A risk score model to predict incidental gallbladder cancer in patients scheduled for cholecystectomy

Abstract

Background: The aim was to construct a risk score model to predict IGBC, developed through statistical modelling, based on our groups previous findings. Methods: Data from cholecystectomies registered in the Nationwide Swedish Register for Gallstone Surgery were analysed. The derivation cohort underwent surgery 2007–2014 and the validation cohort 2015–2016. From the multivariable logistic regression in the derivation cohort, an additive risk model, based on the odds ratio, was constructed and validated. The scoring model´s ability to predict IGBC was estimated by AUROC and Hosmer-Lemeshow (HL) test. Results: The derivation cohort consisted of in total 36,355 patients, including 215 with IGBC, who underwent cholecystectomy. Age (≤41 years 0 points, 42–66 years 10p, ≥67 years 50p), female gender (3.5p), previous cholecystitis (1.5p) and the combination jaundice without acute cholecystitis (2.5) and acute cholecystitis without jaundice (1.5p) were significantly associated with IGBC. The accuracy of the model was assessed both on the derivation cohort, and in the validation set of 9,948 patients, including 42 with IGBC. The AUROC scores for predicting IGBC were 0.80 (CI:0.77–0.83) in the derivation cohort and 0.78 (CI:0.72–0.84) in the validation cohort, with a good calibration (HL test, p = 0.19). Conclusion: We present the first risk score model to predict IGBC. The model is based on five easily registered clinically relevant variables. This validated model may help to optimize treatment strategies in high risk patients.