Our findings support the hypothesis that parental reports of sleep concerns in children with ASD can be confirmed not only by comprehensive questionnaires but by objective measures that include both PSG and actigraphy. Children classified as poor sleepers (as compared to good sleepers) had a longer sleep latency on both PSG and actigraphy, and more sleep fragmentation and WASO on actigraphy. Children with ASD who were classified as good sleepers showed comparable (or, in some cases, better sleep) compared to TD children on objective and subjective sleep parameters. Furthermore, we identified associations between the sleep phenotype and daytime behavior, with poor sleepers with ASD showing more hyperactivity and compulsive and ritualistic behavior than good sleepers with ASD. Objective actigraphic measures also related the sleep phenotype to daytime behavior, with WASO correlated with hyperactivity and sleep fragmentation correlated with restricted/repetitive behaviors. These associations will require confirmation in larger studies.
Our work is the first to simultaneously assess the ability of PSG and actigraphy to distinguish sleep patterns in children with ASD based on parental concern. We had expected to find that actigraphy and PSG were comparable in distinguishing sleep patterns. Our results show that actigraphy not only is comparable to PSG, but it also provides additional information in the areas of nighttime activity (e.g., FI and WASO). While PSG has traditionally measured arousals and awakenings as indicators of fragmentation and restlessness, it appears that movement itself, regardless of whether it results in an arousal or awakening, may impact adversely on sleep and daytime behavior. Further research is warranted into the nature and impact of these movements on the sleep and behavior of these children. The reason for the increased WASO depicted by actigraphy relative to PSG is unclear. WASO, measured with actigraphy, represents the sum of all wake epochs during the sleep period and reflects the number of minutes that exceed the sensitivity threshold and are scored as wake As PSG and actigraphy use different algorithms to detect wake epochs, it is possible that the differences in WASO reflect methodological differences in PSG and actigraphy Further analysis of fragmentation and WASO in larger samples, with attention to the number and durations of awakenings may help delineate these differences.
In prior studies, the ability of actigraphy to differentiate sleep patterns in children with ASD has been mixed. Allik et al (2006)
, collecting one week of actigraphy in 32 high functioning children with ASD ages 8–12 years, and age and gender matched typically developing children, demonstrated prolonged sleep latencies in the children with ASD. The 19 children with sleep problems also demonstrated prolonged sleep onset delay compared to those without sleep problems. Night wakings did not differ in either the total group of ASD compared to controls, or the ASD subgroups compared to each other. In a large group of preschoolers (n = 194) that included children with ASD, developmental delay (DD), and those who were TD, parental reports of sleep problems were higher in the children with ASD and DD compared to the TD group (Goodlin-Jones et al., 2008
). However, while one week of actigraphy and sleep diaries were significantly positively correlated, mean values differed, with diaries underreporting sleep latency and WASO compared with actigraphy. A subsequent paper by the same group combining video with actigraphy examined night wakings in a subset of that group (n=58) (Sitnick et al., 2008
). They concluded that actigraphy had poor agreement in detecting nocturnal awakenings compared to video recordings. Two other studies that included older children with ASD also found discrepancies between parent report and actigraphy, although in contrast to Goodlin-Jones et al., reported that parents appeared to over report
sleep problems in their children.(Hering et al., 1999
; Wiggs and Stores, 2004
) With the exception of an early morning arousal time, Hering et al (1999)
found that 72 hours of actigraphy did not differ in children with ASD and parental sleep concerns (n = 8; ages 3–12 years) as compared to age and gender-matched controls. Wiggs and Stores (2004)
compared 38 ASD children with parentally-reported “sleeplessness” and 24 ASD children without a parentally-reported sleep problem, ages 5–16 years. Five nights of actigraphy were recorded. The two groups did not differ on actigraphic parameters, although sleeplessness was defined broadly and included reluctance to go to bed, insistence on sleeping with someone else, and excessive daytime sleepiness, in addition to sleep onset delay and night wakings.
