Abstract 18584: Predictive Accuracy Of The Pooled Cohort Equations For Estimating Atherosclerotic Cardiovascular Disease Risk Among Women Exposed To Potentially Cardiotoxic Breast Cancer Therapies

Abstract

Introduction: Risk estimation is a critical component in clinical decision making for preventive cardiovascular care. Suboptimal identification and management of cardiovascular disease (CVD) risk, may contribute to excess CVD morbidity and mortality among breast cancer (BC) patients and survivors. This study aimed to assess the performance of the Pooled Cohort Equations (PCEs) in predicting atherosclerotic cardiovascular disease (ASCVD) events among a real-world population of adult women treated for BC. Methods: Using electronic medical record data collected between Jan. 1, 2011 and Jan. 31, 2023, this external validation study retrospectively identified a cohort of BC women, age 40-79, without history of ASCVD in the one year prior to initiation of a potentially cardiotoxic BC therapy. For all cohort members with cholesterol and blood pressure measurements recorded within one year prior to the index cancer therapy exposure (n=1,239), a 5-year ASCVD risk estimate was calculated. Predictive accuracy of the risk estimate was assessed using a Poisson model-based framework for calibration, including evaluation of calibration-in-the-large, the calibration slope, and calibration plots. To account for potential biases related to informative censoring, calibration was assessed among an inverse probability of censoring weighted (IPCW) pseudopopulation with stabilized and unstabilized weights, and a complete case cohort (n=306). Results: Among the IPCW pseudopopulation, calibration-in-the-large, expressed as a standardized incidence ratio, was 1.85 (95% CI: 1.73, 1.98), indicating that, on average, women in this subpopulation experience a significantly higher number of ASCVD events than expected from baseline PCE risk estimates. The calibration slope, estimated at 0.55 (95% CI: 0.45, 0.66), and evaluation of calibration plots, further indicate particularly poor performance of the PCEs at low and high levels of estimated risk. Findings were consistent across the assessed populations. Conclusions: When calculated prior to initiation of a potentially cardiotoxic BC therapy, the PCEs significantly underestimate 5-year ASCVD risk for BC women, indicating this commonly used ASCVD risk estimate is poorly calibrated for this subpopulation.