Epidemiological, methodological, and statistical characteristics of network meta-analysis in anaesthesia: a systematic review

Network meta-analyses (NMAs) combine direct and indirect estimates to provide mixed (or network) estimates of effect sizes. The scientific rigour of the conduct and reporting of anaesthesia NMAs is unknown. This review assessed the epidemiological, methodological, and statistical characteristics of anaesthesia NMAs. We searched four databases for anaesthesia NMAs and developed a 64-item checklist to evaluate NMAs. For 29 binary items, we defined compliance as ‘the ratio of NMAs that was awarded a ‘yes’ for that item, divided by the total number of NMAs. The compliance of such binary items was reclassified as very low (≤25%), low (26–50%), fair (51–75%), and high (>75%). We amalgamated findings from 29 key items to provide specific recommendations (post hoc). We compared the compliance of NMAs in anaesthesia across 26 items, with that of cancer NMAs and Cochrane NMAs, and analysed improvement over time (post hoc). Among 62 included NMAs, compliance was low (26–50%) for protocol registration, use of PRISMA-NMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses for NMA), publication bias assessment, evidence appraisal, reporting of Bayesian methodology and consistency evaluation. Compliance was very low (≤25%) for bias assessment, biostatistician involvement, search specialist, and use of predefined important differences. Anaesthesia NMAs need improvement in their conduct and reporting. Anaesthesia journals should mandate the registration of protocols and reporting of NMAs using PRISMA-NMA. Authors should carefully assess publication bias, and use updated bias assessment tools, and evidence appraisal methods designed for NMAs. PROSPERO CRD42021227608.