Melatonin in autism spectrum disorders: a systematic review and meta-analysis

Authors


Dr Daniel Rossignol at 3800 West Eau Gallie Blvd, Melbourne, FL, 32934, USA. Email: rossignolmd@gmail.com

Abstract

Aim  The aim of this study was to investigate melatonin-related findings in autism spectrum disorders (ASD), including autistic disorder, Asperger syndrome, Rett syndrome, and pervasive developmental disorders, not otherwise specified.

Method  Comprehensive searches were conducted in the PubMed, Google Scholar, CINAHL, EMBASE, Scopus, and ERIC databases from their inception to October 2010. Two reviewers independently assessed 35 studies that met the inclusion criteria. Of these, meta-analysis was performed on five randomized double-blind, placebo-controlled studies, and the quality of these trials was assessed using the Downs and Black checklist.

Results  Nine studies measured melatonin or melatonin metabolites in ASD and all reported at least one abnormality, including an abnormal melatonin circadian rhythm in four studies, below average physiological levels of melatonin and/or melatonin derivates in seven studies, and a positive correlation between these levels and autistic behaviors in four studies. Five studies reported gene abnormalities that could contribute to decreased melatonin production or adversely affect melatonin receptor function in a small percentage of children with ASD. Six studies reported improved daytime behavior with melatonin use. Eighteen studies on melatonin treatment in ASD were identified; these studies reported improvements in sleep duration, sleep onset latency, and night-time awakenings. Five of these studies were randomized double-blind, placebo-controlled crossover studies; two of the studies contained blended samples of children with ASD and other developmental disorders, but only data for children with ASD were used in the meta-analysis. The meta-analysis found significant improvements with large effect sizes in sleep duration (73min compared with baseline, Hedge’s g 1.97 [95% confidence interval {CI} CI 1.10–2.84], Glass’s Δ 1.54 [95% CI 0.64–2.44]; 44min compared with placebo, Hedge’s g 1.07 [95% CI 0.49–1.65], Glass’s Δ 0.93 [95% CI 0.33–1.53]) and sleep onset latency (66min compared with baseline, Hedge’s g−2.42 [95% CI −1.67 to −3.17], Glass’s Δ−2.18 [95% CI −1.58 to −2.76]; 39min compared with placebo, Hedge’s g−2.46 [95% CI −1.96 to −2.98], Glass’s Δ−1.28 [95% CI −0.67 to −1.89]) but not in night-time awakenings. The effect size varied significantly across studies but funnel plots did not indicate publication bias. The reported side effects of melatonin were minimal to none. Some studies were affected by limitations, including small sample sizes and variability in the protocols that measured changes in sleep parameters.

Interpretation  Melatonin administration in ASD is associated with improved sleep parameters, better daytime behavior, and minimal side effects. Additional studies of melatonin would be helpful to confirm and expand on these findings.

Abbreviations
ASD

Autism spectrum disorders

ASMT

Acetylserotonin methyltransferase

6-SM

6-Sulphatoxymelatonin

What this paper adds

  •  Physiological levels of melatonin and/or melatonin derivates are commonly below average in ASD and correlate with autistic behavior.
  •  Abnormalities in melatonin-related genes may be a cause of low melatonin levels in ASD.
  •  This study indicates that treatment with melatonin significantly improves sleep duration and sleep onset latency in ASD.

Autism spectrum disorders (ASD) are a heterogeneous group of neurodevelopmental disorders that share a common behavioral definition. ASD is defined by impairments in communication and social interaction accompanied by restrictive and repetitive behaviors.1 An estimated one out of 110 individuals in the United States is currently affected with ASD.2 Some studies have reported that children with ASD, as a group, have a higher prevalence of sleep abnormalities than typically developing children,3–11 with a prevalence ranging from 40% to 86%.7,8,12–14 These abnormalities include taking longer to fall asleep (longer ‘sleep onset latency’),15–17 frequent night-time awakenings,16,17 and reduced sleep duration.13,18 Sleep problems have been associated with disruptive daytime behavior or medical abnormalities in some children with ASD. For example, sleep disorders were significantly associated with both mood disorders and gastrointestinal abnormalities in one study of 160 children with ASD.19 Sleep problems in individuals with ASD have also been associated with poor social interaction,17,20,21 increased stereotypy,21 problems in communication,17,22 and overall autistic behavior.15,21 Some studies have also reported an association between sleep problems and developmental regression in ASD,3,5,23–25 although this association has not been found in every study.19,26 Furthermore, abnormalities in genes associated with circadian rhythms have been reported in ASD.27,28 About two decades ago, Chamberlain and Herman hypothesized that abnormalities in melatonin secretion may play a role in the development of ASD.29 Interestingly, individuals with Smith–Magenis syndrome can manifest autistic behaviors and have an abnormal melatonin circadian rhythm;30 this syndrome has responded to melatonin treatment.31 Additionally, individuals with Angelman syndrome can manifest both autistic behaviors32 and sleep abnormalities,33 and melatonin has been shown to improve insomnia in this syndrome,34,35 although the melatonin dose must sometimes be kept low as these individuals can have slow melatonin metabolism owing to decreased CYP1A2 enzyme activity.34

