Multivariate meta-analysis TSD

Downloads

TSD 20: Multivariate meta-analysis of summary data for combining treatment effects on correlated outcomes and evaluating surrogate endpoints (PDF, 1.2MB) – October 2019 – Updated December 2022: Trivariate meta-analysis model in product normal formulation with unstructured covariance has been checked and replaced in version 2.

TSD 20 one-page summary (PDF, 549KB)

About

The TSD describes the use of multivariate meta-analytic methods for synthesis of data from multiple correlated outcomes as well as for evaluation of surrogate endpoints, in the context of HTA.

Health technology assessment (HTA) agencies such as the National Institute for Health and Care Excellence (NICE) require evidence synthesis of existing studies to inform their decisions. For example; gathering information about the best available treatments with respect to multiple efficacy and safety outcomes. However relevant studies may not provide direct evidence about all the treatments or outcomes of interest. Studies that do not provide direct evidence about a particular outcome or treatment of interest are often excluded from a meta-analysis evaluating that outcome or treatment.

This exclusion is unwelcome, especially if participants are otherwise representative of the population, clinical settings and condition of interest. Research studies require considerable costs and time, and involve precious patient participation, and simply discarding them could be viewed as research waste. Statistical models for multivariate and network meta-analysis address this by simultaneously analysing multiple outcomes and multiple treatments, respectively. This allows more studies to contribute toward each outcome and treatment comparison.

This Technical Support Document (TSD) describes the key methods of multivariate meta-analysis and their extensions to network meta-analysis of multiple outcomes. We focus on the use of multivariate meta-analytic methods for combining data from multiple correlated outcomes with the aim of including all relevant evidence and borrowing of information across outcomes, in particular when not all of the relevant studies report an outcome of interest. We also devote considerable attention to the use of multivariate meta-analysis for the purpose of surrogate endpoint evaluation.

We describe and apply the methods in the Bayesian framework of estimation and provide WinBUGS code for a Bayesian analysis using Markov chain Monte Carlo (MCMC) simulation.