慢性乙型肝炎患者肝脏炎症程度预测模型的研究  

Prediction model for inflammation degree in chronic hepatitis B patients:a retrospective study

作　　者：王娟霞[1,2] 陈馨悦 魏世博 孙新策 朱浩宇 连唐悠悠 杜雨峰 WANG Juan-xia;CHEN Xin-yue;WEI Shi-bo;SUN Xin-ce;ZHU Hao-yu;LIANTANG You-you;DU Yu-feng(The Second Clinical Medical College of Lanzhou University,,Lanzhou 730030,China;不详)

机构地区：[1]兰州大学第二临床医学院 [2]兰州大学第二医院感染性疾病科 [3]兰州大学公共卫生学院

出　　处：《中国实用内科杂志》2023年第5期390-395,共6页Chinese Journal of Practical Internal Medicine

基　　金：甘肃省自然科学基金项目(22JR5RA1000);兰州大学第二医院萃英学子科研培育计划项目(CYXZ2021-55)。

摘　　要：目的利用慢性乙型肝炎患者(CHB)的血清生化指标建立肝脏炎症程度的无创预测模型,并评估该模型的诊断价值。方法采用回顾性研究,选取2012年10月至2021年12月在兰州大学第二医院住院行肝穿检查的CHB患者1446例,根据肝活检结果显示的炎症程度将研究对象分为对照组(G1期、G2期)和病例组(G3期、G4期),采用LASSO回归进行自变量的筛选,使用限制性立方样条函数判断因变量与自变量间是否存在曲线关系,使用logistic回归进行预测模型的建立。结果使用LASSO回归对单因素分析中具有统计学意义的指标进行筛选,筛选出谷氨酰转肽酶(GGT)、天冬氨酸转氨酶(AST)、凝血酶原活动度(PTA)、年龄、收缩压和白蛋白6个指标,其中GGT和AST与肝脏炎症风险呈曲线关系。用上述6个变量预测的ROC曲线面积(AUC)为0.745(95%CI 0.721~0.768),并与纳入15个自变量的模型的预测价值差异无统计学意义(P=0.643)。结论使用GGT、AST、PTA、年龄、收缩压和白蛋白这6个指标对慢性乙型肝炎患者的肝脏炎症程度具有很好的预测价值。Objective To establish a non-invasive predictive model for the inflammation degree in patients with chronic hepatitis B(CHB)based on serum biochemical markers,and to evaluate the diagnostic value of the established predictive model.Methods In this retrospective study,a total of 1446 patients with CHB who underwent liver biopsy in the Second Hospital of Lanzhou University from October 2012 to December 2021 were included.According to the degree of inflammation assessed by liver biopsy,participants were divided into control group(G1,G2)and case group(G3,G4).LASSO regression model was used for independent variable selection.We used restricted cubic splines to examine if nonlinear associations existed between independent variable and dependent varible.Logistic regression was used to establish predictive models.Results By using LASSO regression model,six covariates including GGT,AST,PTA,age,systolic blood pressure and albumin were selected for prediction model,and GGT and AST had a nonlinear association with liver inflammation risk.The area under the ROC curve was 0.745(95%CI:0.721-0.768),and this model showed nonsignificant difference from the including 15 independent variable(P=0.643).Conclusions The established regression model including GGT,AST,PTA,age,systolic blood pressure and albumin performs well in predicting the degree of liver inflammation in CHB patients.

关 键 词：慢性乙型肝炎 肝脏炎症 预测模型 诊断 

分 类 号：R512.62[医药卫生—内科学]