Original Article

Sleep Differences Among Children With Autism Spectrum

Disorders and Typically Developing Peers: A Meta-analysis
Marilisa G. Elrod, MD, PhD, Bradley S. Hood, MD

ABSTRACT: Objective: Sleep problems such as difficulties in sleep initiation, nighttime awakening, and
shortened sleep time are often subjectively reported in children with autism spectrum disorder (ASD).
However, results of objective studies have been mixed. Our goal was to evaluate the existing data from
objective measures using a systematic approach to identify and describe the differences in sleep parameters
by comparing total sleep time (TST), sleep latency (SL), and sleep efficiency (SE) in children with ASD with
those of typically developing (TD) peers. Methods: Studies that used objective measures such as actigraphy or
polysomnography (PSG) to describe the sleep parameters of TST, SL, and SE in children with ASD compared
with children with TD were identified. A meta-analysis was performed for the 10 studies that met inclusion
criteria with evaluation of differences in means using random effects models. A total of 343 children with
ASD and 221 children with TD were included. Assessments for sources of heterogeneity and publication bias
were undertaken. Results: TST for children with ASD was on average 32.8 minutes less per day (95% con-
fidence interval [CI]: 16.6–49.0 minutes) than their TD peers. Average SL was 10.9 minutes longer (95% CI:
6.7–15.0 minutes), and average SE was 1.9% less (95% CI: 0.7%–3.1%) than their TD peers. Notable het-
erogeneity was found within studies for TST, and mild heterogeneity was found for SE. Concurrent intellectual
disability was a moderator of TST. Children with ASD and normal intelligence had a small and nonsignificant
decrease in TST as compared with TD peers, whereas those with ASD and intellectual disability (ID) had
a significant decrease in TST as compared with TD peers. The magnitude of the difference in mean SL and SE
increased as compared with TD peers as age increased. Studies that used PSG and those that did not include
children on medications were more likely to report mean decreases in SE. Conclusions: Children with ASD
have small but measurable objective differences in their sleep parameters that are consistent with subjective
reporting. Children with ASD have shorter TST, longer SL periods, and decreased SE as compared with TD
peers. Concurrent ID, medication use, method of data collection, and age of subjects significantly moderated
these results. The decrease in TST in children with ASD and normal intelligence was not significant as
compared with TD peers, suggesting that ID may help explain the shortened TST in children with ASD.
(J Dev Behav Pediatr 36:166–177, 2015) Index terms: autism, sleep, sleep parameters, sleep latency, sleep efficiency, total sleep time, polysomnography,
actigraphy.

S leep problems in children with autism spectrum dis-

order (ASD) are ubiquitously reported in the literature.1–5
Poor sleep can be significantly impairing for a child with
ASD and his or her caregivers. Associations have also
been shown between sleep problems and intensified
symptoms of autism and between sleep problems and
decreased daytime cognitive and adaptive functioning.6,7
From the Department of Developmental-Behavioral Pediatrics, Madigan Army
Medical Center, Tacoma, WA.
Received August 2014; accepted December 2014.
Disclosure: The authors declare no conflict of interest.
Supplemental digital content is available for this article. Direct URL citations
appear in the printed text and are provided in the HTML and PDF versions of this
article on the journal’s Web site (www.jdbp.org).
The views expressed are those of the authors and do not reflect the official policy
of the Department of the Army, the Department of the Navy, the Department of
Defense, or the U.S. Government.
Address for reprints: Marilisa G. Elrod, MD, PhD, Department of Developmental-
Behavioral Pediatrics, Madigan Army Medical Center, ATTN: MCHJ-CLP-D, Joint
Base Lewis McChord, WA 98431; e-mail: marilisa.g.elrod.mil@mail.mil.
Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved.
Parents of children with ASD have been shown to have
higher levels of stress than both parents of typically de-
veloping (TD) children, and those with other dis-
abilities.8 This increased stress has been linked to the
severity of behavioral symptoms and their children’s
poor sleep.9,10 Also, parents of children with autism have
been shown to have poorer sleep quality and more
fragmented sleep than other parents.11,12
There have been many attempts to quantify the prev-
alence of sleep abnormalities and characterize sleep
behaviors and sleep quality in this population. However,
many of these studies have been subjective and often
lacked validated sleep questionnaires or strict diagnostic
criteria for autism.1,13 In general, parent-reported sleep
problem rates collected by survey are found to range from
50% to 80% in children with ASD compared with 9% to
50% for comparison groups.14–26 The most common sleep
complaints are related to bedtime resistance, sleep initia-
tion, nighttime awakening, and shortened sleep time.19
Early morning and night waking and shortened total sleep
duration have been reported.18,22,27 However, in a report