Melatonin is an endogenous neurohormone produced predominantly in the pineal gland.36 In individuals with normal vision, the secretion of melatonin increases shortly after darkness, peaks in the middle of the night, and falls slowly during the early morning hours. Melatonin is synthesized from l-tryptophan through several metabolic intermediates, most notably serotonin. As depicted in Figure S1, (supporting information published online only) l-tryptophan is converted into 5-hydroxytryptophan (5-HTP) and then into serotonin (5-hydroxytryptamine) in two metabolic steps. Serotonin is then converted into N-acetylserotonin and finally into melatonin in two additional metabolic steps that involve two enzymes: arylalkylamine N-acetyltransferase and acetylserotonin methyltransferase (ASMT).37 A major metabolite of melatonin is 6-sulphatoxymelatonin (6-SM). Although melatonin is best known for its regulatory role of the circadian rhythm,38,39 it is also a potent antioxidant,40 has anti-inflammatory properties,41 is involved in the immune response,42–44 and helps regulate synaptic plasticity.45,46 Abnormalities in melatonin pathways have been reported in circadian rhythm47,48 and non-circadian rhythm disorders such as diabetes49,50 and ASD.51

Melatonin is commonly used for insomnia in children, has a favorable side-effect profile, is inexpensive and readily available, and is often efficacious for sleep abnormalities.52 Interestingly, a small number of individuals with sleep problems and intellectual disability have been reported to have an initial positive response to melatonin that wanes over time;53 in these individuals this phenomenon may be related to slow melatonin metabolism, possibly due to decreased CYP1A2 enzyme activity.53 Compared with placebo, melatonin has been shown to improve sleep in children with insomnia with and without attention-deficit–hyperactivity disorder.54,55 Several controlled studies have reported that melatonin improves sleep in children with ASD.56–58 However, a systematic, comprehensive review of this recently evolving literature has not been published. Furthermore, no meta-analysis has been published on the effects of melatonin in ASD. In this article, we systematically review melatonin in ASD and examine studies that report on the following: concentrations or physiology of melatonin or melatonin-related metabolites in ASD; melatonin-related genes in ASD; correlation of the concentration of melatonin and melatonin metabolites with ASD behaviors; the prevalence of melatonin usage in ASD; physician recommendations concerning the use of melatonin in ASD; melatonin treatment studies in ASD and meta-analytic statistics of placebo-controlled studies to examine the effect size of treatment; and the potential side effects of melatonin in ASD.

Method

Search strategy

For this review, we included individuals with autistic disorder, Asperger syndrome, Rett syndrome, and pervasive developmental disorders, not otherwise specified. We performed a computer-aided search of PubMed, Google Scholar, CINAHL, EMBASE, Scopus, and ERIC databases from their inception to October 2010 using the search terms ‘autism’, ‘autistic’, ‘ASD’, ‘Asperger’, ‘Rett’, ‘pervasive’, and ‘PDD’ in all combinations with ‘melatonin.’ The references cited in identified publications were also searched to locate additional studies. Figure S2 (supporting information published online only) depicts the publications identified during the search process.

Study selection

Both reviewers screened titles and abstracts of all potentially relevant publications. Studies were initially included if they (1) involved individuals with ASD and (2) reported unique data on melatonin or melatonin-related factors. After screening all records, 35 publications were identified that met these inclusion criteria.5,8,51,56–87Table SI (supporting information published online only) reports the characteristics of these 35 publications. Five of these studies reported on a blended sample of individuals with ASD and individuals with other neurologic abnormalities.57,60,73,74,86 Two57,86 of these five studies were placebo controlled and the authors were contacted to obtain the data on the individuals with ASD in order to include them in the meta-analysis: one study86 contained 51 children with a neurodevelopmental disability, including 16 children with ASD, and the authors of the study provided information on the ASD subgroup,88 while the other study57 comprised 12 children with ASD and/or fragile X syndrome. The data were obtained from graphs reported in the publication on the five children with ASD only (not those with fragile X syndrome) as the data were not accessible to the original study authors. In all of the studies where melatonin was administered as a treatment, the study participants had a sleep abnormality.

