Abstract
IMPORTANCE: There is currently no consensus on the indication for cholecystectomy in patients with uncomplicated gallstone disease. OBJECTIVE: To report on the development and validation of a multivariable prediction model to better select patients for surgery. DESIGN, SETTING, AND PARTICIPANTS: This study evaluates data from 2 multicenter prospective trials (the previously published Scrutinizing (In)efficient Use of Cholecystectomy: A Randomized Trial Concerning Variation in Practice [SECURE] and the Standardized Work-up for Symptomatic Cholecystolithiasis [Success] trial) collected from the outpatient clinics of 25 Dutch hospitals between April 2014 and June 2019 and including 1561 patients with symptomatic uncomplicated cholelithiasis, defined as gallstone disease without signs of complicated cholelithiasis (ie, biliary pancreatitis, cholangitis, common bile duct stones or cholecystitis). Data were analyzed from January 2020 to June 2020. EXPOSURES: Patient characteristics, comorbidity, surgical outcomes, pain, and symptoms measured at baseline and at 6 months' follow-up. MAIN OUTCOMES AND MEASURES: A multivariable regression model to predict a pain-free state or a clinically relevant reduction in pain after surgery. Model performance was evaluated using calibration and discrimination. RESULTS: A total of 1561 patients were included (494 patients in 7 hospitals in the development cohort and 1067 patients in 24 hospitals in the validation cohort; 6 hospitals included patients in both cohorts). In the development cohort, 395 patients (80.0%) underwent cholecystectomy. After surgery, 225 patients (57.0%) reported that they were pain free and 295 (74.7%) reported a clinically relevant reduction in pain. A multivariable prediction model showed that increased age, no history of abdominal surgery, increased visual analog scale pain score at baseline, pain radiation to the back, pain reduction with simple analgesics, nausea, and no heartburn were independent predictors of clinically relevant pain reduction after cholecystectomy. After internal validation, good discrimination was found (C statistic, 0.80; 95% CI, 0.74-0.84) between patients with and without clinically relevant pain reduction. The model had very good overall calibration and minimal underestimation of the probability. External validation indicated a good discrimination between patients with and without clinically relevant pain reduction (C statistic, 0.74; 95% CI, 0.70-0.78) and fair calibration with some overestimation of probability by the model. CONCLUSIONS AND RELEVANCE: The model validated in this study may help predict the probability of pain reduction after cholecystectomy and thus aid surgeons in deciding whether patients with uncomplicated cholelithiasis will benefit from cholecystectomy.