166 | www.jdbp.org Journal of Developmental & Behavioral Pediatrics

Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
 by Schreck and Mulick28 using parental survey methods in
a population of children aged 5 to 12 years, overall sleep
duration was found to be similar between children with
ASD and children with TD. Hours of sleep per night were
also similar between groups of children with TD, children
with autism, and children with Asperger’s disorder in
a report by Polimeni et al.16
Objective data using actigraphy and polysomnography
(PSG) have shown mixed results.13,21,29,30 Some authors
have suggested that differences between parental report
and objective findings may represent parental over-
sensitivity to sleep problems with the increased demands
of caring for a child with ASD.31 Other studies in-
vestigated the relationship between objective and sub-
jective measures of sleep with evidence of concordance
between the 2.2,32,33
This systematic review and meta-analysis aims to
summarize the objective information regarding the dif-
ferences in sleep parameters in children with ASD
compared with their TD peers by focusing on measures
that capture the most often described complaints of
prolonged sleep latency (SL), decreased total sleep time
(TST), and poor sleep efficiency (SE) with frequent and
prolonged awakenings. Efforts are made to provide
clinically relevant estimates of differences in sleep
parameters between children with ASD and TD peers.
METHODS
Search Strategy for Identification of Studies
Three separate search engines were used to identify
candidate studies, including PubMed, SCOPUS, and Google
Scholar using the search terms “Autism,” “sleep,” “sleep
duration,” “sleep problems,” and “Asperger Syndrome.”
The ancestry method was used, and the use of general
search engines helped to identify gray literature. Inclusion
criteria were determined to be the following: published in
English, published since 1994 (during the time the DSM-IV
was in use), include children, contain 1 control group of
typically developing (TD) peers, include measures of total
sleep time (TST) and/or sleep latency (SL) and/or sleep
efficiency (SE) according to standard definitions, and use
methods to determine these measures, which were ob-
jective (such as actigraphy or PSG) instead of or in addition
to surveys from individuals or caregivers. The following
definitions were used and confirmed as consistent for each
study: TST— the time in minutes from sleep onset to the
end of sleep minus the time spent in wakefulness after
initial sleep onset; SL— the time in minutes from bed-
time to the start of sleep; Sleep Efficiency (SE)— the
percentage of time from sleep onset to the end of sleep
spent in a sleep state as opposed to a wakeful state.
After removing duplicates, 201 abstracts were
screened for inclusion. Ninety-nine articles were further
scrutinized. Of these, 75 were excluded based on study
design (e.g., case series, descriptive studies, review arti-
cles, or other study designs that did not include objective
measures of sleep). The full texts of the remaining 24
Vol. 36, No. 3, April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
articles were reviewed in detail. Eight were excluded
because they contained adult subjects, 2 did not use
a control group of TD peers, 2 looked at longitudinal data
from previously studied groups, and 1 reported sleep
parameters that were determined by parental report.
One was excluded as a result of the “Leave One Out”
analysis. Ten studies were ultimately selected for in-
clusion in the meta-analysis. Refer to flow diagram in
Figure 1 for details.
Moderating Factors
Several factors were identified a priori as having po-
tential to explain variability between the results of the
previous studies. These were (1) inclusion of children in
the autism group with intellectual disability, (2) inclusion
of children in either group using potentially sleep alter-
ing medications, (3) inclusion of children in either group
with a seizure disorder, (4) method of data collection,
that is, actigraphy versus polysomnography (PSG), (5)
age of study participants, and (6) year of publication.
Intellectual Disability
Intellectual disability (ID) is hypothesized to be in-
dependently associated with alterations in sleep macro-
structure.34 Reviews of the literature in the past have
estimated the prevalence of ID in the population of
children with autism to be 50% to 70%.35 However,
recent changes in diagnostic criteria to include all per-
vasive developmental disorders into a single diagnosis of
autism spectrum disorder likely renders these previous
figures an overestimate, as many children with normal
intelligence were previously given diagnoses of Asperg-
er’s disorder or pervasive developmental disorder not
otherwise specified rather than autism spectrum disor-
der (ASD). The most recent figures from the Centers for
Disease Control based on 2010 data estimate the preva-
lence of ID in children with ASD at 31%.36
Sleep Altering Medications
It is well known that psychotropic medications alter
sleep patterns. Many children with ASD are treated with
stimulants, selective serotonin reuptake inhibitors, and
atypical neuroleptics. The most common stimulant used,
methylphenidate, is known to prolong SL and decrease
nighttime awakenings in children.37 Effects of SSRIs on
sleep architecture of children is extrapolated from adult
studies, in which SSRI use was associated with varied
effects, including prolonged SL and decreased SE.38 An-
tiepileptic medications also cause various perturbations
of sleep architecture, which are compounded by the
alterations that can be caused by the underlying seizure
disorder. Atypical neuroleptics are known to cause
sedation and therefore may shorten SL.39
Method of Data Collection
PSG is the current gold standard in measuring sleep
macrostructure, but it is expensive and cumbersome.
Wrist actigraphy measures wrist movement to detect
sleep and wake states and is better tolerated as data are
collected in the patient’s normal sleeping environment
where multiple nights of recordings can be collected.
Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved. 167
 Figure 1. Identification of independent studies for inclusion in the meta-analysis.
Although actigraphy has become increasingly common
in pediatric studies in the recent decade, there is still
a lack of standard scoring rules. Both methods use
a procedure in which the sleep period is divided into
time segments, called epochs, and each epoch is cate-
gorized as wakefulness or sleep according to either
expert opinion or set parameters. Actigraphy has been
shown to have a high accuracy in classifying epochs as
sleep across all pediatric age groups and devices when
compared with PSG; unfortunately, a significant weak-
ness is its poor ability to detect epochs of motionless
wakefulness interspersed among epochs of sleep.40 This
may overestimate TST and SE because both depend on
the identification and quantification of wakefulness
during the total sleep duration.
Seizure Disorder
Epilepsy is prevalent in children with ASD and is asso-
ciated with decreased IQ and increased age. According to
a recent study by Viscidi et al, prevalence is 12% for children
of 2 to 17 years with ASD, but 26% for those of 13 years and
168 Alterations in Sleep Parameters in Children with Autism
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
older. For 1 SD increase in IQ, prevalence of seizure dis-
orders dropped by 47%.41 Seizure activity and sleep are in-
tricately related. It has been established that epilepsy can
cause measurable alteration in sleep structure to include
longer SL, poor SE, and a higher number of awakenings.42
Age of Study Participants
As children develop, sleep patterns change. This is
most dramatic in the infant and toddler age ranges when
rapid development occurs but continues into adoles-
cence. A recent meta-analysis of objective sleep param-
eters for ages 5 to 18 years by Ohayon et al aiming to
establish normative sleep values for children and adults
found that TST decreased over the childhood period only
when measured on school days. Otherwise, TST
remained consistent until the end of adolescence. SL and
SE did not seem to significantly change across childhood
and adolescence.43
Year of Publication
Although the diagnostic criteria of DSM-IV remained
constant throughout the time period queried, diagnostic
Journal of Developmental & Behavioral Pediatrics
 practices did not. A recent Danish study described the
incidence of new cases of autism from 1995 to 2010 and
reported an increase in incidence across all diagnostic
subtypes (but especially those with normal IQ) that was
most pronounced in individuals from 14 to 20 years of
age.44 This reflects the general trend of increased di-
agnosis that is believed to be partially because of in-
creased clinical awareness for the disorder. This more
sensitive diagnostic approach would result in a larger
spectrum of severity of the cases of autism in the later
studies’ samples. Additionally, the quality of the studies
may have changed over the time period queried as
researchers improved on earlier studies or better tech-
nology became available.
Methodologic Quality Assessment
The authors independently reviewed the methodol-
ogy of each of the studies and coded all moderator var-
iables. Sources of potential bias were evaluated and
reported with decision of whether to include the study
in the pooled analysis made by the evaluators based on
these assessments per the Preferred Reporting Items for
Systematic Reviews and Meta-Analyses statement.45 A
consensus approach was used with differences resolved.
No quality score was assigned according to recom-
mendations of Sanderson et al46 for the assessment of
observational studies as there is no consensus on how to
weight various aspects of the evaluation to generate
a meaningful summary measure that can be included in
the statistical analysis.47
Data Extraction
Data for each study included demographics of the
study and control groups, diagnoses of the study and
control and groups, sample sizes for each group,
summary measures for TST, SL, and SE, and coding of the
moderator variables as described above.
All means and SDs for study groups consisting of
children with ASD were combined using a pooled SD
calculated as if the combined group had never been di-
vided for each individual article.47 For example, studies
by Goldman et al and Malow et al had separated the
children with ASD into 2 groups: good sleepers and poor
sleepers. In each of these studies, the summary statistics
for the poor and good sleeper group parameters were
combined and compared with the control group
parameters. This was possible because all subgroups
were formed after initial recruitment had occurred. See
Table 1 for the study groups that were combined and
included in the analysis.
One actigraphy study had separated weekday from
weekend data.24 The data were recombined into meas-
ures including both weekend and weekday results
commensurate with other studies included in the meta-
analysis. One study performed both actigraphy and PSG
for the study and control groups.32 Only the data from
PSG were used as this is the accepted gold standard and
actigraphy was performed only on a subset of children
who also underwent PSG.
Statistical Analyses
Calculation of Mean Differences
The differences in means between sleep parameters
in the study group of children with ASD and TD were
examined as the primary outcome measures to better
interpret results and assess clinical significance
(X ASDi 2X TDi , for i 5 TST, SL, SE).53 SDs were not as-
sumed to be equal in the calculation of pooled variance.
Standardized mean differences in the form of Hedges’ g
were also calculated and reported. Hedges’ g was

