基于实验室常规指标建立肝纤维化Fisher判别模型

Constructing Fisher's discriminant model for liver fibrosis based on laboratory conventional markers

目的利用实验室常规指标建立慢性乙型肝炎(chronic hepatitis B,CHB)患者的肝纤维化诊断模型,寻找肝穿刺活检的非创性替代手段。方法选取174例CHB患者,行肝活检组织学分期(S),检测血清肝纤维化和肝功能指标,Spearman等级相关法分析肝活检组织学分期与实验室指标的关系,建立Fisher判别模型,回代法验证模型的诊断性能。结果 (1)肝纤维化组织学分期(S0→S4)与透明质酸(HA,r=0.569)、层黏连蛋白(LN,r=0.428)、血清Ⅲ型前胶原(PC-Ⅲ,r=0.405)、Ⅳ型胶原(C-Ⅳ,r=0.395)、丙氨酸氨基转移酶(ALT,r=0.445)、天冬氨酸氨基转移酶(AST,r=0.412)、γ-谷氨酰转肽酶(γ-GT,r=0.317)、总胆红素(TB,r=0.396)呈正相关,与血小板(PLT,r=-0.427)、血清白蛋白(ALB,r=-0.382)呈负相关(P<0.05或P<0.001)。(2)Fisher逐步判别分析显示,LN、HA、ALT、PLT入选以肝组织学分期为结果变量的判别函数,判别函数式:①Function 1=0.006HA+0.016LN+0.003ALT-0.007PLT-2.636;②Function 2=0.007LN+0.004ALT+0.021PLT-4.763。(3)函数诊断无纤维化(S0)、肝纤维化(S1～S3)和肝硬化(S4)的灵敏度分别是75.00%、78.30%和77.50%,总体一致率为77.59%,Kappa=0.610。结论基于实验室常规指标建立的Fisher判别模型,能为CHB患者肝纤维化分期的判断提供中等强度的诊断效能,可部分替代肝活检。

Objective To construct a diagnostic model for liver fibrosis based on laboratory conventional markers in patients with chronic hepatitis B （CHB）, so as to discover a non-invasive substitute for liver biopsy. Methods A total of 174 patients with CHB were enrolled in this study. Liver biopsy and tissue pathology checking were assessed. The fi- brosis stages, serum liver fibrosis markers and liver function indexes were measured. The relationship between fibrosis stages and the laboratory conventional markers was analyzed by Spearman＇ s rank analysis. Diagnostic model for liver fi- brosis was constructed by Fisher＇ s discriminant analysis. The diagnostic ability of model was verified by substitution method. Results （1） The liver fibrosis stages （S0→S4） was positively correlated with the hyaluronie acid （HA,r = 0. 569 ）, laminin （ LN, r = 0. 428 ）, type Ⅲ pre-collagen （ PC- Ⅲ, r = 0. 405 ）, type IV collagen （ C-IV, r = 0. 395 ）, al- anine aminotransferase （ ALT, r = 0. 445 ） , aspartate aminotransferase （ AST, r = 0. 412 ） , y-glutamyl transpeptidase （ y- GT, r = 0.317 ）, total bilirubin （ TB, r = 0. 396）, and negatively correlated with platelet （ PLT, r = - 0. 427 ） and serum albumin （ ALB, r = - 0. 382） （P 〈 0.05 or P 〈 0. 001 ）. Fisher＇ s dlscriminant analysis indicated that diagnostic model of liver fibrosis was constructed from LN, HA, ALT and PLT. Discriminant function：①Function 1 =0. 006HA ＋0. 016 LN ＋0.003ALT -0.007PLT -2.636;②Function 2 =0.007LN ＋0. 004ALT ＋0.021PLT -4.763. （3） The sensitivity of predicting non-fibrosis （ S0）, liver fibrosis （ S1 -S3 ） and liver cirrhosis （$4） using function were 75.00% , 78.30% and 77.50% , respectively, the total accuracy of this function was 77.59% , Kappa = O. 610. Conclusion Fisher＇ s discriminant model based on laboratory conventional markers show a reliable diagnostic value for liver fibrosis pathological stages in patients with CHB, which could be used to replace liver biopsy.