Extra-pulmonary vein drivers as a potential mechanism of the antiarrhythmic drug-resistant atrial fibrillation after catheter ablation: Digital twin-guided virtual drug test

Background Accurate prediction of the most effective antiarrhythmic drug (AAD) following atrial fibrillation (AF) ablation offers substantial clinical value. Nevertheless, in vivo pharmacologic testing and driver mapping present significant technical and ethical challenges. Purpose To evaluate whether combined restitution and driver responses to Virtual AADs in digital twins predict recurrence after pulmonary vein isolation (PVI). Methods We included 226 patients (60.7±10.1 years; 23.9% female; persistent AF: 64.2%) who took AADs within 3 months after PVI due to early recurrence. Left atrial (LA) digital twins were reconstructed from electroanatomical maps acquired during AF ablation (>1,000 points). Under baseline and Virtual AAD conditions (amiodarone, dronedarone), high-density restitution was quantified by Smax (the maximal slope of the action potential duration restitution curve). After Virtual AF induction, we identified the stable phase singularities (PS) that persisted for ≥2 s within a 10–mm–diameter region. Stable PS area was expressed as the percentage of the LA surface area covered by stable PS regions. We assessed changes in mean Smax and stable PS area and their associations with Clinical 3-year atrial arrhythmia (AA) recurrence under AADs. Results In digital twin modeling, both stable PS presence (95.6% to 72.8%) and stable PS area (2.83% [0.76-6.39] to 0.83% [0.00-4.00]) were reduced under Virtual amiodarone and dronedarone conditions (both p<0.001). Less-reduced stable PS area with Virtual AADs (absolute reduction ≤1.4 percentage points [≤50% of the cohort median baseline stable PS area]) was associated with a higher clinical 3-year AA recurrence (log-rank p=0.018, aHR 1.70 [1.15-2.51], p=0.007) in the overall cohort. Less-reduced stable PS area predicted 3-year AA recurrence with an AUC of 0.707 (95% CI 0.639–0.775). Among the patients treated clinically with amiodarone (50.9%), Less-reduced stable PS area by Virtual amiodarone consistently predicted worse rhythm outcome (log-rank p=0.018). Although the overall mean Smax increased (1.03 [0.79-1.35] to 1.22 [0.94-1.61]) in both Virtual amiodarone and dronedarone conditions (both p<0.001), Virtual AAD-induced ΔSmax was not associated with the rhythm outcome of AAD use (log-rank p=0.420). Conclusion Among patients who required AADs for AA recurrence after PVI, limited suppression of stable PS area under Virtual AADs, but not restitution response, predicts 3-year Clinical AA recurrence. These findings suggest that persistent extra-PV PS drivers may contribute to AAD-resistant AA recurrence after PVI.