Early Sleep Differences in Young Infants with Autism Spectrum Disorder

Abstract

Objective:

Children with autism spectrum disorder (ASD) experience greater sleep challenges than their neurotypical peers, but sleep patterns for infants later diagnosed with ASD are unknown. This study examined differences in total sleep duration and proportion of sleep experienced at night within the first six months of life among infants later diagnosed with ASD, infants who demonstrated subclinical characteristics of ASD and were classified as exhibiting the broad autism phenotype (BAP), and their typically developing (TD) peers. In addition, associations between infant sleep variables and developmental outcomes at 24 months were explored.

Method:

Participants included 79 infants enrolled in a prospective, longitudinal study of the early development of ASD. Between 1 week and 6 months of age, participants completed a monthly retrospective 24-hour sleep log. At 24 months, participants received a comprehensive diagnostic evaluation, including the ADOS-2, MSEL and Vineland-II, and were clinically characterized as ASD, BAP, or TD.

Results:

When accounting for the influence of age, infants later diagnosed with ASD slept less within the 24-hour period than infants in TD or BAP groups from 0 to 6 months (P=.04). Percentage of sleep experienced during nighttime hours did not significantly differ between groups from 0 to 6 months (P=.25). Greater nighttime sleep percentage at 6 months predicted higher receptive language (P<.001) and fine motor scores (P<.0001) at 24 months. Total sleep duration at 6 months did not predict any developmental outcomes at 24 months.

Conclusion:

Findings suggest that differences in sleep may occur among autistic individuals earlier in life than previously documented and have cascading effects on development.

Children with autism spectrum disorder (ASD) are more likely to experience sleep disturbance than their neurotypical peers. Specifically, between 50 to 80% of autistic children experience challenges with falling asleep, reduced sleep duration, and frequent wakings at night^1,2 as compared to 10 to 30% of neurotypical children.^3,4 Due to the high prevalence of sleep problems in autistic youth, some researchers suggest that ASD and disrupted sleep may result from similar underlying neurodevelopmental mechanisms and possibly belong to a shared etiology.⁵ To better understand the progression of sleep problems in ASD, it is critical to determine when differences in sleep-wake organization first emerge in development.

Emerging evidence suggests that pre-diagnosed infants and young children with ASD demonstrate differences in parent-reported sleep characteristics than their neurotypical peers during the first three years of life.^2,5,6 For example, the frequency of nighttime wakings at 12 months of age has been found to predict higher scores on the Modified Checklist for Autism in Toddlers (M-CHAT) at 24 months of age.⁶ Meanwhile, others did not find significant differences in nighttime sleep duration in pre-diagnosed autistic children at 6 months or 18 months, but did find significantly reduced nighttime sleep duration at 30 months.² This difference remained significant through the end of data collection at the age of 11 years,² suggesting that such sleep differences persist through childhood and possibly beyond.

Other researchers have identified difficulty falling asleep at bedtime among infants, children, and adolescents with ASD.^7,8 For example, MacDuffie et al⁵ found that infants who were later diagnosed with ASD demonstrated greater parent-reported challenges with falling asleep at 6 and 12 months.⁵ These early sleep onset challenges were associated with atypical hippocampal development, suggesting a neurological basis for sleep problems in autistic youth. To explain these sleep disturbances in older samples, many theorize that autistic children and adolescents synthesize less melatonin than their neurotypical peers as indicated by lower melatonin and melatonin-derivatives in salivary and urinary samples of autistic children and adolescents aged 4 to 18 years.⁹ As melatonin is critical for modulating circadian rhythm, or the body’s internal clock that regulates sleeping, waking, and feeding states,¹⁰ it is likely that differences in circadian rhythmicity and organization is a major contributing factor to the characteristic insomnia seen in ASD populations.¹¹

Differences in infant circadian rhythm development in ASD is especially intriguing as the first few months of life among neurotypical samples is characterized by rapid changes in melatonin production and circadian rhythm development. Specifically, newborn infants are unable to synthesize their own melatonin as the neural structures responsible for the production of melatonin production remain undeveloped until approximately 3 months of age.¹² During this critical period, infants transition from polyphasic sleep organization (e.g., sleeping for evenly distributed periods in the day or night) into a circadian rhythm of sustained wakefulness in the day and uninterrupted sleep at night.¹³

