Identification and impact of outcome selection bias in meta‐analysis

Ref ID 88
First Author P. Williamson
Journal STATISTICS IN MEDICINE
Year Of Publishing 2005
URL https://pubmed.ncbi.nlm.nih.gov/15580591/
Keywords • Cochrane
• Multiplicity
• Statistical
• General medical
Problem(s) • Poor consideration of publication bias
• Multiplicity of outcomes and lack of pre-specification for outcome reporting
Article Type Empirical
Article Subtype Cross-sectional survey/Methodological systematic review
First Author Country United Kingdom
Aim To assess within-study selective reporting of outcomes, defined as the selection of a subset of the original variables recorded for inclusion in publication of trials from a cohort of meta-analyses on the Cochrane Library .
Level of Investigation Descriptive
Summary of Findings A cohort of 9 included Cochrane reviews that had been previously flagged as being potentially prone to publication bias were re-examined for within-study selective reporting. The pooled estimate (95 per cent CI) for mortality of −0:36 (−0:80; 0:08) suggested an opposite effect of treatment, but results were based on data from only five of the nine trials. Imputation of missing data, under the assumption of selective nonreporting, accounted for some of the asymmetry in the funnel plot. No conclusion changed as a result of this extreme sensitivity analysis, thus the meta-analyses of the primary outcomes appear to be robust to potential within-study selection bias.
Number of systematic reviews included 9
Number of eligible systematic reviews assessed 9
Treatment impacted Yes
Treatment impacted description
Interpretation impacted No
Interpretation impacted description