Eleven studies (n=150) were included in the review: 8 randomised controlled trials (4 parallel group trials and 4 crossover trials) and 3 uncontrolled test-retest studies.
Ratings for methodology.
Scores for the included studies ranged from 1 to 8. Eight studies were rated to be of stronger-than-average design, and 3 of less-than-average design.
Clinician ratings (5 studies, n=61).
A test of homogeneity of effect sizes for these 5 studies was significant (chi-squared 49.88, d.f.=5, p<0.0001). A cluster of 3 studies (n=36) with methodology ratings greater than or equal to the average gave a weighted Hedge's g of 0.43 (z=1.73, p<0.05).
Parent ratings (9 studies, n=125).
A test of homogeneity of effect sizes for these 9 studies was significant (chi-squared 79.16, d.f.=8, p<0.0001). Cluster analysis identified 3 clusters, of which the largest included 6 studies all with methodology scores greater than 6. For this group of 6 studies (n=84), the weighted adjusted average Hedge's g was 0.75 (z=4.58, p<0.0001).
Teacher ratings (6 studies, n=89).
A test of homogeneity of effect sizes for these 6 studies was significant (chi-squared 124.09, d.f.=5, p<0.0001). A cluster of 3 studies (n=48) with stronger-than-average methodology ratings gave a weighted Hedge's g of 0.56 (z=2.68, p<0.004).
Overall ratings from all sources.
The test of homogeneity of effect sizes for overall ratings from all 11 studies was significant (chi-squared 141.95, d.f.=10, p<0.0001). Six studies (n=84), with stronger-than-average methodology, gave a weighted average Hedge's g of 0.58 (95% confidence interval: 0.27, 0.89, p<0.001, z=3.66).
Side-effects.
The most commonly reported side-effects were sedation in 9 out of 10 studies, and irritability in 6 out of 10 studies.