Hi Andreas,

This is a tricky one. I'm not sure if the mean change scores and proportions can be combined; I'd probably consider the following options:

- Assuming that it wouldn't depart from your protocol to do so, report two meta-analyses, one using the mean change in scores, and the other using the proportions. It would be interesting to see whether the results of the two analyses align or contradict.

- Reach out to the authors of the papers which report proportions to see if they can provide you with before/after means, or vice versa.

I think I'd need to know more about the included studies to advise further - I haven't come across methods for effectively estimating a mean change from a proportion or vice versa. If you find any, I'd be interested in hearing about them.

Alice

I would contact the authors. Otherwise I might combine them as two seperate meta-analyses, call this a 'sub-group' analysis based on how tehy measured the main outcome variable.

Thank you Alice.

I am of the same opinion as you, that these two cannot be combined together, so I simply did not include the studies reporting mean differences.

I agree, I should contact authors, however, most of these papers are relatively old and it will probably be a waste of time.

If some of the studies report continuous outcomes (calculating mean difference) and some other studies proportions (prevalence), the result can not be shown together, that's for sure. However, if you want to keep both sorts of studies together in one article, you can show as subgroups, alternatively, can separate into two meta-analyses.

Dr. Asiful

Thank you for your reply Asiful.