original PCE indicated stronger positive risk prediction (ΔBIC > 10) than the revised PCE, despite overlapping calibration curves.

Conclusions:

• Both original and revised PCE exhibit poor calibration and discrimination for ASCVD risk in a real-world patient population.
• The revised PCE did not improve the accuracy of ASCVD risk assessment compared to the original PCE.
• Findings suggest limitations in current PCE models for accurately identifying high-risk individuals for ASCVD.