Heterogeneity and Subgroup Analysis in Network Meta-Analysis

Abstract

Comprehensive healthcare decision-making requires a comparisons of the relevant competing treatment options for a particular disease state. Randomized controlled trials (RCTs) are considered the most credible evidence to obtain insight into the relative treatment effects of a medical intervention. However, an individual RCT rarely includes all competing interventions of interest. Typically, the evidence base consists of multiple RCTs where each of the available studies compares a subset of all the competing interventions of interest. If each of these trials has at least one intervention in common with another trial such that the evidence base can be represented with one connected network, a network meta-analysis (NMA) can provide relative treatment effects between all competing interventions of interest (see the network diagram in Fig. 18.1) (Ades 2003; Bucher et al. 1997; Dias et al. 2013a, 2018a; Hutton et al. 2015; Jansen et al. 2011, 2014; Lumley 2002; Lu and Ades 2004; Salanti et al. 2008). A NMA can be considered a generalization of conventional pairwise meta-analysis (Dias et al. 2018b, c). Rather than synthesizing the findings of multiple RCTs each comparing the same intervention with the same control, with a NMA we are simultaneously synthesizing the findings of multiple pair-wise comparisons across a range of interventions and obtaining estimates of relative treatment effects between all competing interventions based on direct and/or indirect evidence. Even if there was a conclusive RCT that included all competing interventions of interest, the available RCTs comparing a subset of the interventions provide relevant evidence as well. A NMA allows to estimate relative treatment effects based on the totality of the RCT evidence base.