Table 1

Factors that impact on the meta-analysis summary estimate under the FE and the RE models

ScenarioImpact on results
Heterogeneity variance is estimated to be 0The results are identical under the two meta-analyses models
Heterogeneity is greater than 0 and the studies’ size is unrelated to its effect sizeThe two models yield similar mean summary estimates. The CI of the RE meta-analysis is wider than that of the FE meta-analysis
Small studies are associated with more extreme (eg, more beneficial for the active intervention) effect sizes (small study effect)Under the FE meta-analysis, the summary estimate tends to be pooled towards the summary effect from the larger studies (usually towards the no effect line). The influence of the big studies is less pronounced under the RE meta-analysis, resulting in a less conservative result
A mega-trial is included in the meta-analysisUnder the FE meta-analysis, the summary estimate tends to be pooled towards the effect of the mega-trial. The influence of the mega-trial is less pronounced under the RE meta-analysis
  • FE, fixed-effect; RE, random effect.