Flavia Cortesi MD, Center of Pediatric Sleep Disorders, Department of Pediatrics and Developmental Neuropsychiatry, University of Rome ‘La Sapienza’, via dei Sabelli 108, 00185 Rome, Italy. Tel.: +390644712293; fax: +39064957857; e-mail: firstname.lastname@example.org
Although melatonin and cognitive–behavioural therapy have shown efficacy in treating sleep disorders in children with autism spectrum disorders, little is known about their relative or combined efficacy. One hundred and sixty children with autism spectrum disorders, aged 4–10 years, suffering from sleep onset insomnia and impaired sleep maintenance, were assigned randomly to either (1) combination of controlled-release melatonin and cognitive–behavioural therapy; (2) controlled-release melatonin; (3) four sessions of cognitive–behavioural therapy; or (4) placebo drug treatment condition for 12 weeks in a 1 : 1 : 1 : 1 ratio. Children were studied at baseline and after 12 weeks of treatment. Treatment response was assessed with 1-week actigraphic monitoring, sleep diary and sleep questionnaire. Main outcome measures, derived actigraphically, were sleep latency, total sleep time, wake after sleep onset and number of awakenings. The active treatment groups all resulted in improvements across all outcome measures, with moderate-to-large effect sizes from baseline to a 12-week assessment. Melatonin treatment was mainly effective in reducing insomnia symptoms, while cognitive–behavioural therapy had a light positive impact mainly on sleep latency, suggesting that some behavioural aspects might play a role in determining initial insomnia. The combination treatment group showed a trend to outperform other active treatment groups, with fewer dropouts and a greater proportion of treatment responders achieving clinically significant changes (63.38% normative sleep efficiency criterion of >85% and 84.62%, sleep onset latency <30 min). This study demonstrates that adding behavioural intervention to melatonin treatment seems to result in a better treatment response, at least in the short term.
Autism spectrum disorders (ASD) are lifelong neurodevelopmental disorders, characterized by markedly abnormal or impaired social interaction and communication, restricted interests and stereotypical behaviours. Core deficits of ASD and their underlying neurophysiology and neurochemistry may predispose children with ASD to intrinsic stressors that threaten sleep.
According to available studies, sleep disorders, consisting mainly of problems of sleeplessness, have been reported clinically in an estimated 40–80% of children (Honomichl et al., 2002; Krakowiak et al., 2008; Souders et al., 2009). Although sleep disorders often remain untreated in ASD, the severity and frequency of these disorders associated with high levels of maternal stress, negative attitudes to the child and increased rates of child behaviour problems and autistic symptoms suggest the need for effective intervention (Williams et al., 2004).
However, sleep problems might occur as a result of complex interactions between biological, psychological, social/environment and family factors, including child-rearing practices that are not conducive to good sleep (Richdale and Schreck, 2009). It has been reported that ASD children often display a preference for unusual bedtime routines which may be maladaptive in terms of promoting good sleep (Henderson et al., 2011). The increasing recognition of the mediating role of behavioural factors in insomnia has led to the development of non-pharmacological interventions for clinical management in this population.
A recent literature review of behavioural interventions for sleep problems in ASD children concluded that they have been considered efficacious in treating sleep onset and maintenance problems (Vriend et al., 2011). However, only small studies or case reports have been performed, with inclusion of children having a variety of diagnoses, not limited to ASD, and without collection of objective sleep data.
There are efficacious approaches to behavioural interventions in children with neurodevelopmental disabilities (Moon et al., 2011; Weiskop et al., 2005), and as suggested by Vriend et al. (2011): ‘promoting positive sleep hygiene when combined with standard extinction is likely to improve problems initiating and maintaining sleep in children with ASD’.
To our knowledge, no studies have compared directly the efficacy of melatonin and cognitive–behavioural therapy (CBT), singly or combined, for sleep disorders in children with ASD.
Study objectives were, first, to determine the relative efficacy of the three active treatments with placebo, and secondly to compare the combination therapy with either melatonin or CBT alone. We hypothesized that all three active treatments would be superior to the placebo, and that combination therapy would be superior to either melatonin or CBT alone.
Patients were referred from 2007 to 2010 to the Department of Pediatrics and Developmental Neuropsychiatry, University of Rome ‘La Sapienza’.
