Users’ guides to assessing multiple treatment comparison meta-analysis

Article type
Authors
Mills E1, Ioannidis J2, Thorlund K3, Schunemann H3, Puhan M4, Guyatt G3
1University of Ottawa, Canada
2Stanford University, USA
3McMaster University, Canada
4Johns Hopkins Bloomberg School of Public Health, USA
Abstract
Background: The Users’ Guides to the medical literature have provided frameworks for assessment of primary studies and systematic reviews that have proved useful to clinicians, investigators and methodologists. Multiple treatment comparison meta-analysis (MTC) (network meta-analysis) is a rapidly expanding field. There are few if any systematic approaches to assessing the credibility of MTC.

Objectives: To develop a Users’ Guide for MTC

Methods: Using the established Users’ Guide framework (validity; results; applicability) the authors conducted an iterative process of suggesting criteria, feedback, and modification to arrive at a framework all authors found acceptable.

Results: Criteria include those associated with conventional meta-analysis (Validity: sensible question, comprehensive search, assessment of risk of bias in primary studies, reproducible assessment of eligibility and risk of bias; Were the results similar from study to study; Applicability: consideration of all patient-Important outcomes, credibility of subgroup effects, overall quality of the evidence). Issues specific to MTC include: (i) Assessing the geometry of the network and the amount of evidence available. Severe imbalance in terms of the number of included trials, patients and/or events for each intervention (e.g. due to selected, interest-driven comparisons or publication bias), greatly weakens confidence in estimates of effect in the underrepresented comparisons, as inferences will be driven largely from the evidence on few treatments and comparisons. (ii) Assessing consistency in direct and indirect comparisons. When results of direct and indirect comparisons differ (incoherence) explanations can include chance; differences in patients, interventions, or outcomemeasurement in direct and indirect comparisons; bias in direct comparisons; bias in indirect comparions. (iii) Were all relevant management options included? (iv) Were the results robust to sensitivity analysis?

Conclusions: Structured assessment criteria can aid users of MTC meta-analysis in determining credibility and implications for practice and further research.