What is new?
- •
Since 2009, published systematic reviews that report mixed treatment comparison (MTC) have increased rapidly. These reviews are difficult to identify because databases offer no search method to separate these from the far more common reviews reporting pair-wise meta-analysis and because authors use various names when reporting MTCs. Previous reviews identified very small numbers of systematic reviews reporting MTCs up to 2008, so this increase since 2009 has not been previously reported. In view of the increasing use of this method, a consensus on terminology and standards for conduct and reporting would be timely.
Evidence-based health care in individual patient care or in the development of health policy requires synthesis of trial data to identify the best available evidence. Systematic reviews of health interventions aim to identify all trial data relevant to answering a clinical question. Individual clinical trials rarely have sufficient power to provide definitive answers to clinical questions, so methods of meta-analysis have been used to combine data from similar trials to achieve sufficient power. Standard pair-wise meta-analysis can only be applied to trials comparing directly the interventions of interest. Trials commonly compare a new intervention with placebo or with an existing leader in the field. Over time, several interventions may be the leader. Many clinical questions involve more than two potential interventions, and not all these interventions will have been compared directly with each other. Primary research to compare all the interventions is commonly not available, which has left health practitioners and makers of health policy with an incomplete series of pair-wise comparisons and consequently important gaps in the available evidence.
Data from trials can also be synthesized to make indirect comparisons between interventions. As with pair-wise meta-analysis, heterogeneity among the included trials is a potential source of bias to be considered when assessing the validity of the results of indirect comparison meta-analysis. Edwards et al. [1] reviewed the use of various methods of indirect comparison in systematic reviews and reported that the mixed treatment comparison (MTC) method was the only one, without relying on a single common comparator that could estimate the difference between interventions and provide a measure of the uncertainty around that estimate. The MTC, sometimes referred to as multiple treatment comparison or network meta-analysis, is a method that can be used to estimate the effects of multiple interventions, by combining both indirect and direct comparisons. This requires a network in which each intervention shares a comparator with at least one other. The MTC provides a single synthesis of direct and indirect comparisons between all interventions in the network.
Fig. 1 shows an example of a network for an MTC comparing five interventions, A–E. The solid lines show where trial results of direct comparisons exist. In this example, there is no single common comparator for all the interventions, but each intervention shares at least one comparator with another intervention in the network.
Types of network include, not exclusively, the star structure, in which only one intervention has been directly compared with each of the others, and single and multiloop structures, which contain direct comparisons between, respectively, one or more sets of at least three interventions.
MTC methodology is based on assumptions regarding similarity and consistency across all pair-wise sets of trials included from within the network of interventions being studied. There are formal and informal ways to assess the validity of these assumptions both statistically and clinically. Song et al. [2] outlined these assumptions and described inconsistency in both assessment and reporting of assumptions in indirect comparisons.
The origins of MTC methodology can be traced back to 1990 [3]. Both frequentist and Bayesian approaches to MTC are used, but according to the 2011 ISPOR Task Force report, “Bayesian methods have undergone substantially greater development” [4]. Bayesian methods can, for example, rank the effects of interventions by the order of probability that each is best. Lu and Ades [5] summarized key stages in the development toward current Bayesian MTC methodology and presented models that are commonly cited in reviews as the background to current use of this methodology. Further publications have continued development of the methodology [6], [7], [8], [9]. Because MTC meta-analysis of trials identified by systematic reviews potentially provides clinicians and policy makers with the maximum use of the available evidence to compare multiple interventions, this research was undertaken to summarize the currently available evidence of this form.