Development of a preoperative prediction model for non-curative resection in intrahepatic cholangiocarcinoma involving the hepatic hilum.

BACKGROUND: For intrahepatic cholangiocarcinoma involving the hepatic hilum, accurate preoperative assessment of curative resection feasibility remains challenging. Some patients undergo extensive surgery without achieving the intended oncologic benefit. This study aimed to develop and internally validate a strictly preoperative model for individualized prediction of non-curative resection risk in this population. METHODS: This single-center retrospective study included 333 patients with intrahepatic cholangiocarcinoma involving the hepatic hilum. The primary endpoint was non-curative resection. Model development was based exclusively on variables available before surgery. Prespecified multivariable logistic regression models were compared mainly using the Akaike information criterion, with discrimination and prediction error as supportive metrics. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC), calibration using calibration analysis and the Brier score, and internal validation using 1000 bootstrap resamples. The study was conducted and reported in accordance with the TRIPOD statement. RESULTS: Among 333 patients, 80 (24.0%) underwent non-curative resection. The final model included five preoperative predictors: albumin modeled using restricted cubic splines, ln(CA19-9 + 1), Bismuth type IV, vascular involvement extent, and intrahepatic metastasis. The model showed strong discrimination, with an apparent AUC of 0.908 (95% CI, 0.870-0.946). The apparent Brier score was 0.092, and the Hosmer-Lemeshow test showed no evidence of poor fit (P = 0.188). After bootstrap validation, the optimism-corrected AUC was 0.895, the calibration slope was 0.905, and the Brier score was 0.100. Risk stratification separated patients into low-, intermediate-, and high-risk groups, with observed non-curative resection rates of 1.8% (2/111), 11.7% (13/111), and 58.6% (65/111), respectively. CONCLUSIONS: We developed and internally validated a preoperative prediction model for non-curative resection in patients with intrahepatic cholangiocarcinoma involving the hepatic hilum. The model showed strong discrimination, acceptable calibration, and clinically meaningful risk stratification, and may support individualized preoperative assessment and multidisciplinary decision-making before major hepatobiliary surgery. External validation is required before broader clinical implementation.