Table 1. Description of Studies Using Actigraphy or Polysomnography to Assess Sleep Parameters in Children with Autism and Typically Developing
Controls
Sample Nights Mean Subjects Intellectual Seizure
ASD Study Sizes Used in Age, Taking Disability Disorder
First Author Country Group (ASD/TD) Method Analysis yr SD Medications Included Included
Hering et al31 Israel PS 8/8 ACT 3 8.0 2.7 Yes Yes Yes
Elia et al48 Italy ASD 13/5 PSG 1 8.9 2.2 No Yes No
Allik et al 24
Sweden AS/HFA 32/32 ACT 7 10.9 1.3 No No No
Malow et al33 USA PS 1 GS 19/9 PSG 1 6.7 2.0 No No No
Bruni et al 49
Italy AS 1 ASD 18/12 PSG 1 12.4 3.0 No Yes No
Miano et al2 Italy ASD 16/18 PSG 1 9.8 2.7 No Yes No
Goodlin-Jones USA ASD 68/69 ACT 7 3.6 0.9 Yes Yes No
et al13
Goldman et al32 USA PS 1 GS 42/16 PSG 1 ACT 2 6.2 1.9 No No No
Souders et al50 USA ASD 57/37 ACT 10 7.4 2.0 Yes Yes Yes
Buckley et al51 USA ASD 60/15 PSG 1 4.6 1.8 Yes Yes Yes
Giannotti et al52,a Italy REG 1 NREG 40/12 PSG 1 5.4 2.9 No Yes No
aThe Gianotti study was included in the systematic review but not in the formal meta-analysis because of statistical issues. ACT, actigraphy; AS, Asperger syndrome; ASD,
autism spectrum disorder; GS, good sleepers; HFA, high functioning autism; NREG, nonregressed autism; PS, poor sleepers; PSG, polysomnography; REG, regressed
autism; SD, standard deviation; TD, typically developing.

Vol. 36, No. 3, April 2015 Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved. 169

Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
 selected to account for potential bias introduced with
the inclusion of studies with small sample sizes.54 In-
terpretation of Hedges’ g follows that of Cohen’s d with
effect sizes of 0.2 considered small, 0.5 considered
medium, and 0.8 considered large.
Calculation and Analysis of Pooled Differences
in Means
Each study’s differences in means were weighted
according to the inverses of their variances including both
within and between-studies variance with random-effects
models used to calculate the pooled differences in means.
Ninety-five percent confidence intervals (95% CIs) were
calculated for each study result and for the pooled estimates.
Statistical analyses were performed using Comprehensive
Meta-analysis (version 2.2; Biostat, Englewood, NJ) and R (R
Foundation for Statistical Computing, Vienna, Austria).
Heterogeneity was assessed using the Cochran Q sta-
tistic (a # .10) and the I2 statistic, which indicates the
percentage of variation in the effect size estimate attrib-
utable to heterogeneity rather than sampling error. An I2
less than 30% indicates mild heterogeneity, whereas an I2
greater than 50% suggests notable heterogeneity.55
Random-effects models were used in all analyses of
pooled mean difference in sleep parameters between
children with ASD and TD to account for variability in
methodology, setting, and populations. Random-effects
meta-regression analyses (using unrestricted maximum
likelihood estimation) and moderator analyses (using
mixed effects models) were performed to determine if
the potential moderating factors described above could
explain variability across studies. A post hoc power
analysis for the categorical moderators between groups
differences was performed using the methods contained
in Hedges and Pigott56 with the estimate of the non-
centrality parameter based on the mixed-model variance
of the group mean effect size estimates for each group
and a 5 .05. “Leave-one-out” analyses were conducted by
iteratively deleting each study and calculating the result-
ing pooled mean difference.57
Analysis of Publication Bias
Publication bias was assessed visually using funnel
plots and also by using the commonly used regression
procedure to evaluate for asymmetry developed by Egger
et al.58 We adjusted for potential publication bias using
the trim-and-fill method by Duval and Tweedie.59 This
method evaluates the funnel plot for asymmetry with the
assumption that this asymmetry indicates unpublished
studies in the area with fewer studies. The theoretically
missing studies are imputed, and the pooled estimate is
recalculated including these studies. This adjusted result
can be used as a sensitivity analysis to assess the impact of
the assumed publication bias on the pooled estimate.60
RESULTS
Description of the Studies Included
Eleven studies were selected for inclusion in the sys-
tematic review, whereas 10 were selected for formal
170 Alterations in Sleep Parameters in Children with Autism
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
meta-analysis (Table 1). In addition to evaluating the re-
quired sleep parameters in the population of children
with ASD, 3 studies included an additional study group of
children with developmental delays or Fragile X.13,48,51
Three studies also evaluated differences between chil-
dren with autism and typical peers in nonrapid eye
movement sleep microstructure in the form of cyclic
alternating pattern.2,49,52 Other studies attempted to
correlate parental complaints with objective sleep
data.2,31–33
Giannotti et al52 stated that all parameters showed
positive skew and log transformed the data for analysis,
although the authors reported raw means and SDs. This
study was not ultimately included in the formal meta-
analysis because of significant influence on the parame-
ter mean differences found in the “leave-one-out” analysis.
This resulted in a total of 343 children with ASD and 221
TD children included in the meta-analysis.
Recruitment
Most studies used convenience sampling to generate
the index group, recruiting from subspecialty clinics.
Souders et al50 made an effort to randomly select indi-
viduals from their Regional Autism Center to provide
a sample with which to estimate prevalence of sleep
disorders. Reporting of the method of recruitment of
control subjects was relatively poor overall with only 2
studies providing information detailing a randomized
selection process while matching for age.31,50
Age Considerations
The age of participants across studies ranged from 2
years to 19 years of age. Most studies were performed
using school-aged subjects with control children
matched for age. However, both Goodlin-Jones et al and
Buckley et al studied younger children and purposely
chose controls that were 6 to 12 months younger than
the index group to attempt to match for developmental
age rather than chronological age.13,51 Goodlin-Jones
et al also included nap times for their young participants.
Establishment of ASD Diagnosis
Nine of the 11 studies performed diagnostic reas-
sessment of the index group to ensure correct categori-
zation of children with ASD versus TD. Evaluations
included an Autism Diagnostic Observation Scale in 7 of
the studies,13,32,33,49–52 the Autism Diagnostic Interview-
Revised in 4 studies,13,49,51,52 the Child Autism Rating
Scale in 2 studies,2,48 and the Asperger Syndrome Di-
agnostic Scale in 1 study.50 The remaining 2 studies re-
lied on clinical diagnoses made before recruitment of the
participants.24,31
Consideration of Intellectual Disability
Some authors attempted to control for this con-
founder, whereas others specifically recruited subjects
with intellectual disability (ID) or global developmental
delay for analysis. Only Buckley et al51 provided an av-
erage nonverbal intelligence quotient (IQ) of 56.92 (SD
17.96) for their study participants with ASD. Goodlin-
Jones et al13 provided an average score of 60.3 (SD
Journal of Developmental & Behavioral Pediatrics
 18.2) for the group of children with ASD on the Mullen
Scales of Early Learning (a cognitive ability measure).
Medication Use
Most studies excluded children taking psychotropic
or sleep-altering medications or required subjects to be
off of medications for an extended period of time (at
least 2 weeks) before collection of sleep data. However,
3 of the studies did include children on medications at
the time of data collection. Hering et al31 reported 1
child in the autistic group on carbamazepine. Goodlin-
Jones et al13 did not exclude children taking medications
but did not report a list of medications taken. Souders
et al did not require children to be medication free and
included children on medications to treat allergies,
asthma, and gastrointestinal complaints, as well as stim-
ulant medications, SSRIs, atypical neuroleptics, anti-
convulsants, and medications to aid sleep such as
melatonin and clonidine. The use of these medications
was significantly higher in the ASD group as compared
with the TD cohort.50 In the study by Buckley et al,
subjects were also allowed to continue any medications
including the night of the PSG recording. Only 1 child in
the TD group took an antihistamine, whereas 11 of the
children in the ASD group took SSRIs, atypical neuro-
leptics, anticonvulsants, and medications to aid sleep.51
Measurement Procedures
Actigraphy was performed using various commer-
cially available actigraphs with different softwares. Two
studies specifically used an actigraph with software that
used the Sadeh algorithm, which is the algorithm with
normative data on the pediatric population.31,50 PSG was
performed in sleep laboratories except for 1 study,
where it occurred on an inpatient pediatric ward.51
The “first-night effect” in PSG, where a patient’s first 2
consecutive nights of recordings in the sleep laboratory
fundamentally differ because of accommodation to sur-
roundings and testing, has been well documented in
children and adults and more recently in children with
autism.61 For this reason, all but 2 of the studies using
PSG excluded the first night from analysis.32,51
Some authors experienced difficulties in sleep moni-
toring in the population of children with ASD due to
sensory issues and intolerance to change and new
settings. Hering et al31 were unable to collect data on
one-third of their original sample because of subjects
removing the actigraph from their wrists due to sensory
intolerance. Souders et al50 piloted a new method of
collecting actigraphy data by placing the actigraph in
a soft cotton pocket hidden in the sleeve of a fitted
pajama top.
PSG monitoring differed between the studies but at
a minimum included electroencephalogram (ranging
from 3 to 21 leads), electrooculogram, and electromyo-
gram. Only 2 studies included pulmonary monitoring to
detect sleep disordered breathing.32,33 Other studies
excluded children with reported snoring.49,52 Scorers of
the PSGs were reported as blinded to subject’s study
group in 4 of the 7 studies using PSG.32,33,51,52 Scorer
Vol. 36, No. 3, April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
identity was reported as sleep technologist in 1 study,32
sleep technologist with sleep physician review in 1
study,33 physician in 2 studies,2,49 and qualified sleep
physician in 2 studies.51,52
Evaluation for Baseline EEG Abnormalities
All but 3 of the studies excluded children with seiz-
ures.31,50,51 Most studies excluded children on the basis
of medical history, but the 4 studies performed in Italy
required both normal neuroimaging and a normal EEG
for inclusion.2,48,49,52
Leave-One-Out Analysis
Sensitivity analyses indicated that only the Giannotti
study unduly influenced the pooled mean differences of
the sleep parameters studied between the group of
children with autism and the group of typical controls
(Table, Supplemental Digital Content 1, http://links.lww.
com/JDBP/A73). On visual inspection of histograms and
funnel plots, it was a clear outlier. Therefore, it was not
included in the formal meta-analysis.
Magnitude of Mean Differences and Main Effect Sizes
Forest plots for each outcome measure along with
summary measures based on random effects models are
presented in Figure 2A–C. Study design and quality
seemed to improve with time. Therefore, the studies
were ordered by year of publication to observe for
a possible trend.
Table 2 provides summary information for each out-
come’s pooled analysis to include the number of studies
included in the analysis, the pooled weighted differences
in means and Hedge’s g with 95% confidence intervals,
and Cochran’s Q and I2 values for evaluation for hetero-
geneity. In general, all pooled measures were statistically
significant with effect sizes ranging from the small to
medium range. Statistically significant heterogeneity of
results was found in TST but not for SL or SE. I2 values
indicated mild heterogeneity among the studies for sleep
efficiency and notable heterogeneity among the studies
for TST.
Impact of Moderator Variables for Total Sleep Time
The relationship between autism and TST was mod-
erated by ID. Studies that included children with ID in
the index group showed a pooled difference in means in
TST of 245.0 minutes (95% CI: 262.6 to 228.6) from
the typical controls. However, the remaining 3 studies
that excluded children with ID showed a much smaller
and nonsignificant difference in TST between the group
of children with autism and the control group of 27.3
minutes (95% CI: 221.5 to 7.0) (Table 3).
Method of data collection and inclusion of children on
psychotropic medication or with seizure disorders were
not found to be moderators. The continuous variables of
mean subject age (slope [95% CI]: 20.26 [26.27 to
5.758]; p 5 .933) and publication year (slope [95% CI]:
1.53 [23.61 to 6.67]; p 5 .560) were also not found to be
moderators, although confidence intervals were wide.
Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved. 171
 Figure 2. Forest plots for mean differences and combined mean differences between groups of children with autism spectrum disorders (AUT) and
typically developing peers based on random effects models for (A) total sleep time (TST), (B) sleep latency (SL), and (C) sleep efficiency (SE). Studies are
arranged in chronological order according to date of publication. Study-specific mean differences are denoted by black boxes with 95% confidence
intervals (95% CI) shown as black lines. Box sizes are proportional to individual study weights. The combined mean differences are represented by black
diamonds, where diamond widths correspond to 95% CI bounds.

