Improvement needed in the network geometry and inconsistency of Cochrane network meta-analyses: a cross-sectional survey

• Published in: Journal of clinical epidemiology 2019-09, Vol.113, p.214-227
• Main Authors: Gao, Ya, Ge, Long, Ma, Xueni, Shen, Xiping, Liu, Meng, Tian, Jinhui
• Format: Article
• Language: English
• Subjects: AMSTAR 2; Bias; Cancer; Cochrane library; Correlation analysis; Cross-Sectional Studies; Data collection; Epidemiology; Funding; Geometry; Humans; Internal Medicine; Library collections; Meta-analysis; Meta-Analysis as Topic; Meta-epidemiology; Network meta-analyses; PRISMA-NMA; Quality; Quality assessment; Research Design - standards; Research Design - statistics & numerical data; Research Report - standards; Software; Studies; Surveys and Questionnaires; Systematic Reviews as Topic
• ISSN: 0895-4356; 1878-5921
• DOI: 10.1016/j.jclinepi.2019.05.022

AbstractObjectivesThe aim of the study was to investigate the general characteristics and methodological and reporting quality of network meta-analyses (NMAs) published in the Cochrane library. Study Design and SettingWe conducted a comprehensive search of the Cochrane library in April 2018 and included 42 NMAs. We used the Assessing the Methodological Quality of Systematic Reviews (AMSTAR) 2 to assess methodological quality and Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA)-NMA for reporting quality. Stratified analysis and correlation analysis were conducted to explore the factors that might affect the quality. ResultsA total of 42 NMAs investigated 29 topics. The compliance of PRISMA-NMA was moderate. Only 26.2% NMAs described the geometry of network, 64.3% presented the network plot, and 33.3% fully assessed the inconsistency. The overall methodological quality was low. Only 11.9% NMAs explained the selection of study designs, and 40.5% investigated the publication bias. The compliance of PRISMA-NMA was higher with the increase of the AMSTAR 2 compliance rates (Spearman's ρ = 0.630, P = 0.000). NMAs with statistical or epidemiological authors often better reported the titles ( P = 0.032). Compared with nonfunding NMAs, nonindustry funding NMAs often better reported data collection process ( P = 0.028), planned methods of analysis ( P = 0.034), and synthesis of results ( P = 0.028). ConclusionThe quality still needs to be further improved, especially referring to the assessment of publication bias, the geometry of network, and assessment and exploration of inconsistency.