Abstract
Background/Objectives: Pancreatic cancer carries a five-year relative survival of 13%. Polygenic risk scores (PRS) rank inherited susceptibility across common variants. Prior syntheses pooled PRS association and discrimination, but none asked whether a PRS changes a clinical decision. We synthesized evidence on the clinical utility of PRS for pancreatic cancer risk assessment. Methods: In this PROSPERO-registered review (CRD42024511326), we searched four databases through May 11, 2026, for studies evaluating a PRS for primary pancreatic cancer risk in adults, alone or with clinical predictors. We recorded discrimination as the prerequisite for utility and organized utility across five prespecified domains: threshold-based accuracy, incremental value, calibrated absolute risk, decision-analytic output, and economic impact. We synthesized narratively and appraised risk of bias with PROBAST. Results: Forty-three studies met the inclusion and exclusion criteria, including 25 from the UK Biobank. Ten of 16 standalone discrimination estimates fell between 0.56 and 0.64. Adding PRS to a clinical model increased discrimination by at most 0.05 in general populations. Combined PRS and clinical models exceeded 0.80 only in three diabetes-focused studies. Six studies reported calibration, four reclassification, two decision-curve net benefit, none cost-effectiveness, and 20 had no translational milestone. Thirteen studies were at high risk of bias, with 11 in the analysis domain. Conclusions: Pancreatic cancer PRS discriminates modestly alone and adds most where prior risk is highest. Actionable use awaits an externally validated absolute-risk model with a prespecified threshold and a decision analysis.