Hi all,

If one finds group differences in a meta-analysis and would like to see how much a severity score on a particular scale (e.g., depression, BDI) is associated with those group differences, should the scores from the patients groups in the included studies be used in the meta-regression or would it make sense to calculate  group difference (patients minus controls) scores for the studies and used that in the meta-regression. Basically, we are interested in understanding to what extent the group differences we see are associated with the severity of psychopathology.

Looking at previous meta-analyses, it seems that most of the time the mean score from only the patient group is used. Thanks in advance for your help on this.

Stephane

Dear Stephane,

I think this is a very good question, though probably far beyond SDM.

My thoughts on this (please correct and/or expand them !!!) are:

- A regression by group differences in BDI scores would somehow assume that disorder-related brain changes represent a continuous in the population, with some healthy controls already showing some of these abnormalities. Conversely, a regression for patients' BDI scores would somehow assume that disorder-related brain changes are not present in healthy controls, independently of their BDI scores.

- Many studies may not report BDI scores in healthy controls, making a regression by group differences in BDI scores infeasible.

Hope this helps,

Joaquim