DNN-derived electrocardiographic age is associated with the atrial fibrillation risk in a Chinese population

Objective To assess whether DNN-derived electrocardiographic age (ECG-age) and its difference from chronological age (Δage) are associated with odds of atrial fibrillation (AF) in a Chinese population. Methods In a 1:1 sex-matched case control study, 1,574 patients with AF from Zhongda Hospital and 1,574 community controls from Nanjing were included. ECG-age was estimated using a validated deep neural network trained on the Brazilian CODE cohort. Participants were classified as accelerated, normal, or decelerated aging if Δage was greater than, within, or less than the model’s mean absolute error (MAE). Logistic regression and restricted cubic spline (RCS) models assessed associations between Δage and odds of AF. Results Mean ECG-age exceeded chronological age in both groups (AF: 76.68 ± 7.65 vs. 75.30 ± 10.37 years; controls: 64.21 ± 8.26 vs. 63.35 ± 7.55 years). Logistic regression analysis revealed that each 5-year increased in Δage was associated with a 19.5% increased risk of AF (OR = 1.195, 95% CI: 1.117–1.279). Accelerated aging was associated with a 79.9% increased risk (OR = 1.799, 95% CI: 1.377–2.354), whereas decelerated aging conferred a 32.5% decreased risk of AF (OR = 0.675, 95% CI: 0.481–0.894). RCS analysis demonstrated a U-shaped nonlinear association between Δage and odds of AF (P for nonlinear < 0.001). Conclusions DNN-derived ECG-age is highly correlated with chronological age in AF, and Δage is also a novel indicator of cardiac aging. Accelerated aging increases odds of AF. Given wide availability and economic ECG, ECG-age could be a promising AI-based novel biomarker for odds of AF in the clinical practice.