Atrial fibrillation (AF) related atrial remodeling may persist during sinus rhythm (SR) and manifest as subtle P-wave abnormalities. Athletes exhibit increased AF risk and structural atrial adaptation. This study investigates whether P-wave features from SR ECG can identify latent AF electrophysiological signatures by distinguishing athletes with prior AF from those without.
23 male athletes with prior AF (AF-SR; 56.30 ± 5.63 y) and 20 male non-AF athletes (SR-SR; 56.55 ± 5.25 y) underwent 10-min resting 12-lead SR ECG (180 Hz). A 10-s segment per recording was analysed; P-waves were annotated on V3 and aligned across leads, yielding 328 beats. P-wave features were extracted and aggregated across beats; group differences were tested using non-parametric statistics with effect sizes. Classification was evaluated with logistic regression and random forest using five-fold cross-validation.
Compared with SR-SR controls, AF-SR athletes showed higher P-wave area variability (4.14 [3.80, 4.73] vs 5.94 [5.24, 6.34] mv.mv, p < 0.0001), P-wave energy variability (0.10 [0.085, 0.13] vs 0.20 [0.14, 0.22] mV², p = 0.001), P-wave amplitude variability (0.024 [0.022, 0.026] vs 0.035 [0.030, 0.038] mV, p = 0.001). Mean P-wave biphasic index (0.13 [0.10, 0.15] vs 0.09 [0.06, 0.10], p = 0.002). Random forest outperformed logistic regression (AUC 0.837 vs 0.815), with similar accuracy (0.77), higher F1 (0.81 vs 0.80) and sensitivity (0.83 vs 0.78), but lower specificity (0.70 vs 0.75).
P-wave phenotyping during SR distinguishes athletes with and without AF history. Subtle abnormalities in P-wave magnitude, variability, and morphology may reflect latent AF-related atrial substrate even when in SR. In addition, machine learning models demonstrated promising discrimination, supporting the value of P-wave analysis as a noninvasive approach for AF-risk indication in athletic populations.