Prognostic value of deep learning-derived body composition in advanced pancreatic cancer—a retrospective multicenter study

•We show the applicability of a deep learning-based workflow for body composition analysis in three German cancer centers.•Automatically extracted body composition markers from routine CT scans were comparable between cohorts.•CT-derived sarcopenia and myosteatosis were associated with OS in the pooled cohort.•In a subgroup, body composition markers were associated with anemia, hypoproteinemia, and inflammation.