Deep learning-derived abdominal adiposity phenotypes and 28-day mortality in sepsis.
Authors
Affiliations (6)
Affiliations (6)
- Transplant Research Center, Research Institute for Convergence of Biomedical Science and Technology, Pusan National University Yangsan Hospital, Yangsan, South Korea.
- Division of Allergy, Pulmonary and Critical Care Medicine, Department of Internal Medicine, Pusan National University School of Medicine, Yangsan, South Korea.
- Department of Convergence Medical Sciences, Pusan National University School of Medicine, Yangsan, South Korea.
- Department of Statistics, Changwon National University, Changwon-si, South Korea.
- Transplant Research Center, Research Institute for Convergence of Biomedical Science and Technology, Pusan National University Yangsan Hospital, Yangsan, South Korea. [email protected].
- Division of Allergy, Pulmonary and Critical Care Medicine, Department of Internal Medicine, Pusan National University School of Medicine, Yangsan, South Korea. [email protected].
Abstract
Body mass index does not distinguish visceral from subcutaneous adiposity, which may have different prognostic implications during sepsis. We used automated computed tomography (CT) analysis to evaluate the associations of abdominal adipose tissue distribution with 28-day mortality. In this single-center retrospective cohort study of 1107 adults with sepsis, abdominal CT scans were analyzed using a pretrained nnU-Net-based segmentation pipeline. Subcutaneous adipose tissue cross-sectional area (SAT CSA), visceral adipose tissue (VAT) CSA, the VAT/SAT ratio, adipose tissue attenuation, and psoas muscle measures were quantified at L3-L4. Cox proportional hazards models estimated hazard ratios (HRs) per 1-standard deviation (SD) increase using clinically selected covariate adjustment. Exploratory BMI-stratified models included formal interaction testing. Among 1107 patients, 300 (27.1%) died within 28 days during the index hospitalization. Obesity (BMI ≥ 25 kg/m²) was associated with lower mortality (adjusted HR 0.62; 95% CI, 0.41-0.94, P = 0.023), as was greater SAT CSA (adjusted HR 0.83; 95% CI, 0.69-1.00, P = 0.048). A higher VAT/SAT ratio was associated with higher mortality (adjusted HR 1.14; 95% CI, 1.01-1.28, P = 0.031). In exploratory BMI-stratified analyses, greater SAT CSA was associated with lower mortality among underweight patients (adjusted HR 0.37; 95% CI, 0.16-0.83), and a higher VAT/SAT ratio was associated with higher mortality among patients with obesity (adjusted HR 1.48; 95% CI, 1.18-1.85); however, interactions with BMI category were not significant (P for interaction = 0.145 and 0.127, respectively). Among psoas muscle measures, only a higher low-attenuation muscle proportion remained associated with mortality after adjustment (adjusted HR 1.14; 95% CI, 1.01-1.28, P = 0.033). Greater SAT CSA and a higher VAT/SAT ratio showed opposing adjusted associations with 28-day mortality. Automated CT-derived body composition may complement BMI-based risk characterization in sepsis, although the subgroup and skeletal muscle findings require external validation.