ObjectiveTo establish a diagnosis model combining four serum markers of hepatic ifbrosis (HA, LN,PⅢNP,CⅣ) and assess the diagnostic performance of model aimed to discriminate between patients with and without signiifcant hepatic ifbrosis in patients with hepatitis B.Methods Total of485 patients with hepatitis B were randomly divided into an estimation group (324 cases) and a validation group (161 cases). The serum levels of HA, LN, PⅢNP and CⅣ were measured by radiommunoassay in all patients and liver biopsy was used as the gold standard. A binary logistic regression model was established combining four serum markers and applied to the validation group to test its accuracy. The diagnostic value of the model was assessed by the receiver operating characteristic (ROC) curves.ResultsBinary logistic regression identiifed HA, PⅢNP and CⅣ as independent predictors of ifbrosis. We constructed a model named FSM combining HA, PⅢNP and CⅣ that proved useful to identify patients with or without significant hepatic fibrosis in patients with hepatitis B. FSM=1.56 ×ln(HA)+0.92 ×ln(PⅢNP)+1.9 ×ln(CⅣ)-8.18. The area under the ROC curve (AUC) was 0.85 for predicting signiifcant ifbrosis. In validation group, the AUC was also 0.85 for predicting significant fibrosis. Using optimized cutoff values, the sensitivity and negative predictive value of predicting the absence of significant fibrosis (FSM<10.8) were 88.31% and 69%,while the speciifcity and positive predictive value of predicting the presence of signiifcant ifbrosis (FSM≥12.6) were 93.55% and 95.3%. Comparing the FSM diagnostic model and HA, the Youden’s index, positive likelihood ratio, sensitivity, speciifcity and accuracy of FSM model were higher than those of HA, and AUC had signiifcant difference between the two (Z=2.06,P<0.05).ConclusionsThe FSM model established in this study improved the diagnostic performance. It can make the 65%-70% patients without signiifcant hepatic ifbrosis to avoid “liver biopsy”. As a noninvasive method, it can be employed for monitoring the progression of hepatic ifbrosis.%目的：构建以“肝纤四项”为变量的诊断模型，并探讨该诊断模型对明显肝纤维化的诊断效能。方法选取485例乙型病毒性肝炎患者，随机分为建模组324例和验证组161例，采用放射免疫分析法检测建模组和验证组患者血清中HA、LN、PⅢNP、CⅣ的含量；采用二元Logistic回归分析法，以肝组织活检作为“金标准”，构建以“肝纤四项”为变量的诊断模型；采用受试者工作特征（ROC）曲线，评价该诊断模型的诊断效能；用独立的验证组检验该模型的诊断效率。结果在二元Logistic回归分析中，确定了3项与研究终点相关的独立预测指标（HA、PⅢNP和CIV），由这3个指标建立了诊断模型FSM，FSM=1.56 ln（HA）＋0.92 ln（PⅢNP）＋1.90 ln（CⅣ）－8.18。FSM用于预测明显肝纤维化具有较高的诊断价值，ROC曲线下面积（AUC）为0.85。该模型应用于验证组，其诊断明显肝纤维化的AUC同样为0.85。根据ROC曲线，用于预测明显肝纤维化时，FSM取10.8为排除界值，敏感性和阴性预测值分别为88.31%和69%；取12.6为诊断界值，特异性和阳性预测值分别为93.55%和95.3%。比较FSM模型与HA的诊断效能，FSM模型的约登指数、阳性似然比、敏感性、特异性和准确率均高于HA，两者的AUC差异有统计学意义（Z =2.06，P ＜0.05）。结论本研究所建立的“肝纤四项”联合检测FSM模型，一定程度上提高了诊断效能。将该模型应用于预测明显肝纤维化，可使65%～70%无或轻度肝纤维化乙型肝炎患者避免“肝组织活检”。作为一种非创伤性检查手段，该模型可用于动态监测乙型肝炎患者肝纤维化进程。
展开▼