Data extraction and statistical analysis

For treatment studies, two types of values were derived: prevalence values and comparisons of changes in sleep parameters with melatonin treatment. A mean prevalence value was computed by dividing the total number of participants with a certain characteristic pooled for all studies by the number of participants evaluated for all studies. A 95% confidence interval (CI) was then calculated assuming a Bernoulli distribution.89 We followed PRISMA 2009 guidelines.90

For the meta-analysis, outcome variables included total sleep duration, number of night-time awakenings, and sleep onset latency. We performed meta-analysis only on (1) studies that used a randomized double-blind, placebo-controlled design, (2) individuals with ASD and not other developmental disorders, and (3) studies that reported quantitative data on total sleep time, sleep onset latency, and/or number of night-time awakenings. Five studies met these criteria for meta-analysis.56–58,78,86 The methodological quality of these five studies was independently assessed by each reviewer using the Downs and Black checklist.91 The intraclass correlation coefficient (using average measures with a two-way random effect model and an absolute agreement definition) between the two reviewers was calculated using the Statistical Package for the Social Sciences (version 19.0, SPSS Inc., Chicago, IL, USA). The following statistics were computed for the comparisons between sleep parameters with melatonin treatment and baseline sleep parameters and sleep parameters with placebo treatment: (1) the pooled means with confidence intervals, (2) the effect size using two different methods, and (3) the homogeneity statistic Q in order to assess the between study variation.92 Two different approaches were used to calculate the effect size: Glass’s Δ, which is based only on the control standard deviation,93 and Hedge’s g, which uses the mean sample standard deviation of the two conditions being compared.94 The data generated from this analysis are depicted as forest plots in Figure 1. Lastly, to assess for potential publication bias, we inspected funnel plots, a scatter plot of treatment effect (i.e. the mean difference) against the study precision (i.e. inverse of the standard error). Funnel plots were constructed using meta-analysis with interactive explanations (MIX: version 2.0; BiostatXL, Yamato-shi, Japan;95 see Fig. S3, supporting information published online only). A linear regression conducted on the funnel plot values was used to assess quantitatively the presence of significant funnel plot asymmetries. The statistical threshold for all studies was set at an alpha of 0.05.

Figure 1.

 Forest plots showing the effect of melatonin on sleep parameters in autism spectrum disorders (ASD) compared with baseline and placebo. The effect size derived from studies that compared melatonin with baseline (pre-melatonin treatment) measurements and placebo treatment are presented separately in each panel. The effect size using both Hedge’s g and Glass’s Δ are presented. Each graph includes the homogeneity statistic Q in the lower left corner and a measure of statistical significance of the effect size, the z-score, in the lower right corner. A separate bar chart within each graph depicts the weight to which each study contributed to the final effect size calculation. (a) Melatonin demonstrates a significant effect on sleep duration compared with both baseline and placebo using both effect size calculations. (b) Melatonin demonstrates a significant effect on sleep onset latency compared with both baseline and placebo using both effect size calculations. (c) Melatonin demonstrates no significant effect on night-time awakenings as compared with both baseline and placebo using both effect size calculations. CI, confidence interval.

Results

Included studies

The computerized search initially identified 68 unique publications after duplicates were removed. An additional five publications were identified by searching references. After screening and assessment, 35 references met the inclusion criteria as depicted in Figure S2. The selected studies were separated into two main themes: studies examining the biochemistry or physiology of melatonin in ASD and studies examining melatonin as a treatment in ASD. These studies are discussed separately.

Studies of the biochemistry or physiology of melatonin in ASD

Studies measuring concentrations of melatonin or melatonin derivatives in ASD

Table SII (supporting information published online only) lists specific details of the nine studies51,59,77,79–81,84,85,87 reporting melatonin or melatonin metabolite concentrations in ASD, with all reporting at least one abnormality. Seven studies51,59,77,79,81,85,87 reported lower melatonin or melatonin metabolite concentrations in individuals with ASD than in healthy individuals or laboratory reference ranges. Two of these studies reported a significantly lower mean melatonin concentration measured between 12 am and 4 am in individuals with ASD compared with healthy individuals.77,81 One controlled study reported a plasma melatonin level at or below 50% of the mean control melatonin concentration in 65% of individuals with ASD (significance not reported).51 Another study reported that 63% of individuals with ASD had a 6-SM level at or below 50% of the mean 6-SM level found in healthy individuals (significance not reported).85 The remaining three studies59,79,87 reported that melatonin or 6-SM levels were lower than laboratory reference values (significance not reported in these studies).

Two studies reported that daytime melatonin levels were significantly higher in the group with ASD than in healthy individuals.81,84 However, no study reported an elevated melatonin or melatonin derivative at night in the group with ASD compared with healthy individuals. Four studies51,77,79,81 reported an abnormal melatonin circadian rhythm in individuals with ASD compared with healthy individuals. Two studies51,80 measured serotonin levels. In one of these studies, an inverse relationship between platelet serotonin and urinary 6-SM was found,80 while the other study reported concomitantly elevated plasma serotonin and depressed plasma melatonin and ASMT activity in children with ASD compared with healthy individuals.51

