[Development and validation of a preoperative prediction model for very early recurrence after radical resection of intrahepatic mass-forming cholangiocarcinoma].

Objective: To develop and validate a nomogram model based on CT imaging features for predicting the very early recurrence (VER) of patients with intrahepatic mass-forming cholangiocarcinoma (IMCC) after radical resection. Methods: This is a retrospective multicenter case series study. Retrospective analysis of clinic data was conducted in IMCC patients who underwent curative resection and contrast-enhanced CT at three independent institutions from Jiangsu province between January 2009 and December 2019 (institution 1: the First Affiliated Hospital with Nanjing Medical University; institution 2: Yancheng First People's Hospital; institution 3: Changzhou First People's Hospital). A total of 282 patients were included. A preoperative nomogram was developed based on a training cohort of 179 IMCC patients who were collected from institution 1. In the training cohort, univariate and multivariate Logistic regression analysis were used to construct the nomogram. The constructed model was validated in an independent external dataset (103 IMCC patients received surgical treatment at institution 2 and institution 3). The predictive efficacy of the nomogram model was evaluated using the receiver operating characteristic curve and its area under the curve (AUC), calibration curve, and decision curve analysis (DCA), and was compared with the AJCC 8th edition staging system. Results: The preoperative clinical-imaging prediction model was constructed based on the albumin-bilirubin grading and three CT imaging features (tumor size, multi-nodular type, and arterial phase enhancement pattern). In the training cohort, the predictive efficacy of the preoperative clinical-imaging model (AUC=0.819, 95%CI: 0.756 to 0.883) was significantly higher than that of the AJCC 8th edition staging system (AUC=0.707, 95%CI: 0.633 to 0.782)(P=0.006); in the external validation cohort, the predictive efficacy of the preoperative clinical-imaging model (AUC=0.766, 95%CI: 0.672 to 0.861) was slightly better than that of the AJCC 8th edition staging system (AUC=0.709, 95%CI: 0.611 to 0.808), but the difference was not statistically significant (P=0.370). The calibration curve indicated that the predicted probabilities of the clinical-imaging nomogram model were in good agreement with the actual observed values. The DCA showed that this model had better clinical net benefit compared to the AJCC 8th edition staging system. Conclusions: The preoperative albumin-bilirubin grading and three CT imaging features, including tumor size, multi-nodular type, and arterial phase enhancement pattern, are independent risk factors for postoperative VER in IMCC patients. The clinical-imaging nomogram model constructed based on the albumin-bilirubin grading and these three imaging features can accurately predict postoperative VER in IMCC patients before surgery, providing a reference for the selection of individualized treatment strategies.