Diagnostic test accuracy network meta-analysis methods: A scoping review and empirical assessment
2University of Toronto; St. Michael’s Hospital, Unity Health Toronto, Canada
3Test Evaluation Research Group, Institute of Applied Health Research, University of Birmingham, UK
4Department of Hygiene, Social-Preventive Medicine and Medical Statistics, Medical School, Aristotle University of Thessaloniki, Greece
5Institute of Medical Biometry and Statistics, Faculty of Medicine and Medical Center - University of Freiburg, Germany
6Department of Primary Education, School of Education, University of Ioannina, Greece
7Test Evaluation Research Group, Institute of Applied Health Research, University of Birmingham; NIHR Birmingham Biomedical Research Centre, University Hospitals Birmingham NHS Foundation Trust and University of Birmingham, UK
Background: The first and crucial step in the intervention selection process is the diagnosis of the patient’s condition. In healthcare settings, diagnostic tests are commonly used to confirm or exclude a target condition (disease). Most diagnostic test accuracy (DTA) meta-analyses evaluate the accuracy of a single index test at a time; however, there may be several index tests available for a target condition. Comparative accuracy of index tests for a certain target condition is important for decision-making. To compare the accuracy of multiple index tests, several comparative meta-analysis models have been developed over the past few years.
Objectives: To present and discuss methods that have been recognized in our recent scoping review for comparing at least three index tests coming from a variety of studies within a single framework.
Methods: We conducted a scoping review up until March 3, 2021, to identify methodological and application papers reporting a model developed specifically for diagnostic test accuracy network meta-analysis (DTA-NMA) or a hierarchical meta-regression method for comparing at least three index tests. We reviewed and summarized the characteristics of the collected methods and the application papers and compared DTA-NMA and hierarchical meta-regression methods empirically, using the collected application papers.
Results: Publications of comparative meta-analyses with three or more index tests have increased steeply over the past few years. Our research resulted in 49 articles plus 1 companion report: 9 methodological (describing 11 different DTA-NMA models) and 40 application papers of DTA-NMAs or meta-regression. Most of the application papers employed a meta-regression model. In the empirical assessment, meta-regression produced on average higher estimates of specificity than DTA-NMA models while sensitivity estimates were comparable.
Conclusions: DTA-NMA models may yield different results. The inclusion of different thresholds for test positivity, design differences, and software familiarity will determine which DTA-NMA model is appropriate.
Patient, public and/or healthcare consumer involvement: With a good and accurate diagnosis, patients have the best chance of achieving a positive health outcome based on clinical decisions made after understanding their health issue correctly.