Artificial intelligence applied to electrocardiographic images for the risk stratification of cancer therapeutics-related cardiac dysfunction

Abstract Background Traditional risk stratification strategies for cancer therapeutics-related cardiac dysfunction (CTRCD) rely on serial monitoring by specialized imaging, limiting their scalability and affordability. Purpose To examine an artificial intelligence (AI)-enhanced electrocardiographic (AI-ECG) surrogate for imaging-based risk biomarkers, and its association with CTRCD. Methods Between 2013-2023, across 5 hospitals of a large U.S.-based health system, we identified patients with breast cancer or non-Hodgkin lymphoma (NHL) who received anthracyclines (AC) and/or trastuzumab (TZM), as well as a (negative) control cohort of patients receiving immune checkpoint inhibitors (ICI). We deployed a validated AI model of left ventricular systolic dysfunction (LVSD) to ECG images (≥0.1=positive screen) (ref 1) and explored the association between the model’s predictions and: i) global longitudinal strain (GLS) measurements within 15 days; ii) future CTRCD (defined as a new cardiomyopathy, heart failure, or left ventricular ejection fraction (LVEF) decrease to <50%), as well as a more specific secondary outcome of LVEF decrease <40%. In a negative control analysis of patients receiving ICI, we correlated baseline AI-ECG LVSD predictions with downstream myocarditis. Results Higher AI-ECG LVSD predictions were associated with progressively worse GLS (-18% [IQR: -20 to -17%] for AI-ECG predictions <0.1, to -12% [-15 to -9%] for ≥0.5, p<0.001, n=7,271) (A). In 1,308 patients receiving AC/TZM (age 59 [IQR 49-67] years, 999 [76.4%] women, 80 [IQR 42-115] follow-up months) a positive baseline AI-ECG LVSD screen was associated with higher incidence of CTRCD (n=404 events, adjusted HR 2.22 [95%CI: 1.63-3.02]) and/or LVEF <40% (n=72 events, adjusted HR 4.76 [95%CI: 2.62-8.66]) (B). In a negative control analysis of 2,056 patients receiving ICI (age 65 [IQR 57-73] years, 913 (44.4%) women, follow-up 63 [IQR 28-99] months) AI-ECG predictions were not associated with downstream myocarditis (adjusted HR 1.36 [0.47-3.93]). Conclusions AI applied directly to baseline ECG images may stratify the risk of CTRCD associated with anthracycline or trastuzumab exposure.