Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by a persistent impairment in reciprocal social communication and social interaction, and restricted, repetitive patterns of behavior, interests, or activities. Symptoms are present from early childhood and significantly affect daily functioning (1). ASD is four times more common in males than in females, and its worldwide reported prevalence approaches 1% of the population (2–4). Comorbidities are very common in ASD. Indeed, over 70% of individuals diagnosed with ASD have concurrent somatic or psychiatric conditions (5, 6). Sleep disorders are among the most common associated disorders in this population, with prevalence rates ranging from 50 to 80% (7, 8).

Studies based on parent-reported sleep problems have shown that the most common complaints are related to bedtime resistance, sleep initiation, nighttime awakening, and shortened sleep time (9). In accordance with these findings, studies using objective measurements of sleep quality parameters have revealed that children with ASD, compared to typically developing children (TD), display increased sleep onset latency, decreased sleep efficiency, as well as an increased number and duration of night wakings (10, 11). Research also provides support for the idea that sleep disturbances are strongly associated with daytime functioning in children with ASD. In a recent study, Mazurek and Sohl (12) showed that sleep disturbances significantly account for behavioral dysregulation, notably inattention, impulsivity, irritability, and physical aggression (12). They also showed that night wakings have the most consistently strong association with daytime behavior problems. Other authors found that sleep disturbances are significantly correlated with both internalizing and externalizing symptoms in children with ASD, using the Pediatric Behavior Scale (13). Studies have further suggested that ASD children who experience sleep problems have also cognitive impairment. For example, decreased sleep duration in children with ASD was correlated with nonverbal communication deficits (14), lower overall intelligence, adaptive functioning, and socialization skills (15, 16).

Finally, it is well characterized that sleep disturbances worsen quality of life of both children with ASD and their families (17, 18). Levin and Scher reported that sleep problems contributed significantly to maternal stress (19).

If not treated early, sleep disorders persist from infancy to adolescence (20). Thus, it is essential to address sleep problems in children with ASD, in order to favorably impact not only nocturnal symptoms, but also their daily functioning as well as overall quality of life of these children with ASD and those of their caregivers.

In order to efficiently design sleep interventions and overall medical care, it is essential to assess sleep quality parameters in children with ASD. Polysomnography (PSG) is the gold standard for sleep quality assessment but, aside from the cost involved and its limited availability, it can be challenging and often impossible to conduct PSG in this population (21). Indeed, polysomnographic recording may be compromised because many children with ASD present sensory abnormalities, and thus may not tolerate electrodes on their scalp or face (11). As an alternative, actigraphy (ACT) has been used as a non-invasive, objective, and cost-effective assessment tool for the study of rest-activity cycles as a proxy to sleep–wake rhythms. It has become a tool of choice to assess sleep quality in children with ASD. However, to the best of our knowledge, no study has yet investigated the validity of ACT compared to PSG for the measurement of sleep in children with ASD.

The aim of this study was to compare the agreement of actigraphy (MotionWare 8®—CamNtech MotionWare 1.1.20) with gold standard polysomnography in children diagnosed with ASD.

To the best of our knowledge, this is the first study to demonstrate the validity of actigraphy for the assessment of sleep in children with ASD, compared to polysomnography, the gold standard. The comparison of PSG and ACT included two sets of state-of-the-art comparisons and analyses: the agreement analysis of the four sleep parameters and an epoch-by-epoch agreement analysis.

Sleep parameter agreement analyses revealed, on the basis of ICC, a good agreement between PSG and ACT for the assessment of SL and TST, and a moderate agreement for WASO and SE. Similarly, we observed a clinically satisfactory equivalence for the measurement of SL and TST but not for WASO between ACT and PSG, within the pre-set range of acceptable deviation. SE was significantly equivalent in the two methods when using the less conservative range of acceptability of 10% (p < 0.01) but not when using the conservative range of 5% (p = 0.25). The Bland-Altman plots revealed that, for almost all participants, the differences between ACT and PSG fell between [-2SD; +2SD] limits of agreement, and there were no visible trends indicating that ACT performed differently with respect to SL, TST, WASO, or SE values.

