Dear Editor, In a recent article of this journal, Doyen et al. [2] provided a narrative review of melatonin use for sleep-disturbed children with autism spectrum disorders. In this very comprehensive work, the authors usefully summarized the pharmacokinetics of melatonin and its physiopathology in autism spectrum disorders, and discussed clinical trials in this population. In their very complete selection of publications, Doyen et al. considered 17 articles, from single case reports to randomized controlled trials. It must be noted, however, that eight of these studies mainly involved children with non-autistic neurodevelopmental disabilities (without providing subgroup analyses), or concerned children with Rett syndrome, a genetic disease whose affiliation to the autism spectrum is much questioned nowadays [1], and in which sleep problems generally differ from those seen in autism [5]. For these reasons, in a recent systematic review on the same topic, we limited our documentary sources to 12 of these 17 citations [4]. However, despite these discrepancies in the selection of literature, we came to the same conclusions than Doyen et al. that, although not evidencebased for now (considering the small number of randomized controlled trials and their methodological limits), melatonin treatment for sleep problems in individuals with autism spectrum disorders has received very encouraging results. We also agreed with Doyen et al. that more studies are needed, particularly for the collection of long-term follow-up data [3]. In Doyen et al.’s article, the section which is devoted to the analysis of clinical trials is entitled ‘‘Melatonin substitution in autistic subjects with sleep disturbance’’ (our emphasis), and we now would like to focus our comment on whether melatonin treatment in this context really is a substitution or not. First, it must be noticed that, whereas melatonin deficiency has been repeatedly reported in individuals with autism spectrum disorders, its association with sleep disturbance in this population has never been demonstrated for now. Second, within all the trials reviewed by Doyen et al.—and if we put aside Rett syndrome—no patient with an autism spectrum condition had a documented melatonin deficiency. Finally, 11 of these 17 studies included melatonin doses escalation up to 5 mg per day and more, which—as indicated by Doyen et al. later in their paper—lead to melatonin blood levels incommensurable to the physiological ones. For all these reasons, we believe that melatonin use in sleep-disturbed individuals with autism spectrum disorders cannot yet be considered as a substitution treatment. In fact, as Doyen et al. wrote in their discussion, the mechanisms by which melatonin improves sleep in many individuals with autism spectrum disorders remain unknown. Although some authors have hypothesized that these beneficial effects may proceed from the chronobiotic properties of melatonin, this hypothesis has not been proven for now and some arguments suggest that hypnotic effects may be involved (like for example the fact that the clinical benefits of melatonin are often seen from the first day of treatment). Indeed, it is well demonstrated that F. Guenole J.-M. Baleyte CHU de Caen, Service de psychiatrie de l’enfant et de l’adolescent, 14000 Caen, France