Implications of Different Evidence-Summaries for Contextualizing and Interpreting Findings of Meta-Analysis of Diagnostic Tests

• Activity Number: 65
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Anja Zgodic*+ and Thomas Trikalinos and Ingram Olkin and Christopher Schmid and Joseph Lau and Issa Dahabreh
• Companies: Brown University and Brown University and Stanford University and Brown University and Brown University and Brown University
• Keywords: meta-analysis ; diagnostic tests ; evidence summaries ; interpretation ; decision model

Abstract

To be informative, systematic reviews and meta-analyses should ask meaningful questions and facilitate the interpretation of findings in specific contexts. The interpretation of random effects meta-analysis models of medical tests is not straightforward, and can have important implications for conclusions because most such models summarize test performance by means of patient-relevant clinical outcomes. Using as example a meta-analysis of tests for diagnosing breast cancer recurrence (PET, PET/CT, conventional workup), we discuss the interpretation of four evidence summaries: the random effects meta-analysis mean of the true positive and false positive rate (TPR and FPR); the predicted TPR and FPR for an implementation of the tests in a new setting; and the estimated TPR and FPR in a setting similar to that of a specific study, either disregarding or incorporating information from other meta-analysis studies. We find that the meta-analysis mean is rarely appropriate for describing the anticipated TPR and FPR in a new setting. We demonstrate the impact of choosing alternative estimates in a decision model that calculates the comparative effectiveness of the tests.