Studies examining polymorphisms in melatonin-related genes in ASD

Five studies51,61,63,67,76 examined genes that code for melatonin receptors (MTNR1A, MTNR1B, and GPR50) or enzymes involved in melatonin synthesis (alkylamine N-acetyltransferase and ASMT). Four of these studies51,63,67,76 examined the ASMT gene, which codes for the last enzyme involved in melatonin synthesis. ASMT abnormalities in the group with ASD were reported in all four studies. One study reported a partial duplication of ASMT in 6 to 7% of individuals with ASD compared with 2% of healthy individuals (p=0.003).67 In another study, several single-nucleotide polymorphisms in ASMT were reported in 2.8% of individuals with ASD compared with 0.5% of healthy individuals.76 One study reported several single-nucleotide polymorphisms, including a novel R319X stop mutation, in ASMT in 2.6% of individuals with ASD compared with 1.8% of healthy individuals.63 Finally, one study reported that two single-nucleotide polymorphisms in ASMT (rs4446909 and rs5989681) were significantly more frequent in individuals with ASD than in healthy individuals (p<0.001) and were associated with a significant decrease in ASMT enzymatic activity and lower plasma melatonin levels.51 Furthermore, in this study, decreased ASMT activity was significantly correlated with lower plasma melatonin levels in individuals with ASD but not in healthy individuals.51 The one study that examined the arylalkylamine N-acetyltransferase gene did not find any abnormalities.76 Variants in genes (MTNR1A and MTNR1B) that code for melatonin receptors 1A and 1B were reported in two studies,61,76 with one study reporting a variant in 2.8% of individuals with ASD compared with 0% of healthy individuals 76 and the other study reporting that variants were not significantly different compared with healthy individuals.61 Finally, two variants in G protein-coupled receptor 50 (GPR50), which codes for a melatonin-related receptor, were found to be associated with ASD in one study, but this finding did not hold up after correction for multiple comparisons,61 and a second study found no significant association.76 Overall, these findings suggest that genetic abnormalities in enzymes involved in melatonin metabolism and/or melatonin receptor function could contribute to lower melatonin concentrations or an altered response to melatonin in a small percentage of individuals with ASD.

Correlations between melatonin metabolites and ASD findings

Four studies reported a correlation between levels of melatonin or melatonin metabolites and ASD symptoms or clinical findings.51,59,81,85 For example, night-time urinary excretion of 6-SM was reported to be inversely correlated with the severity of impairments in verbal communication and play85 as well as daytime sleepiness in individuals with ASD.59 In one study, a lower mean serum melatonin level was associated with an abnormal electroencephalogram in individuals with ASD.81 Finally, lower ASMT activity was associated with hyperactivity but was not found to be related to Autism Diagnostic Interview – Revised scores in another study.51 These studies suggest that melatonin metabolism is directly or indirectly related to certain autistic behavior.

Studies examining melatonin as a treatment in ASD

Studies reporting the prevalence of melatonin usage in ASD

Three survey studies,8,65,70 involving a total of 1071 individuals, reported a prevalence of melatonin use in individuals with ASD. The prevalence ranged from 2.98 to 10.8%70 with an mean prevalence of 7.2% (95% CI 5.6–8.7%).

Studies reporting prevalence of physician recommendations of melatonin in ASD

Three survey studies60,69,82 reported the prevalence of recommendations for melatonin usage in ASD among 2483 physicians although two studies also included several pediatric conditions other than ASD.60,82 The overall prevalence of physicians recommending the use of melatonin from these studies was 32.4% (95% CI 30.6–34.2%) with a range of 24.982 to 39%.60 In one study, 22% of physicians did not feel ‘knowledgeable’ enough about melatonin to recommend its use, 14% discouraged melatonin use, and 39% were accepting of its use if the child was already taking melatonin.69 These studies indicate that, even though melatonin is not approved as a treatment for insomnia in ASD by the US Food and Drug Administration, a relatively large percentage of physicians recommends melatonin treatment for individuals with ASD.

Studies of the effects of melatonin treatment on sleep in ASD

A total of 18 studies5,56–58,64,66,68,71–75,77–79,83,86,87 reported the effects of melatonin treatment on sleep parameters in individuals with ASD (Table SI). One study reported on melatonin usage in individuals with ASD who were over 18 years of age68 and another reported the effects of melatonin in individuals with ASD aged between 3 and 28 years.73 The remaining 16 studies5,56–58,64,66,71,72,74,75,77–79,83,86,87 involved individuals aged between 2 and 18 years, but none of these studies reported treatment effects by age group (e.g. age up to 6y, 6–12y, or 12–18y). The dosage of melatonin in these studies ranged from a low of 0.75mg66 to a high of 15mg86 with rare use of 25mg.74 The length of melatonin usage in these studies ranged from 14 days83 to over 4 years.74 In order to identify sleep abnormalities, studies used a variety of subjective (parent-report questionnaires and sleep diaries) and objective (actigraphy) measures. Actigraphy involves a sensor (typically worn on the wrist) that measures movement as a surrogate for wakefulness and is felt to be more accurate and objective than other measures,10,78 especially when used in conjunction with sleep diaries.52

