Development and validation of a novel risk score for predicting post-ERCP pancreatitis in patients with choledocholithiasis: a retrospective study.

introductionPost-endoscopic retrograde cholangiopancreatography (ERCP) pancreatitis (PEP) remains the most common serious complication of ERCP, with variable incidence across patient populations. Choledocholithiasis patients represent a distinct subgroup with unique risk profiles, yet existing prediction models lack indication-specific validation. We aimed to develop and internally validate a practical risk score for predicting PEP in patients undergoing ERCP for choledocholithiasis. METHODOLOGY: This single-center retrospective cohort study included 500 consecutive patients who underwent therapeutic ERCP for choledocholithiasis at Liuzhou People’s Hospital between January 2019 and December 2024. PEP was defined according to Cotton consensus criteria. Candidate predictors were selected based on literature review and clinical availability. Multivariable logistic regression with least absolute shrinkage and selection operator (LASSO) regularization was employed for variable selection. The final model was internally validated using 500 bootstrap iterations with optimism correction. Model performance was assessed through discrimination (C-statistic), calibration (calibration slope, Hosmer-Lemeshow test), and overall fit (Brier score). A simplified integer-based risk score was developed for clinical application.

resultsPEP occurred in 76 patients (15.20%), with mild, moderate, and severe cases representing 72.37%, 23.68%, and 3.95% of events, respectively. Five independent predictors were retained in the final model: prior post-ERCP pancreatitis (adjusted odds ratio [aOR] 4.38, 95% confidence interval [CI] 1.92–10.03), asymptomatic choledocholithiasis presentation (aOR 6.57, 95% CI 3.52–12.25), difficult cannulation (aOR 2.50, 95% CI 1.43–4.39), pancreatic duct instrumentation (aOR 1.93, 95% CI 1.12–3.33), and precut sphincterotomy (aOR 3.43, 95% CI 1.66–7.08). The model demonstrated acceptable discrimination with an optimism-corrected C-statistic of 0.760 (95% CI 0.708–0.817) after bootstrap internal validation. Calibration was satisfactory (Hosmer-Lemeshow p = 0.967; calibration slope 0.939). The derived risk score stratified patients into low (0–1 points; 6.92% PEP rate), moderate (2–3 points; 21.85% PEP rate), and high-risk (≥ 4 points; 44.44% PEP rate) categories.

conclusionsWe developed and internally validated a parsimonious, choledocholithiasis-specific risk score demonstrating robust discrimination and calibration for PEP prediction. This practical tool enables individualized risk stratification in patients undergoing ERCP for common bile duct stones. The clinical utility of risk-stratified prophylaxis requires evaluation through external validation studies and prospective trials assessing patient outcomes before clinical implementation can be recommended.