Currently, I am conducting a meta-analysis on the association between BMI and prognosis. I discovered that the studies used various cut-offs. However, only two studies are eligible for dose-response meta-analysis (DRMA) as only these studies reported >1 cut-off values. Regarding this study:

  1. Is it possible to perform a dose-response meta-analysis when only 2 studies are included?
  2. Is there any statistical software that is able to perform dose-response meta-analysis other than STATA and R?
  3. Regarding the preliminary analysis, we would like to perform a two-class analysis by comparing studies reporting 1 cut-off values (i.e. >=30 vs <30, >=25 vs <25). Can you pool all the cut-offs in the analysis? Or do you have to group the analyses based on the cut-offs?

Any help will be much appreciated. Thank you very much


I do not see any reason in principle why having only two studies would be a fatal drawback.

I am not aware of software other than R to do this.

If you have studies which use various cut-offs then I would perform a single analysis and put cut-off into the model as a moderator. This would make it a meta-regression. It does not get round the problem that this is an ecological analysis and what the moderator is telling you is not the effect of being above 30 (or 25) but the effect of being enrolled in a study where that cut-off was used. Since I imagine the choice of cut-off was made dependent on the overall obesity of the population I think this may be a serious inferential problem.

  • $\begingroup$ Thank you very much for the help. I have re-screened the included studies, and find that there were 2 studies reporting the odds ratio of BMI and death as a continuous variable. Can I combine these two studies with the generalized least squares-estimated data in a random-effects meta-analysis to assess the linear dose-response relationship? $\endgroup$ – Gilbert Aug 24 '20 at 16:41

Your Answer

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

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