Controversy exists about the appropriate role of medical, psychosocial and combined modalities of therapy in ADHD. Previous meta-analysis have examined the efficacy of drug or school-based treatments individually. We investigated whether meta-analysis could help summarize the evidence of the relative efficacy of drug vs psychosocial vs combined modalities of treatment. To achieve quality control and comparability of primary studies, the following inclusion/exclusion criteria were used on a preliminary list of several hundred articles identified through electronic database and handsearching: publication after 1980 to coincide with DSM-III and subsequent criteria for diagnosis of ADD(H) or ADHD; subjects aged 0-19 years with ADD(H) or ADHD, unselected to the presence of specific comorbid diagnoses; randomized trials; at least one week of stimulant medication or a course of psychosocial intervention; outcome ascertained with widely-used standardized rating scales designed to measure the core behaviours of ADHD; and data presented in a form suitable for meta-analysis.

Twenty-four studies (drug only 19, psychosocial only 2, combination 3) representing 945 subjects met these criteria. A series of quality assessment questions were used to assess the validity of each randomized trial included in the analysis. Data combined across studies using a random effects model, producing a point estimate and 99% confidence intervals, revealed the following: (1) in drug studies treatment was more effective than placebo(ES=-0.998, CI= -1.325, -0.67), psychosocial intervention alone was no more effective than control (ES=-0.398, CI=-1.552, 0.756), and combination treatment was no more effective than drug (ES=0.405, CI=-.200, 1.01), psychosocial (ES=-0.318, CI=-0.957, 0.322) or control (ES=-0.794, CI=-0.2374, 0.786).

Serious difficulties were encountered in this study due to the limited number of acceptable studies of non-drug interventions and great heterogeneity in methodologies across studies, particularly in subject selection, control conditions, specific interventions and outcome measures, rendering meta-analysis of limited value/usefulness. The results of a large scale prospective randomized trial, such as the current MTA study, are required to shed light on the relative efficacy of various modalities of therapy.