Event

Network meta-analysis of diagnostic test accuracy (DTA-NMA) extends the bivariate random-effects framework for sensitivity and specificity to K index tests, posing a 2K-variate modeling problem involving within-test sensitivity–specificity correlation, between-test dependence in paired designs, and threshold-related heterogeneity. Drawing on a scoping review of 50 publications, this talk critically compares existing approaches, including bivariate/HSROC meta-regression, arm-based hierarchical models (ANOVA and beta-binomial), contrast-based and latent-class models accommodating imperfect reference standards, and variance-component models for multiple thresholds, with emphasis on model assumptions, covariance structures, identifiability, and estimation. Results from an empirical comparison across 32 published diagnostic networks will illustrate differences in model performance and practical implications for inference. We will also examine transitivity and consistency in the joint modeling of sensitivity and specificity, discuss methodological challenges in assessing inconsistency and simultaneous test ranking. Real-data applications will highlight implications for model selection and identify outstanding statistical challenges in DTA-NMA.

Areti-Angeliki (Argie) Veroniki, PhD, is an Associate Professor of Health Services, Policy and Practice and of Biostatistics at the Brown University School of Public Health. Her research focuses on developing and evaluating statistical methods for evidence synthesis, including network meta-analysis of interventions and of diagnostic test accuracy, with the aim of providing patients, healthcare providers and policy-makers with reliable evidence for decision-making. She is Co-Chair of the Cochrane Methods Executive, Co-Convenor of the Cochrane Statistical Methods Group, and Statistical Editor of the Journal of Clinical Epidemiology.

This seminar will be held via Zoom only and will be recorded.

Time and date

Thursday, 26 November 9:30 – 10:30am AEDT

Zoom

Please click this URL to join
Or, go to https://monash.zoom.us/join and enter

Meeting ID: 812 6178 5701

Passcode: 868079