Advanced meta-analysis of diagnostic test accuracy: dealing with heterogeneity and test comparisons

Article type
Macaskill P1, Takwoingi Y2
1University of Sydney
2Screening and Diagnostic Tests Methods Group
1. To explain the recommended statistical models used for diagnostic test accuracy meta-analysis and the concepts underlying these methods.
2. To illustrate the methods and their interpretation using examples.

Description: This workshop will cover advanced statistical methods for the meta-analysis of studies of diagnostic test accuracy that report estimates of sensitivity and specificity. We will cover the recommended hierarchical models and the relationship between them: 1) the bivariate model, which focuses on the estimation of a summary estimate of sensitivity and specificity; and 2) the Hierarchical Summary ROC model of Rutter and Gatsonis, that focuses on the estimation of a summary ROC (receiver operating characteristic) curve. Both methods will be illustrated by means of examples, and the rationale for which approach to adopt will be discussed. The use of these models for investigating sources of heterogeneity and also comparing the performance of alternative tests through the inclusion of study level covariates will be described and discussed.

This workshop is designed to explain and discuss the concepts and methods covered in the statistics chapter of the Cochrane Handbook for Diagnostic Test Accuracy reviews. Participants will interpret examples and accompanying computer output to facilitate understanding of key concepts and issues, and the use of RevMan to display results. The examples will also be used to motivate group discussions.