Abstract
BACKGROUND: Pancreatic cancer remains the fourth leading cause of cancer-related deaths in the USA despite its lower incidence, primarily due to late-stage diagnosis. While early detection could double survival rates, screening the general population is not cost-effective due to low disease prevalence and technical limitations. SUMMARY: This review examines the relationship between diabetes and pancreatic cancer, highlighting how diabetes types differently impact cancer risk. New-onset diabetes triples pancreatic cancer risk compared to the general population, while long-standing diabetes doubles it. Several prediction models have been developed to identify high-risk individuals among new-onset diabetes patients, with recent models achieving AUCs up to 0.91. Current biomarkers like CA 19-9 show improved utility when combined with other clinical parameters, though they remain inadequate for general population screening. Cost-effectiveness analysis suggests that screening becomes viable when 3-year cancer incidence exceeds 2% and 25% of cases are detected at a localized stage. KEY MESSAGES: (1) New-onset diabetes presents a stronger risk factor for pancreatic cancer than long-standing diabetes. (2) Multiple prediction models show promise but face challenges with missing data and cross-population validation. (3) Integrated approaches combining clinical parameters, biomarkers, and machine learning offer the most promising path forward for early detection. (4) Current detection rates fall below cost-effectiveness thresholds, highlighting the need for improved screening strategies.