Mapping the characteristics of network meta-analyses on drug therapy: A systematic review

BACKGROUND: Network meta-analysis (NMA) is a new tool developed to overcome some limitations of pairwise meta-analyses. NMAs provide evidence on more than two comparators simultaneously. This study aimed to map the characteristics of the published NMAs on drug therapy comparisons. METHODS: A systematic review of NMAs comparing pharmacological interventions was performed. Searches in Medline (PubMed) and Scopus along with manual searches were conducted. The main characteristics of NMAs were systematically collected: publication metadata, criteria for drug inclusion, statistical methods used, and elements reported. A methodological quality score with 25 key elements was created and applied to the included NMAs. To identify potential trends, the median of the publication year distribution was used as a cut-off. RESULTS: The study identified 365 NMAs published from 2003 to 2016 in more than 30 countries. Randomised controlled trials were the primary source of data, with only 5% including observational studies, and 230 NMAs used a placebo as a comparator. Less than 15% of NMAs were registered in PROSPERO or a similar system. One third of studies followed PRISMA and less than 9% Cochrane recommendations. Around 30% presented full-search strategies of the systematic review, and 146 NMAs stated the selection criteria for drug inclusion. Over 75% of NMAs presented network plots, but only half described their geometry. Statistical parameters (model fit, inconsistency, convergence) were properly reported by one third of NMAs. Although 216 studies exhibited supplemental material, no data set of primary studies was available. The methodological quality score (mean 13·9; SD 3·8) presented a slightly positive trend over the years. CONCLUSION: The map of the published NMAs emphasises the potential of this tool to gather evidence in healthcare, but it also identified some weaknesses, especially in the report, which limits its transparency and reproducibility.