Although circadian rhythm maturation has been extensively studied in typically developing (TD) infants, questions remain if this process differs among infants later diagnosed with ASD. For example, studies have demonstrated associations between infant sleep challenges and concurrent and future delays in executive attentional control^14-16 and language development^17,18 in TD infants. The role of infant sleep on subsequent development for those with ASD, and possible cascading effects of disrupted infant sleep on neurodevelopmental trajectories of ASD, is unknown.¹⁹ Most notably, it remains unclear at what age infants later diagnosed with ASD exhibit decreased daytime sleep and increased nighttime sleep duration. By tracking 24-hour sleep wake cycles across the first months of life, researchers can infer atypical and/or delayed circadian rhythm development among infants later diagnosed with ASD.

In this prospective longitudinal study, we aimed to examine parent perceptions of sleep-wake organization from 0 of 6 months of age among infants with and without ASD. Additionally, we explored the potential impact of nighttime sleep percentage and total sleep duration within the 24-hour period at 6 months on later autistic trait severity and developmental outcomes at 24 months of age.

Methods

Participants

Participants included 79 infants who were enrolled prior to 6 months of age in a prospective, longitudinal study of early development of ASD. 10 participants were later diagnosed with broad developmental delay (see below) and were excluded from analysis, resulting in a final sample size of 69 participants. Participants were classified as either having a low familial likelihood (LL) or high familial likelihood (HL) of developing ASD. LL participants included infants without familial history of ASD in first- or second-degree relatives. HL participants had an older full-biological sibling with a confirmed ASD diagnosis. The status of the older sibling was confirmed through clinical review of a diagnostic evaluation report along with scores on the Social-Responsiveness Scale-2²⁰ and the Social Communication Questionnaire.²¹ If the diagnostic report was not completed within the past 5 years or did not include documented Autism Diagnostic Observation Schedule (ADOS) assessment with calibrated severity scores (CSS) in the ASD range, then the older sibling completed a diagnostic evaluation at the clinic. Exclusion criteria included first-born infants, gestational age prior to 35 weeks, major visual/hearing impairment, non-febrile seizure disorder, confirmed genetic syndrome, and substantial pre- or peri-natal medical complications including low birthweight, and any condition requiring the infant to be monitored in the NICU for longer than 24 hours.

At the age of 24 months, all participants received an evaluation for ASD in adherence to the accepted reference standard for diagnosis. A clinical best estimate (CBE) diagnostic procedure, described below, was used to determine the diagnostic status of all participants. From this process, 40 (51%) participants were determined to be typically developing (TD), 20 (25%) participants were diagnosed with ASD, and 9 (11%) exhibited the broad autism phenotype (BAP). BAP is defined as the subclinical characteristics that are frequently seen in family members of children with ASD.²² For this study, a classification of BAP included the subclinical presence of restrictive and repetitive interests and/or social-communication challenges.²³ Participants with developmental delay who did not exhibit core autistic traits (n=10), as judged by expert clinicians and a comprehensive evaluation per the CBE procedure, were excluded from analysis. Participant demographic characteristics are displayed in Table 1.

Table 1.

