Reasons or excuses for avoiding meta-analysis in forest plots

Ref ID 474
First Author J. P. Ioannidis
Journal BMJ
Year Of Publishing 2008
URL https://pubmed.ncbi.nlm.nih.gov/18566080/
Keywords • General medical
• Heterogeneity
• Statistical
• Expertise
Problem(s) • Poor execution of narrative synthesis
• Lack of statistical expertise in handling of quantitative data
Article Type Editorial
Article Subtype Discussion piece
First Author Country United States
Aim To assess common reasons employed in systematic reviews for not producing a summary effect in reviews which did produce Forest plots from issue 4 of the Cochrane Database of Systematic Reviews (2005).
Level of Investigation Descriptive
Summary of Findings 135 of Cochrane reviews (8%) had 559 forest plots with no summary estimate. Reasons provided for avoiding quantitative synthesis typically revolved around heterogeneity. However, these typically revolve around statistical heterogeneity, such as the I squared statistic, which is not adequate to fully explain or explore heterogeneity. Conducting random effects meta-analysis, meta-regression or Bayesian meta-analysis are methods highlighted by the authors as potential methodological approaches to heterogenous data. Specifying clearer reasons would improve transparency of the systematic reviewers' implicit judgments about heterogeneity and the ability to meta-analyse.
Number of systematic reviews included 135
Number of eligible systematic reviews assessed 1739
Treatment impacted No
Treatment impacted description
Interpretation impacted Not Applicable
Interpretation impacted description