ROC曲线法评价简易无创模型预测乙型肝炎相关肝硬化

Appraisal of simple non-invasive models for predicting hepatitis B associated cirrhosis by ROC curve method

目的评价简易无创模型AAR（AST-to-ALT ratio）、APRI（AST-to-platelet ratio index）、SPRI（spleen-to—platelet ratio index）、API（age—platelet index）、ASPRI（age-spleen—to-platelet ratio index）预测乙型肝炎相关肝硬化的临床价值。方法慢性乙型肝炎170例，其中病理诊断为非肝硬化138例，肝硬化32例。参照原始文献构建预测肝纤维化程度的简易无创模型。统计分析采用SPSS13．0软件。简易无创模型在非肝硬化与肝硬化病人之间的比较采用两独立样本的t检验。简易无创模型预测肝硬化的评价采用ROC曲线法。结果肝硬化病人的平均AAR、SPRI、API、ASPRI显著大于非肝硬化病人（P=0．000、0．009、0．000、0．005），平均APRI相近于非肝硬化病人（P=0．223）。AAR、APRI、SPR／、API、ASPRI预测肝硬化的ROC曲线下面积为0．81、0．62、0．71、0，73、0．74，均显著大于诊断参考下面积（P=0．000、0．033、0．000、0．000、0．000）。结论AAR、APRI、SPRI、API、ASPRI均有预测乙型肝炎相关肝硬化的价值；其中，AAR的实践效能最大，APRI实践效能最小。

Objective To appraise the clinical efficacy of simple non-invasive models of AST-to-ALT ratio （AAR）, AST-to-platelet ratio index （APRI）, spleen-to-ptatelet ratio index （SPRI）, age-platelet index （API）, age-spleen-to-platelet ratio index （ASPRI） for predicting hepatitis B associated cirrhosis. Methods A total of 170 consecutive patients with chronic hepatitis B were retrospectively studied, in which 138 patients and 32 patients were pathologically diagnosed as non-cirrhosis and cirrhosis. The simple non-invasive models were calculated as described originally. SPSS 13.0 was used for statistical analysis. Independent sample t tests were used for comparisons of the simple non-invasive models between patients with non-cirrhosis and cirrhosis. ROC curve was used for appraisement of simple non-invasive models to predict cirrhosis. Results The average AAR, SPRI, API, ASPRI in patients with cirrhosis were significantly higher than those in patients with non-cirrhosis （P=0.000, 0.009, 0.000, 0.005） , but the average APRI in patients with cirrhosis was similar to that in patients with non-cirrhosis （P=0.223）. The areas under ROC curve of AAR, APRI, SPRI, API, ASPRI for predicting the cirrhosis were 0.81, 0.62, 0.71, 0.73, 0.74, respectively, which were larger than those under the diagnosis reference line （P=0.000, 0.033, 0.000, 0.000, 0.000, respectively）. Conclusions All of the simple non-invasive models of AAR, APRI, SPRI, API, ASPRI can be used for predicting hepatitis B associated cirrhosis; Among them, AAR has the most practical efficacy and APRI the least practical efficacy for predicting hepatitis B associated cirrhosis.