Sample Characteristics by Diagnostic Group

	ASD N = 20a	BAP N = 9b	TD N = 40c	Test Statistic
Familial Likelihood				p<.00011
High Likelihood	19 (95%)	7 (78%)	9 (23%)
Low Likelihood	1 (5%)	2 (22%)	31 (77%)
Sex				p=.31131
Male	15 (75%)	6 (77%)	22 (55%)
Female	5 (25%)	3 (33%)	18 (45%)
Child’s Race				p=.48141,d
White	11 (55%)	2 (22%)	31 (78%)
Black	3 (15%)	0 (0%)	2 (5%)
Asian	1 (5%)	0 (0%)	0 (0%)
More than One Race	1 (5%)	0 (0%)	3 (7%)
Ethncty				p=.59381
Hispanic or Latino	0 (0%)	0 (0%)	3 (8%)
Not Hispanic or Latino	16 (80%)	2 (22%)	34 (85%)
Household Income				p=.00351
<$40,000	5 (25%)	0 (0%)	0 (0%)
$40,000-$80,000	2 (10%)	1 (11%)	5 (12%)
$80,000-$125,000	4 (20%)	1 (11%)	11 (28%)
>$125,000	4 (20%)	0 (0%)	20 (50%)
Maternal Education				p=.00021
High School Degree	6 (30%)	1 (11%)	0 (0%)
College Degree	9 (45%)	1 (11%)	16 (40%)
Graduate Degree	2 (10%)	4 (44%)	19 (48%)
ADOS Total CSS, M (SD)	6.11 (2.47)	3.00 (1.07)	1.69 (1.33)	p<.00012
MSEL T-Scores3
Visual ReceptionΨ, Ω	49.56 (17.89)	58.88 (12.03)	62.00 (9.08)	p=.00672
Receptive LanguageΨ,ǂ	37.61 (18.98)	58.88 (11.57)	57.06 (9.31)	p<.00012
Expressive LanguageΨ,ǂ	37.50 (16.84)	54.00 (14.54)	54.69 (10.22)	p=.00012
Fine MotorΨ,ǂ	43.61 (13.28)	56.25 (6.54)	53.88 (11.13)	p=.00582

^{a =}(a) 4 children had unknown race/ethnicity data; 5 children had unknown Household Income data; 2 children had missing outcome data

^{b =}7 children had unknown race/ethnicity/household income data, 1 child had missing outcome data

^{c =}4 children had unknown race/household income data; 3 children had unknown ethnicity data; 7 children had missing outcome data

^{d =}The following race categories were combined into one category for analysis: Asian, More than One Race, and Other.

^{1 =}Fisher’s Exact Test were used to examine differences between diagnostic groups due to the small expected values.

^{2 =}One-way ANOVA

^{3 =}MSEL scores are represented by t-scores with an average of 50 and a standard deviation of 10

^{Ψ =}Post-Hoc Analysis using Tukey’s Honest Significant Difference (HSD) tests revealed significant differences between ASD and TD group means in MSEL Domain scores.

^{Ω =}Post-Hoc Analysis using Tukey’s HSD tests revealed significant differences between TD and BAP group means in MSEL Domain scores.

^{ǂ =}Post-Hoc Analysis using Tukey’s HSD tests revealed significant differences between ASD and BAP group means in MSEL Domain scores.

Procedure

Infants were seen in the laboratory for a battery of clinical assessments and experiments monthly from 1 week to 6 months of age. As part of this battery, parents completed a 24-hour sleep log about the infant’s previous day and night. If the respondent was not present with the infant during part of the 24-hour period and could not retrieve this information from another caregiver, they were instructed to not complete the form. At 24 months, all infants completed an evaluation for ASD in adherence to the accepted reference standard for diagnosis. Psychologists who performed the evaluation were masked to the likelihood status of each participant.

Measures

Infant Sleep Questionnaire.

At each visit, information on infant sleep/wake cycle was collected through a retrospective 24-hour sleep log. Parents were first asked if the past 24 hours represented a typical sleep pattern for their infant. If so, they indicated the behavioral state of their infant for each 30-minute increment in the past 24 hours. Behavioral states included: awake and alert, crying, feeding, or sleeping. Parents were instructed to select that state that best represented the majority of that 30-minute interval. In order to measure circadian system maturation, the percentage of time spent asleep at night within the total amount of time spent asleep in the 24-hour period was calculated. It should be noted that the 24-hour sleep log was developed internally and has not yet been tested for reliability or validity.

Clinical Best Estimate Diagnosis.

At 24 months, a CBE diagnosis was determined by a trained, licensed psychologist with expertise in infant development and ASD after administering a comprehensive diagnostic evaluation. This evaluation included administration of the Mullen Scales of Early Learning (MSEL), the Vineland Adaptive Behavior Scales, Second Edition (Vineland-II), and the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). The MSEL is a direct assessment used to determine developmental and adaptive functioning in toddlers.²⁴ The Vineland-II is parent-report clinical interview to assess the child’s adaptive behavior, which can be defined as day-to-day activities necessary to take care of oneself and successfully interact with others.²⁵ The ADOS-2 is a semi-structured assessment of social skills, communication abilities, and repetitive behaviors and restricted interests commonly seen in ASD.²⁶ The Calibrated Severity Score (CSS) was calculated as a measure of autistic trait severity for each participant.

