AI-Powered ECG Predicts Heart Risk Before Cancer Treatment in 31,000 Patients
A large retrospective study tests whether AI-enhanced ECGs can flag cardiac risk before chemotherapy, potentially protecting cancer patients from heart damage.
Summary
Chemotherapy drugs like anthracyclines and immune checkpoint inhibitors can damage the heart, sometimes causing life-threatening cardiac dysfunction. This completed study from Severance Hospital in South Korea examined over 31,000 cancer patients to determine whether an AI-enhanced electrocardiogram — taken before chemotherapy begins — could predict who is most at risk for developing cancer therapy-related cardiac dysfunction (CTRCD). Researchers also tracked whether changes in AI-ECG scores between pre- and post-chemotherapy readings could serve as early warning signals. The study compared AI-ECG performance against existing risk models, looking at outcomes including heart failure, severely reduced heart function, and death within 12 to 24 months of treatment. If validated, this approach could offer a simple, low-cost screening tool to personalize cardiac monitoring during cancer care.
Detailed Summary
Cancer treatments save lives, but they can silently damage the heart. Anthracyclines, HER2 inhibitors, and immune checkpoint inhibitors are among the most effective cancer therapies available — and among the most cardiotoxic. Cancer therapy-related cardiac dysfunction (CTRCD) affects a meaningful proportion of patients and can lead to heart failure, reduced quality of life, and even death. Early, accurate risk stratification before treatment begins could allow clinicians to tailor monitoring and preventive strategies.
This large retrospective study from Yonsei University's Severance Hospital enrolled 31,486 adult cancer patients treated between May 2006 and November 2022. All participants received at least one of three cardiotoxic regimens — anthracyclines, HER2-targeted agents, or immune checkpoint inhibitors — and had an ECG performed within 90 days before chemotherapy. A subset with a post-chemotherapy ECG within 90 days was included in trajectory analysis to evaluate whether changes in AI-derived ECG scores could predict subsequent cardiac events.
The primary outcome was the development of CTRCD within 12 months after the last treatment cycle, and no more than 24 months after the first. Secondary outcomes included severe CTRCD, defined as left ventricular ejection fraction below 40%, and all-cause mortality. Patients with pre-existing heart failure, cardiomyopathy, or myocarditis, or baseline LVEF under 40%, were excluded.
The study's scale — over 31,000 patients across nearly 17 years — positions it as one of the most comprehensive evaluations of AI-ECG for cardio-oncology to date. If AI-ECG proves accurate, it could transform pre-treatment cardiac screening by offering a fast, inexpensive, and widely available alternative to more resource-intensive imaging.
Caveats apply: this is a retrospective, single-center observational study, limiting generalizability. No results have yet been published beyond the abstract, so predictive performance metrics remain unknown. The summary is based on the abstract only.
Key Findings
- AI-enhanced ECG was tested as a pre-chemotherapy tool to predict cardiac dysfunction risk in 31,486 patients.
- Study covered three major cardiotoxic regimens: anthracyclines, HER2 inhibitors, and immune checkpoint inhibitors.
- Changes in AI-ECG scores between pre- and post-chemo readings were evaluated as early cardiac warning signals.
- Primary outcome was CTRCD development within 12–24 months; secondary outcomes included severe LV dysfunction and death.
- If validated, AI-ECG could offer a simple, low-cost alternative to echocardiography for cardiac risk screening.
Methodology
Retrospective observational cohort study of 31,486 adult cancer patients at a single academic medical center in South Korea, spanning May 2006 to December 2022. Participants required a baseline ECG within 90 days before chemotherapy and a baseline LVEF ≥40%. A subset with post-chemotherapy ECG data was included for trajectory analysis.
Study Limitations
This is a retrospective, single-center study conducted at a Korean academic hospital, which may limit generalizability to other populations and health systems. No results data are available beyond the abstract, so the predictive accuracy and clinical utility of AI-ECG cannot yet be assessed. The summary is based on the abstract only, as the full study data have not been publicly reported.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.