Abstract 13572: Electrocardiographic Aging Derived From Artificial Intelligence is Associated With the Risk of Early- and New-Onset Atrial Fibrillation: A Multi-National Cohort Study

Background: The application of artificial intelligence (AI) algorithms to ECG provides promising age prediction methods. We investigated whether the age discrepancy between AI-predicted age from ECG (AI-ECG age) and chronological age, known as the AI-ECG age gap or electrocardiographic aging (ECG-aging), could predict atrial fibrillation (AF) risk.

Methods: We developed an AI-ECG age prediction model using a single-center dataset (1,533,042 ECGs from 689,639 participants) and validated it using five independent, multi-national datasets (637,177 ECGs from 230,838 participants). The AI-ECG age gap was calculated in two cohorts from South Korea and the UK. Based on this age gap, participants were classified into two study groups: Normal ECG-aging (Normal EA, age gap <7 years) and Early ECG-aging (Early EA, age gap ≥7 years). We assessed the predictive capability of ECG-aging for early- and new-onset AF risk.

Results: In two cohorts followed up for 5.9±3.7 and 3.0±1.6 years with 37,208 and 40,989 participants, respectively, the mean AI-ECG age with mean age gap was 47.4±12.4 (-0.1±6.0) and 68.4±7.8 (4.7±8.7) years, respectively. The early EA group was associated with increased risk of new-onset AF with HR of 1.86 (95% confidence interval, 1.46-2.38) and 1.88 (1.54-2.29), and early-onset AF with OR of 2.23 (1.79-2.77) and 1.52 (1.34-1.73), compared to the normal EA group in each cohort, respectively. Furthermore, there was an increased risk of both early- and new-onset AF with an increasing AI-ECG age gap.

Conclusions: ECG-aging derived from the AI model was associated with the risk of early- and new-onset AF, indicating its potential as a risk predictor for AF in primary prevention.