There is increasing interest in comparative meta-analysis of diagnostic studies which evaluates the accuracy of a test relative to an alternative test. Researchers have attempted to develop a network meta-analysis (NMA) like approach for evaluating two or more tests simultaneously.
Several approaches have been proposed in recent years (e.g. Trikalinos 2014, Ma 2015, Menten 2015, Dimou 2016 and Hoyer 2016) . They share some common features but there are also conceptual and statistical differences among them.
In this study, we aim to give a comprehensive review of the currently proposed approaches for comparative meta-analysis of DTA studies. We compared the following features with respect to conducting a comparative meta-analysis: 1) input data requested from primary studies; 2) assumption of common threshold; 3) focus on absolute or relative accuracy; 4) inclusion of indirect evidence; 5) detecting inconsistency; and, 6) statistical package of implementation.
We also investigated and compared the performance of these approaches using simulation. The simulation is done by 1) generating summary test results data (2×2×2 or 2×2 tables) based on assumptions of these approaches; and, 2) generating original test results from individual patients, which is more close to real world but in this case the assumptions on the distributions of sensitivity and specificity may not be valid.