Left atrial wall thickness measured by a machine learning method predicts AF recurrence after pulmonary vein isolation

Abstract

Abstract Background Left atrial (LA) remodeling plays a significant role in the progression of atrial fibrillation (AF). Although LA wall thickness (LAWT) has emerged as an indicator of structural remodeling, its impact on AF outcomes remains unclear. We aimed to determine the association between LAWT and AF recurrence after pulmonary vein isolation (PVI). Methods Single-center registry of patients enrolled for radiofrequency PVI from 2016 to 2018. In all cases, a pre-ablation CT scan was performed within less than 48 hours. Mean LAWT was retrospectively determined by a semi-automated machine learning method with minimal human intervention (ADAS 3D®). Additionally, regional tissue thickness was assessed in four locations: roof and inferior, posterior, and anterior walls. In a subgroup of patients, a pre-ablation cardiac magnetic resonance (CMR) was also performed within the same week. LA functional parameters and fibrosis, using 3D delayed gadolinium enhancement, were analyzed. The primary endpoint was AF recurrence after a 3-month blanking period. Results A total of 439 patients (mean age 61±12years, 62% male, 78% with paroxysmal AF) were included. The mean LAWT was 1.4±0.2mm (ranging from 0.9 to 1.9mm). Software processing duration was 8.2±0.4 min, and the mean human input time was 1.3±0.1 min. There was no correlation between LAWT and CT-derived LA-indexed volume (Spearman –0.01, p=0.845). During a median follow-up of 5.8 (IQR 4.9–6.6) years, 238 patients (54%) had an AF relapse. LAWT was an independent predictor of recurrence, after adjusting for known confounders including age, non-paroxysmal AF, LA-indexed volume, and chronic kidney disease (adjusted HR 6.49 [95% CI 2.70-15.49], p<0.001). AF recurrence rates were 11%, 15%, and 21%/year across terciles of increasing LAWT (log-rank p<0.001) – Figure. The posterior LAWT revealed the strongest association with the study endpoint (HR 1.93 [95% CI 1.24-3.02], p=0.004). In the cohort of 62 patients with both CT and CMR, LAWT showed weak correlations with LA ejection fraction and LA coupling index (Spearman <0.25; p=0.054 and p=0.093, respectively), and a moderate correlation with LA fibrosis (Spearman 0.468; p<0.001). Conclusions Mean LAWT, easily assessed by commercially available machine learning software, is an independent predictor of AF recurrence after PVI in the long-term follow-up. This association is mainly driven by the posterior LAWT. Whether patients with increased LAWT should receive tailored therapy deserves further investigation.Figure