Abstract: Background　Esophageal variceal bleeding is a fatal complication of cirrhotic portal hypertension. Early diagnosis and treatment of esophageal varices can significantly reduce the mortality. However，there is still a lack of a unified noninvasive model for predicting the risk of esophageal variceal bleeding. Objective　To develop a prediction nomogram for high-risk esophageal varices in cirrhosis patients. Methods　Participants were cirrhosis patients selected from Affiliated Jinhua Hospital，Zhejiang University School of Medicine from January 2018 to October 2020. Laboratory examination indices were collected within 24 hours after admission. Gastroscopy was performed for evaluating the degree of esophageal varices. Enhanced abdominal CT was conducted. The potential predictors subsets of high-risk esophageal varices were screened by LASSO regression and 10-fold cross-validation method，and determined by using multivariate Logistic regression analysis，then were used for developing a prediction nomogram for high-risk esophageal varices using the R software. ROC curve analysis was used to evaluate the predictive ability of the model. Internal validation was performed using a bootstrap method. The discrimination of the nomogram was determined by calculating the average consistency index（C-index）. Calibration curve was used to assess the calibration of the nomogram. Results　A total of 204 cases were enrolled，including 92 with high-risk esophageal varices and 112 with low-risk esophageal varices. The diameter of the left gastric vein〔OR=2.174，95%CI（1.547，3.056），P<0.001〕，platelet count 〔OR=0.970，95%CI（0.950，0.991），P<0.001〕，splenic thickness〔OR=1.061，95%CI（1.003，1.123），P=0.039〕，and ascites〔OR=3.091，95%CI（1.768，5.404），P<0.001〕 were independent predictors for high-risk esophageal varices. The AUC of the nomogram for predicting high-risk esophageal varices was 0.949〔95%CI（0.921，0.978）〕，and the C-index was 0.951 in internal validation. The calibration curve showed that the predictive results were in good consistency with the actual results. Conclusion　This nomogram built based on variables of left gastric vein，platelet，splenic thickness and ascites has good discrimination and accuracy，which may be helpful for screening high-risk esophageal varices in cirrhosis patients.

Key words: Liver cirrhosis, Esophageal and gastric varices, Nomograms, Forecasting, LASSO regression

摘要： 背景　食管静脉曲张破裂出血是肝硬化门静脉高压最致命的并发症，早期诊断和治疗食管静脉曲张能显著降低肝硬化门脉高压患者的病死率。然而，目前临床上尚缺乏统一的预测食管静脉曲张破裂出血风险的无创模型。目的　构建肝硬化患者高危食管静脉曲张的列线图预测模型。方法　选取2018年1月至2020年10月在浙江大学医学院附属金华医院确诊为肝硬化的患者为研究对象，收集其入院24 h内的实验室检查指标，并进行胃镜检查，评估食管静脉曲张程度。采用CT扫描仪对患者进行腹部增强CT扫描。采用LASSO回归模型、十折交叉验证法获得高危食管静脉曲张的最佳风险预测因子子集，并采用多因素Logistic回归模型确定高危食管静脉曲张的风险预测因子。根据风险预测因子，应用R软件构建列线图预测模型。采用受试者工作特征（ROC）曲线评估列线图预测模型的预测能力。采用Bootstrap法进行模型内验证，计算一致性指数（C-index）以评估模型区分度；绘制校准曲线以评估模型的校准度。结果　共纳入204例患者，其中低危食管静脉曲张92例，高危食管静脉曲张112例。胃左静脉内径〔OR=2.174，95%CI（1.547，3.056），P<0.001〕、血小板计数〔OR=0.970，95%CI（0.950，0.991），P<0.001〕、脾厚径〔OR=1.061，95%CI（1.003，1.123），P=0.039〕、腹腔积液〔OR=3.091，95%CI（1.768，5.404），P<0.001〕是高危食管静脉曲张的独立风险预测因子。构建的列线图预测模型预测肝硬化患者发生高危食管静脉曲张的ROC曲线下面积为0.949〔95%CI（0.921，0.978）〕；经内部验证，该模型的C-index为0.951；校准曲线显示该模型预测结果与实际结果的一致性良好。结论　本研究基于胃左静脉内径、血小板计数、脾厚径、腹腔积液4项独立风险预测因子构建出肝硬化患者高危食管静脉曲张的列线图预测模型，并具有良好的区分度与校准度，有助于在肝硬化患者中筛选出高危食管静脉曲张者。

关键词: 肝硬化, 食管和胃静脉曲张, 列线图, 预测, LASSO回归