All studies

Out of these 18 studies, 12,64,66,68,72–75,77,79,83,86,87 totaling 349 individuals, reported the percentage of individuals with ASD who experienced an improvement in sleep with melatonin. This percentage ranged from 67%77 to 100%,64,68,72,75,79,87 with an overall improvement rate of 84.2% (95% CI 81.4–88.9%). The two studies that used actigraphy reported sleep improvements with melatonin in 92%86 to 93%83 of individuals, although one86 of these studies contained a blended sample of individuals with and without ASD. In order to reduce the effects of individuals without ASD, we examined the four non-case series66,68,77,83 that contained only individuals with ASD (142 individuals) and found an overall sleep improvement rate of 84.7% (95% CI 78.7–90.6%). One additional article reported the results of a parental online survey by the Autism Research Institute; although this was not a formal study and, therefore, was not included in the aforementioned analysis, this survey found that out of 1105 children with ASD, 65% had an improvement with melatonin usage and 27% had no change after melatonin usage.62

Uncontrolled studies

Of the 18 studies examining melatonin treatment, 13 were uncontrolled5,64,66,68,71–75,77,79,83,87 and ranged in size from one child64,72,75,87 to 107 children.66 Because these studies were uncontrolled, they were not included in the meta-analysis. All 13 studies reported some type of improvement in sleep with melatonin use. Improvements were described in total sleep duration,5,68,71,72 number of night-time awakenings,5,64,68,75 and sleep onset latency.5,64,71,79,83,87 Four studies reported general improvements in overall sleep without a further description.66,73,74,77 One study reported a significant decline in night terrors within 2 days of starting melatonin75 and two studies reported a decrease in screaming in the middle of the night.79,87 Four studies reported worsening of sleep when melatonin was stopped or the dose was lowered,5,64,79,83 and one study reported that improvements with melatonin were maintained at both 12- and 24-month follow-ups.5 Four studies reported improvements with melatonin when other sleep medications had previously failed.64,68,79,87

Controlled studies

The remaining five studies56–58,78,86 were randomized, double-blind, placebo–controlled crossover studies, the details of which are reported in the supplementary material (Table SIII, supporting information published online only). Four studies56–58,86 included children with non-syndromic ASD, whereas one study78 contained nine children with syndromic ASD (Rett syndrome). The mean score on the Downs and Black checklist as performed by both reviewers for these studies was 19.9 of 31 (SD=1.66; range 17–22) and the intraclass correlation coefficient between the two reviewers was 0.870 (p=0.033). Three studies compared both a baseline and a placebo with melatonin treatment56,58,86 while two studies compared only melatonin treatment with placebo.57,78 Furthermore, for one of these studies,86 the data provided by the study authors88 (described in the ‘Method’ section) for children with ASD only did not include baseline measurements. Four56,58,78,86 of the five studies had a washout period between the two treatment arms that varied from 3 to 5 days86 to 1 week56,78 to 1 month,58 while one study57 did not contain a washout period. A shorter washout period (or no washout period) can result in a smaller effect size. However, the length of the washout period was not related to the effect size of the study.

Compared with placebo, these studies reported significant improvements in total sleep duration,56–58,78,86 number of night-time awakenings,56 and sleep onset latency.56–58,78,86Figure 1 depicts the effect sizes (both Glass’s Δ and Hedge’s g) for the changes in sleep parameters with melatonin treatment, for each study and all studies combined, compared with placebo and baseline. The inclusion or exclusion of the one study containing syndromic ASD78 did not significantly change the results of the meta-analysis. The overall effect sizes of the differences between melatonin treatment and both baseline and placebo were significant and large for sleep duration and sleep onset latency, although the effect size was about 50% larger for sleep onset latency. The overall effect size for the number of night-time awakenings was significant when compared with placebo only when using the Hedge’s g effect size calculation with an alpha probability of 0.04. Overall, sleep duration was 73 and 44 minutes longer during melatonin treatment than with baseline and placebo respectively, and sleep onset latency was 66 and 39 minutes shorter during melatonin treatment than during baseline and placebo respectively.

The Q statistic indicated that the effect size varied significantly across clinical studies. This was predominantly due to the consistently large effect size for all parameters in the Garstang and Wallis study.56 In fact, for most sleep parameters, the effect size of the Garstang and Wallis study56 was three to four times greater than in most other studies. The exception to this was the impressive effect size of the study by Wirojanan et al.57 for sleep onset latency. The remainder of the studies demonstrated smaller effect sizes, especially for the treatment versus placebo comparison. Some heterogeneity was evident in the funnel plots for sleep duration and sleep onset latency (see Fig. S3), and the study with the most precision (i.e. Garstang and Wallis56) had the largest effect size, resulting in a visually asymmetric funnel. However, linear regression demonstrated non-significant slope coefficients (p>0.10), suggesting that this asymmetry was not statistically significant. In addition, publication bias normally results in an asymmetric funnel with smaller studies demonstrating larger effect sizes, rather than smaller effect sizes.96 Thus, these analyses did not suggest publication bias.

