[Establishment and evaluation of noninvasive diagnostic models for liver fibrosis in patients with chronic hepatitis B].

Objective: To establish the model of liver fibrosis based on noninvasive indices, and to investigate the diagnostic value of this model. Methods: A total of 838 patients with chronic hepatitis B (CHB) who underwent liver biopsy in our hospital from March 2003 to October 2013 were selected, and the results of blood tests and B-ultrasound were collected. The correlation between these indices and liver fibrosis stage was analyzed. A logistic regression analysis was performed to establish a predictive model, and the value of this model was examined in validation group. The t-test, Mann-Whitney U non-parametric test, and chi-square test were used for data analysis. A Spearman rank correlation analysis was used for bivariate correlation analysis, and a dichotomous logistic stepwise regression analysis was used for multivariate analysis. Results: In the model group, a model (FV) consisting of age, platelet count (PLT), γ-glutamyl transferase (GGT), albumin/globulin ratio (A/G), and splenic square area (SSA) was established. The areas under the receiver operating characteristic curve (AUROCs) of the model FV were 0.892, 0.910, and 0.915, respectively, in diagnosing significant liver fibrosis (S2-4), progressive liver fibrosis (S3-4), and early-stage liver cirrhosis (S4), with sensitivities of 77.6%, 83.7%, and 86.0%, respectively, specificities of 89.7%, 84.5%, and 83.7%, respectively, and accuracy of 82.1%, 84.2%, and 84.2%, respectively. There were no significant differences in AUROCs between the validation group and the model group (Z = 0.360, 0.885, and 0.046, all P > 0.05). In all patients, FV had significantly higher AUROCs in the diagnosis of liver fibrosis than FIB4 index and S index (Z = 4.569/3.423, 5.640/4.709, and 4.652/4.439, all P < 0.05). With < 0.374 and ≥ 0.577 as the cut-off values for the exclusion and diagnosis of significant liver fibrosis, 61.1% (512/838) of all patients could avoid liver biopsy, and the accuracy was 92.6% (474/512). Conclusion: The noninvasive model based on age, PLT, GGT, A/G, and SSA can accurately predict liver fibrosis degree in patients with CHB with good reproducibility; therefore, it can be used for dynamic monitoring of liver fibrosis degree in clinical practice.目的： 从非创伤性指标中建立肝纤维化模型，并对模型的诊断价值进行评价。 方法： 选择2003年3月至2013年10月入院并进行肝活组织检查的慢性乙型肝炎患者838例，收集患者的血液及B超等检查结果。分析这些指标与肝纤维化分期的相关性，用logistic回归分析等建立预测模型，并在验证组中检验模型的价值。对数据进行t检验、Mann-Whitney U非参数检验或χ(2)检验；双因素相关分析采用Spearman等级相关分析，多因素分析采用二分类logistic逐步回归分析。 结果： 在模型组，建立了由年龄、血小板（PLT）、γ-谷氨酰转移酶(GGT)、白蛋白球蛋白比值（A/G）、脾脏面积（SSA）五项指标构成的纤维化评分模型FV。FV诊断显著纤维化（S2～4）、进展期纤维化（S3～4）、早期肝硬化（S4）的受试者工作特征曲线下面积（AUROC）分别为0.892、0.910、0.915，灵敏度和特异度分别为：77.6%和89.7%、83.7%和84.5%、86.0%和83.7%，准确率分别为82.1%、84.2%、84.2%。验证组与模型组的AUROC差异无统计学意义(Z值分别是0.360、0.885、0.046，P值均>0.05)。在全部病例中，FV诊断上述三种纤维化程度的AUROC均优于FIB-4指数、S指数（Z值分别是4.569、3.423，5.640、4.709，4.652、4.439，P值均<0.05）。如果以< 0.374和≥0.577作为除外和诊断显著纤维化的界值，可使61.1%（512/838）的患者避免肝活组织检查，准确率为92.6%（474/512）。 结论： 基于年龄、PLT、GGT、A/G、SSA的非创伤性模型可较准确地预测慢性乙型肝炎患者纤维化程度，并有良好的重复性，便于临床肝纤维化程度的动态监测。.