肝硬化合并静脉曲张出血患者内镜治疗后再出血风险的模型建立与验证

Establishment and validation of a model for the risk of rebleeding after endoscopic treatment in patients with cirrhosis and variceal bleeding

目的探讨肝硬化合并食管胃静脉曲张出血患者内镜治疗后再出血风险的相关危险因素,构建并验证列线图模型价值。方法采用回顾性分析方法,选取2022年1月至2023年1月在石家庄市人民医院治疗的肝硬化合并静脉曲张出血接受内镜治疗患者,将所有患者随机分为训练集和验证集,采用多因素logistic回归模型分析发生再出血的危险因素,在危险因素的基础上建立列线图预测模型。采用受试者工作特征曲线(ROC曲线)、Hosmer-Lemeshow检验和决策曲线分析对列线图模型进行验证。结果研究共纳入肝硬化合并食管胃静脉曲张出血内镜治疗患者153例,有38例(25.49%)发生再次出血。单因素分析结果显示体重指数(BMI)、肝硬度、门静脉直径、血红蛋白、血小板计数、谷氨酰转移酶(GGT)、Child-Pugh分级、终末期肝病评估模型(MELD)评分是影响肝硬化合并食管胃静脉曲张出血内镜治疗患者再出血的因素,差异具有统计学意义(P<0.05)。多因素logistic回归分析显示BMI、肝硬度、血红蛋白、血小板计数、MELD评分是肝硬化合并食管胃静脉曲张出血内镜治疗患者再出血的危险因素,差异有统计学意义(P<0.05)。基于上述危险因素构建线段式列线图模型,各因素赋分,其中总分为-4~4,对应的风险率范围为0.02~0.97,分值越高预测肝硬化合并食管胃静脉曲张出血内镜治疗患者再出血风险越大。训练集模型ROC曲线下面积为0.873(95%CI 0.807~0.939),验证集ROC曲线下面积为0.772(95%CI 0.584~0.959)。训练集和验证集校准曲线经Hosmer-Lemeshow检验结果提示,校正曲线趋近于理想曲线,无明显差异(χ^(2)=3.231、2.397,P=0.919、0.934)。临床决策曲线分析显示,该模型具有临床实用性。结论本研究构建并验证了肝硬化合并食管胃静脉曲张出血患者内镜治疗后再出血风险的列线图模型,具有较好的准确性、校准度以及临床实用性。

Objective To explore the relevant risk factors for rebleeding after endoscopic treatment in patients with liver cirrhosis complicated with esophageal and gastric variceal bleeding,and to construct and validate the value of a nomogram model.Methods A retrospective analysis method was used to select patients with cirrhosis complicated with variceal bleeding who received endoscopic treatment in Shijiazhuang People's Hospital from January 2022 to January 2023.All patients were randomly divided into a training set and a validation set.A multivariate logistic regression model was used to analyze the risk factors for recurrent bleeding,and a nomogram prediction model was established based on the risk factors.The nomogram model was verified using receiver operating characteristic curve(ROC curve),Hosmer-Lemeshow test,and decision curve analysis.Results A total of 153 patients with cirrhosis combined with esophageal and gastric variceal bleeding who underwent endoscopic treatment were included in this study,of which 38 cases(25.49%)experienced recurrent bleeding.The results of univariate analysis showed that body mass index(BMI),liver hardness,portal vein diameter,hemoglobin,platelet count,glutamyltransferase(GGT),Child Pugh grading,and model for end-stage liver disease(MELD)score were the factors affecting rebleeding in patients with liver cirrhosis complicated with esophageal and gastric variceal bleeding treated with endoscopy,and there were statistically significant differences(P<0.05).Multivariate logistic regression analysis showed that BMI,liver hardness,hemoglobin,platelet count,and MELD score were risk factors for rebleeding in patients with liver cirrhosis complicated with esophageal and gastric variceal bleeding treated with endoscopy,and there were statistically significant differences(P<0.05).Based on the above risk factors,a line segment nomogram model was constructed,with scores assigned to each factor.The total score was-4 to 4,and the corresponding risk rate range was 0.02 to 0.97.The higher the score,the greater the risk of rebleeding in patients with liver cirrhosis combined with esophageal and gastric variceal bleeding who undergo endoscopic treatment.The area under the ROC curve of the training set model was 0.873(95%CI:0.807-0.939),and the area under the ROC curve of the validation set was 0.772(95%CI:0.584-0.959).The results of the Hosmer-Lemeshow test on the calibration curves of the training and validation sets indicated that the calibration curves approached the ideal curve without significant differences(χ^(2)=3.231,2.397,P=0.919,0.934).Clinical decision curve analysis shows that the model had clinical practicality.Conclusion A nomogram model for the risk of rebleeding in patients with liver cirrhosis complicated with esophageal and gastric variceal bleeding after endoscopic treatment is constructed and validated,which has good accuracy,calibration,and clinical practicality.