Development and validation of a CT-measured body composition radiomics model for prognostic assessment in resected pancreatic adenocarcinoma.

Pancreatic ductal adenocarcinoma (PDAC) carries a dismal prognosis due to high post-resection recurrence, yet current prognostic tools inadequately capture systemic tumor-host interactions. This study developed and validated a novel CT-based body composition radiomics model for predicting recurrence-free survival (RFS) in 191 PDAC patients undergoing curative resection. Using a temporal validation design (training cohort: 2019-2022, n = 142; validation cohort: 2016-2018, n = 49), we extracted 1,688 radiomics features from adipose and muscle tissues at the L3 vertebral level. Following a standardized feature selection protocol (variance filtering, correlation reduction, and Cox regression), we constructed fat- and muscle-specific radiomics scores, combining them into a unified risk stratification system. The combined model effectively identified high-risk patients with significantly reduced RFS (HR = 6.455, p < 0.001), achieving consistent performance across cohorts (C-index: 0.71 training, 0.69 validation). This approach quantifies metabolic alterations in body composition, providing a clinically actionable tool to guide personalized therapy decisions-including intensified neoadjuvant regimens for high-risk patients or standard surveillance for low-risk individuals.