Network meta-analysis: making best use of the evidence for better decisions

Activity: Talk or presentationInvited talk


Meta-analyses are often used to synthesise evidence from multiple studies in order to decide which treatment is most effective (or cost-effective) out of several alternatives but often more than two treatments are available for a condition of interest. Separate meta-analyses of each pair of treatments is data- and time-inefficient and can lead to conflicting conclusions. Network meta-analysis (NMA) is commonly used to compare multiple treatments by combining all the studies making comparisons of any two or more treatments of interest. Thus, NMA makes best use of the available data by combining direct and indirect evidence on all comparisons of interest, by simply extending the familiar assumptions in pairwise (two-treatment) meta-analysis.

By jointly synthesising evidence from multiple sources, a degree of ‘evidence redundancy’ can be created in a network of treatment comparisons, and this can be used to inform additional parameters of interest. In addition, the typical Bayesian hierarchical modelling framework used to implement NMAs can also be used to mitigate some common issues caused when treatment networks are sparse, e.g. when many treatments but few studies are available for a particular decision-problem.

We will discuss the underlying assumptions in NMA and describe (1) models that attempt to estimate and adjust for bias due to study design characteristics and (2) “class models” where an overall relative treatment effect for a class, or group, of treatments is assumed, whilst retaining individual treatment effects which may have their own specific costs and adverse events.

Data requirements, assumptions and limitations will be discussed with reference to some real examples.
Period30 Sept 2021
Event titleExploring the Limits of Advanced Meta-analysis
Event typeWorkshop
LocationGermanyShow on map
Degree of RecognitionInternational