Dear Joaquim

I am considering to conduct an ES-SDM analysis given its ability to combine t maps and coordinates. The meta-analysis I wish to conduct will involve treatment studies—some studies will have between-group designs (Randomised Treatment A group vs Treatment B group) while others will have within-subjects or paired designs (such as cross-over trial designs; or simple ABAB etc), and I can envision some studies may have mixed designs (Randomised to Treatment A then Treatment B at a low dose; or, Treatment A then Treatment B at a high dose).

I am wondering whether it is possible to combine contrasts from each study design type, without loss of power or failure to account for differential error terms. If it is possible, then could you please provide some guidance on how to do this? If it is not possible, would you please be able to confirm a way around this or that this is a current limitation?

Thank you.

Faithfully,
Tim

Dear Tim,

of course, if you had enough studies, you could conduct one meta-analysis for two-sample studies, one meta-analysis for one-sample studies, and then check whether the findings overlap.

However, I guess that the problem is that you don't have so many studies. Then my view is that you could mix the studies, with the understanding that this may be an extra source of heterogeneity. In other words, the effect size of one-sample studies may be different from the effect-size of two-sample studies, just as the effect size of patients with a mild disorder may be different from the effect-size of patients with a severe disorder. In the SDM table specify both n1 and n2, and for one-sample studies type the number of subjects in n1 and 0 (i.e. zero) in n2.

Hope this helps,