How does artificial intelligence-derived electrocardiographic age and sex predict recurrence outcomes after atrial fibrillation catheter ablation?

Abstract Aims We previously reported that the artificial intelligence (AI)-enabled electrocardiographic (ECG)-age gap and ECG-sex misclassification impact the rhythm outcome after atrial fibrillation catheter ablation (AFCA). To explore the mechanisms by which the ECG-age gap and sex-misclassification affect AFCA outcomes, we systemically evaluated the association between clinical, imaging, and electrophysiological indicators with ECG-age and sex. Methods We developed (1,730,222 ECGs from 662,246 participants) a residual network (ResNet)-based model for chronological age and biological sex prediction and validated it on independent multinational datasets. Then, we calculated ECG age and sex from a pre-procedural sinus rhythm ECG among 4,707 patients who underwent de-novo AFCA. We categorized pre-procedural AI-ECG age gap into two groups: aged-ECG (≥10 year) and normal ECG age (10 year) groups based on the mean absolute ECG age gap error in the validation datasets. ECG sex misclassification was defined as an ECG sex probability of more than 50% for the opposite sex. Results The ResNet-based AI-ECG age model successfully reproduced the chronological age on the training datasets (959,514 ECGs); CODE-15% (r=0.78) and MIMIC4 (r=0.69) cohorts. Compared with normal ECG age, patients with aged ECG were associated with persistent AF (age, sex adjusted odds ratio OR 1.25, 95% confidence interval CI 1.20-1.30), heart failure (OR 1.13, 95% CI 1.10-1.16), H2FPEF score (OR 1.44, 95% CI 1.27-1.62), left atrial (LA) voltage (OR 0.88, 95% CI 0.83-0.94), LA volume index (OR 95.38, 95% CI 25.74-353.50), and LA epicardial adipose tissue volume (OR 44.74, 95% CI 19.00-105.33). Patients with ECG sex misclassification were associated with persistent AF (OR 1.04, 95% CI 1.00-1.08) and heart failure (OR 1.04, 95% CI 1.01-1.07). Aged ECG and ECG sex misclassification had an additive effect on the association with persistent AF (OR 1.37, 95% CI 1.27-1.49), heart failure (OR 1.23, 95% CI 1.15-1.31), H2FPEF score (OR 1.73, 95% CI 1.32-2.25), and LA volume index (OR 172.05, 95% CI 11.03-2693.82). Conclusions Aged ECG and ECG sex misclassification reflects LA remodeling, AF chronicity, and comorbid heart failure, which might help in predicting rhythm outcomes after AFCA.