I have run two separate random effects meta-analyses to estimate summary effect sizes (risk ratios). Both analyses are based on randomized controlled trials examining same treatment for identical populations. The only difference between the two meta-analyses are that the first model includes studies that follow up participants for 12 months or less (17 studies), while the second model includes studies that follow up participants for 18 months or more (20 studies).

I am interested in whether the treatment effect diminishes with increased follow-up time. The second model gives a lower summary effect size than the first, however the difference is small and the confidence intervals overlap. Furthermore, I am uncertain how to approach a comparison in this scenario. Are there any formalized statistical tests? Is meta-regression preferable? I have searched for literature on comparisons in this case without finding satisfying answers, so input on this would be much appreciated.


1 Answer 1


It is perfectly possible to do this in a number of different ways. You can either directly compare your two summary estimates or you can do a meta-regression with follow-up time as a moderator variables. There is an extended discussion of this with code in R on Wolfgang Viechtbauer's pages here.

Note that I am assuming the two sets of trials are independent, that is to say they are not the same trial but reporting twice at the two different time points. If my assumption is false then life becomes more difficult.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.