Combining the double-blind, placebo-controlled studies in a meta-analysis was limited by the method used to collect the sleep parameter data. The two studies with the largest effect sizes for the treatment versus placebo condition were very similar as they both used sleep diaries kept by the parents.56,58 The three remaining studies inferred sleep parameters using wrist actigraphy by quantitatively measuring each child’s movements,57,78,86 with two of these studies combining actigraphy data with a parent sleep diary.57,78 Overall, the effect of melatonin on sleep duration and sleep onset latency was greater for the studies that used only sleep diaries.

The reason for a greater effect size for studies using parental report (i.e. sleep diaries) compared with actigraphy is not known, but it appears that these two techniques are different in several ways. For example, McArthur and Budden78 analysed the reliability of the sleep diary data provided by parents. They reported that the data obtained from actigraphy were much more reliable at monitoring long-term sleep–wake cycles than the diaries. Whereas the average data loss from not wearing the actigraphy monitor was 8.4%, the average diary data loss was 57.9%. In addition, parents reported filling in the diaries with sleep parameter data estimated from memory several days after the sleep events in instances when they forgot to consistently maintain the sleep logs. While the mean sleep duration and onset latency were somewhat similar across studies regardless of the two methods used, the mean number of night-time awakenings was clearly much larger for the studies that used actigraphy and much lower for studies that used the diary method. This discrepancy was echoed in the analysis of McArthur and Budden,78 in which the two methods for measuring the number of night-time awakenings differed by an order of magnitude (diary 0.6 vs actigraphy 13.3). This suggests that each of these two methods measured something different for the number of night-time awakenings. Since some studies56 did not require parents to go into the children’s room at night to check if they were awake, it is also possible that the parents were not aware of minimal night-time awakenings that actigraphy was sensitive enough to record.78 Alternatively, it is difficult to know if participants were truly awake during the night if they were not observed. For example, actigraphy could mistake parasomnias (e.g. sleep walking) for night-time awakenings. Even though there was significant variation in the number of night-time awakenings, the effect size was very small for night-time awakenings for almost every study except Garstang and Wallis,56 suggesting that regardless of the method of measuring this sleep parameter, melatonin probably had little effect on night-time awakening.

Studies of the effects of melatonin treatment on daytime behavior in ASD

Six studies reported that the night-time administration of melatonin led to improvements in daytime behavior in some children with ASD.5,56,58,64,74,83 Improvements included less behavioral rigidity, ease of management for parents and teachers,56 better social interaction, fewer temper tantrums, less irritability, more playfulness,74 better academic performance, and increased alertness.64 One study noted a significant improvement in overall daytime behavior as measured by the Developmental Behavior Checklist when comparing melatonin with placebo.58 However, a meta-analysis of the effects of melatonin on daytime behavior could not be performed as only two of the six studies56,58 were placebo controlled and only one58 reported results for melatonin compared with placebo (the other study56 reported only that several parents and teachers noted better daytime behavior with melatonin use).

Studies examining side effects of melatonin in ASD

Twelve studies5,58,64,66,68,71,73,74,78,79,83,86 examined the potential side-effects of melatonin in individuals with ASD. In addition, one parental survey reported the prevalence of ‘worse behavior’ with melatonin.62 Seven of these studies reported that melatonin use was not associated with any side-effects,5,64,68,71,74,78,79 even with over 4 years of use.74 Two studies reported mild side-effects in a small number of children,86 including three of 107 children with ASD who had morning drowsiness or increased enuresis.66 One study reported that side-effects were not significantly different from those reported with a placebo.58 Another study reported mild morning tiredness in two children, headache in one child, and tiredness, dizziness, and diarrhea in one child resulting in study discontinuation.83 The use of melatonin in 21 children with ASD who also had epilepsy was not associated with any increase in seizure activity.66 In two studies, the use of melatonin was frequently combined with multiple other psychotropic medications without any adverse events.66,68 Four studies reported evidence of a loss of effect of melatonin in a small number of individuals,66,73,74,78 including two studies that used relatively high melatonin doses66,74 and one study that reported loss of positive effect at 4 weeks of treatment.78 A large parental survey reported that, out of 1105 children with ASD, 8% had ‘worse behavior’ after melatonin use.62 Interestingly, in one study of 50 individuals with a sleep and developmental disorder (including 27 with ASD), mild adverse effects were reported in 34% of the participants. These adverse effects included morning drowsiness, night-time awakening, and excitement before going to sleep;73 however, this adverse event rate is dramatically higher than any of the other studies in this review. Overall, no serious adverse events were reported in any of the studies.

Discussion

Biochemistry and physiology of melatonin in ASD

One of the objectives of this systematic review and meta-analysis was to review the biochemical characteristics of melatonin metabolism in ASD. Abnormalities in the levels of melatonin or melatonin derivatives in ASD were found in nine studies, and four studies reported abnormalities in the circadian melatonin rhythm in ASD. Five studies reported abnormalities in genes in a small percentage of individuals with ASD, which could contribute to the decreased production of melatonin or adversely affect melatonin receptor function. Four studies reported a correlation between concentrations of melatonin or melatonin metabolites and ASD behaviors or clinical findings.