Impact of Moderator Variables for Sleep Latency found to be a moderator (Table 3). The meta-regression
Inclusion of children with ID, on psychotropic medi- of mean subject age on difference in mean SL showed an
cations, or with seizure disorders was not found to be increase in difference in mean SL as age increased (slope
moderators. Method of data collection was also not [95% CI]: 1.51 [0.13 to 2.90]; p 5 .032). Publication year

172 Alterations in Sleep Parameters in Children with Autism Journal of Developmental & Behavioral Pediatrics
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
 Table 2. Pooled Differences in Means and Effect Sizes with Homogeneity Statistics for Sleep Variables
Difference in Means
With 95% CI
Total sleep time (minutes, k 5 10) 232.8 (249.0 to 216.6)
Sleep latency (minutes, k 5 9) 10.9 (6.7 to 15.0)
Sleep efficiency (%, k 5 10) 21.9 (23.1 to 20.7)
Pooled difference in means and Hedges g were calculated using random effects models. Heterogeneity statistics for differences in means were calculated using fixed
effects models. CI, confidence interval; df, degrees of freedom; k, number of studies included.
(slope [95% CI]: 21.72 [23.61 to 0.16]; p 5 .07) was not
found to be related to difference in means but
approached statistical significance.
Impact of Moderator Variables for Sleep Efficiency
Method of data collection was found to be a modera-
tor. Pooling of the studies that used actigraphy did not
result in a statistically significant difference in means in
SE between children with ASD and TD peers, whereas
those who used PSG yielded a pooled difference in
means of 23.5% (95% CI: 25.3 to 21.6; p , .001). In-
clusion of children on psychotropic medications also
was a moderating factor for difference in mean SE.
Studies that included children taking medications did not
show a significant pooled difference in mean SE,
whereas those that excluded children on medication did
(22.9%, 95% CI: 24.3 to 21.4; p ,.001). Inclusion of
children with ID and with seizure disorders was not
found to be moderators (Table 3).
The meta-regression of mean subject age on differ-
ence in mean SE showed an increase in magnitude of
difference in mean SE as age increased (slope with 95%
CI: 20.32 [20.62 to 20.02]; p 5 .034). Publication year
(slope with 95% CI: 0.11 [20.28 to 0.51]; p 5 .566) was
not found to be related to difference in means.
Post Hoc Power Analysis of the Comparison of
Subgroups Within Categorical Moderators
Power, in this case, can be thought of as the sensi-
tivity to detect differences between subgroups if they
exist. This has no bearing on the tests where a signifi-
cant result has been determined such as the difference
between the subgroups of children with ASD and ID and
those with ASD without ID. However, it provides con-
text for the comparisons that were not found to be
significant. To interpret the nonsignificant tests in the
moderator analysis as evidence that effects do not differ
across subgroups, it should be clear that power is suf-
ficient to detect meaningful differences. The power
analysis is presented in Table 4. Two choices of the
smallest meaningful difference are presented with evi-
dence of variable power. Power for any moderator
analysis is expected to be lower than that for the main
effect tested. However, it is noted that across all sleep
parameters, the test of SZD as a moderator was
underpowered.
Vol. 36, No. 3, April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hedges g Heterogeneity
With 95% CI p Qdf p I2
20.55 (20.82 to 20.28) ,.001 28.39 .001 68.2
0.41 (0.18 to 0.63) ,.001 8.98 .355 9.7
20.28 (20.45 to 20.10) .003 12.99 .165 30.5
Publication Bias
The funnel plots of each parameter (in which each
study’s effect size was plotted against the standard error)
were inspected for asymmetry. Marked asymmetry was
noted in the case of SE suggesting that studies with
negative findings may not have been published. This was
confirmed by the regression intercept approach. The
trim-and-fill adjusted pooled mean difference (95% CI)
with 3 imputed studies was found to no longer be sta-
tistically significant at 21.2 (22.6 to 0.2). This indicates
that results for sleep efficiency were sensitive to the
effects of possible publication bias (Figure, Supplemental
Digital Content 2, http://links.lww.com/JDBP/A74 and
Table, Supplemental Digital Content 3, http://links.lww.
com/JDBP/A73).
DISCUSSION
This systematic review found that, on average, chil-
dren with ASD had shorter TST, longer SL, and lower SE
than their TD peers. Standardized effect sizes were small
to medium. Although these differences are not extreme,
they are noticeable to parents, who have frequently
reported such differences to clinicians and in studies
including parental sleep questionnaires. This amount of
sleep disturbance may be enough to alter parents’ sleep
quality and add stress and also exacerbate the behavioral
symptoms of autism.
The meta-analysis was performed with the goal of
making the results as generalizable as possible to the
population of children with ASD by including all di-
agnoses within the spectrum and using random effects
models. Despite the limitations of the moderator analysis
because of small sample sizes and a small number of
available studies resulting in low power for some com-
parisons, the results are believed to be consistent with
the subjective reporting of the underlying changes in
sleep parameters as experienced by parents. In fact, the
improved power of the combined analysis allowed the
confirmation of what had been subjectively reported in
underpowered studies used in this analysis. For example,
the parents in Miano et al reported prolonged SL, shorter
TST, and more interrupted sleep patterns. Only TST was
found to be significantly reduced in the underpowered
objective arm of the study.2 In the small study (n 5 8
children with ASD) by Hering et al, where parental
reporting of nighttime arousals and early morning
Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved. 173
 Table 3. Categorical Moderator Analysis
Within Group
Heterogeneity Effect of Moderator
2
Moderator No. of Studies D (95% CI) p Qdf Within p I Qdf Between p
Total sleep time (min)
ID
Included (240/164) 7 245.0 (262.6 to 228.6) ,.001 12.46 .054 51.6 11.51 .001
Excluded (93/57) 3 27.3 (221.5 to 7.0) .316 1.02 .599 0.0
Medications
Yes (193/129) 4 235.7 (258.0 to 213.4) .002 6.93 .076 56.3 0.11 .810
No (140/92) 6 231.7 (256.1 to 27.3) .011 20.05 .001 75.0
Method
Actigraphy (165/146) 4 222.1 (238.3 to 25.8) .008 5.63 .131 46.8 1.61 .201
PSG (168/75) 6 242.9 (270.5 to 215.4) .002 20.05 .001 74.9
SZD
Included (125/60) 3 234.5 (275.6 to 6.6) .100 6.62 .038 69.5 0.011 .925
Excluded (208/161) 7 232.3 (251.5 to 213.1) .001 21.76 .001 72.4
Sleep latency (min)
ID
Included (232/156) 6 10.2 (3.4 to 17.1) .003 8.35 .139 16.5 0.291 .588
Excluded (93/57) 3 12.9 (6.0 to 19.9) ,.001 0.12 .974 0.0
Medications
Yes (187/121) 3 6.6 (22.1 to 15.2) .136 4.82 .090 58.4 2.221 .136
No (140/92) 6 14.6 (8.6 to 20.5) ,.001 1.65 .902 0.0
Method
Actigraphy (157/138) 3 10.3 (4.7 to 15.9) ,.001 3.52 .175 42.6 0.161 .691
PSG (168/75) 6 12.4 (3.8 to 20.9) .005 5.25 .390 4.1
SZD
Included (117/52) 2 6.1 (211.0 to 23.2) .482 3.01 .716 66.3 0.271 .603
Excluded (208/161) 7 10.8 (6.1 to 15.6) ,.001 5.96 .436 0.0
Sleep efficiency (%)
ID
Included (240/164) 7 22.2 (24.1 to 20.2) .028 12.16 .060 50.3 0.011 .939
Excluded (93/57) 3 22.1 (23.7 to 20.5) .012 0.22 .907 0.0
Medications
Yes (193/129) 4 20.6 (21.8 to 0.6) .314 3.03 .385 1.3 5.641 .018
No (140/92) 6 22.9 (24.3 to 21.4) ,.001 4.15 .533 0.0
Method
Actigraphy (165/146) 4 20.9 (22.0 to 0.2) .113 3.33 .350 8.7 5.491 .019
PSG (168/75) 6 23.5 (25.3 to 21.6) ,.001 3.95 .571 0.0
SZD
Included (125/60) 3 21.0 (23.9 to 1.9) .498 3.02 .229 32.3 0.581 .446
Excluded (208/161) 7 22.2 (23.7 to 20.8) .003 9.66 .142 37.6
A mixed effects method was used with random effects models used to combine studies within each subgroup, and fixed effects models used to combine groups and yield
the overall effect. Sample sizes for the group of children with autism/typically developing peers are provided for each subgroup. CI, confidence interval; D, pooled
difference in means; DF, degrees of freedom; ID, intellectual disability; min, minutes; PSG, polysomnography; SZD, seizure disorder.

