Multivariate meta-analysis helps examine the impact of outcome reporting bias in Cochrane rheumatoid arthritis reviews

Ref ID 494
First Author G. Frosi
Journal JOURNAL OF CLINICAL EPIDEMIOLOGY
Year Of Publishing 2015
URL https://www.jclinepi.com/article/S0895-4356(14)00495-8/fulltext
Keywords • Cochrane
• Multiplicity
• Pre-specification
• Rheumatology
Problem(s) • Multiplicity of outcomes and lack of pre-specification for outcome reporting
Article Type Empirical
Article Subtype Meta-epidemiological analysis
First Author Country United Kingdom
Aim To assess the impact of bias from selective outcome reporting (driven by significance and/or direction of the effect size) in Cochrane systematic reviews of disease modifying antirheumatic drugs for rheumatoid arthritis and to demonstrate how multivariate meta-analysis can examine its impact. 155 trials from 21 reviews were assessed.
Level of Investigation Analytical
Summary of Findings Outcome reporting bias was highly suspected 22% of the 1,118 evaluable outcomes from 155 assessable trials. Multivariate meta-analysis and univariate results sometimes differed importantly. The maximum change in treatment effect estimate between MVMA and univariate meta-analysis approach was found to be 176% for one of the outcome considered.
Number of systematic reviews included 21
Number of eligible systematic reviews assessed 56
Treatment impacted Yes
Treatment impacted description A multivariate meta-analysis approach was applied to a review comparing Auranofin with placebo for the treatment of rheumatoid arthritis to examine whether the summary results and conclusions differed to those from the original meta-analyses performed by the review authors.
Interpretation impacted Yes
Interpretation impacted description Although there were no changes in statistical significance for any of the other outcomes between the univariate and multivariate meta-analysis approaches, potentially clinical important differences were found between the treatment effect estimates. There were larger benefit differences in treatment effect favouring Auranofin using the MFMA approach for pain (0.06 difference) but smaller benefit differences in treatment effect for all the other outcomes. The authors conclude that outcome reporting bias has the potential to affect the conclusions in meta-analyses.