Accelerated Pancreatic Volume Loss as a Potential Pre-Diagnostic Imaging Biomarker for Pancreatic Cancer Risk...

Background/Objectives: Treatment for pancreatic ductal adenocarcinoma (PDAC) depends on early detection, yet most present at an inoperable stage. Focal pancreatic parenchymal atrophy (FPPA) is a prodromal imaging feature lacking objective quantification. We evaluated whether longitudinal, AI-measured pancreatic volume (PV) change is associated with subsequent PDAC. Methods: In this retrospective case-control study, we analyzed longitudinal CT from 40 PDAC patients and 61 controls, each with ≥5 years of pre-index imaging. PV was measured by AI segmentation and compared as a subject-specific rate of change, with robustness assessed via multivariable, mixed-effects, and sensitivity analyses. Results: PDAC patients showed more rapid PV decline than matched controls (5-year reduction 9.71 vs. 1.23 mL; p = 0.013), persisting after excluding diabetes and adjustment. A mixed-effects model confirmed faster decline, consistent across truncation windows but absent near diagnosis. Adding the marker to a basic covariate model raised the area under the curve (AUC) from 0.679 to 0.823. Exploratory transcriptomic profiling of microdissected FPPA identified 131 concordantly altered genes across PanIN- and invasive-PDAC-adjacent comparisons, with an immune-related signal present adjacent to non-invasive PanIN and larger adjacent to invasive PDAC. Conclusions: Patients who later developed PDAC showed greater long-term PV decline than controls. Exploratory tissue profiling raises the hypothesis that the same process underlying FPPA may also contribute to PV loss. Because the imaging and molecular analyses only partially overlapped and included no within-patient imaging-transcriptomic correlation, the proposed tissue-level correlate remains an unproven hypothesis requiring same-patient validation. This marker requires validation before use as a risk-enrichment signal, not a screening test.