Abstract 4362916: Performance of the PREVENT CVD risk prediction score in breast cancer survivors

Abstract

Background: Breast cancer survivors face an increased risk of cardiovascular disease (CVD) due to shared risk factors and cardiotoxic treatments such as anthracyclines and trastuzumab. The performance of CVD risk prediction models among breast cancer survivors is understudied. Goals: To evaluate the performance of the American Heart Association PREVENT equations in breast cancer survivors and determine if recalibration or inclusion of cardiotoxic treatment enhances prediction while considering the effect of age at breast cancer diagnosis. Methods: We included 10,403 women diagnosed with first primary unilateral breast cancer between 2002 and 2019 within the NCI-Kaiser Permanente (KP) Breast Cancer Survivors Cohort (KP Colorado, Georgia, and Washington members aged 30-79 years, survived ≥1 year without second cancer or prevalent CVD). PREVENT risks for total CVD, heart failure (HF), coronary heart disease (CHD), and stroke were calculated using CVD risk factors at breast cancer diagnosis. The primary outcome was total CVD; deaths from other causes were treated as competing events. Discrimination was evaluated using time-dependent area under the receiver operating characteristic curve (AUC t ). Calibration was assessed by the estimated-to-observed risk ratio (E/O). We evaluated the performance of 3 models: PREVENT, recalibrated PREVENT and modified PREVENT which included age and chemotherapy. Results: Over a median follow-up of 4.5 years, 734 women developed CVD events. The AUC t (95% CI) of PREVENT equations were 0.74 (0.71-0.77) for CVD, 0.76 (0.73-0.79) for HF, 0.75 (0.71-0.78) for CHD, and 0.79 (0.76- 0.82) for stroke.The PREVENT equations markedly underestimated the risk of total CVD (E/O 0.56; 95%CI 0.51 to 0.61), HF (E/O 0.70; 0.62 to 0.78), CHD (E/O 0.34; 0.30 to 0.38), and stroke (E/O 0.40; 0.35 to 0.45). As shown for total CVD (Figure), recalibration to the baseline risk of the study population improved calibration, but underestimation persisted in patients aged <45 years (E/O 0.47; 0.31 to 0.72) and those treated with anthracyclines and/or trastuzumab (E/O 0.71; 0.61 to 0.82). The modified PREVENT (Figure) was well calibrated. Conclusion: The PREVENT equations underestimate CVD risk in breast cancer survivors, especially in young patients and those treated with anthracyclines and/or trastuzumab. Our findings have important implications for tailored CVD prevention in breast cancer survivors.