Association Between Publication Characteristics and Treatment Effect Estimates: A Meta-epidemiologic Study

Area of Science:

• Medical research methodology
• Evidence synthesis
• Biostatistics

Background:

• The impact of including unpublished or non-English trials on meta-analysis outcomes is not well-established.
• Existing evidence is often contradictory, necessitating further investigation.

Purpose of the Study:

• To compare treatment effects between published and unpublished randomized controlled trials (RCTs).
• To compare treatment effects between RCTs published in English versus other languages.
• Utilizing a meta-epidemiologic approach to analyze these comparisons.

Main Methods:

• Analysis of Cochrane reviews published between March 2011 and January 2017.
• Inclusion of RCTs within meta-analyses comprising at least three trials with binary efficacy outcomes.
• Extraction of trial characteristics and assessment of publication status and language by independent reviewers.

Main Results:

• A significant proportion of trials (93.7%) were published, with most (92.6%) in English.
• Published trials demonstrated larger treatment effects compared to unpublished trials (ROR: 0.90; 95% CI: 0.82 to 0.98).
• Trials published in languages other than English showed larger treatment effects than English-language trials (ROR: 0.86; 95% CI: 0.78 to 0.95).

Conclusions:

• Meta-analyses tend to report larger treatment effects for published trials over unpublished ones.
• Within published trials, those in non-English languages exhibit greater treatment effects than English-language publications.
• These findings suggest potential publication and language biases influencing meta-analysis results.