Machine learning model to predict early recurrence in patients with perihilar cholangiocarcinoma planned treatment with curative resection: a multicenter study

作者：Xiang Wang, Li Liu, Zhi-Peng Liu, Jiao-Yang Wang, Hai-Su Dai, Xia Ou, Chengcheng Zhang, Ting Yu, Xing-Chao Liu, Shujie Pang, Haining Fan, Jie Bai, Yan Jiang, Yanqi Zhang, Zi-Ran Wang, Zhiyu Chen, Aiguo Li · 发表于：Journal of Gastrointestinal Surgery · 年份：2024 · DOI：10.1016/j.gassur.2024.09.027 · 被引用次数：6 · 研究领域：Cholangiocarcinoma and Gallbladder Cancer Studies、Gallbladder and Bile Duct Disorders、Pancreatic and Hepatic Oncology Research

BACKGROUND: Early recurrence is the leading cause of death for patients with perihilar cholangiocarcinoma (pCCA) after surgery. Identifying high-risk patients preoperatively is important. This study aimed to construct a preoperative prediction model for the early recurrence of patients with pCCA to facilitate planned treatment with curative resection. METHODS: This study ultimately enrolled 400 patients with pCCA after curative resection in 5 hospitals between 2013 and 2019. They were randomly divided into training (n = 300) and testing groups (n = 100) at a ratio of 3:1. Associated variables were identified via least absolute shrinkage and selection operator (LASSO) regression. Four machine learning models were constructed: support vector machine, random forest (RF), logistic regression, and K-nearest neighbors. The predictive ability of the models was evaluated via receiving operating characteristic (ROC) curves, precision-recall curve (PRC) curves, and decision curve analysis. Kaplan-Meier (K-M) survival curves were drawn for the high-/low-risk population. RESULTS: Five factors: carbohydrate antigen 19-9, tumor size, total bilirubin, hepatic artery invasion, and portal vein invasion, were selected by LASSO regression. In both the training and testing groups, the ROC curve (area under the curve: 0.983 vs 0.952) and the PRC (0.981 vs 0.939) showed that RF was the best. The cutoff value for distinguishing high- and low-risk patients was 0.51. K-M survival curves revealed that...