The therapeutic intervention whose effect I need to meta-analyse has several regional variants, that amount to small-to-medium differences in how the therapy is administered to patients.

Now, there is of course a trade-off between the number of studies (data points) that are meta-analysed, and the precision of the meta-analysed summary measure. What determines whether I should

  • (i) pool across these variants to cast a wider net and increase my estimate's precision (reduce CIs), versus
  • (ii) carry out separate meta-analyses for each identifiable study variant, in order for the effect to be more meaningful within each variant, but at a cost of course to the precision of each of those?
  • 1
    $\begingroup$ Have you thought of meta-regression using region as a moderator? $\endgroup$ – mdewey Feb 20 '19 at 16:57
  • $\begingroup$ Yeah, thanks again for suggesting that in the other thread. From what I understand, meta-regression is suitable to continuous parameters such as length of therapy in weeks. Whereas here we're dealing with a categories, e.g. therapy variant A1 vs variant A2. Is meta-regression (or perhaps use of a moderator variable) still the way to go, or am I right that altogether separate meta-analyses need to be done? $\endgroup$ – z8080 Feb 21 '19 at 12:34
  • 1
    $\begingroup$ The moderator can be either categorical or continuous. $\endgroup$ – mdewey Feb 21 '19 at 13:21

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.