Inclusion criteria were: (i) age 4–10 years; (ii) DSM-IV-TR diagnosis of Autistic Disorder confirmed by the Autistic Diagnostic Interview–revised (ADI-R) (Lord et al., 1994) and the Autism Diagnostic Observation Schedule–generic ADOS-G (Lord et al., 2000); (iii) mixed sleep onset and maintenance insomnia, defined as a sleep onset latency (SOL) and wake after sleep onset (WASO) >30 min that occurred on 3 or more nights a week. We chose these criteria because most statements about pathological SOL and sleep maintenance in children define a cutoff of 20–30 min. As sleep patterns change with age, we recruited subjects in a narrow age range, avoiding the pitfall of many studies that included a wide range of age. These criteria had to be fulfilled 3 months prior to screening and again during the 14-day run-in period of parental daily recording of sleep data; and (iv) absence of other serious neurological, psychiatric or medical conditions.
In order to rule out associated pathologies and structural alterations of the central nervous system, high-resolution banding karyotype, brain magnetic resonance imaging and neurometabolic screening, multi-modal evoked responses and prolonged awake and sleep electroencephalography (EEG) recordings were performed. No subjects were obese and parents did not report breathing disturbances during sleep or periodic limb movement disorders at the time of the study. All subjects were drug-free for at least 6 months prior to the beginning of the study and throughout the study. In order to exclude the associated behaviours and psychiatric comorbidities seen frequently in ASD children, parents completed the Child Behavior Checklist (CBCL; Achenbach and Rescorla, 2000) during the initial screening. Children reporting a T-score ±70 on any of the syndrome scales and three composite scales were not included in this study. Children currently in psychotherapy or receiving other behavioural interventions, as well as families currently in family therapy, were not enrolled.
Sleep behaviour was assessed using the Children’s Sleep Habits Questionnaire (CSHQ) (Owens et al., 2000) a 33-item parent questionnaire. It includes items relating to a number of key sleep domains grouped conceptually into subscales reflecting: (i) bedtime resistance; (ii) sleep onset delay; (iii) sleep duration; (iv) sleep anxiety; (v) night-wakings; (vi) parasomnias; (vii) sleep-disordered breathing; and (viii) daytime sleepiness. A higher score is indicative of more disturbed sleep. In the present study, the CSHQ showed adequate internal consistency (Cronbach’s alpha 0.80).
Each child was monitored with an actigraph in the zero crossing mode from the Ambulatory Monitoring Inc. (Ardsley, NY, USA). Actigraphy, a non-invasive method used to study sleep–wake patterns by assessing movement, provides continuous monitoring of activity level that can be translated to valid estimates of sleep–wake measures. Sleep–wake cycles were scored with the Action W software program. Information is accumulated over a fixed 1-min time interval, considered as an epoch. Sleep epochs were determined based on the Sadeh algorithm. It was highly recommended that parents recorded simultaneously a sleep diary to edit the actigraphic data for potential artefacts and failures. If parents noted on the sleep diary that their child was sick or had an unusual night, the night’s data were discarded and a minimum of 7 nights were necessary to obtain reliable data. Sleep measurements obtained through actigraphy were SOL (time from the parents’ note of lights out to the actigraphically measured first sleep onset), total duration of sleep (actual sleep time, excluding sleep latency and wakening after sleep onset), number of night-wakings and WASO. The number of night-wakings was scored manually and defined as at least 5 min in duration per episode. Sleep efficiency [SE, the ratio of total sleep time (TST) to total time in bed × 100] was also calculated.
The sleep diary is a standardized weekly form that asks the parents to note specific sleep-related events on a daily basis. At bedtime, and during night-wakings of which parents were aware, the parents were instructed to check on the child every 5 min to track whether he or she was awake or asleep, otherwise they noted their child’s sleep in 30-min time blocks.
After baseline assessment, participants were assigned randomly to one of four conditions. Randomization was performed using a computer-generated schedule independent of treatment personnel. Subjects in the melatonin and placebo groups did not know they were receiving active therapy, nor did their clinicians.
