Deep Learning-Derived Body Composition Analysis Predicts Long-Term Mortality After Transcatheter Aortic Valve Replacement.

OBJECTIVE: To examine the association between body composition metrics derived from preprocedural computed tomography (CT) angiography and all-cause mortality after transcatheter aortic valve replacement (TAVR). PATIENTS AND METHODS: We included patients who underwent TAVR between September 1, 2011 and November 30, 2023 at a single academic center. Skeletal muscle (SM), subcutaneous adipose tissue (SAT), visceral adipose tissue (VAT), and intermuscular adipose tissue areas (cm2), as well as SM index (SMI; cm2/m2), were quantified from CT angiography using a validated U-Net-based deep learning model. Associations between each parameter and 3-year all-cause mortality were assessed using multivariable Cox proportional hazards models adjusted for clinical covariates, with adjusted hazard ratios (aHRs) expressed per 1-SD increase. RESULTS: Among 2642 patients (median age, 80.0 years [interquartile range, 74.0-85.0 years]; 1572 were men [59.5%]), median follow-up was 2.8 years, and 74.8% survived to 3 years. Lower SM, SAT, VAT, and SMI (analyzed as continuous variables) were independently associated with higher 3-year all-cause mortality (SM: aHR, 0.831; 95% CI, 0.762-0.906; SAT: aHR, 0.847; 95% CI, 0.775-0.926; VAT: aHR, 0.826; 95% CI, 0.762-0.896; SMI: aHR, 0.832; 95% CI, 0.763-0.907; all P≤.001). Restricted cubic spline analysis showed increased mortality risk below threshold values of the following-SM<128 cm2, SAT<161 cm2, VAT<104 cm2, and SMI<41 cm2/m2; sex-specific thresholds were also derived. CONCLUSION: Reduced SM and adipose tissue reserves are independently associated with increased mortality after TAVR. Automated CT-derived body composition assessment may improve preoperative risk stratification and guide clinical decision making in TAVR candidates.