Abstract 4364578: Performance of Traditional Cardiovascular Risk Scores in Predicting Mortality Among Cancer Survivors

Background: Cardiovascular disease(CVD) is a leading cause of non-cancer death among cancer survivors, attributable to cardiotoxic therapies, systemic inflammation, and conventional cardiovascular risk factors. Prevailing risk prediction tools including ASCVD, PREVENT, and Framingham’s Score were developed in general population and may not reflect cancer specific risks. Hypothesis: We aimed to evaluate their predictive performance for CVD mortality in cancer survivors and identify optimal cutoffs for this high-risk population. Methods: We conducted a retrospective cohort study using NHANES 2001–2018, linked with mortality data, including adults aged 30–79 years with a prior cancer diagnosis and free of baseline cardiovascular disease. We applied PREVENT exclusion criteria to ensure cohort comparability. Cardiovascular mortality was the primary endpoint. Risk discrimination of ASCVD, PREVENT, and Framingham Score was evaluated using ROC analysis and AUC comparisons for traditional 7.5%, 20% high risk, and Youden-optimized thresholds. Results: Among 1,074 cancer survivors, conventional ASCVD thresholds (7.5% and 20%) showed suboptimal discrimination (AUCs: 0.572 and 0.661, respectively). In contrast, a Youden-optimized ASCVD cutoff (29.76%) significantly enhanced discrimination (AUC = 0.721, p < 0.001). PREVENT original threshold yielded moderate performance (AUC = 0.647), 20% cutoff (AUC = 0.661); a Youden-derived optimal cutoff (11.01%) modestly improved discrimination (AUC = 0.680, p = 0.3265). Framingham scores at standard thresholds demonstrated moderate discrimination (AUC = 0.564), with marginal improvement at a Youden-optimized cutoff (35.64%) (AUC = 0.683, p < 0.001). Optimized thresholds consistently outperformed conventional cutoffs, underscoring the necessity for recalibrated, cohort-specific risk stratification in cancer survivors. Conclusion: Standard cardiovascular risk prediction tools, including ASCVD, PREVENT, and Framingham's showed inadequate discrimination for cardiovascular mortality in cancer survivors, even after threshold optimization. Our analysis revealed that optimized, cohort-specific cutoffs improved predictive performance. Our findings underscore a critical gap and an imminent need for cardio-oncology risk prediction. It is imperative to develop specific models that integrate oncologic exposures and cardiovascular risk to improve long-term outcomes in this expanding cancer survivor population.