A diagnosis of ASD was provided if the participant’s performance in the evaluation met DSM-5 criteria. Participants were considered to be typically developing if they did not exhibit symptoms of ASD and their MSEL scores did not indicate developmental delay. Participants who demonstrated atypical, subthreshold traits associated with ASD, as determined by clinician observation during the evaluation were classified as exhibiting the broad autism phenotype (BAP). This determination is similar to other longitudinal infant sibling studies.^22,27 Occasionally, participants did not demonstrate ASD symptoms, but exhibited significant developmental delays and were excluded from analysis. Significant developmental delays were defined as scores that fell two standard deviations below the average score in one domain of the MSEL or one and a half standard deviations below the average score in two or more domains of the MSEL.

Statistical Analysis

Descriptive statistics of demographic characteristics, including sex, race/ethnicity, household income, and developmental outcomes were computed and compared using ANOVA and Fishers exact tests (see Table 1). The primary aim of this study was to compare early sleep trajectories across infants who were later diagnosed with ASD, infants categorized as BAP, and typically developing infants. Specifically, linear mixed models were used to examine the effect of age and diagnosis on 24-hour sleep duration and percentage of sleep at night between ASD, TD, and BAP infants across the first six months of life (see Table 2). Next, the potential impact of nighttime sleep percentage and total sleep duration within the 24-hour period at 6 months on later autistic trait severity, represented by ADOS-2 CSS scores, and developmental outcomes across MSEL domains of visual reception, receptive language, expressive language, and fine motor skills were examined using multiple regressions (see Table 3).

Table 2.

Main Effects and Interaction of Age and Diagnosis in Linear Mixed Models

	Main Effect of Diagnosis	Main Effect of Age	Interaction of Diagnosis and Age
Night Sleep Percentage	0.19	58.02**	1.39
Total Sleep Duration	1.52	14.07**	3.19*

^{* =}p < .05

^{** =}p <.01

Table 3.

Pearson Coefficients Between Sleep Variables at 6 Months and Developmental Outcomes at 24 months

	Autistic Trait Severity	Visual Reception	Expressive Language	Receptive Language	Fine Motor
Night Sleep Percentage	.29	.04	.44*	.43*	.51**
Total Sleep Duration	.38ǂ	.03	−.06	−.12	−.12

^{ǂ =}p < .10

^{* =}p < .05

^{** =}p <.01

It is well established in the literature that maternal education^28,29 and income²⁸ are positively associated with infant sleep duration. Further, maternal education has been shown to be strongly associated with both income and socioeconomic status.³⁰ As other studies investigating similar relationships used maternal education alone to establish SES groups,^29,31 all analyses were run with maternal education as a covariate.

Results are presented as model-based means and are accompanied by test statistics and p-values. Unless otherwise noted, significance was assessed at the 0.05 level. All significance tests were two-sided. SAS v. 9.4 was used to conduct statistical analyses.

Results

Descriptive statistics are displayed in Table 1. Overall, participants in the ASD and TD groups significantly differed in regard to familial likelihood, maternal education, household income, and outcome ADOS-2 and MSEL scores. As expected, the ASD group included significantly more infants with high familial likelihood, demonstrated elevated ADOS-2 CSS scores, and showed lower scores across all MSEL subscales. The ASD group included more participants with lower maternal education and household income. Maternal education was used as a covariate in the linear mixed models, but did not emerge as significant for either nighttime sleep percentage (F_2,40=2.67, P=.08) or total sleep duration (F_2,40=.62, P=.55).

Participants had the opportunity to complete 6 sleep logs over the course of the study. On average, participants completed the sleep log 4 times across the first 6 months of life. A small number of participants only completed the sleep log once (n=3), and most participants completed the sleep log between 2 and 5 times (n=56).

In regard to the percentage of time infants spent asleep through the night (e.g., 7:00pm to 7:00am) from 0 to 6 months, age was a significant predictor (F_1,139=58.02, P<.0001), but the effect of diagnosis was nonsignificant (F_2,40=.19, P=.83). There was no observed interaction between age and diagnosis (F_2,139=1.39, P=.25) (see Figure 1; see Table 2). Across all infants, nighttime sleep percentage significantly increased by about 18% from 0 to 6 months (P<.001) with no significant differences in rate of this increase between diagnostic groups.

