Accuracy of the END-PAC Model in Predicting the Risk of Developing Pancreatic Cancer in Patients with New-Onset Diabetes: A Systematic Review and Meta-Analysis

Abstract

Objectives: To investigate the performance of the END-PAC model in predicting pancreatic cancer risk in individuals with new-onset diabetes (NOD). Methods: The PRISMA statement standards were followed to conduct a systematic review. All studies investigating the performance of the END-PAC model in predicting pancreatic cancer risk in individuals with NOD were included. Two-by-two tables, coupled forest plots and summary receiver operating characteristic plots were constructed using the number of true positives, false negatives, true negatives and false positives. Diagnostic random effects models were used to estimate summary sensitivity and specificity points. Results: A total of 26,752 individuals from four studies were included. The median follow-up was 3 years and the pooled risk of pancreatic cancer was 0.8% (95% CI 0.6–1.0%). END-PAC score ≥ 3, which classifies the patients as high risk, was associated with better predictive performance (sensitivity: 55.8% (43.9–67%); specificity: 82.0% (76.4–86.5%)) in comparison with END-PAC score 1–2 (sensitivity: 22.2% (16.6–29.2%); specificity: 69.9% (67.3–72.4%)) and END-PAC score < 1 (sensitivity: 18.0% (12.8–24.6%); specificity: 50.9% (48.6–53.2%)) which classify the patients as intermediate and low risk, respectively. The evidence quality was judged to be moderate to high. Conclusions: END-PAC is a promising model for predicting pancreatic cancer risk in individuals with NOD. The score ≥3 should be considered as optimum cut-off value. More studies are needed to assess whether it could improve early pancreatic cancer detection rate, pancreatic cancer re-section rate, and pancreatic cancer treatment outcomes.