Participants underwent 7 consecutive nights of actigraphic evaluations at baseline and at 12-week reassessment. Similarly, CSHQ and sleep diary were completed at baseline and readministered after 12 weeks of treatment. For safety reasons, haematological parameters were monitored exclusively at baseline and at 12-week assessment.
Sample size was determined on the basis of the ability to detect a difference of at least 30 min [standard deviation (SD) 45] in actigraphically recorded total sleep time between four treatment groups from baseline to the 12-week assessment. A sample size of 32 subjects in each group (yielding a total number of 128) provided 90% power to detect an effect size of 0.6667 among the four groups. These calculations were based on an alpha of 0.05. Assuming a 25% dropout rate, we planned to recruit 160 children. They were enrolled and assigned randomly to melatonin (40), CBT (40), combined melatonin and CBT (40) or placebo (40).
The treatment protocol was described in detail and written informed consent was obtained.
Building upon previous research (Montgomery et al., 2004; Weiskop et al., 2005), a treatment program was designed to reduce insomnia in ASD. After the baseline evaluation, each family receiving CBT attended four weekly individual treatment sessions, each lasting approximately 50 min, and administered at the outpatient university clinic. Treatment consisted of a sleep-focused multi-factorial intervention that involved cognitive, behavioural and educational components. The cognitive component was focused to help recognize distorted sleep cognition, and aimed at changing dysfunctional beliefs and attitudes about sleep. The behavioural and educational components consisted of a set of oral instructions about management of a child’s sleep, identification of patient-specific maladaptive sleep condition and implementation of methods for replacing poor sleep habits with more appropriate behaviours (Table 1).
Table 1. Cognitive–behavioural therapy (CBT)
Additionally, parents attended twice-monthly individually tailored CBT sessions. The focus of maintenance sessions was on consolidating treatment strategies learned during initial therapy and developing methods for coping for residual insomnia.
Provided general information about nature, function and regulation of sleep, sleep need and the differentiation of normal and pathological sleep changes. The importance of sleep, good sleep habits and sleep as a learned behaviour were explained to parents
Parents were instructed to create a consistent sleep schedule, to avoid long and late napping, to establish a positive bedtime routine, including parent and child engaging in enjoyable activities before the child went to bed; to allow the child to fall asleep alone in his/her bed (awake but drowsy); to allow the child to sleep in his/her bed for the whole night. Parents were instructed to limit their involvement in the sleep initiation process gradually, to avoid physical contact and gradually increase the interval length of their interventions
Entailed reviewing instructions, identification of individual difficulties, implementation of methods for replacing the remaining maladaptive sleep behaviours with more appropriate sleep habits
Reviewed therapy instructions and provided information on prevention of insomnia relapse
CBT sessions were led by experienced clinical psychologists, and all sessions were audiotaped and reviewed with principal investigators to ensure adherence to protocol. Treatment completion was defined using a minimum adequate trial cutoff (baseline and at least two treatment sessions).
Participants assigned to the melatonin group were given orally a 3-mg controlled-release (CR) dose of melatonin and administered at approximately 21:00 h. Dose changes were not permitted. In Italy melatonin is not a licensed drug, and it is sold as a food supplement in a variety of preparations.
However, the product used in this study contains a high-purity melatonin preparation (99.9%), formulated to provide 1-mg fast-release and 2-mg controlled-release (CR)-melatonin over a prolonged 6-h period after ingestion (Armonia® Retard 3 mg; Nathura, Montecchio Emilia, Italy). This product has been regularly registered in the list of food supplements of the Italian Ministry of Health (cod. 08 29284 Y).
We used CR-melatonin because of its advantage for multi-disabled patients who experienced sleep maintenance problems (Garstang and Wallis, 2006; Giannotti et al., 2006; Wasdell et al., 2008). No behavioural recommendations regarding sleep were given during these short meetings and the main focus was on encouraging participation. Participants met at the outpatient clinic every 2 weeks for a 15-min meeting to report adverse effects and to obtain the dosage pills for the following 2 weeks.
Participants in this group were treated according to a protocol identical to those receiving active medication. As with melatonin, the placebo was made in the identical formulation, and there were no differences in appearance, smell or flavour between the active and inactive pills. An active treatment was offered after the12-week assessment.
