Using dissertations to examine potential bias in child and adolescent clinical trials.

The major youth psychotherapy meta-analyses have relied on published studies, which may have led to biased effect size estimates. To examine this possibility, the authors compared 121 dissertations with 134 published studies and found the following: (a) few differences on individual methodological variables, but, overall, stronger methodology in dissertations; (b) no differences in the steps taken to ensure treatment integrity; and (c) a mean dissertation effect size less than half that of published studies. The effect size difference remained robust across tests controlling for all reliable method differences. The findings suggest that dissertations are so strong, both methodologically and clinically, that they warrant inclusion in child psychotherapy meta-analyses and that previous meta-analyses, by excluding them, may have overestimated treatment effects.