CMR-based analysis of atrial epicardial adipose tissue aids identification of a low-risk group for the presence of LA fibrosis

Abstract Introduction There is growing evidence that occurrence of atrial fibrillation (AF) is influenced by epicardial adipose tissue, which triggers left atrial (LA) remodeling through an inflammatory reaction, leading to dilatation and fibrosis. Identification of patients with advanced stage LA remodeling and fibrosis resulting in low voltage zones (LVZ) can contribute to individualized decision making in pulmonary vein isolation (PVI) treatment. Objective To establish a prediction model utilizing cardiac magnetic resonance (CMR) derived atrial epicardial adipose tissue quantification for the presence of LVZ during electroanatomic mapping that corresponds to LA fibrosis. Methods 66 patients admitted for primary AF ablation routinely underwent a preprocedural CMR scan. Dedicated adipose tissue enhancing sequences (2D fat/water separation) using short axis slices covering the LA were employed to create a 3D epicardial adipose tissue (EAT) model reconstruction. In addition, an LA angiography using gadolinium contrast was performed to create a 3D LA model. Model segmentation and volume quantification were performed with dedicated software (Figure 1). In the subsequent PVI an electroanatomic map of the LA was created to assess the presence of LVZ, defined by point measurements ≤0,5mV on the LA wall. Results General patient characteristics were 55% male gender, mean age 68 ± 9.8 years, median BMI 28 (IQR 26-31), median LVEF 60% (IQR 56-64), paroxysmal AF present in 38% of patients. LVZ was present in 37.9% patients during mapping (n=25). Patients were divided into LVZ positive (LVZ+) and LVZ negative (LVZ-) groups. LA volume (LAV) and LAV/BSA (LAVi) were higher in LVZ+ [145.3±32.7 ml and 75.6±17.9 ml/m² (n=25)] compared to LVZ- patients [120.8±31.7 ml and 58.1±14.8 ml/m² (n=41)], each p<0.05. LA-EAT volume (LA-EATV) and LA-EATV/BSA (LA-ETAVi) were higher in LVZ+ [64.2±23.5 ml and 33.0±11.3 ml/m² (n=20)] compared to LVZ- patients [52.7±16.7 ml and 25.2±7.8 ml/m² (n=38)], each p<0.05 (Figure 2a/b). A moderate correlation with the presence of LVZ is present for LAVi (r=0.47, p<0.001, n=66) and LA-EATVi (r=0.31, p=0.003, n=58). Both parameters are independent (r=0.194, p=0.145). A correlation with LVZ size can be described for LAVi (r=0.552, p<0.01) and LA-EATVi (r=0.303, p=0.05). A binomial logistic regression model using a combination of the two independent CMR-derived parameters LAVi and LA-EATVi (log odds (LVZ) = -6.691 + 0.054*LAVi + 0.086*LA-EATVi) was able to predict the presence of LVZ with an area under the ROC curve of 0.82 (p<0.001) and Nagelkerke's R² of 0.36 (Figure 2c). Overall accuracy in classification was 75.9% with a sensitivity of 55% and a specificity of 87%. Odds ratio for LAVi were 1.06 [1.02-1.10] and for LA-EATVi 1.09 [1.02-1.18]. Summary: This CMR-based combined prediction model utilizing LA volume and LA adipose tissue volume allows identification of low-risk patients for the presence of LA fibrosis.Figure 1:3D visualisation of structures