Meta-analysis for Diagnostic Devices
View Presentation View Presentation
*Martin Ho, FDA/CDRH 


Meta-analysis of diagnostic test studies facilitates our understanding of a particular diagnostic test by combining different assessments of the test's accuracy from a collection of completed studies. Although typically these studies were not originally designed as a group of experiments in an identical setting, various statistical techniques have been proposed to reduce the bias caused by study heterogeneity retrospectively. In this presentation, some fundamental principles that the CDRH uses to assess the quality of a meta-analysis will be highlighted and elaborated. Like any sound analysis of a prospective experiment, a valid meta-analysis requires a prospectively specified design to evaluate the diagnostic test based on retrospective results. Selected examples will be provided as a context for discussion.