OBJECTIVES: 5-year survival rate for breast cancer is relatively high. Cancer survivors face an elevated risk of cardiac complications due to cancer treatments. In this study, machine-learning models were developed to predict cardiovascular events in female breast cancer patients undergoing adjuvant therapy.

METHODS: We selected breast cancer patients from a retrospective dataset of Taipei Medical University Clinical Research Database and Taiwan Cancer Registry between January 2004 and December 2020. Patients were monitored at the date of prescribed chemo- and/or -targeted therapies until cardiac adverse events occurred during a year. Clinical features were used, including demographics, comorbidities, medications, and lab values.

RESULTS: Random forest and gradient boosting achieved the highest AUC (around 0.80). Regarding other metrics, random forest had an overall better performance compared to gradient boosting (i.e., precision 0.26 versus 0.16; recall: 0.68 versus 0.82; and F1-score: 0.38 versus 0.28). The most important features were pre-existing cardiac disease, hypertension, cerebrovascular disease, beta blockers, and angiotensin II receptor blockers.

CONCLUSIONS: Prediction models were built for cardiac risk assessment among breast cancer patients. These prediction models offer potential approaches for cardio-oncology clinical practice. Further research is necessary to determine the feasibility of applying the tool in the clinical setting and explore whether this tool could improve care and outcomes.