应用R软件Metatron程序包实现诊断准确性试验Meta分析及程序包汇总比较

摘要

There is a connatural negative correlation between sensitivity and specificity of diagnostic test accuracy (DTA). For avoiding the influence of this negative correlation on DTA review results, many scholars propose bivariate model. The bivariate model keeps the two-dimensional nature of original data, so it can produce summary estimated value of the sensitivity and specificity and negative correlation value through parameter fitting for analyzing the accuracy of DTA. At present, the most representative bivariate model is linear mixed bivariate model proposed by Reitsma et al. Metatron package is a package developed based on above mentioned model for implementing DTA Meta-analysis. The specialized packages for DTA Meta-analysis in R software were compared in this paper for users to choose.%诊断准确性试验（diagnostic test accuracy，DTA）的灵敏度与特异度之间存在固有的负相关性，为避免二者间负相关性对诊断试验结果的评价产生影响，很多学者提出了双变量模型，因其保留了原始数据的二维结构特性，双变量模型通过参数拟合可以得到灵敏度和特异度的综合估计量值及二者之间负相关的值，从而对诊断试验的准确性进行综合性分析。当前最具代表的是由Reitsma等提出的线性混合双变量模型，Metatron程序包正是基于此模型所研发的用于DTA Meta分析的程序包，同时本文将对R软件中专用于DTA Meta分析的程序包做出比较，便于使用者选择。