Development and Internal Validation of a Preoperative Risk Stratification Model for Difficult Laparoscopic Cholecystectomy in Acute Cholecystitis

Background: Difficult laparoscopic cholecystectomy (DLC) in patients with acute cholecystitis (AC) is associated with greater operative complexity and increased perioperative morbidity, yet reliable preoperative risk stratification remains limited. This study aimed to develop and internally validate a preoperative prediction model for DLC and to compare its predictive performance with that of an alternative C-reactive protein–albumin–lymphocyte (CALLY)-based model. Materials and Methods: We retrospectively analyzed consecutive adult patients who underwent laparoscopic cholecystectomy (LC) for AC at a single center between 1 January 2022 and 31 December 2025. DLC was defined as operative time >90 min, conversion to open surgery, subtotal cholecystectomy or a Nassar grade ≥4. Preoperative clinical, laboratory and ultrasonographic variables were evaluated using multivariable logistic regression. Model performance was assessed by discrimination, calibration, Brier score, decision curve analysis and bootstrap internal validation. Results: Among 8530 screened patients, 7736 met the inclusion criteria and 1736 (22.4%) experienced DLC. In the primary multivariable model, Tokyo Guidelines 2018 (TG18) grade II, TG18 grade III and preoperative C-reactive protein (CRP) were independently associated with DLC. The model demonstrated modest discrimination (area under curve (AUC), 0.617; optimism-corrected AUC, 0.612) with a Brier score of 0.167. In the alternative model, TG18 grade III and a lower CALLY index remained independent predictors; however, predictive performance did not improve (AUC, 0.580; Brier score, 0.171). In the sensitivity analysis excluding operative time, severe DLC occurred in 378 patients (4.9%) and none of the candidate predictors remained independently associated with this endpoint. Conclusions: TG18 severity grade and preoperative CRP were independently associated with the composite DLC endpoint, which was predominantly driven by prolonged operative time. Discrimination was limited and no independent associations were identified for severe DLC when operative time was excluded. The alternative CALLY-based model showed lower discrimination than the conventional model. These findings do not support using the model to guide clinical decisions regarding severe operative difficulty.