Melatonin is synthesized from serotonin through two reactions. Abnormalities in one of the genes involved in this pathway (ASMT) were identified in four studies in a small number of individuals with ASD. Previous studies have reported that some individuals with ASD have elevated blood serotonin (hyperserotoninemia),97–103 which theoretically would lead to an elevated melatonin level. However, the reviewed studies suggest that a majority of individuals with ASD have low levels of melatonin or melatonin derivatives.51,85 In fact, in one reviewed study, children with ASD who had an elevated platelet serotonin also had a lower urinary 6-SM level.80 The reason for this paradoxical finding may be secondary to abnormalities in ASMT function. For example, one reviewed study reported significantly elevated mean serotonin along with lower mean melatonin and ASMT activity in individuals with ASD compared with healthy individuals.51 Additional studies examining melatonin, serotonin, and ASMT activity levels would be helpful to expand on these findings.

Furthermore, four studies reported evidence of a loss of melatonin effect in a small number of individuals.66,73,74,78 Some of these studies described this loss of effect as tolerance, and because of this the dose of melatonin was increased over time. However, other investigators have described a loss of effect with melatonin in some individuals without ASD owing to slow melatonin metabolism, which leads to elevated melatonin levels in the daytime, even when melatonin was not recently given, as well as a loss of the normal melatonin circadian rhythm; this problem responds to a reduction in melatonin dose rather than an increase in dose.53 Interestingly, significantly higher daytime levels of melatonin have been reported in some individuals with ASD compared with healthy individuals 81,84 as has an abnormal melatonin circadian rhythm.51,77,79,81 Thus, additional studies are needed to determine whether the loss of melatonin effect observed in some individuals with ASD is related to melatonin metabolism and/or abnormalities in circadian rhythms, and to examine the relationship between genetic polymorphisms and these factors.

Since the concentration of melatonin is low in some individuals with ASD, and because some individuals with ASD have abnormalities in genes involved in melatonin synthesis, the use of melatonin may function to replace a deficiency. In the future, routine testing for melatonin levels may be helpful, but currently this testing is not readily available. Overall, the findings of this review and meta-analysis suggest that abnormalities in the physiology and concentrations of melatonin, as well as certain genetic abnormalities, may play a role in both sleep and behavioral problems in individuals with ASD.

Treatment effects of melatonin in ASD

Another objective of this review was to examine the clinical characteristics of melatonin treatment in children with ASD. Three studies reported that the prevalence of melatonin use in the general population of individuals with ASD was about 7%. Three surveys indicated that a recommendation for melatonin use is relatively common in ASD, with approximately 32% of physicians recommending its use, while one survey reported that 14% of physicians discouraged its use. Six studies reported improved daytime behavior with the use of melatonin at night-time. A total of 18 studies reported the effects of melatonin treatment in individuals with ASD who had a sleep abnormality. All of these studies reported improvements in sleep parameters, including improvements in overall sleep, sleep duration, sleep onset latency, and night-time awakenings. A meta-analysis of the five randomized, double-blind, placebo-controlled, crossover studies demonstrated an overall significant improvement in sleep duration (73min compared with baseline; 44min compared with placebo) and sleep onset latency (66min compared with baseline; 39min compared with placebo) but no overall significant improvement in the number of night-time awakenings. Reported side-effects in these studies were minimal to none.

In the meta-analysis, the effect size was larger for sleep onset latency than for sleep duration, which might be related to the half-life of fast-release melatonin. Fast-release melatonin helps children to fall asleep faster, but controlled-release melatonin helps to maintain sleep better.104 A few studies used controlled-release melatonin, and some used a combination of controlled-release and fast-release, while others used only fast-release melatonin (Table SI). Some studies did not note which form of melatonin was used, but given the special formulation required for controlled-release melatonin it is likely that these studies used fast-release melatonin. Because of an extended release system, it is possible that controlled-release melatonin might have a better effect on sleep duration than fast-release melatonin. Conversely, fast-release melatonin might help sleep onset latency better than controlled-release melatonin. Unfortunately, there were not enough studies that used controlled-release melatonin to compare changes in sleep parameters with fast-release melatonin. Additional studies are needed to examine the differences in effect between controlled-release and fast-release melatonin in ASD.

