Longitudinal body composition changes distinguish pancreatic ductal adenocarcinoma from controls in a pre-diagnostic CT cohort.
Authors
Affiliations (8)
Affiliations (8)
- Department of Gastroenterology and Hepatology, Leiden University Medical Center, Leiden, the Netherlands; Department of Gastroenterology and Hepatology, Mayo Clinic, Jacksonville, FL, United States. Electronic address: [email protected].
- Department of Gastroenterology and Hepatology, Leiden University Medical Center, Leiden, the Netherlands; Department of Gastroenterology and Hepatology, Mayo Clinic, Jacksonville, FL, United States.
- Department of Gastroenterology and Hepatology, Leiden University Medical Center, Leiden, the Netherlands.
- Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, MN, United States.
- Department of Gastroenterology and Hepatology, Leiden University Medical Center, Leiden, the Netherlands; Department of Gastrointestinal Oncology, Netherlands Cancer Institute, Amsterdam, the Netherlands.
- Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, MN, United States; Department of Artificial Intelligence and Informatics, Mayo Clinic, Rochester, MN, United States.
- Department of QHS, Mayo Clinic, Rochester, MN, United States.
- Department of Gastroenterology and Hepatology, Mayo Clinic, Jacksonville, FL, United States.
Abstract
Pancreatic ductal adenocarcinoma (PDAC) is often diagnosed at an advanced stage, limiting curative treatment options. Although prior studies have characterized longitudinal body composition changes in PDAC, it remains unclear whether these pre-diagnostic trajectories differ from matched controls or distinguish PDAC cases from non-PDAC individuals. In this retrospective, multicenter case-control study, adults diagnosed with PDAC between 2002 and 2020 with abdominal CT imaging within 36 months before diagnosis were identified. Controls were selected from CT-imaged individuals and matched 2:1 to cases on age, sex, and diabetes status. Deep learning-based segmentation quantified skeletal muscle index (SMI), visceral fat index (VFI), and subcutaneous fat index (SFI). Cohort-level trajectories were evaluated using locally estimated scatterplot smoothing and linear mixed-effects models. For individual-level analyses, body composition rates of change were estimated and evaluated for discrimination. Cohort-level analyses included 1825 cases and 3650 controls. Compared with controls, PDAC cases exhibited steeper declines in SFI (-0.53 cm<sup>2</sup>/m<sup>2</sup>/month, p < 0.001) and VFI (-0.37 cm<sup>2</sup>/m<sup>2</sup>/month, p < 0.001) over 36 months preceding diagnosis, whereas SMI differences emerged within the final 12 months (-0.28 cm<sup>2</sup>/m<sup>2</sup>/month, p = 0.01). Individual-level analyses (581 cases, 581 controls) showed longitudinal changes in SMI, SFI, and VFI moderately discriminated cases from controls (AUC 0.76; 95% CI, 0.73-0.78), with performance driven largely by changes occurring closer to diagnosis. In a CT-imaged cohort, PDAC is associated with measurable longitudinal body composition changes prior to diagnosis that differ from matched controls. These changes demonstrate moderate discrimination to distinguish cases from controls but appear to reflect processes occurring closer to clinical diagnosis.