There is a great need for treatment outcome research in child and adolescent psychiatry, yet research to date has been of variable quality. Studies have been characterised by the use of small and heterogenous samples, and multiple outcome measures. Sample size is naturally addressed in meta-analysis by the pooling of studies. The problem of multiple outcome measures has been handled in a number of ways in meta-analysis. One approach is to present the effect sizes on each individual outcome measure, another is to report a grand mean effect size of all measures. With reference to one published meta-analysis of the efficacy of tricyclic antidepressants in the treatment of depression in children and adolescents, and with reference to work in progress examining the efficacy of tricyclics and psychostimulants in the treatment of attention deficity hyperactivity disorder, we present an alternative approach to multiple outcome measures.

In this approach we develop a hierarchy of outcome measures ranked on the basis of their published reliability, applicability to child and adolescent research, previous track record in psychopharmacological research, construct validity, and agreement with clinical interview. For each study the highest ranking, or "best available" measure is used in the pooling of results. We argue that in an emerging field where study quality is often marginal, this approach ensures that the best quality data possible can be included in the meta-analysis.