Seven studies noted the use of pharmaceutical-grade melatonin or that the melatonin was provided by the hospital pharmacy.5,56,58,68,78,83,86 Since melatonin is available without a prescription, it is possible that the quality of the melatonin supplement used by children with ASD in the general community (over the counter) may vary from that used in the studies. However, one double-blind placebo-controlled study noted significant improvements with a form of melatonin readily available over the counter.57

Potential mechanisms of melatonin treatment in ASD

The potential mechanism of melatonin improvement in individuals with ASD is not clear. It may act to align the circadian rhythm in children with ASD or work as a hypnotic or sedative agent.52 Alternatively, some children with ASD appear to be deficient in melatonin and, therefore, exogenous melatonin may replace a deficiency. Additionally, melatonin may also ameliorate certain physiological abnormalities that are reported in some individuals with ASD. For example, melatonin is a potent antioxidant40 and may be beneficial since individuals with ASD, as a group, have been shown to be under higher oxidative stress and have reduced levels of antioxidants compared with healthy individuals.105–114 Furthermore, melatonin has anti-inflammatory properties,36 and evidence of inflammation has been described in some individuals with ASD.115–120 Melatonin may also have a positive effect on the immune system in ASD. For example, a dysfunctional natural killer cell system has been noted in some individuals with ASD,121–124 and melatonin is known to have a positive effect on natural killer cell production.43 Melatonin also stimulates the production of CD4+ cells,43 and several studies have reported lower levels of certain populations of CD4+ cells in individuals with ASD.125–128

Furthermore, since melatonin helps regulate synaptic plasticity,45,46 abnormalities in melatonin pathways could play a role in the development of ASD. For example, an imbalance in the excitatory (glutamatergic) and inhibitory (GABAergic) neurotransmitter systems has been implicated in the pathogenesis of ASD, with a relative increase in the glutamatergic neurotransmitter system.129 Since GABAergic neurons are inhibitory, they may be especially important between the ages of 12 and 30 months as this window of development corresponds to an overproduction in excitatory neurotransmitters and receptors.130,131 Thus, without proper GABAergic neuronal function, the brain may be highly susceptible to excitotoxicity during this developmental period. Interestingly, this is the age range when developmental regression most commonly occurs in ASD.132 In animal models, melatonin is known to enhance brain GABAergic transmission133,134 and, therefore, a lack of melatonin production could contribute to an imbalance in the glutamatergic to GABAergic systems. Further studies are needed to investigate these possibilities.

Safety of melatonin

The reviewed studies indicate that melatonin is safe and inexpensive, and has been used extensively in children with ASD, sometimes for years at a time.74 Seven of the reviewed studies reported that melatonin did not have any side-effects, while the remaining studies reported transient or mild side-effects in a small number of individuals. Even when combined with other psychotropic medications, melatonin was well tolerated and had no apparent interactions with these medications. Furthermore, in one study, melatonin did not increase seizure activity in children with ASD who were already taking seizure medication,66 and no studies were identified in individuals with ASD who reported an increase in seizure activity with melatonin. In fact, one study reported that a lower mean serum melatonin level was associated with an abnormal EEG in individuals with ASD.81

Limitations of this review

Some reviewed studies suffered from limitations, including small sample sizes (some were case reports) and variations in protocols for measuring changes in sleep parameters. Five of the reviewed studies contained a mixture of individuals with ASD and individuals with other developmental disabilities. However, we found a similar prevalence of sleep improvements with melatonin when we included case reports and studies containing individuals with other developmental disabilities as when we excluded these studies. In addition, we were able to obtain data on the children with ASD for two of these studies.

The meta-analysis was limited by inconsistencies across studies in the method used to collect the sleep parameter data. For example, some studies relied on a subjective parental report of changes in sleep parameters (parent-report questionnaires and sleep diaries) while others used more objective measures (actigraphy) and some used a combination of both. Additional larger controlled studies of melatonin treatment using a combination of subjective and objective measures of sleep parameters would be helpful to confirm and expand on the findings of this review.

Interestingly, several studies reported improvements with melatonin when other sleep medications had previously failed.64,68,79,87 However, none of the reviewed studies compared the effects of melatonin with other sleep medications, so conclusions about the effectiveness of melatonin compared with other sleep medications cannot be drawn.

Conclusions

This systematic review found that, when measured, levels of melatonin or melatonin derivatives are often below average in individuals with ASD compared with healthy individuals. Furthermore, the physiology of melatonin is abnormal in many individuals with ASD, which in some cases correlates to ASD symptoms. Some individuals have abnormalities in genes involved in melatonin production or receptor function. The meta-analysis found that the use of melatonin in ASD is associated with significantly improved sleep parameters (sleep duration and sleep onset latency). Furthermore, melatonin appears to improve daytime behavior in some individuals with ASD and has minimal to no side effects. However, additional studies are needed to examine melatonin metabolism in ASD, including the relationship between melatonin and serotonin in ASD. Some studies reported a higher daytime melatonin level in children with ASD, which could be related to slower melatonin metabolism; this finding needs further investigation. Studies examining optimal effective dosing,53 optimal timing of melatonin dosing,53 the potential differential effects of fast-release and controlled-release melatonin, and its long-term adverse effects and safety in ASD135 are also needed. Additional studies looking at the effects of melatonin on other sleep parameters in ASD, such as early morning awakening, are needed.135 Finally, studies examining the mechanism of melatonin in possibly ameliorating certain findings associated with ASD, including oxidative stress, inflammation, immune dysregulation, and abnormal neurotransmitter function are needed.

References

References are available online as Supporting Information.

Ancillary