Structural Cardiac Remodeling in Atrial Fibrillation

This study sought to evaluate preablation computed tomography angiography (CTA) for atrial and epicardial features to predict atrial fibrillation (AF) recurrence after ablation.Structural atrial remodeling is a process associated with occurrence or persistence of AF. Different anatomical imaging features have been proposed to influence atrial remodeling both negatively and positively as substrate for AF.Patients with nonvalvular AF underwent cardiac CTA before pulmonary vein isolation at 2 high-volume centers. Left atrial (LA) and right atrial volumes, LA wall thickness (LAWT), and epicardial adipose tissue volume and attenuation were evaluated. Additional subanalyses of electroanatomical maps were made. Follow-up was performed for at least 12 months, including subanalysis of repeated cardiac CTA studies. Interrater variability was assessed.Of 732 patients, 270 (36.9%) had AF recurrence after a mean of 7 months. CT analysis revealed larger indexed LA volume (47.3 mL/m2 vs 43.6 mL/m2; P = 0.0001) and higher mean anterior (1.91 mm vs 1.65 mm; P < 0.0001) and posterior (1.61 mm vs 1.39 mm; P = 0.001) LAWT in patients with AF recurrence. Epicardial adipose tissue volume in patients with AF recurrence was higher (144.5 mm³ vs 128.5 mm³; P < 0.0001) and further progressed significantly in a subset of 85 patients after 2 years (+11.8 mm2 vs -3.5 mm2; P = 0.041). Attenuation levels were lower, indicating a higher lipid component associated with AF recurrence (-69.1 HU vs -67.5 HU; P = 0.001). A total of 103 atrial voltage maps were highly predictive of AF recurrence and showed good discriminatory power for patients with low voltage >50% and LAWT (1.55 ± 0.5 mm vs 1.81 ± 0.6 mm; P = 0.032). Net reclassification improvement (NRI) showed a significant incremental benefit (NRI = 0.279; P < 0.0001) when adding LAWT to established risk models.Atrial wall thickness, epicardial fat volume, and attenuation are associated with AF recurrence in patients undergoing ablation therapy.