Abstract 13612: Using Pre-Operative Cardiac Computed Tomography and Echocardiography to Predict New-Onset Atrial Fibrillation Following TAVR

Abstract

Introduction: New-onset atrial fibrillation (NOAF) following transcatheter aortic valve replacement (TAVR) has been associated with higher morbidity and mortality. Numerous studies have focused on clinical variables for NOAF prediction. In this study, we examine pre-operative cardiac computed tomography (cCT) and echocardiographic measurements for NOAF prediction following TAVR. Methods: This retrospective study includes 939 patients without pre-existing atrial fibrillation who underwent transfemoral TAVR between 2015 and 2020. Clinical data and standard pre-operative echocardiographic and cCT measurements were recorded. A machine learning method was used to derive left atrial volume and measures of epicardial adipose tissue from cCT. NOAF by 30-days was determined based on continuous electrocardiographic monitoring prior to hospital discharge and routine ECG at 2 and 4 weeks following TAVR. In a two-step analysis, random survival forest (RSF) was used to select variables of highest importance to NOAF prediction without assuming statistical models, followed by logistic regression analysis of the selected variables. Results: Among 939 patients, 47 (5%) experienced NOAF. Patients with NOAF had larger left atrial volume index (68.11 vs 56.96 cm 3 /kg/m 2 ; p<0.001) but similar epicardial adipose tissue volume (71.78 vs 70.98 cm 3 ; p>0.1) and epicardial adipose tissue attenuation (-81.16 vs -82.77 Hounsfield Units; p>0.1). Among 28 covariates selected by RSF for variable importance, 4 covariates were predictive of NOAF in the multivariable logistic regression, including sinotubular junction short axis >34mm [OR 5.687 (2.271-14.241); p=0.0002], maximum aortic valve pressure gradient [OR 1.037 (1.004-1.072); p=0.0266], mean aortic valve pressure gradient [OR 0.939 (0.894-0.986); p=0.0115], and body mass index < 22kg/m 2 [OR 0.331 (0.150-0.729); p=0.006]. Model concordance was 0.75. Conclusions: Pre-operative cardiac computed tomography and echocardiographic measurements including sinotubular junction enlargement and aortic valve pressure gradient predict new-onset atrial fibrillation following TAVR.