awakenings were significantly different in parents of stated that “Parental oversensitivity to sleep disturbances
children with ASD as compared with TD controls but of the autistic children may explain this phenomenon.”31
was not confirmed by actigraphic monitoring, it was Although more data are needed to confirm the size of the

174 Alterations in Sleep Parameters in Children with Autism Journal of Developmental & Behavioral Pediatrics
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
 Table 4. Posthoc Power for Categorical Moderator Analysis heterogeneity between studies were SL and TST on
Power to Detect a Difference Between nonschool days.43 The difference in mean SL between
Subgroups children with ASD and TD peers seems to be greater in
the school aged children compared with the preschool-
Total sleep time 30 min 40 min
aged children. This effect seems to be stable in the
ID 0.756 0.943 school years. This could reflect the high prevalence of
Medications 0.428 0.659 bedtime resistance in the preschool-aged population
Method 0.451 0.688 estimated at nearly 30% tapering off to 13% by middle
SZD 0.254 0.408 school, which could obscure underlying differences in
SL in the younger ages.63 Age also was related to the
Sleep latency 10 min 15 min
difference in SE, with the magnitude of the mean dif-
ID 0.519 0.854
ference increasing as the subjects’ ages increased.
Medications 0.464 0.801 Method of data collection was found to moderate the
Method 0.462 0.819 differences in mean SE but not the differences in mean
SZD 0.198 0.382 TST. This is an interesting finding as both of these
Sleep efficiency 2% 3% parameters rely on the identification and measurement
of sleep epochs within episodes of sleep, which is
ID 0.337 0.636
a known weakness of actigraphy. This may be explained
Medications 0.565 0.889
by the lower power to detect meaningful differences in
Method 0.449 0.784 the moderator analyses for method of data collection and
SZD 0.228 0.444 TST. The point estimates of the pooled mean differences
Power calculated using the mixed-model variances of the group mean effect size for TST in the actigraphy versus the PSG subgroups
estimates for each group and a 5 .05. ID, intellectual disability; min, minutes; provide some evidence that actigraphy may be inferior at
SZD, seizure disorder.
detecting the underlying difference, although it did not
reach the level of statistical significance.
differences between the sleep parameters of children The variable approach to inclusion of children on
with ASD as compared with TD peers, this meta-analysis medications only moderated the difference in mean SE.
substantiates the existence of the differences that have Studies that included children on medications were less
been subjectively reported. likely to report significant mean differences in SE.
The amount of heterogeneity among the studies in the However, for TST and SL, medications seemed to have
measurement of TST was high and at least partially re- little effect. Indeed, the study by Souders et al50 did not
lated to the decision to include children with ID. Taylor show normalization of sleep for the children on medi-
et al7 have shown a correlation between sleep duration cations to treat sleep-related concerns. They described
and full scale, and verbal IQ scores in that sleep duration the use of sleep medications rather as a predictor of
decreases with IQ. Therefore, by selecting children for sleep disturbances as these children’s sleep parameters
the study group with normal IQ results, one may also be were more affected than the other children not taking
selecting those who have more normal sleep duration. medications. Parents of these children perceived the
Although the number of studies that did not include medications as helping and improving the quality of their
children with intellectual disabilities was small, the lives. This again is evidence that parents are sensitive to
pooled mean difference in TST was not statistically sig- even minor changes in their children’s quality of sleep
nificant (27.3 minutes; 95% CI: 221.5 to 7.0; p 5 .316). such as slightly improved SE. It also may speak to the
Note that for SL and SE, the presence of ID was not difficulty in pharmacologically targeting what may be
a moderating factor. These findings are also consistent intrinsic biological or genetic abnormalities that result in
with data from studies of adults with autism and normal alteration in neuronal pathways in ASD.64
intelligence where TST was not different from matched It has been hypothesized that hyperarousal related to
controls, but SL was prolonged and SE was decreased.62 increased anxiety and perseverative symptoms of autism
Thus, a child or an adult with ASD would take longer to is to blame for the difficulties in sleep initiation and
fall asleep and then also to achieve the same TST as their maintenance in people with autism. However, in adults,
TD peers because of decreased SE. The interaction be- Limoges et al62 showed changes in sleep parameters
tween IQ, autism, and decreased TST requires additional without subjective complaints and without an increase
evaluation that was not possible with the limited data in cortisol levels. This may point to a more fundamental
available. alteration in neurobiology as a cause of the differences in
Very little heterogeneity between studies was found SL and SE. The persistence of the sleep alterations into
in the measurement of difference in SL. There seemed to the adult years also supports this.
be little effect if it was measured by actigraphy or PSG. The results of this meta-analysis should be inter-
This is consistent with the meta-analysis that described preted in the context of its limitations. First, the final
normative sleep values across the life span, where the meta-analyses were conducted with 10 studies with
only measures in children that did not show significant relatively small sample sizes. Although it was

Vol. 36, No. 3, April 2015 Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved. 175

Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
 sufficiently well powered for the main effect estima-
tion, it was not well powered for the moderator analysis
as is common in meta-analyses. The small number of
studies also precludes the evaluation of more sophisti-
cated models and interactions between moderators.
Second, the small sizes of some of the studies make
the assumption of normality tenuous. Meta-analytic
methods rely on the principle of the central limit the-
orem where the distribution of the mean of the mean
differences approaches the normal distribution. This
convergence occurs irrespective of the distribution of
the underlying distributions from which the mean dif-
ferences originated. However, this is not as robust for
small sample sizes in individual studies. Many of the
studies with small sample sizes used nonparametric
statistics as not to assume a normal distribution (al-
though they all reported means and SDs). Regardless,
the results which were most heavily weighted in the
analysis were not unduly influenced by the smaller
studies with larger SDs. We used a small sample size
correction in our estimation procedure. We also re-
moved 1 study that reported skewness and would have
had a significant impact on the size of the pooled dif-
ferences in means reported, although not the in-
ference. It is difficult to determine if skewness affected
the remainder of the studies, as this was not reported.
The best way to evaluate this and to fully combine the
available data would be to collate individual participant
data to alleviate the reliance on reported summary
measures.
Third, there was a large amount of heterogeneity be-
tween studies in the parameter of TST likely related to
the variable inclusion of children with ID. It would be
interesting to evaluate the results based on average IQ of
participants in the group of children with autism. How-
ever, most studies did not provide this information. Au-
tism may only slightly decrease TST, whereas ID may be
the most important factor either in combination with
ASD or alone. More research is needed to understand the
interaction between ID and ASD and their effect on sleep
parameters. The pooled difference in mean TST between
the group of children with ASD and their TD peers
should be interpreted with this heterogeneity in mind.
Finally, possible publication bias was evident in the
reporting of SE. It is difficult to draw inference from this
isolated finding as there was little evidence of possible
publication bias for TST or SL, and these parameters
were derived from the same 10 studies. It cannot be
ignored that the imputed studies by the trim-and-fill
method did change the inference showing that our
measure of change in SE may not be robust to publica-
tion bias.
ACKNOWLEDGMENTS
The authors would like to thank Dr Sean Meagher, MD, Dr
Christine Erdie-Lalena MD, Dr Bonnie Geneman, MD, Mr Michael
Elrod, MA, and Ms Kathleen Knorr, LICSW for their thoughtful
reading of the article.
176 Alterations in Sleep Parameters in Children with Autism
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
REFERENCES
1. Sivertsen B, Posserud MB, Gillberg C, et al. Sleep problems in
children with autism spectrum problems: a longitudinal
population-based study. Autism. 2012;16:139–150.
2. Miano S, Bruni O, Elia M, et al. Sleep in children with autistic
spectrum disorder: a questionnaire and polysomnographic study.
Sleep Med. 2007;9:64–70.
3. Baker E, Richdale A, Short M, et al. An investigation of sleep
patterns in adolescents with high-functioning autism spectrum
disorder compared with typically developing adolescents. Dev
Neurorehabil. 2013;16:155–165.
4. Anders TF, Iosif AM, Schwichtenberg AJ, et al. Six-month sleep-
wake organization and stability in preschool-age children with
autism, developmental delay, and typical development. Behav
Sleep Med. 2011;9:92–106.
5. Ming X, Walters AS. Autism spectrum disorders, attention
deficit/hyperactivity disorder, and sleep disorders. Curr Opin Pulm
Med. 2009;15:578–584.
6. Schreck KA, Mulick JA, Smith AF. Sleep problems as possible
predictors of intensified symptoms of autism. Res Dev Disabil.
2004;25:57–66.
7. Taylor MA, Schreck KA, Mulick JA. Sleep disruption as a correlate
to cognitive and adaptive behavior problems in autism spectrum
disorders. Res Dev Disabil. 2012;33:1408–1417.
8. Hastings R, Kovshoff H, Ward N, et al. Systems analysis of stress and
positive perceptions in mothers and fathers of pre-school children
with autism. J Autism Dev Disord. 2005;35:635–644.
9. Tomanik S, Harris GE, Hawkins J. The relationship between
behaviours exhibited by children with autism and maternal stress.
J Intellect Dev Disabil. 2004;29:16–26.
10. Doo S, Wing YK. Sleep problems of children with pervasive
developmental disorders: correlation with parental stress. Dev Med
Child Neurol. 2006;48:650–655.
11. Meltzer LJ. Brief report: sleep in parents of children with autism
spectrum disorders. J Pediatr Psychol. 2008;33:380–386.
12. Cotton S, Richdale A. Brief report: parental descriptions of sleep
problems in children with autism, Down syndrome, and Prader–
Willi syndrome. Res Dev Disabil. 2006;27:151–161.
13. Goodlin-Jones BL, Tang K, Liu J, et al. Sleep patterns in preschool-
age children with autism, developmental delay, and typical
development. J Am Acad Child Adolesc Psychiatry. 2008;47:
930–938.
14. Honomichl RD, Goodlin-Jones BL, Burnham M, et al. Sleep patterns
of children with pervasive developmental disorders. J Autism Dev
Disord. 2002;32:553–561.
15. Richdale AL, Schreck KA. Sleep problems in autism spectrum
disorders: prevalence, nature, & possible biopsychosocial
aetiologies. Sleep Med Rev. 2009;13:403–411.
16. Polimeni MA, Richdale AL, Francis AJ. A survey of sleep problems
in autism, Asperger’s disorder and typically developing children.
J Intellect Disabil Res. 2005;49:260–268.
17. Miano S, Ferri R. Epidemiology and management of insomnia in
children with autistic spectrum disorders. Paediatr Drugs. 2010;
12:75–84.
18. Mayes SD, Calhoun SL. Variables related to sleep problems in
children with autism. Res Autism Spectr Disord. 2009;3:931–941.
19. Krakowiak P, Goodlin-Jones B, Hertz-Picciotto I, et al. Sleep
problems in children with autism spectrum disorders,
developmental delays, and typical development: a population-
based study. J Sleep Res. 2008;17:197–206.
20. Liu X, Hubbard JA, Fabes RA, et al. Sleep disturbances and
correlates of children with autism spectrum disorders. Child
Psychiatry Hum Dev. 2006;37:179–191.
21. Wiggs L, Stores G. Sleep patterns and sleep disorders in children
with autistic spectrum disorders: insights using parent report and
actigraphy. Dev Med Child Neurol. 2004;46:372–380.
22. Richdale AL, Prior MR. The sleep/wake rhythm in children with
autism. Eur Child Adolesc Psychiatry. 1995;4:175–186.
Journal of Developmental & Behavioral Pediatrics
