Twenty-five controlled studies with a total of 1,325 participants were included in the review.
When all 25 studies were combined, there was a statistically significant moderate to large ES (mean ES 0.95, SD 0.69, P<0.01) that suggested a beneficial effect for all the psychodramatic techniques under investigation.
The results for the four different techniques differed significantly from each other (P<0.01). When the results were combined by technique, the ESs for role reversal (100 effect measures from 10 studies: mean ES 0.93, SD 0.74) and doubling (118 effect measures from 5 studies: mean ES 1.29, SD 0.51) were moderate to large, whilst the ES of multiple techniques (46 effect measures from 6 studies: mean ES 0.42, SD 0.30) was small to moderate, and the results for role-playing (17 effect measures from 4 studies: mean ES 0.17, SD 0.68) showed very little improvement.
There was a statistically significant difference between studies of only one gender and studies of mixed gender groups (P=0.01): mixed gender groups tended to report higher ESs (mean ES 0.99, SD 0.75) than single gender groups (mean ES 0.72, SD 53).
There was no statistically significant difference between studies with students as their participants (mean ES 0.89, SD 0.77) and studies of clinical populations (mean ES 1.01, SD 0.60). There was also no statistically significant difference between studies that involved a single session (mean ES 0.88, SD 0.78) and those that involved multiple sessions (mean ES 1.02, SD 0.61).
In the assessment of publication bias, the authors calculated that it would need 1,057 statistically insignificant results to challenge the findings of this meta-analysis. They concluded that given the paucity of research in this area, the studies included in the meta-analysis did not represent a selection bias.