[Skip to Navigation]
Sign In
Comment & Response
July 2, 2019

Random-Effects Assumption in Meta-analyses

Author Affiliations
  • 1Department of Public Health, University of Helsinki, Helsinki, Finland
JAMA. 2019;322(1):81. doi:10.1001/jama.2019.5439

To the Editor Repeated clinical trials on a single medical intervention are never exact replications because patients differ and there are usually some variations in the intervention and outcome measurements. Such variations can lead to inconsistent results between studies on the same intervention.

Drs Serghiou and Goodman1 proposed random-effects meta-analysis as a solution to summarize inconsistent results. However, a long-term concern in meta-analysis has been the apples-and-oranges comparison problem when studies that are too different are combined into a single estimate of effect. Random-effects meta-analysis can exacerbate this problem.

Add or change institution