The differences between these studies and ours may be attributable to several factors. First, in our study, actigraphy data were collected in a controlled environment, rather than in the child’s natural setting. Second, we used a restricted age range, with a sample limited to higher functioning children without intellectual disability, epilepsy, or psychiatric conditions requiring psychotropic medications. Both of these factors may have contributed to the integrity of our data and decreased variability in actigraphic measurements. In one study of actigraphy reliability, preschoolers and adolescents showed poorer reliability than 4 and 5 year olds. (Acebo et al., 1999
). Third, our groups of children with and without parentally reported sleep problems showed significant differences on the CHSQ, with comprehensive sleep histories confirming parentally reported sleep concerns. Fourth, PSG was performed in all children to evaluate for co-existing sleep disorders. Finally, differences in actigraph equipment and scoring algorithms among studies may have affected our results.
Polysomnography remains the gold standard for detecting sleep disorders such as obstructive sleep apnea, and, when combined with video and EEG, parasomnias and seizures. Therefore, we recommend PSG be performed in any child in whom the clinical suspicion of one of these disorders is high. It is notable that we found a relatively low prevalence of sleep disorders other than insomnia in our cohort, with only one child having obstructive sleep apnea and another having sleep terrors. Both were suspected to have these disorders based on their sleep histories. None had REM behavior disorder or sleep-related epileptic seizures. REM behavior disorder has been reported previously in a case series of children with ASD, although this series differed from ours in containing children with coexisting intellectual disability (Thirumalai et al., 2002
). REM behavior disorder can also be caused by some of the antidepressants used in this population to treat coexisting psychiatric disorders. Epilepsy has also been associated with intellectual disability in autism (Tuchman and Rapin, 2002
), and in a cohort of children taking psychotropic medications with coexisting intellectual disability, parasomnias were more prevalent (Ming et al., 2008
). Future studies examining sleep in a more heterogenous group of children with coexisting intellectual disability or taking psychotropic medications will be necessary to determine the true prevalence of sleep disorders other than insomnia in this population.
Our study had several strengths. Our participants were a well-defined group of children with ASD, whose clinical diagnosis was confirmed with the ADOS. Parental sleep concerns were confirmed by sleep histories, and all had video-EEG-PSG to exclude sleep disorders such obstructive sleep apnea, parasomnias, and sleep-related seizures. They were free of psychotropic medications and did not have intellectual disability or epilepsy. This has allowed us to establish a baseline sleep phenotype, free of confounders, in this population. Our staff were successful in obtaining recordings in all children, although the higher functioning nature of our cohort may have contributed to this success. The characteristics of our cohort may also be viewed as a limitation, in that results may not be generalizable to all children with ASD. Another important strength of our study, but also a limitation, is that actigraphy data were collected in a controlled setting, where precise measurements of “lights off” could be recorded and the integrity of the recordings could be ensured. In the home setting, with competing stressors, parents may not remember to activate event markers or even to place the watch on the child. Future studies will be necessary to confirm our findings in the home setting. Of note, we have documented a significant decrease in actigraphic sleep latency in a cohort of children whose parents received behavioral sleep education (Reed et al., in press
). Another limitation of our study is that night wakings were relatively in common, both by history and by objective findings. In one large study surveying 210 children with ASD, night wakings were more frequent in children with ASD and coexisting intellectual disability (Williams et al., 2004
). Had our sample contained children with mixed cognitive levels, it is possible we may have detected more night wakings. Finally, we used subjective parentally-reported measures of daytime behavior to differentiate the ASD-PS and ASD-GS groups. However, it is noteworthy that parents did not globally report problems in daytime behavior—inattention and hyperactivity were associated with sleep concerns. Inattention and hyperactivity have also been reported in association with sleep disorders in typically developing children (Chervin et al., 2002
; Gottlieb et al., 2003
In summary, our results suggest that parentally based sleep concerns in children with ASD are substantiated by objective parameters, and selectively influence measures of daytime behavior. Actigraphy and PSG are complementary objective measures for defining sleep that substantiate parental report. Defining the phenotype of sleep in ASD, its relation to daytime behavior, and appropriate measurement modalities provides the foundation for focused studies of sleep pathophysiology and targeted interventions in this population.