Epicardial adipose tissue volume stratifies atrial fibrillation subtypes in reduced ejection fraction heart failure: a CT-based predictor

Background Epicardial adipose tissue (EAT) is involved in atrial fibrillation (AF) pathogenesis. Its specific role in stratifying AF subtypes among patients who also have reduced ejection fraction heart failure is not well defined. Objective We aimed to compare EAT volume between AF patients with and without reduced ejection fraction heart failure, and to assess its link with AF subtypes. Methods In this retrospective study, 224 AF patients (2018–2023) were classified into CAH group (combined AF and heart failure, LVEF < 50%, n = 90) and AF group (AF without heart failure, LVEF ≥50%, n = 134). Cardiac computed tomography (CT) was used to quantify EAT volume. Results Overall EAT volume was similar between the CAH and AF groups (152.3 ± 61.6 vs. 166.8 ± 73.2 mL, P = 0.12). However, patients with persistent AF (PsAF) had significantly larger EAT volumes than those with paroxysmal AF (PaAF) in the entire cohort (166.9 ± 71.0 vs. 145.1 ± 61.0 mL, P = 0.04). This difference was driven entirely by the CAH subgroup, where PsAF patients had greater EAT than PaAF patients (160.2 ± 59.5 vs. 127.8 ± 13.4 mL, P = 0.03); no such difference existed in the AF subgroup. EAT volume correlated with left atrial (LA) diameter overall ( r = 0.35, P < 0.05). While EAT volume predicted PsAF with high sensitivity but very low specificity in all patients (AUC = 0.60, sensitivity = 97.8%, specificity = 1.6%), its performance was more balanced within the CAH subgroup (AUC = 0.66, sensitivity = 86.8%, specificity = 45.5%). Notably, in this CAH group, LA diameter lost its predictive value for PsAF (AUC = 0.58, P = 0.28). Conclusion In patients with heart failure, increased epicardial adipose tissue (EAT) volume is linked to the progression from paroxysmal to persistent atrial fibrillation (AF). Measuring EAT volume may therefore serve as a useful imaging biomarker to identify those at higher risk of AF progression, which could help guide more personalized management.