Figure 1.

Trajectories of parent-reported 24-hour sleep duration from 0-6 months for infants later characterized as ASD, Broad Autism Phenotype (BAP), and Typically Developing (TD)

Regarding sleep duration within the 24-hour period, there was a significant interaction between age and diagnosis (F_2,139=3.19, P=.04) and a main effect of age (F_1,139=14.07, P<.001), but no main effect of diagnosis (F_2,40=1.52, P=.23) (see Figure 2; see Table 2). Regardless of diagnosis, infants showed an average decrease in total sleep duration of 2.35 hours from 0 to 6 months.

Figure 2.

Trajectories of parent-reported proportion of sleep at night from 0-6 months for infants later characterized as ASD, Broad Autism Phenotype (BAP), and Typically Developing (TD)

When assessing the relationship between both types of sleep variables (i.e., 24-hour sleep duration and nighttime sleep percentage) and developmental skills, the following trends emerged. Across all participants, nighttime sleep percentage at six months significantly predicted 24-month fine motor skills (F_2,168=10.95, P<.0001, $R_{adjusted}^2=.10$) and receptive language (F_2,168=9.61, P<.001, $R_{adjusted}^2=.09$) (see Table 3). Nighttime sleep percentage did not significantly predict 24-month autistic trait severity, expressive language, or visual reception skills (see Table 3). Total sleep duration within the 24-hour period among 6-month-old participants did not significantly predict 24-month fine motor skills, receptive language, expressive language, visual reception, or autistic trait severity (see Table 3).

Discussion

Results of this longitudinal prospective study suggest observable differences in sleep trajectories for infants later diagnosed with ASD. Infants later diagnosed with ASD slept less within a 24-hour period, but did not spend a smaller percentage of sleep time at night across the first six months of life in comparison to their neurotypical peers. These findings imply that although autistic children experience less overall sleep at 6 months of age, the organization of their sleep-wake cycles across the 24-hour period (i.e., proportion of sleep experienced at night) is not significantly different from neurotypical children. The authors view 24-hour sleep duration as a reflection of global neurological maturation, while the nighttime sleep percentage may reflect an approximate estimate of circadian system maturation.³² Thus, the results of this study suggest that while infants later diagnosed with ASD demonstrate similar circadian rhythm development, overall neurological development may be delayed in comparison to their typically developing peers and contribute to reduced sleep duration in infancy. Similar findings of reduced sleep duration have been observed in older children aged 2 to 11 years^1,2,, suggesting that this sleep difference may emerge earlier in life than previously considered.

Nighttime sleep percentage in infancy was associated with aspects of later development in this sample. Specifically, 6-month-old infants who spent a greater proportion of time asleep at night also demonstrated more advanced fine motor skills and receptive language at 24 months. In contrast, 24-hour sleep duration was not associated with any developmental outcomes at 24 months. This finding was surprising as infants later diagnosed with ASD experienced reduced sleep duration within the 24-hour time period, yet sleep duration at 6 months did not predict autistic trait severity on the ADOS-2. Other studies demonstrated similarly mixed findings.^33,34,35 For example, Schwichtenberg et al. (2013) also did not find associations between ADOS CSS scores and parent-reported sleep challenges at 36 months. However, this study did find strong, positive correlations between sleep challenges and parent-reported autistic traits.³⁴ Thus, stronger associations between infant and future autistic traits may have been identified in our sample if parent-reported measures of trait presentation had been included in addition to behavioral observations made by a trained clinician. Regardless, our findings suggest that the organization of infant sleep-wake cycles, rather than the amount of sleep, may provide insight into the relationship between infant sleep and skill-based neurodevelopmental outcomes, including fine motor and receptive language skills, in early childhood.

