Background: network meta-analyses have increased in popularity in systematic reviews as they compare multiple treatments for the same clinical condition. Publishing systematic review protocols is a fundamental part of Cochrane Systematic Reviews to ensure transparency and reproducibility. As current Review Manager software does not support network meta-analyses, the reporting of network meta-analyses as well as their protocols lacks standardization, although guidance on preparing systematic review protocols with network meta-analyses has been proposed recently.
Objectives: to evaluate how Cochrane Systematic Review protocols with network meta-analyses are reported.
Methods: we searched protocols of systematic reviews with network meta-analyses published in 2018 in the Cochrane Database of Systematic Reviews. We excluded protocols on prognosis or diagnostic test accuracy. We assessed the characteristics and reporting of methodologies relevant to network meta-analyses in the protocols. We reported frequencies for each characteristic and reporting item.
Results: 27 systematic review protocols with network meta-analyses published in 2018 fulfilled the inclusion criteria. Most (93%) were protocols for intervention reviews and two (7%) were protocols for overviews of systematic reviews. They were from 16 different Cochrane Groups, and 63% were from Europe. We identified only 20 protocols (74%) as protocols for network meta-analyses from the title. Overall, 93% explained the need of network meta-analyses, 67% reported the considerations of transitivity assumption in the inclusion criteria, and 56% clearly differentiate interventions of direct interest for practice and interventions that provide indirect evidence, with only four (15%) presenting a conceptional network plot to illustrate the 'nodes' of interest. With regards to evidence synthesis methods and statistical analyses, 70% reported prespecified potential effect modifiers for the evaluation of transitivity, 74% reported measures for relative ranking, 93% reported the approaches of network meta-analysis, 70% reported the assumptions on heterogeneity variances, 85% reported methods to assess statistical inconsistency, 85% reported methods to assess potential reporting bias and 78% reported possible sources of important heterogeneity and inconsistency for network meta-analyses.
Conclusions: Cochrane Systematic Review protocols with network meta-analysis need improvement, especially regarding transitivity assumptions. Implementing existing guidance could potentially assist authors, reviewers and editors in preparing protocols of network meta-analyses and help readers in critically appraising published protocols.
Patient or healthcare consumer involvement: we will invite patients and healthcare consumers to comment on this research project, help with dissemination of the results and increase its accessibility from the patients’ perspective.