Bayesian multivariate meta-analysis for modelling surrogate endpoints in HTA

Abstract: Bayesian statistics provides flexible framework for modelling complex data structures by allowing multiple parameters to be modelled simultaneously in a convenient way. Multivariate meta-analysis (MVMA) allows to model jointly treatment effects on multiple correlated outcomes. Despite of substantial methodological developments in this area  and many advantages of this approach to evidence synthesis, MVMA has not been widely used. This talk aims to introduce the concepts of Bayesian methods for efficient synthesis of evidence in HTA. MVMA methodology will be introduced and its application illustrated in scenarios for its use for purpose of (i) combining data on correlated outcomes, such as surrogate endpoints, from diverse sources of evidence, (ii) predicting treatment effect on a target clinical endpoint from treatment effects on surrogate endpoints, and (iii) exploiting such prediction framework to inform decision modelling, based onexamples in multiple sclerosis and cancer.