Combined CBT and melatonin
Participants in this group received both melatonin and CBT.
To assess compliance with CBT recommendations, parents were asked how many days per week they followed elements of the CBT regimen. Children who missed taking more than 20% of the medication and who did not follow the CBT recommendation were considered non-compliant. In all groups, an 80% compliance rate was required for continued participation in the study.
The study protocol was approved by the ethics committee of the University ‘La Sapienza’ of Rome.
Outcome measures and clinical significance of changes
The primary outcomes were the between-group differences in the mean change from baseline to endpoint, including TST, SOL, SE and WASO, as inferred from the actigraph data.
These variables were averaged over 7 nights for each assessment phase. Furthermore, to assess the clinical significance of these changes, we utilized Morin et al. criteria (1999) examining group differences in rates of percentage of patients achieving a SOL of 30 min or less, or a reduction of SOL by 50% and a SE in the normative level of >85%.
We coded and computerized all data and performed statistical analysis using the statistica version 10 package for Windows (StatSoft Inc., Tulsa, OK, USA).We calculated descriptive statistics (mean ± SD) for all continuous variables and frequencies which were generated for non-continuous variables. Pearson’s correlation was calculated to determine the baseline level of association between parent-completed somnologues and actigraph-generated sleep measures. Analysis of variance or Pearson’s χ2 tests were used to examine demographic and clinical variables at pre-treatment. Bonferroni adjustments of the P-values (alpha level of significance set to 0.05) were performed due to the multiple comparisons. The Bonferroni adjustments were as follows: nine component measures of both CSHQ 0.05/9 = 0.005, and four actigraphically derived variables 0.05/4 = 0.01.
Before analysis, log-transformations were performed on variables to normalize distribution. To reduce the likelihood of type 1 error, Greenhouse–Geisser correction, which adjusts the numerator and denominator degrees of freedom in repeated-measures designs, was applied.
A repeated-measure analysis of variance (rm-anova), adjusted for baseline mean values, was performed to determine whether or not there was a difference in response among treatments, using time and treatment (as a predictor of outcome) using the groups as the between-subject factors comprising four levels and the time as the within-subject factors comprising two levels to determine differences over time for the principal outcome measures. Because of developmental changes, we included also age and sex as covariates. Post-hoc comparisons between groups were performed using the Sheffé test. Effect sizes were judged according to the following commonly used cutoffs: r = 0.10 small effect, r = 0.30 medium effect and r = 0.50 large effect.
We also examined group differences in rates of clinically significant improvement as assessed by the percentage of participants achieving a normal sleep status at the 12-week assessment on actigraphic data using the two-tailed Fisher’s exact test.
More than 185 children met inclusion criteria. One hundred and sixty patients were randomized; of these, 144 completed the study, but only 134 were suitable for analyses (Fig. 1). Thus, the results are based on the following subjects: 35 in the combined therapy, 34 in MLT, 33 in CBT and 32 in the placebo condition. Notably, there were no significant differences in the demographic or clinical variables between subjects who completed the study and those who dropped out.
The average reported duration of insomnia was 2.4 years (SD = 1.7) and 72% of the patients had had symptoms for longer than 2 years. Analyses of sleep diaries show a significant correlation with actigraph measures of total sleep time (r = 0.75), sleep latency (r = 0.83) and WASO (r = 0.70).
Analyses of variance were conducted on the baseline-dependent actigraphic and CSHQ measures. anovas were not significant, and reflected that all the dependent measures were similar across the conditions at pre-treatment. In addition, demographic variables were examined with anova and chi-square tests. Again, the four groups did not vary significantly, indicating that the groups were equivalent at entry into the study (Table 2).
Table 2. Summary of demographic characteristics for the 160 patients assigned randomly to treatment
*Low socioeconomic status (SES) was defined as an index of 3 or less on the Hollingshead Two-Factor Index of Social Position, which ranges from 1 to 5, based on the education and occupation of the head of the household. anova, analysis of variance; COMB, combination; MLT, melatonin; PL, placebo; SD, standard deviation.
