I am looking at a meta-analysis where a subgroup analysis has been performed, as it is important to consider the treatment effect within 2 subgroups of interest, as well as the overall pooled result.

Little heterogeneity was detected in the overall pooled result (I2=32%), but when the studies are subgrouped, there is substantial heterogeneity in one of the subgroups (I2=66%) and no heterogeneity in the other subgroup (I2=0%).

In the protocol the authors said they would use random effects when moderate heterogeneity was observed (I2>40%). Currently they have performed the analysis using fixed-effects, but should I recommend they switch to random-effects so that the treatment effect in the subgroup with substantial heterogeneity is appropriately estimated?

  • $\begingroup$ Are you serving as a referee for the stated study? $\endgroup$ – Ayalew A. Dec 14 '18 at 14:40
  • $\begingroup$ My typical approach is to report both fixed and random effect results, and if they are discrepant consider the findings hypothesis generating but not conclusive. $\endgroup$ – Joe_74 Dec 17 '18 at 13:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.