What is Network Meta-Analysis?
A standard pairwise meta-analysis can compare the efficacy or safety of exactly two treatments that have been directly compared in head-to-head clinical trials. However, in practice, there are often many potential treatments for a single disease. Policy-makers, physicians and patients need to be able to select the best treatment from amongst the many potential options. Network meta-analysis is an extension of standard pairwise meta-analysis that can be used to simultaneously compare any number of treatments.
In the simplest possible example of a network meta-analysis, there are two treatments of interest that have been directly compared to a common comparator, but not to each other. Usually the common comparator will be placebo or a standard treatment. This situation is illustrated by Figure 1A. The circles indicate treatments and the lines connect treatments that have been directly compared in head-to-head clinical trials. In this example, treatments A and C have both been directly compared to treatment B, but there are no trials that compare treatments A and C.
There is no limit to the number of treatments, trials and patients that can be included in a network meta-analysis. However, in order to conduct a standard network meta-analysis, the treatments should form a connected network, such that there is path from each treatment to every other treatment in the network.
Key Takeaways
- Network meta-analysis (NMA) extends standard pairwise meta-analysis, allowing simultaneous comparison of multiple treatments by combining direct and indirect evidence in a connected treatment network.
- Indirect treatment comparison and mixed treatment comparison are both types of NMA: indirect comparisons use networks without loops, while mixed treatment comparisons combine direct and indirect evidence in networks with loops.
- Valid NMA depends on an adequate, systematically identified evidence base, and on trials being sufficiently similar with respect to potential treatment effect modifiers.
- Key structural assumptions for NMAs are homogeneity (within-comparison consistency), transitivity/similarity (validity of indirect comparisons), and consistency (agreement between direct and indirect evidence).
In Figure 1, diagrams A, B and C all illustrate connected networks. However, diagram D illustrates a disconnected network – there are no trials that connect treatments E and F to the rest of the network. We will look at how to handle disconnected networks in a future blog post.
Standard terminology
You may be familiar with other network meta-analysis terms, such as indirect treatment comparison and mixed treatment comparison. Different groups have used these terms in different ways, but, in this blog, we will follow the definitions proposed by the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Task Force on Indirect Treatment Comparisons (1).
Following the ISPOR task force definitions, ‘indirect treatment comparison’ and ‘mixed treatment comparison’ are both types of network meta-analysis. Indirect treatment comparison is used to describe the analysis of networks that do not contain any loops; mixed treatment comparison is used to described the analysis of networks that do contain loops. In Figure 1, diagrams A and C both illustrate indirect treatment comparisons, and diagram B illustrates a mixed treatment comparison. In diagram B, treatments A and B can be compared using direct evidence from AB trials, or indirect evidence via the BC and AC trials – hence, overall, the evidence for A versus B is a mix of the direct and indirect evidence.
A short history of network meta-analysis
A simple method for indirect treatment comparison was first introduced by Bucher et al. in 1997 (2); Bayesian methods suitable for indirect treatment comparison or mixed treatment comparison were published by Lu and Ades in 2004 (3). Since then, the use of network meta-analysis has risen steadily (4) and many health technology assessment agencies now accept network meta-analysis, including England’s National Institute of Health and Care Excellence (NICE).
Statistical methods for network meta-analysis are continually being developed and Quantics can provide advice on the best choice of methodology for both simple and complex networks. Our recent work includes using a three-level hierarchical network meta-analysis model to account for different doses and classes of drugs. This project included over 50 studies and 30 treatments and we applied network meta-analysis to several different outcomes. For another recent project, data was available from several different time points. We used a longitudinal network meta-analysis model to simultaneously synthesize all of the data and provide estimates of efficacy at different time points.
Further information
We hope you’ve found our first post informative. If there are any particular topics you’d like to see covered in the future then please let us know and make sure you have a look at our publications, presentations and posters which you can find here.
If you are looking for an in depth intro to network meta-analysis YHEC and Quantics run joint courses – currently available in-house, on request,
Find out more about our joint training program with YHEC
References – An introduction to network meta-analysis (mixed treatment comparison / indirect treatment comparison)
(1) National Institute for Health and Care Excellence. Guide to the methods of technology appraisal. 2013.
(2) Hutton B et al. The PRISMA Extension Statement for Reporting of Systematic Reviews Incorporating Network Meta-analyses of Health Care Interventions: Checklist and Explanations. Annals of Internal Medicine. 2015; 162: 777-784.
(3) Higgins JP et al. The Cochrane Collaboration’s tool for assessing risk of bias in randomised trials. BMJ. 2011; 343: d5928.
(4) Egger M et al. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997; 315: 629-634.
(5) Mavridis D et al. A selection model for accounting for publication bias in a full network meta‐analysis. Statistics in Medicine. 2014; 33(30): 5399-412.
(6) Trinquart L et al. A test for reporting bias in trial networks: simulation and case studies. BMC Medical Research Methodology. 2014; 14: 112.
(7) Higgins J et al. Measuring inconsistency in meta-analyses. BMJ. 2003; 327: 557-559.
(8) Dias S et al. Checking consistency in mixed treatment comparison meta-analysis. Statistics in Medicine. 2010; 29(7-8): 932–944.
(9) Salanti G. Indirect and mixed-treatment comparison, network, or multiple-treatments meta-analysis: many names, many benefits, many concerns for the next generation evidence synthesis tool. Res Synth Methods. 2012; 3: 80–97.
