Sponsors
MTC: Mixed Treatment Comparison
Mixed Treatment Comparison (MTC), or network meta-analysis is a generalization of pair-wise meta-analysis to enable the analysis of complex evidence networks involving more than two alternative treatments.
As part of our research, we have been working to automate network meta-analysis. We provide the GeMTC software to automatically generate MTC analysis models that can be run in MCMC software (either BUGS or JAGS). In addition, ADDIS provides the automated estimation of the generated models.
About Mixed Treatment Comparison (MTC)
Mixed Treatment Comparison (MTC), or Network Meta-Analysis is a technique to meta-analyze more than two drugs at the same time. Using a full Bayesian evidence network, all indirect comparisons are taken into account to arrive at a single, integrated, estimate of the effect of all included treatments based on all included studies.
In network meta-analysis, because of the more complex evidence structure, we can assess inconsistency of evidence, in addition to heterogeneity within a comparison. Whereas heterogeneity represents between-study variation in the measured relative effect of a pair of treatments, inconsistency can only occur when a treatment C has a different effect when it is compared with A or B (i.e., studies comparing A and C are systematically different from studies comparing B and C). Thus, inconsistency may even occur with normal meta-analysis, but can only be detected using a network meta-analysis, and then only when there are closed loops in the evidence structure.
For more information about assessing inconsistency, see G. Lu and A. E. Ades (2006), Assessing evidence inconsistency in mixed treatment comparisons, Journal of the American Statistical Association, 101(474): 447-459. doi:10.1198/016214505000001302. An alternative method called node-split analysis can also be used, see S. Dias, N. J. Welton, D. M. Caldwell and A. E. Ades (2010), Checking consistency in mixed treatment comparison meta-analysis, Statistics in Medicine, 29 (7-8, Sp. Iss. SI): 932-944. doi:10.1002/sim.3767.
If there is no relevant inconsistency in the evidence, a consistency model can be used to draw conclusions about the relative effect of the included treatments. Using normal meta-analysis, we could only get a subset of the confidence intervals for relative effects we derive using network meta-analysis. Network meta-analysis gives a consistent, integrated picture of the relative effects. The Bayesian approach allows us to do even more with the data, and can be used to estimate the probability that each of the treatments is the best, the second best, etc. Rank probabilities sum to one, both within a rank over treatments and within a treatment over ranks.
Our research
We work on automating the tedious aspects of network meta-analysis. We hope this will enable network meta-analysis for a broader audience. In addition, we want to enable decision modeling based on network meta-analysis and since decision making often involves many criteria, it is critical that a large number of analyses can be carried out with minimal effort.