摘要
目的采用常规实验室指标构建肝纤维化数学诊断模型，并评价其对慢性乙型肝炎（CHB）肝纤维化肝硬化的预测价值。方法2005年3月至2008年5月在福建医科大学附属第一医院肝病中心确诊的CHB患者共391例，同期进行肝组织病理学检查及常规实验室检查，应用多因素非条件Logistic回归分析相关指标，构建不同肝纤维化分期的预测模型并用受试者工作特征（ROC）曲线评价其诊断价值。结果经Spearman等级相关分析，筛选出年龄、血小板（PLT）、国际标准化比值（INR）、总胆红素、白蛋白（ALB）、天冬氨酸转氨酶、γ-谷氨酰转肽酶（GGT）、总胆汁酸及胆碱酯酶（CHE）与肝纤维化程度相关的指标（P〈0．01），进行单因素方差分析与多因素非条件Logistic回归分析，筛选出Model-1（S≥2）、Model-2（S≥3）、Model-3（S=4）组包括PLT、INR、ALB、GGT、CHE的肝纤维化独立预测因子，最终构建数学模型并计算肝纤维化评分（fibrosis score，FS）；Model-1、Model-2和Model-3中Fs的ROC曲线下面积分别为0．784、0．768、0．806，最优截断点分别约为7．09、5．67、3．65，灵敏度分别为67．4％、75．0％、71．4％，特异度分别为79．3％、67．7％、78．5％，阳性预测值分别为87．9％、83．9％、88．1％，阴性预测值分别为52．2％、55．0％、55．2％，准确率分别为71．1％、72．9％、73．7％。结论由常规实验室检查指标构建的肝纤维化预测模型对CHB肝纤维化肝硬化有较好的诊断价值。
Objective To build a mathematical model for diagnosing liver fibrosis progression by using conventional laboratory indicators, and to evaluate its clinical value of predicting hepatic fibrosis and hepatocirrhosis in chronic hepatitis B. Methods Liver biopsy and routine laboratory tests were performed in 391 patients with chronic hepatitis B. Using Multiple logistic regression to analyse evidently relevant indicators, then the models predicting for different stages of liver fibrosis were built and analyzed by receiver operating characteristic （ ROC ） curve. Results Age, platelet （ PLT ）, international rate （ INR ）, total bilirubin, albumin （ ALB ）, aspartate aminotransferase, gamma-glutamyhranspeptidase （ GGT ）, total bile acid and cholinesterase （CHE） were correlated with liver fibrosis stage. Multiple Logistic regression analysis showed that PLT, INR, ALB, GGT and CHE were independent predictors of three models（ S ≥2, S ≥ 3, S = 4）. We finally built the predicting models and got Fibrosis scores （FS）. ROC curve analysis revealed that the area under the curve was O. 784 in model-1 （ S≥2 ）, 0. 768 in model-2 （ S≥ 3 ） and 0. 806 in model-3 （S =4）. A FS1 cutoff point of 7. 09 had 67.4% sensitivity, 79. 3% specificity and 71.1% accuracy in Model-1. A FS： cutoff point of 5.67 had 75.0% sensitivity, 67. 7% specificity and 72. 9% accuracy in Model-2. A FS3 cutoff point of 3.65 had 71.4% sensitivity, 78.5% specificity and 73.7% accuracy in Modelo3. Conclusion The mathematical models using conventional laboratory indicators have fairly well value for predicting hepatic fibrosis progressing in chronic hepatitis B.
出处
《中华医学杂志》
CAS
CSCD
北大核心
2009年第33期2349-2352,共4页
National Medical Journal of China
