Multiple logistic regression model to predict bile leak associated with subtotal cholecystectomy

Abstract

Background: There are no prediction models for bile leakage associated with subtotal cholecystectomy (STC). Therefore, this study aimed to generate a multivariable prediction model for post-STC bile leakage and evaluate its overall performance.

Methods: We analysed prospectively managed data of patients who underwent STC by a single consultant surgeon between 14 May 2013 and 21 December 2021. STC was schematised into four variants with five subvariants and classified broadly as closed-tract or open-tract STC. A contingency table was used to detect independent risk factors for bile leakage. A multiple logistic regression analysis was used to generate a model. Discrimination and calibration statistics were computed to assess the accuracy of the model.

Results: A total of 81 patients underwent the STC procedure. Twenty-eight patients (35%) developed bile leakage. Of these, 18 patients (64%) required secondary surgical intervention. Multivariable logistic regression revealed two independent predictors of post-STC bile leak: open-tract STC (odds ratio [OR], 7.07; 95% confidence interval [CI], 2.191-25.89; P = 0.0170) and acute cholecystitis (OR, 5.449; 95% CI, 1.584-23.48; P = 0.0121). The area under the receiver-operating characteristic curve was 82.11% (95% CI, 72.87-91.34; P < 0.0001). Tjur's pseudo-R² was 0.3189 and the Hosmer-Lemeshow goodness-of-fit statistic was 4.916 (P = 0.7665).

Conclusions: Open-tract STC and acute cholecystitis are the most reliable predictors of bile leakage associated with STC. Future prospective, multicentre studies with higher statistical power are needed to generate more specific and externally validated prediction models for post-STC bile leaks.

Keywords: Bile leak; Logistic regression; Multivariable model; Prediction; Risk factors; Subtotal cholecystectomy.