Publisher's Synopsis
Comparative effectiveness reviews (CERs) often aim to compare the benefits and harms of multiple available approaches for treating a health condition with the ultimate goal of informing clinical practice and other decisionmaking. To this end, analysts conducting CERs aim to find studies conducting direct head-to-head comparisons. However, direct head-to-head evidence on competing interventions is often scant. As a result, several methods to conduct indirect comparisons have been proposed. These include meta-regression, logistic regression, the Bucher method, and, more recently, Bayesian mixed treatment comparison (MTC) meta-analysis.MTC meta-analysis is a relatively new methodology. Various other terms have been used to describe the approach, including multiple treatment comparisons and network meta-analysis. Terminology has evolved to where most experts in that field now refer to the broad area of comparison of different treatments as network meta-analysis and restrict the use of MTC to describe methods that explicitly look at combining direct and indirect evidence. One of the most compelling reasons to use MTC meta-analysis is that it allows for the combination of both direct head-to-head and indirect evidence (e.g., placebo-controlled trials) in one modeling framework. The use of all potentially relevant available evidence is an appealing feature for analysts, because other methods rely solely on one type of evidence. In addition, unlike other indirect analysis methods, MTC meta-analysis allows all relevant comparisons to be made through a single analysis, providing the information to calculate an effect size for each comparison of interest and to rank treatments based on the probability of being the best treatment. The main objectives of this report are to contribute to the body of literature on MTC meta-analysis by examining (1a) how results of Bayesian MTC methods compare with several frequentist indirect methods for various types of outcome measures, (1b) how Bayesian MTC methods perform for different types of evidence network patterns, (2) how study-level covariates can be incorporated with Bayesian MTC meta-analysis to explore heterogeneity through meta-regression, and (3) how findings of Bayesian MTC meta-analysis compare for different numbers of studies and different network pattern assumptions. For objectives 1 and 2, we aimed to conduct case studies using data from two recent CERs. For objective 3, we aimed to use simulated data. We address the KQs listed below. KQ 1. How do the results of Bayesian MTC meta-analysis methods compare with those of several frequentist indirect methods? Related questions of interest included the following: For each of the common evidence network patterns, how do the Bayesian MTC methods compare with frequentist indirect methods? How do Bayesian MTC methods perform (e.g., precision, convergence) for different types of evidence network patterns? KQ 2. How can meta-regression be used with Bayesian MTC meta-analysis to explore sources of heterogeneity? KQ 3. How do findings of Bayesian MTC meta-analysis compare for different numbers of studies and network pattern assumptions?
