Same family, different species: methodological conduct and quality varies according to purpose for five types of knowledge synthesis

Objectives: The aim of the study was to characterize methodological conduct, reporting, and quality of five knowledge synthesis (KS) approaches. Study Design and Setting: Retrospective analysis of a convenience sample of five published databases of KS approaches: overview of reviews (n = 74), scoping reviews (n = 494), rapid reviews (n = 84), systematic reviews (n = 300), and network meta-analyses (NMAs; n = 456). Data in the five published databases were abstracted by two reviewers independently, any missing data for this retrospective analysis were abstracted by one experienced reviewer. Methods were appraised using the A MeaSurement Tool to Assess systematic Reviews (AMSTAR) tool. Descriptive analysis was performed. Results: Reporting the use of a protocol ranged from 4% for rapid reviews to 32% for systematic reviews. The use of two reviewers for citation and full-text screening ranged from 20% for scoping reviews to 60% for NMAs. Data abstraction was performed in duplicate for 11% of rapid reviews and 54% of NMAs, and for risk of bias appraisal, this ranged from 6% for scoping reviews to 41% for NMAs. NMAs had the highest median percentage of maximum obtainable AMSTAR score (64%; Q1–Q3:45–73%), while scoping reviews had the lowest (25%; Q1–Q3:13–38%). Conclusion: NMAs consistently scored the highest on the AMSTAR tool likely because the purpose is to estimate treatment effects statistically. Scoping reviews scored the lowest (even after adjusting the score for not relevant items) likely because the purpose is to characterize the literature.