CardioOnco-AI: AI-Empowered Cardiotoxicity Risk Prediction Among Breast Cancer Survivors Using Multi-Site Real-World Data

University Of Minnesota, Minneapolis MN

Investigators

Abstract

PROJECT SUMMARY Advances in cancer prevention, diagnosis, and treatment have dramatically improved long-term survival of those diagnosed with breast cancer. However, this success has been tempered by a parallel increased incidence of adverse outcomes and chronic conditions in breast cancer survivors, including cardiotoxicity due at least in part to cardiotoxic treatment regimens. Current evidence-based guidelines for preventing and controlling cardiotoxicity in breast cancer survivors are broad, and we lack clear guidance for assessing individualized risks of cardiovascular events. Existing cardiovascular disease (CVD) risk prediction models focus on the general population and rely only on a limited number of variables. Thus, there is a critical need to develop novel AI- powered informatics frameworks to build, maintain, and enhance models predicting cardiovascular risk among breast cancer survivors across diverse health systems. Responding to the RFA-FD-25-015, the objective of this application is to develop and validate a scalable, generalizable, and explainable AI-powered informatics framework, CardioOnco-AI, which innovates on the curation and modeling of real world data (RWD) for individualized prediction of cardiotoxicity among breast cancer survivors. Towards this objective, we propose the following specific aims: 1) Data curation - extract and derive cancer phenotype and non-medical determinant of health (nMDoH) variables to create research datasets for cardiotoxicity risk prediction through novel AI- empowered informatics solutions; 2) EHR-based predictive modeling - develop and evaluate novel cardiotoxicity prediction models for breast cancer survivors across two sites; and 3) Evaluation - assess the generalizability of CardioOnco-AI in two large geographically diverse RWD consortia. This project will deliver a novel, generalizable framework that integrates structured and unstructured RWD to improve cardiotoxicity risk prediction in breast cancer survivors. This work directly aligns with FDA’s regulatory science priorities and supports safer post- treatment survivorship care. The tools and methods developed will be adaptable to other sites, data environments, and oncology drug safety surveillance efforts.