肝硬化患者食管胃底静脉曲张无创预测模型的构建及验证

Construction and validation of non-invasive prediction model for gastroesophageal varices in patients with liver cirrhosis

目的构建慢性乙型肝炎(乙肝)肝硬化患者食管胃底静脉曲张无创模型,并在其他肝硬化患者中进行验证。方法回顾性分析2015年1月至2020年12月该中心收治的693例经上消化道内镜检查的肝硬化患者的临床资料,其中乙肝肝硬化465例,非乙肝肝硬化228例。乙肝肝硬化患者根据内镜检查食管胃底静脉曲张结果分为曲张组(151例)和非曲张组(314例)。使用多因素logistic回归模型分析试验队列乙肝肝硬化患者食管胃底静脉曲张的危险因素,并构建预测模型。采用受试者工作特征曲线判断预测模型的效能,并计算曲线下面积(AUC)。在非乙肝肝硬化患者中验证预测模型的预测效能。结果肝脏表面粗糙、脾脏厚度增加、肝硬度值增加是乙肝肝硬化患者食管胃底食管静脉曲张的危险因素(优势比=6.375、1.127、1.182)。构建肝硬化患者食管胃底静脉曲张预测模型公式为预测指数(R)=-6.68+1.852×肝脏表面粗糙情况(粗糙=1,光滑=0)+0.12×脾脏厚度(mm)+0.028×肝硬度值(kPa)。在乙肝肝硬化组患者中R判断食管胃底静脉曲张的AUC为0.935,灵敏度为84.8%,特异度为85.4%,诊断效能优于肝硬度、脾脏厚度,差异均有统计学意义(P<0.05)。在非肝硬化患者中R判断食管胃底静脉曲张的AUC为0.931(95%可信区间:0.897~0.965),灵敏度为75.2%,特异度为91.3%。结论所构建新的预测模型可作为可靠的临床肝硬化患者食管胃底静脉曲张的无创预测模型。

Objective To establish a non-invasive model of gastroesophageal varices in the patients with hepatitis B virus(HBV)related cirrhosis and to verify it in the patients with other cirrhosis.Methods The clinical data of the patients with liver cirrhosis who underwent upper gastrointestinal endoscopy in this center from January 2015 to December 2020 were retrospectively analyzed,including 465 cases of hepatitis B cirrhosis and 228 cases of non-hepatitis B cirrhosis.According to the results of endoscopic examination of esophageal and gastric varices,the patients with hepatitis B cirrhosis were divided into the varices group(n=151)and non-varices group(n=314).The multivariate logistic regression was used to analyze the risk factors of gastroesophageal varices in the patients with hepatitis B cirrhosis in the experiment cohort.The prediction model was constructed.The receiver operating characteristic(ROC)curve was used to evaluate the efficiency of predictive model,and area under the curve(AUC)was calculated.The predictive efficiency was validated in the patients with non-hepatitis B cirrhosis.Results The rough surface of liver,spleen thickness increase and liver hardnessincrease were the risk factors of esophageal varices in hepatitis B cirrhosis(OR=6.375,1.127,1.182).The established prediction model formula of gastroesophageal varices in the patients liver cirrhosis was the prediction index(R)=-6.68+1.852×roughness of liver surface(roughness=1,smoothness=0)+0.12×thickness of spleen(mm)+0.028×hardness of liver(kPa).In the hepatitis B cirrhosis group,AUC of R for judging esophageal and gastric varices was 0.935,the sensitivity was 84.8%,and the specificity was 85.4%.The diagnostic efficiency was better than the liver stiffness and spleen thickness,and the differences were statistically significant(P<0.001).In the non-cirrhotic patients,AUC of R for judging esophageal and gastric varices was 0.931(95%CI:0.897-0.965),the sensitivity was 75.2%,and the specificity was 91.3%.Conclusion The established new prediction model can serve as a noninvasive and reliable prediction model for gastroesophageal varices of the patients with liver cirrhosis in clinic.