Introduction

Network meta-analyses synthesise networks of direct and indirect comparisons of interventions, and enable researchers to simultaneously assess the effects of more than two interventions for the same condition.1 Indirect evidence refers to estimates from different direct meta-analyses with a common comparator and allows for treatment comparisons that have not been directly compared in a clinical trial.1 Guidelines from the National Institute for Health and Care Excellence (NICE)2 and the Cochrane Collaboration3 typically prefer direct evidence from randomised clinical trials and conventional meta-analyses to indirect evidence. However, the WHO have recently begun using network meta-analyses to inform clinical guidelines,4 the global scientific production of network meta-analyses is increasing rapidly5 and some argue that the methodology should represent the highest level of evidence for instructing clinical decision-making.6 Are we witnessing a shift in what constitutes the highest level of medical evidence?

As network meta-analyses are becoming more influential in informing clinicians and decision-makers, we need a thorough discussion of reporting standards and methodological concerns. A previous paper published in Evidence-Based Medicine considered several challenges related to heterogeneity and inconsistency (see box 1 for a description of inconsistency) in network meta-analyses.7 The present analysis considers five additional topics relevant to network meta-analyses: quality of evidence, statistical power, random errors, multiplicity issues and treatment rankings. Our aim is to debate methodological and reporting problems for these topics, and their relevance for evidence-based practice. A glossary of key terms used throughout the article is provided in box 1.