Epoch-by-epoch agreement analyses showed high sensitivity, PVS and accuracy, substantial kappa, but moderate PVW and specificity. These results indicate that ACT has a high ability to identify TS, but lower ability to detect TW. According to Sadeh’s recommendation, one should look for a specificity higher than 0.60, in the epoch-by-epoch ACT/PSG comparisons (36). In our study, we found a moderate specificity of 0.51 with the low sensitivity-threshold setting. With the medium and high sensitivity-threshold settings (results in Supplementary, Table S3 ), we obtained better specificity (respectively, 0.617 ± 0.193 and 0.699 ± 0.18), but all other indicators were lower (specificity and PVW) or equivalent (accuracy, kappa, and PVS), and above all, sleep parameters (TST, WASO, and SE) were less accurate with those settings (results in Supplementary, Table S3 ).

To date, most studies which have investigated the validity of actigraphy, by comparing it to polysomnography, were conducted in healthy young adults (25, 32, 37). However, actigraphy is a particularly useful tool in pediatric research and thus widely employed, even though some devices have not been validated yet in this population. Actigraphy has been validated in the TD pediatric population, e.g., in infants and young children (38–40) and also in adolescents (41). Although results of previous studies may be misleading as different devices and scoring algorithms do not perform equally across age groups (24), the agreement indicators of the present study were within the range of those reported in the literature ([0.88–0.93] for the sensitivity, [0.46–0.77] for the specificity, and [0.84–0.9] for the accuracy) (24, 42). Thus, our results are consistent with what has been reported in the TD pediatric population: actigraphy correctly identifies sleep periods (as denoted by the high sensitivity) but is less accurate in identifying WASO (as denoted by the low specificity). Despite the growing interest in actigraphy research, validation studies in children are still lacking. This is of particular importance as children display different sleep behaviors than adults with children displaying more movements during sleep than adults. Validation studies should take into consideration these differences and examine sleep across different developmental age groups.

Within the pediatric population, there are children for whom sleep disorders are much more prevalent. Indeed, up to 80% of young children with ASD have sleep disorders (43–45) compared to about 25% of TD children (46). These disturbances among the ASD population are three to four times more common and have been related to daytime symptomatology (12–15), hence the importance of measuring sleep in this population.

The vast majority of sleep studies in the ASD population have focused on subjective assessment of sleep (i.e., parent-reported sleep questionnaires or sleep diaries) because of the difficulties performing more objective assessments (e.g., actigraphy, polysomnography). Indeed, in 2015, according to Elrod and Hood’s meta-analysis (7), there were only 10 studies that examined sleep objectively using PSG (11, 47, 48) and/or ACT (49–51). The PSG sleep parameters observed in our study were similar to those described by Buckley et al. (52), who included 60 ASD children aged 2.24–13.1 years. In the present study, compared to theirs, were reported, respectively: a SL of 35.2 min (±33.2) against 39.4 min (±33.2), a WASO of 77.2 min (±56) against 73.5 min (±72.2), and a SE of 80.3% (±10.1) against 80.2% (±12.7). In average, TST was 47 min higher in our sample compared to Buckley et al. (52), but the SD were very large on this measure for both studies (respectively, 91 and 126 min). For actigraphy, results are too device and setting-dependent to allow a direct comparison between studies. For example, in the current literature, reported WASO range from 18 to 88 min (50, 51) (for details see Methodology and Supplementary Data ). As many children with ASD present sensory abnormalities, and thus may not tolerate electrodes on their scalp or face, PSG recording may be compromised. Actigraphy has become more and more popular as an alternative, non-invasive, objective, and cost-effective assessment tool for the study of rest-activity cycles as a proxy to sleep–wake rhythms. It has become a tool of choice for the assessment of sleep quality in children with ASD but has not been validated yet in this population. Our study provides, for the first time, the mandatory evidence for the validity of actigraphy compared to PSG for these assessments in children with ASD, providing thus the basis for improved medical care of sleep disorders in these children. It is essential to validate sleep measures in other specific pediatric populations, as results found in the TD children cannot be extrapolated to them.

As reported by Meltzer et al. (53), it is common to find validation-type studies using inadequate methodology and relying only on correlation analyses, or alternatively focusing more on sensitivity than specificity, leading to deceptive results. To address this, we used multiple comparison and performed recommended agreement analysis methods (27, 28). We also compared both clinical sleep parameters and recorded epochs, to report sensitivity and specificity. Moreover, in the literature, results regarding actigraphy-derived sleep quality parameters in children with ASD are contradictory (10, 50). This may be due to variable wake sensitivity threshold that many studies do not report. Thus, we compared several sensitivity-threshold settings.

