Establish And Evaluate The Pathological Classification For Chronic Hepatitis B In Non-invasive Serology Models

Objective: To evaluate the serology noninvasive diagnostic models for predictingliver fibrosis and inflammation in patients with chronic hepatitis B and providereference for clinic diagnosis.Methods: A total of322chronic hepatitis B patients were enrolled,1iver biopsy wereproceeded to assess the stage of fibrosis and inflammation histologically coupled withlaboratory tests. Cases were stratified by different study endpoint(S≤2with>S2andG≤2with> G2) and randomly divided into to the training group and the validationgroup equally.In the training group，by using one-way analysis of variance andBinary Logistic regression to analyze evidently relevant indicators, we selected outthe independent predictive factors and established the liver fibrosis-predicting modelsB and inflammation-predicting model A. The diagnostic value of these models werevalidated in the validation group.92patients with the normal ALT were selected fromthe total recruited patients to establish the liver fibrosis-predicting models C offibrosis-predicting Model D according to the different study endpoint as well. Thesemodels were analyzed with ROC curve. SPSS17.0software was used to do thestatistics analysis. A p value less than0.05was considered statistically significant.Results: Totally322patients with CHB were enrolled. Binary Logistic regressionanalysis showed that ALB、PT、PLT/ALT、PLT/AST are independent predictors forliver inflammation prediction in models A. ROC curve analysis indicated that theAUC was0.883with the sensitivity of77.4%and specificity of91.8%. In validationgroup, the AUC was0.8with the sensitivity of71.9%and specificity of72.7%. Similarly, the AUC was0.819with the sensitivity of70.4%and specificity of81.2%in fibrosis-predicting model B, while the AUC was0.766with the sensitivity of80.2%%and specificity of68.2%in validation group. In92patients with normal ALT,AST、GLO, HBeAg/AST were independent predictor in model C with the AUC of0.915, sensitivity of73.1%and specificity of96%. PT and PLT were independentpredictors in model D,. in which the AUC was0.885wich the sensitivity of88.20%and specificity of81.3%respectively.Conclusions: The Model A has good diagnostic values for liver inflammation stages,especially for those with inflammation score of more than G2, while model B isvaluable for liver fibrosis evaluation. In CHB patients with normal ALT, thenon-invasive diagnostic model C is valuable for liver fibrosis stage of S<2andS≥2evaluation, while the model D is valuable for liver fibrosis stage of S≤3and S=4evaluation.