血浆氨基酸Logistic回归模型诊断乙型肝炎相关肝纤维化  

Logistic Regresslon Models Based on Plasma Concentrations of Amino Acids for Diagnosis of Hepatitis B Associated Hepatic Fibrosis

作　　者：张占卿[1] 陆伟[1] 王雁冰[1] 贾小芳[2] 张利军[2] 丁荣蓉[1] 周新兰[1] 冯艳玲[3] 

机构地区：[1]上海市公共卫生临床中心肝病二科,201508 [2]上海市公共卫生临床中心科研中心,201508 [3]上海市公共卫生临床中心病理科,上海201508

出　　处：《临床消化病杂志》2012年第5期278-284,共7页Chinese Journal of Clinical Gastroenterology

基　　金：上海市科委医学引导类项目(09411965800);国家"863"高技术研究发展计划生物和医药领域重大重点基金(2006AA02A411)

摘　　要：目的探讨基于氨基酸血浆浓度构建的Logistic回归模型诊断乙型肝炎相关肝纤维化程度的效能。方法 148例经肝组织病理学检查的慢性乙型肝炎患者入选本研究。病理学分期≥S2、≥S3和=S4分别被定义为显著纤维化、严重纤维化和进展期纤维化。基于氨基酸血浆浓度构建的Logistic回归模型的预测概率诊断肝纤维化程度的效能评价采用ROC曲线法。结果基于氨基酸血浆浓度构建的诊断显著纤维化的Logistic归模型不成立。基于氨基酸血浆浓度的诊断严重纤维化的Logistic回归模型的预测概率(predictive probabilities for severe fibrosis,PPs)诊断严重纤维化的ROC曲线下面积为0.745,95%CI为0.665～0.826;基于最佳截断值的预测概率PPs诊断严重纤维化的灵敏度、特异度、准确度分别为0.621、0.793、0.716。基于氨基酸血浆浓度的诊断进展期纤维化的Logistic回归模型的预测概率(predictive probabilities for advanced fibrosis,PPa)诊断进展期纤维化的ROC曲线下面积为0.814,95%CI为0.736～0.892;基于最佳截断值的预测概率PPa诊断进展期纤维化的灵敏度、特异度、准确度分别为0.75、0.75、0.75。结论基于氨基酸血浆浓度构建的Logistic回归模型对乙型肝炎相关的严重和进展期纤维化有一定诊断价值。Objective To explore the efficacy of Logistic regression models based on plasma concentrations of amino acids for diagnosis of hepatitis B associated hepatic fibrosis. Methods 148 patients with chronic hepatitis B with pathological diagnoses of liver tissue were enrolled in present study. The pathological staging ≥s2、≥s3,=s4 were defined as significant, severe, advanced fibrosis, respectively. ROC curve methods were used for appraisement of the Logistic regression models based on plasma concentrations of amino acids for diagnosis of different levers of hepatic fibrosis. Results The Logistic regresion model fordiagnosis of significant fibrosis based on plasma concentrations of amino acids could not be established. The area under ROC curve of predictive probabilities of the Logistic regression model for severe fibrosis （PPs） for diagnosis of severe fibrosis was 0.745, and the 95% CI ofthe AUROC was 0. 665-0.826 ; based on the optimal cutoff, the sensitivity, specificity, accuracy of the PPs for diagnosis of severe fibrosis were 0. 621,0. 793,0.716. The area under ROC curve of predictive probabilities of the Logistic regression models for advanced fibrosis（PPa） for diagnosis of advanced fibrosis was 0. 814, and the 95% CI of the AU- ROC was 0. 736-0. 892; based on the optimal cutoff, the sensitivity, specificity, accuracy of the PPa for diagnosis of dvanced fibrosis were 0. 75 ,0. 75 ,0. 75. Conclusion There could be diagnostic significance of the Logistic regression models based on plasmaconcentrations of amino acids for severe and advanced fibrosis.

关 键 词：乙型肝炎:肝纤维化 肝硬化 氨基酸 LOGISTIC回归 

分 类 号：R575.2[医药卫生—消化系统]