Abstract 4139274: Implications of Electrocardiographic-Based Artificial Intelligence on Patients with Exercise-Induced Cardiomyopathy

Background: Electrocardiography-based artificial intelligence (AI-ECG) validated models that detect cardiac disease are increasingly being applied in clinical practice. The utility of such tools to detect insidious cardiac disease in patients with normal baseline left ventricular ejection fraction (LVEF) who paradoxically develop reduced LVEF during exercise stress echocardiography (ESE), and do not have coronary artery disease (CAD) or hypertension, is unknown. Hypothesis: AI-ECG is useful to diagnose insidious cardiac disease and predict clinical outcomes in patients with exercise-induced cardiomyopathy. Aims: To assess the utility of AI ECG in patients with exercise-induced cardiomyopathy. Methods: Among all ESE performed between January 2003 and December 2022, patients without a hypertensive response to exercise and without CAD (confirmed by coronary angiography within 90 days after ESE), with resting LVEF ≥50% and a paradoxical ≥5% LVEF decrease during ESE were identified. A previously validated AI-ECG algorithm that predicts atrial fibrillation (AF), reduced LVEF, and cardiac amyloidosis (CA) was applied to the baseline ECG closest to the time of ESE. The predicted probability of AF, reduced LVEF, and CA if above the published thresholds, was determined. Results: There were 134 patients with exercise-induced cardiomyopathy who were identified. The mean age of this cohort was 66±10 years, 76% were women and 16% had AF at baseline. Mean LVEF was 58±4% at rest and 43±4% at peak stress. The median follow-up period was 6.8 years (IQR 3.0-12.2). Among patients without a baseline history of AF (n=112), AI-ECG identified 29% with a significant probability of AF, which was associated with a subsequent AF diagnosis at univariable analysis (HR 2.505, 95%CI 1.016-6.177, p=0.046). There were 10 patients with an AI-ECG prediction of reduced LVEF among whom 3 were subsequently hospitalized with HF. AI-ECG was positive for CA in 8 and 3 amongst these had subsequent HF hospitalizations (Table). Conclusions: Baseline AI-ECG may help predict subsequent AF and diagnose preclinical amyloid cardiomyopathy in patients who by resting echocardiography do not seem to have cardiac disease but with exercise develop reduced LVEF.