引用文本：卢思丽, 王佳敏, 程洁, 等. 基于弹性网络-Cox回归的肝细胞癌术后生存预测模型构建[J/CD]. 消化肿瘤杂志（电子版）, 2026, 18(1): 117-130.

作者：卢思丽¹，王佳敏¹，程洁¹，黄安琪¹，彭建新^1,2，何军明^1,2

单位：1. 广州中医药大学第二临床医学院，广东 广州 510405；2. 广东省中医院肝胆外科，广东 广州 510120

Authors：Lu Sili¹, Wang Jiamin¹, Cheng Jie¹, Huang Anqi¹, Peng Jianxin^1,2, He Junming^1,2

Unit：1. The Second Clinical Medical School, Guangzhou University of Chinese Medicine, Guangzhou 510405, Guangdong, China；2. Department of Hepatobiliary Surgery, Guangdong Provincial Hospital of Chinese Medicine, Guangzhou 510120, Guangdong, China

摘要：

目的 建立并验证基于弹性网络-Cox回归的肝细胞癌（hepatocellular carcinoma, HCC）肝切除术后个体化生存预测模型。方法 回顾性分析2013年1月至2022年12月广东省中医院收治的559例接受肝切除术的HCC患者的临床资料。采用分层随机法按7∶3比例将患者分为训练组（n=392）和验证组（n=167）。采用弹性网络回归筛选影响总生存（overall survival, OS）的预后因素，并通过多因素Cox回归构建列线图预测模型。采用一致性指数（concordance index, C-index）、受试者操作特征曲线（receiver operating characteristic curve, ROC曲线）的曲线下面积（area under the curve, AUC）和校准曲线评估模型的区分度和校准度。基于列线图模型总评分和X-tile软件将训练组（低风险234例、中风险103例、高风险55例）和验证组（低风险103例、中风险40例、高风险24例）进行风险分层，并通过Kaplan-Meier生存分析和Log-rank检验验证模型的风险分层能力。结果 弹性网络-Cox回归分析显示，肿瘤直径、肿瘤数目、血管癌栓、手术类型、术中输血、白蛋白（albumin, ALB）水平是HCC患者OS的独立影响因素（均P＜0.05）。列线图模型在训练组和验证组中的C-index分别为0.747和0.732。列线图模型预测训练组1、3、5年OS率的AUC分别为0.804、0.789、0.773，验证组对应为0.799、0.780、0.770；校准曲线显示列线图模型的生存率预测值与实际观察值高度一致。风险分层分析表明，训练组和验证组中不同风险患者的生存差异均有统计学意义（均P＜0.001）。结论 基于弹性网络-Cox回归构建的列线图模型能有效预测HCC肝切除术后患者的OS率，具有较好的临床应用价值。

关键词：肝细胞癌；弹性网络回归；Cox回归；列线图；总生存

Abstract：

Objective To develop and validate an individualized survival prediction model based on elastic net-Cox regression for patients with hepatocellular carcinoma (HCC) after hepatectomy. Method A retrospective analysis was conducted on the clinical data of 559 HCC patients who underwent hepatectomy at Guangdong Provincial Hospital of Chinese Medicine between January 2013 and December 2022. Patients were stratified and randomly divided into a training cohort (n=392) and a validation cohort (n=167) at a ratio of 7∶3. Prognostic factors affecting overall survival (OS) were screened using elastic net regression, and a nomogram prediction model was constructed via multivariate Cox regression. The model's discriminative ability and calibration were assessed using the concordance index (C-index), the area under the curve (AUC) of the receiver operating characteristic (ROC) curve, and calibration curves. Based on the total scores from the nomogram model and the X-tile software, risk stratification was performed in the training cohort (low-risk: n=234, intermediate-risk: n=103, high-risk: n=55) and the validation cohort (low-risk: n=103, intermediate-risk: n=40, high-risk: n=24). The risk stratification capability of the model was validated using Kaplan-Meier survival analysis and the Log-rank test. Result Elastic net-Cox regression analysis identified tumor diameter, tumor number, microvascular invasion, type of surgery, intraoperative blood transfusion, and albumin (ALB) level as independent influencing factors for OS of HCC patients (all P<0.05). The C-indices of the nomogram model were 0.747 in the training cohort and 0.732 in the validation cohort. The AUCs for predicting 1-, 3-, and 5-year OS rates were 0.804, 0.789, and 0.773 in the training cohort, respectively, and 0.799, 0.780, and 0.770 in the validation cohort, respectively. Calibration curves showed high consistency between the nomogram-predicted survival probabilities and the actual observations. Risk stratification analysis demonstrated statistically significant differences in survival among patients in different risk groups in both the training and validation cohorts (all P<0.001). Conclusion The nomogram prediction model based on elastic net-Cox regression effectively predicts OS rates in HCC patients after hepatectomy and demonstrates good potential for clinical application.

Key words：Hepatocellular carcinoma; Elastic net regression; Cox regression; Nomogram; Overall survival

注：网络优先发布