Age, years (SD)
Male sex (%)
White Causasian (%)
Low SES* (%)
Primary caregiver sex (female) (%)
Age, years (SD)
Marital status (married %)
Education, years (SD)
Using SOL, total sleep time, WASO and sleep efficiency as dependent measures, rm-anovas yielded significant time effects and significant group × time interaction effects for all measures (Table 3). Post-hoc comparisons revealed that subjects in all three active groups were more improved than those in the placebo condition at the 12-week assessment (P > 0.01). However, the data suggest a trend for the combination group to yield greater improvement rates than either of its single components (Table 3, Fig. 2).
Table 3. Sleep data actigraphically derived for each treatment condition at 2 assessment points
Sleep measure and Time
Time × Group effect
CBT, Cognitive Behavioural Therapy; PL, Placebo; TST, Total sleep time; SOL, Sleep onset latency; SE, Sleep efficiency; NW, Number of night-wakings; ES, Effect size. %, percentage of improvement.
The same pattern of results were obtained for CSHQ scores, which showed all treated children as more improved in most of the CSHQ subdomains, except for parasomnias and sleep-disordered breathing scales, than those in the placebo condition (Table 4). Similarly to the actigraphic results, post-hoc tests confirmed further that those in the combination group generally improved more than did those in the melatonin group, followed by the CBT group.
Table 4. CSHQ results for each treatment condition at 2 assessment points
However, it should be noted that melatonin therapy alone was more effective than CBT alone in improving bedtime resistance, sleep onset delay, night-wakings and sleep duration subscales. CBT alone seemed to be slightly more effective in reducing sleep anxiety subscale. The effect sizes for all significant comparisons fell into the ‘medium–high’ range. From covariance analysis, it appeared that neither gender nor age influenced sleep significantly.
Finally, the percentage of children who met a standard sleep criterion of SOL 30 min or less or reduction of SOL by 50% at the 12-week assessment was at 84.62% for the combination group, 39.29% for the MLT group and 10.34% for the CBT group. As expected, none of the children in the placebo group met this criterion. There were significantly more patients in the combination group versus the MLT group (P < 0.01) and versus the CBT group (P < 0.001) who met this criterion, and there were significantly more participants in the MLT group than in the CBT group (P < 0.01) that also met the same criterion.
Moreover, the percentage of participants who obtained 85% or more on SE at post-treatment was 63.38% for the combination group, 46.43% for the MLT group and10.34% for the CBT alone group (respectively, <0.001 and <0.01). None of the children in the placebo group met this criterion.
Melatonin was well tolerated, and no adverse effects were reported or observed; blood tests and urinanalysis showed no abnormalities and no one had to discontinue the study because of side effects. Moreover, none of the parents reported a loss of response of the child during the treatment period.
The results of this large, randomized placebo-controlled 12-week trial show effectiveness of both CR-melatonin and CBT in the treatment of sleep insomnia in ASD children. Those who received active treatment were able to maintain sleep more efficiently than those in the placebo group. However, of the three active treatments, melatonin in combination with CBT was the most effective in reducing insomnia symptoms, followed by the MLT alone and then the CBT group compared to the placebo group.
Our findings report the efficacy of CR-melatonin in initiating and maintaining sleep. In this study CBT improved sleep latency slightly, but it was less effective in reducing impaired sleep maintenance. However, combining CBT and CR-melatonin produce additional improvement.
No consensus on the optimal doses of melatonin to promote sleep in children has been established. In a recent open-label escalation study, Malow et al. (2011) reported that in most ASD children 1–3 mg of supplemental melatonin was well tolerated and also improved sleep latency.
In our study, a dose of 3 mg was effective and safe and no adverse effect were reported. It is important to note that we used a CR formulation, which promotes sleep for 6–8 h and was thus also effective in maintaining sleep. Consistent with previous studies (Garstang and Wallis, 2006; Giannotti et al., 2006; Wasdell et al., 2008), our results showed that in all cases melatonin improved sleep and children showed a more regular and appropriate bedtime and a longer sleep duration with a significant reduction of night-wakings.
