Network meta-analysis & models for inconsistency

Abstract: Network meta-analysis (multiple treatments meta-analysis) is an increasingly popular way to combine a body of evidence and try to identify the best intervention. I will give an introduction to statistical models for network meta-analysis and how they are fitted and interpreted. A major assumption of these models is that different sources of evidence are mutually consistent. We can test this assumption using models that allow for inconsistency. I will describe the “design-by-treatment interaction model” and compare it with other inconsistency models in the literature. I will also discuss what can be done when inconsistency is detected.