I'm hoping to include studies that include 3 or more comparison groups (e.g., Control, Treatment 1, Treatment 2) in a meta-analysis using CMA. So far, I've been unable to find any tutorials or videos explaining how to do this.
The only thing I've found is a section from the Cochrane Handbook suggesting that when you have multiple comparison groups, you should collapse these down into two (e.g., Control vs. Treatment 1+ Treatment 2) - but surely this means you can't look at specific differences between treatment types? Or is there a way to do this by collapsing all treatment groups into one group, and using treatment type as a moderator?
Any help would be greatly appreciated!
Dear Rachel Menzies
Yes you have to collapse them
Exactly the same as you are looking for independent variables in linear regression ( you need to insert variables with only two values only).
U can do this only when you are doing network meta-analysis ,,, then u can compare several arms on the condition that they share the same comparator or at least have common treatments
good luck
Your problem can be solved in several different ways, in which one you choose is dependent upon what you want to know, and what data you have. As I understand it, you have some studies with just one treatment, others with two. If that is the case, you must look at the similarity of treatments. If the double treatments are both very similar to the single treatments, you can collapse them, i.e. average the effect and sum the subjects. However, it would be better to just enter them as separate samples in your data. I am assuming here that you are talking about two different treatment groups, not two treatments to the same group, because if that is so, things become very different.
Now, if the treatments are different, you need to do separate analyses of them, or use a moderator code. If you have like three or more different treatments where there are theoretical reasons to believe that the effects should differ, you have a good reason to use meta-regression and do a moderator analysis on this.
It all comes back to the principle of whether the studies are similar enough to be treated as coming from the same population of studies, or not. This is a notoriously difficult question, but if you add together a lot of things/effects and find that you have a lot of heterogeneity in your data, then you might suspect you need to revise your method and dataset, adding moderators.
Hi Rachel,
If it is reasonable to pool/collapse across the treatment groups (e.g., if they are different agents from the same class or similar enough to each other in some other way), then you could do a meta-analysis where you examine effect sizes for control vs. pooled treatment group comparisons. To look at differences between treatment types, one option would be to then do a second meta-analysis where you examine effect sizes for treatment 1 vs. treatment 2 comparisons.
Good luck!
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Sectioning