Frye and Rossignol reply

SIR–We appreciate the comments of Drs Guenole and Baleyte regarding our meta-analysis on the effectiveness of melatonin for sleep-disturbed individuals with autism spectrum disorder (ASD) and their commentary regarding the inclusion of individuals with Rett syndrome. We, too, were concerned with the inclusion of this group and thus specifically addressed this in the results section of the article where we state that ‘The inclusion or exclusion of the one study containing syndromic ASD did not significantly change the results of the meta-analysis.’ Since we did not present the changes in the meta-analysis with the exclusion of children with Rett syndrome, we present them below. The only study of the five double-blind placebo-controlled studies included in the meta-analysis that included patients with Rett syndrome was the study of McArthur et al. This study contained nine (16%) of the 57 participants in the metaanalysis. Removing the McArthur et al. study from the metaanalysis increases the Hedge’s g and Glass’s D effect sizes of sleep duration from 1.08 to 1.20 and 0.94 to 1.02 respectively, with virtually no change in the z-scores of the effect sizes (Hedge’s g z-score did not change from 3.58 and Glass’s delta z-score decreased slightly from 3.01 to 2.96). Similarly, removing the McArthur et al. study from the meta-analysis increases the Hedge’s g and Glass’s D effect sizes of sleep onset from 2.52 to 2.86 and 1.30 to 1.87 respectively, with a small increase in the z-scores (Hedge’s g z-score increased from 9.60 to 10.01 and Glass’s delta z-score increased from 4.12 to 5.18). Finally, removing the McArthur et al. study from the meta-analysis examining the number of night-time awakenings increases the Hedge’s g effect size from 0.50 to 0.57 and the Glass’s D effect size from 0.33 to 0.35, with small changes in the z-scores (Hedge’s g z-score increased from 1.74 to 1.79 and Glass’s delta z-score decreased from 1.11 to 1.06). The decision to include the study with participants with Rett syndrome within the meta-analysis was difficult since, as Drs Guenole and Baleyte point out, the characteristics of sleep disturbances as well as the neuropathology in Rett syndrome are different from those in children with idiopathic ASD. However, according to the current version of the DSM-IV, children with Rett syndrome are still considered to be within the category of pervasive developmental disorders. Removing the patients with Rett syndrome appears to increase the effect sizes of all of the variables analysed in our meta-analysis, suggesting that children with Rett syndrome may have a different and more difficult-to-treat sleep disorder. Our revised analysis supports the removal of Rett syndrome as a pervasive developmental disorder in the forthcoming version of the DSM. Finally, we also agree that further research on melatonin in individuals with both ASD and Rett syndrome is needed.