 23. Gail Williams P, Sears LL, Allard A. Sleep problems in children with
autism. J Sleep Res. 2004;13:265–268.
24. Allik H, Larsson JO, Smedje H. Sleep patterns of school-age children
with asperger syndrome or high-functioning autism. J Autism Dev
Disord. 2006;36:585–595.
25. Patzold LM, Richdale AL, Tonge BJ. An investigation into sleep
characteristics of children with autism and Asperger’s Disorder.
J Paediatr Child Health. 1998;34:528–533.
26. Johnson CR, Turner KS, Foldes EL, et al. Comparison of sleep
questionnaires in the assessment of sleep disturbances in children
with autism spectrum disorders. Sleep Med. 2012;13:795–801.
27. Hoshino Y, Watanabe H, Yashima Y, et al. An investigation on sleep
disturbance of autistic children. Folia Psychiatr Neurol Jpn. 1984;
38:45–51.
28. Schreck KA, Mulick JA. Parental report of sleep problems in
children with autism. J Autism Dev Disord. 2000;30:127–135.
29. Allik H, Larsson JO, Smedje H. Insomnia in school-age children with
Asperger syndrome or high-functioning autism. BMC Psychiatry.
2006;6:18.
30. Øyane NMF, Bjørvatn B. Sleep disturbances in adolescents and
young adults with autism and Asperger syndrome. Autism. 2005;9:
83–94.
31. Hering E, Epstein R, Elroy S, et al. Sleep patterns in autistic
children. J Autism Dev Disord. 1999;29:143–147.
32. Goldman SE, Surdyka K, Cuevas R, et al. Defining the sleep
phenotype in children with autism. Dev Neuropsychol. 2009;34:
560–573.
33. Malow BA, Marzec ML, McGrew SG, et al. Characterizing sleep in
children with autism spectrum disorders: a multidimensional
approach. Sleep. 2006;29:1563.
34. Miano S, Bruni O, Elia M, et al. Sleep phenotypes of intellectual
disability: a polysomnographic evaluation in subjects with Down
syndrome and Fragile-X syndrome. Clin Neurophysiol. 2008;119:
1242–1247.
35. Matson JL, Shoemaker M. Intellectual disability and its relationship
to autism spectrum disorders. Res Dev Disabil. 2009;30:
1107–1114.
36. Baio J. Prevalence of Autism Spectrum Disorders in Children Aged
8 Years: Autism and Developmental Disabilities Monitoring
Network, 11 Sites, United States, 2010. MMWR Surveill Summa.
2014;63:6.
37. Sangal RB, Owens J, Allen AJ, et al. Effects of atomoxetine and
methylphenidate on sleep in children with ADHD. Sleep. 2006;29:
1573.
38. Jindal RD, Friedman ES, Berman SR, et al. Effects of sertraline on
sleep architecture in patients with depression. J Clin
Psychopharmacol. 2003;23:540–548.
39. McConville BJ, Sorter MT. Treatment challenges and safety
considerations for antipsychotic use in children and adolescents
with psychoses. J Clin Psychiatry. 2003;65:20–29.
40. Meltzer LJ, Montgomery-Downs HE, Insana SP, et al. Use of
actigraphy for assessment in pediatric sleep research. Sleep Med
Rev. 2012;16:463–475.
41. Viscidi EW, Triche EW, Pescosolido MF, et al. Clinical
characteristics of children with autism spectrum disorder and co-
occurring epilepsy. PLoS One. 2013;8:e67797.
42. Eriksson SH. Epilepsy and sleep. Curr Opin Neurol. 2011;24:
171–176.
43. Ohayon MM, Carskadon MA, Guilleminault C, et al. Meta-analysis of
quantitative sleep parameters from childhood to old age in healthy
individuals: developing normative sleep values across the human
lifespan. Sleep. 2004;27:1255–1274.
Vol. 36, No. 3, April 2015
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
44. Jensen CM, Steinhausen HC, Lauritsen MB. Time trends over 16
years in incidence-rates of autism spectrum disorders across the
lifespan based on nationwide Danish register data. J Autism Dev
Disord. 2014;44:1808–1818.
45. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for
reporting systematic reviews and meta-analyses of studies that
evaluate health care interventions: explanation and elaboration.
PLoS Med. 2009;6:e1000100.
46. Sanderson S, Tatt ID, Higgins JP. Tools for assessing quality and
susceptibility to bias in observational studies in epidemiology:
a systematic review and annotated bibliography. Int J Epidemiol.
2007;36:666–676.
47. Higgins JPT, Deeks JJ. Chapter 7: selecting studies and collecting
data. In: Higgins JPT, Green S, eds. Cochrane Handbook for
Systematic Reviews of Interventions Version 5.1.0 (Updated
2011). Oxford, UK: The Cochrane Collaboration; 2011. Available
at: http://handbook.cochrane.org/. Accessed May 13, 2014.
48. Elia M, Ferri R, Musumeci SA, et al. Sleep in subjects with autistic
disorder: a neurophysiological and psychological study. Brain Dev.
2000;22:88–92.
49. Bruni O, Ferri R, Vittori E, et al. Sleep architecture and NREM
alterations in children and adolescents with Asperger syndrome.
Sleep. 2007;30:1577–1585.
50. Souders MC, Mason TBA, Valladares O, et al. Sleep behaviors and
sleep quality in children with autism spectrum disorders. Sleep.
2009;32:1566–1578.
51. Buckley AW, Rodriguez AJ, Jennison K, et al. Rapid eye movement
sleep percentage in children with autism compared with children
with developmental delay and typical development. Arch Pediatr
Adolesc Med. 2010;164:1032–1037.
52. Giannotti F, Cortesi F, Cerquiglini A, et al. Sleep in children with
autism with and without autistic regression. J Sleep Res. 2011;20:
338–347.
53. Cummings P. Arguments for and against standardized mean
differences (effect sizes). Arch Adolesc Med. 2011;165:592–596.
54. Hedges LV, Olkin I. Statistical Methods for Meta–Analysis.
Orlando, FL: Academic Press; 1985.
55. Higgins JP, Thompson SG. Quanitfying heterogeneity in a meta-
analysis. Stat Med. 2002;21:1539–1558.
56. Hedges LV, Pigott TD. The power of statistical tests for moderators
in meta-analysis. Psychol Methods. 2004;9:426.
57. Greenhouse JB, Iyengar S. Sensitivity analysis and diagnostics. In:
Cooper HM, Hedges L, eds. The Handbook of Research Synthesis.
2nd ed. New York, NY: Russell Sage Foundation; 2009.
58. Egger M, Smith GD, Schneider M, et al. Bias in meta-analysis
detected by a simple, graphical test. BMJ. 1997;315:629–634.
59. Duval S, Tweedie R. Trim and fill: a simple funnel-plot–based
method of testing and adjusting for publication bias in meta-
analysis. Biometrics. 2000;56:455–463.
60. Peters JL, Sutton AJ, Jones DR, et al. Performance of the trim and fill
method in the presence of publication bias and between-study
heterogeneity. Stat Med. 2007;26:4544–4562.
61. Buckley A, Wingert K, Swedo S, et al. First night effect analysis in
a cohort of young children with autism spectrum disorder. J Clin
Sleep Med. 2013;9:67–70.
62. Limoges É, Mottron L, Bolduc C, et al. Atypical sleep architecture
and the autism phenotype. Brain. 2005;128:1049–1061.
63. Archbold KH, Pituch KJ, Panahi P, et al. Symptoms of sleep
disturbances among children at two general pediatric clinics.
J Pediatr. 2002;140:97–102.
64. Cortesi F, Giannotti F, Ivanenko A, et al. Sleep in children with
autistic spectrum disorder. Sleep Med. 2010;11:659–664.
Copyright Ó 2015 Wolters Kluwer Health, Inc. All rights reserved. 177