Healthy sleep practices also promote sleep and enhance sleep regulation by reducing environment stimulation and behavioural sleep conditioning, which reinforce the association of certain activities and environments with sleep, limit wake-promoting activities and may play a crucial role in sleep promotion. Consistent with previous studies (Montgomery et al., 2004; Moon et al., 2011; Weiskop et al., 2005), our results show that behavioural intervention alone, even though less effective than melatonin, improved sleep mainly by reducing SOL. As already reported, the action of behavioural treatment seems to be mainly on the behavioural components of sleep disorders rather than on the sleep pattern itself. CBT seems to promote good sleep, as suggested by Jan et al. (2008), entraining intrinsic circadian rhythms to the external environmental and 24-h day/night cycle, reducing environmental stimulation and decreasing anxiety by means of appropriate bedtime routine activities.
Recently, Henderson et al. (2011) investigated the relationship between consistent bedtime routine and sleep quality in children with ASD and showed that the ASD group had less consistent general routines, lower sleep quality and higher levels of externalizing behaviour.
A consistent sleep/wake schedule may be particularly important for ASD children, as they are vulnerable to both sleep and circadian rhythm disruptions. However, sleep hygiene practices and CBT by themselves are not sufficient enough to treat insomnia adequately in these children.
In our study, there was a significant trend for the combination group to produce higher improvement on sleep continuity and efficiency than in either melatonin or CBT groups with the strongest treatment response. Our results indicate that the greatest number and percentage of responders in the combination treatment group achieve a clinically significant change (63.38% of children with normative SE criterion of >85% and 84.62% of children with a SOL <30 min). In addition, this group produced fewer treatment dropouts than under other treatment conditions.
The results achieved after short-term melatonin treatment singly or in combination are notable because of the baseline severity and chronicity of insomnia in the current study group. Obtaining these remission rates for combination and MLT alone groups within 3 months is remarkable.
In conclusion, our results extend those of previous investigations, which studied the efficacy of melatonin alone, either in fast or in the CR formulation (Garstang and Wallis, 2006; Giannotti et al., 2006; Wasdell et al., 2008) or CBT alone (Montgomery et al., 2004; Moon et al., 2011; Weiskop et al., 2005) in the treatment of sleep insomnia in children with ASD. The observed advantage of combination therapy over either CBT or melatonin alone suggests that among these effective therapies, combination therapy provides the best chance for a positive outcome. Thus, in agreement with Wirojanan et al. (2009), our results suggest that melatonin can be considered a safe and effective treatment in combination with behavioural therapies and sleep hygiene practices for the management of sleep disorders in children with ASD. The superiority of combination therapy is due most probably to additive or synergistic effects of the two treatments.
Some limitations of this study should be taken into account. First, this trial assessed only the effects of the 12-week treatment, failing to assess mid- and long-term efficacy and treatment gain after withdrawal. Therefore, we do not know exactly when the positive response to treatments appeared and whether or not it would persist during a long-term period. However, in order to obtain better compliance in this population, which may display non-compliant or avoidant behaviour, and whose parents experienced more stress, we decided to conduct just one post-baseline evaluation. In fact, this study had a low discontinuation rate of 10%. Moreover, we did not measure melatonin salivary and plasma level or its urinary metabolite before and during the treatment; thus, in this study we cannot correlate the efficacy of melatonin with the presence of an eventual melatonin deficit. Secondly, the findings cannot be generalized to patients who have disorders that were excluded from our protocol. Thirdly, parents had agreed to comply with the treatment regimen, which implies a motivation that may not be present in all cases. Therefore, the present results may be biased and could over- or underestimate the efficacy of the treatments. Finally, sleep disorders such as obstructive sleep apnoea syndrome or periodic limb movements were ruled out clinically, but no pre-screening with gold standard polysomnography was performed, so we cannot exclude that some of these patients may have been enrolled into the study.
Limitations notwithstanding, this study confirms that melatonin and CBT in combination or as monotherapies appear to be effective for these commonly occurring sleep disorders in children with ASD. The results confirm those of previous studies of melatonin and CBT and, most importantly, show that together they offer children the best chance of a positive outcome.
Our findings indicate that all three treatment options may be recommended, taking into consideration the family’s treatment preferences, treatment availability, cost and time burden. Further analysis of predictors and moderators of treatment response, assessing a long-term effect, may identify who is the most likely to respond to which of these effective alternative. The extent and eventual clinical utility of these approaches should be evaluated by other studies using long-term follow-up and controlling for several confounding factors.