Overall, these findings may point to disrupted, and potentially delayed, circadian rhythm development in ASD and provide additional clues of the role of infant sleep patterns in neurodevelopment. Specifically, melatonin is critical for modulating circadian rhythm and it is possible that a melatonin deficit may present itself in the first year of life and affect infant sleep duration for those with ASD. Although the role of melatonin on sleep has not been extensively researched among infants later diagnosed with ASD, previous research has found reduced melatonin levels among older autistic children who exhibit significant sleep disturbance.¹ In this study, infants later diagnosed with ASD experienced reduced total sleep duration but similar nighttime sleep percentage to TD infants, implying that they experienced similar organization of sleep-wake cycles to TD infants and less overall sleep. Thus, a possible explanation for the results of this study is that early differences in melatonin production and metabolism may disrupt sustained wakefulness during the day and interfere with continued sleep at night for infants later diagnosed ASD. ¹¹ While this theory provides a possible explanation for the differences in total sleep duration observed in this sample, additional research is needed to confirm melatonin levels in samples of infants later diagnosed ASD and its relationship to developmental outcomes. Such research may yield findings that point to the cascading effects of melatonin deficits in infancy on broader development and adaptive functioning later in childhood.

Strengths, Limitations, and Future Directions

To the best of the authors’ knowledge, this study is among the first to examine sleep differences in infants later diagnosed with ASD within the first few months of life. Although there is a large literature base dedicated to exploring developmental outcomes of infants at an elevated likelihood for ASD, the majority of this research begins after 6 months of age and is focused on cognitive or language development. This study is therefore uniquely valuable in determining early differences in sleep duration and nighttime sleep percentage and how these sleep patterns influence clinical outcomes at 24 months. By answering these research questions in a prospective longitudinal design, we also minimize the risk of recall bias associated with retrospective analysis. In addition, all participants had an older sibling in the home. Thus, the impact of another child on the disruption of the infant’s sleep is equated across subgroups in this study and minimizes the risk of additional confounding variables in the participant’s environment.

There are several limitations that must be addressed. Most notably, the results of this study relied on parent report of sleep and may not be as accurate as objective measures of sleep, such as actigraphy and polysomnography.³⁶ Numerous studies, for example, have identified possible environmental influences that reduce the reliability of parent-report measures of infant sleep, such as the quality of parent–child attachment and parental insomnia.^37,38 Furthermore, the sleep diary used in this study was developed internally and therefore has not yet been validated. In addition, the generalizability of the results is a noteworthy limitation as the sample size is small and it is unclear how sleep may differ among first born infants later diagnosed with ASD compared to autistic individuals who have an older sibling diagnosed with ASD.³⁹ It should also be noted that it is possible that, due to the young age of evaluation in this study, some participants may change diagnostic categories later in life. For example, the BAP group may later demonstrate additional autistic traits and thus meet criteria for an ASD diagnosis at later time points. In other studies investigating early identification of ASD, most children with BAP remained characterized as BAP beyond 24 months of age.^40,41 Thus, the authors chose to analyze the BAP participants separately from those with ASD rather than combining the two participant groups.

Future research on infant sleep trajectories within ASD should incorporate objective measures of sleep, in addition to self-report questionnaires. Findings from this study can serve as a launching point for tracking neurodevelopmental trajectories and circadian rhythm development in autism. In addition, future studies should address the mechanisms underlying the association between atypical infant sleep and future autistic trait severity. Although sleep challenges are often overwhelming for children and their parents, there is hope for successful treatment as circadian rhythms are fluid and are greatly influenced by environmental factors. The modifiability of sleep rhythmicity therefore provides exciting opportunities for developing targeted intervention addressing both behavioral challenges at bedtime and reduced sleep duration. With recent technological advancements in genotyping and neuroimaging, scientists are now able to measure sleep and its metabolic pathways with more precision than ever before. This increased academic interest towards disrupted sleep in ASD may lead to improved understanding of sleep not only in neurodiverse communities, but within neurotypical populations as well.

Conclusion

The amount of sleep experienced in the first six months of life was significantly reduced in infants later diagnosed with ASD than infants categorized as BAP or TD. Although nighttime sleep percentage did not significantly differ between groups, greater proportions of sleep experienced in the nighttime at 6 months predicted more advanced fine motor and receptive language skills at 24 months. Meanwhile, total sleep duration at 6 months was not associated with any developmental outcomes at 24 months. The findings of this study suggest that differences in sleep-wake cycles in autism possibly occur earlier in life than previously thought and may lead to future inquiry on the role of sleep in the etiology of ASD. It is critical for researchers to examine sleep differences as a potential early disruption in development as it may lend new insights into the emergence and progression of autistic traits in early childhood.