A particular strength of the present study is that we included ASD children across the spectrum, with and without associated intellectual disability ranging from low to high severity (with respect to DSM-5). This allows to generalize our results to the overall population of ASD children and not only to high functioning ASD children, as is often the case. In our study, we find the same male predominance in ASD population [i.e., six girls and 20 boys (54)]. Our study sample was young, with a mean age of 5.36 years. It is important to study sleep quality in children and detect sleep difficulties as early as possible, especially since untreated sleep difficulties maintain with time and become chronic.

Despite these strengths, there are also some limitations to this study. The main limitation is the relatively small sample size of 26 subjects and high interindividual variability, which may both reduce statistical power, especially regarding equivalence tests. However, the sample size was above the average sample size of previous studies examining the validity of actigraphy against polysomnography reported by Meltzer et al. (53), which was of 18 (range [8–45] participants). Also, the interindividual variability of the present study is similar to that reported in previous studies (11, 14, 55) and did not refrain from finding that actigraphy is a valid method to assess sleep quality parameters in children with ASD, compared to polysomnography. Furthermore, we have included children within the broad spectrum of ASD and did not only focus on a homogeneous group of high functioning children with ASD, thus reflecting a more representative population of children with ASD and sleep disturbances such as those seen in sleep clinics—which is a strength of this study. We also compared only a single night of actigraphy to concurrent PSG recording while it is recommended to collect five to seven nights of recording across devices (56). This can be explained by the fact that it is not feasible to carry-out seven consecutive days of polysomnography in children with ASD, because of sensory abnormalities or associated behavior disturbances. Lastly, it should be noted that the data sampling over the actigraph recording was set to 1-min epochs and retrospectively converted to 30-s epochs by the software to allow comparison with the PSG epoch settings, as done by previous studies (24, 25). Although this has little influence on the comparison of general sleep parameters, it can lead to a small artificial reduction of epoch-by-epoch agreement indicators, including specificity and sensitivity. For a better comparability between PSG and ACT epochs, future studies should insure automated synchrony in timing and epoch lengths.

In addition to these limitations, it is important for researchers and clinicians to understand that sleep/wake scoring depends on specific devices, algorithms, and sensitivity thresholds. The default mode of an actigraph may be valid in one specific population but not in another. For scoring purposes, a daily diary and/or event marker use, when possible, is necessary in order to accurately identify sleep periods.

To our knowledge, our study is the first to validate actigraphy as a method to assess sleep quality parameters compared to polysomnography, using a particularly sound state-of-the-art methodology, in a young population of children with ASD. The results suggest actigraphy to be a valid method to evaluate sleep within a particularly vulnerable population, with a high sensibility and a good reliability for most sleep parameters, including TST and sleep onset latency. With the increasing number of research studies using actigraphy, it is important to have multiple validation studies for each device and each developmental age, across healthy and clinical samples. This study confirms ACT as the objective alternative to assess sleep quality in children with ASD and can provide reliable information to clinicians when investigating sleep to improve quality of life.

The datasets generated for this study are available on request to the corresponding author.

This study was carried out in accordance with the recommendations of the regional French Institutional Review Board with written informed consent from parents of participants and assent was obtained from the child when possible. All parents of subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the regional French Institutional Review Board (CPP Est IV 11/04/2012).

EY-K collected the data, conducted the actigraphy and polysomnography data processing, and wrote the manuscript. ER planned and performed the analyses and contributed to manuscript revision. M-MG collected part of the data and assisted with manuscript review and revisions. NL collected part of the actigraphy and polysomnography data and assisted with manuscript review and revisions. PF collected part of the actigraphy and polysomnography data and assisted with manuscript review and revisions. PB assisted with the study design and the funding and contributed to manuscript review and revisions. CS designed the study, secured funding, collected the data, assisted with data interpretation and critique, as well as contributed to manuscript review and revisions. All authors are responsible for the reported research and have reviewed and approved the final manuscript as submitted.

This work was supported by a grant from the National Hospital Program for Clinical Research (PHRC National Autisme et Sommeil, HUS N°5060, 2012). ER received a grant from SFRMS and EY-K from Ensar Foundation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.