A Predictive Tool for Choledocholithiasis in Patients Undergoing Emergency Cholecystectomy.

Background: Management of acute cholecystitis with emergency laparoscopic cholecystectomy has been established; however, detection and management of concurrent choledocholithiasis are debated. The aim of this study is to develop a more accurate choledocholithiasis predictive model. Materials and Methods: A 9-year audit of emergency cholecystectomies and evaluation of preoperative factors in predictive models. Receiver Operating Curve (ROC) analysis/Youdon Index was used to identify thresholds maximizing these associations for continuous variables. Results: 1601/1828 patients were analyzed. Patients who were diagnosed with choledocholithiasis were more likely to be febrile on admission, have a higher C-reactive Protein, and higher median bilirubin (25.0 μmol/L versus 11.0 μmol/L, P  < .001). When excluding bilirubin, multivariate analysis detected several significant variables, including fever, biliary tree dilatation, or a common bile duct stone seen on ultrasound. When bilirubin was included into the model, bilirubin of 20-39 μmol/L (odds ratio [OR] 2.44, 95% confidence interval [CI]: 1.74-3.44) and ≥40 μmol/L (OR 4.84, 95% CI: 3.40-6.91) were shown to have increased likelihood of choledocholithiasis detection on intraoperative cholangiogram, with the ROC model having a significant C-statistic of 0.796 ( P  < .001). Discussion: A perfect predictive model for concurrent choledocholithiasis in acute cholecystitis does not exist; however, the results from this study are encouraging that high and low predictive groups can be established.