ABSTRACT
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354 11.5 Methods to Evaluate Statistically the Inconsistency
in a Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 11.5.1 Local Approaches for Inconsistency . . . . . . . . . . . . . . . . . . . 357
11.5.1.1 Loop-Specific Approach . . . . . . . . . . . . . . . . . . . . . 357 11.5.1.2 Composite Test for Inconsistency . . . . . . . . . . . . . . 359 11.5.1.3 Node-Splitting Approach . . . . . . . . . . . . . . . . . . . . 360
11.5.2 Global Approaches for Inconsistency . . . . . . . . . . . . . . . . . . 362 11.5.2.1 Lu and Ades Model . . . . . . . . . . . . . . . . . . . . . . . . 362 11.5.2.2 Design-by-Treatment Interaction Model . . . . . . . . . 364 11.5.2.3 Q-Statistic in NMA . . . . . . . . . . . . . . . . . . . . . . . . . 366
11.5.3 Overview of Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368 11.6 Presentation of Data and Results in Network Meta-Analysis . . . . . 369
11.6.1 Presentation of the Evidence Base . . . . . . . . . . . . . . . . . . . . 369 11.6.1.1 Network Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . 369 11.6.1.2 Contribution Plot . . . . . . . . . . . . . . . . . . . . . . . . . . 369
11.6.2 Presenting the Assumptions of the Analysis . . . . . . . . . . . . 371 11.6.2.1 Inconsistency Plot . . . . . . . . . . . . . . . . . . . . . . . . . 371
11.6.3 Presenting the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
11.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374 Appendix 11A.1: Data for the Worked Examples in This Chapter . . . . . . 375 Appendix 11B.1: Software Options for NMA . . . . . . . . . . . . . . . . . . . . . 378 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
ABSTRACT Meta-analysis is now established in comparative effectiveness research as a valid statistical tool, useful for synthesizing evidence from studies that compare two treatments for the same condition. For the majority of diseases, however, more than two treatments are available and trials might compare different subsets of them. When this is the case, a simple, pairwise meta-analysis cannot provide a definite answer to decision makers as to which intervention is associated with the largest benefit for patients with the target condition. In order to address this very common scenario, a statistical tool has been developed called network meta-analysis (NMA). NMA can be used to jointly analyze the totality of evidence in order to provide estimates for all relative treatment effects, to compare treatments that have never been compared head-to-head, and to obtain a ranking of all competing interventions in order to further facilitate the decision-making process. In this chapter, we introduce the basic concepts and discuss the underlying assumptions of NMA. We present alternative methods for fitting models and for assessing the validity of the underlying assumptions. We especially focus on the various statistical methods that can be applied for checking the consistency of the evidence, a key aspect of NMA. We provide worked examples in order to illustrate all methods presented in the chapter. Finally, we discuss various graphical and tabular options for presenting the results.
