The feasibility of using actigraphy to characterize sleep in Rett syndrome

Rett syndrome (RTT) is a rare neurodevelopmental disorder characterized by myriad debilitating health complications and numerous developmental and behavioral problems [1, 2]. RTT affects approximately one in 10,000–15,000 live female births (it is exceedingly rare in males) and is most often caused by mutations in the methyl-CpG-binding protein 2 gene (MECP2) [3]. Most females with RTT (i.e., “classic RTT”) follow a pattern of seemingly typical development from birth until 6–18 months, after which they experience a significant regressive period. The regression involves a loss of acquired skills including purposeful hand movement and language abilities [2]. There are numerous health issues including difficult to manage seizures and autonomic dysfunction marked by cardiac arrhythmias, breathing abnormalities, and peripheral skin temperature anomalies [1, 4].

Considering the numerous health and behavioral issues, it is not surprising that sleep problems are commonly reported in RTT. The presence of an abnormal sleep pattern is a supporting characteristic included in the diagnostic criteria for classic and atypical RTT [5], and 79–85% of caregivers reported their children experienced at least one sleep problem, such as frequent nighttime waking, spells of screaming or laughing at night, or daytime sleepiness [6‐8]. Clinical and preclinical studies have examined melatonin and circadian activity rhythm cycles in RTT. Results indicated abnormal melatonin cycles in a sample of two girls with RTT [9] and abnormal circadian activity rhythms in mice with RTT compared to wild type controls [10, 11].

Understanding the prevalence and presentation of sleep problems in RTT has important clinical implications. A qualitative investigation by McDougall, Kerr, and Espie [12] describes that caregivers of individuals with RTT report their child’s sleep dysfunction affects their sleep and energy levels, marriage, mood, and social life, as well as their child’s mood and energy [12]. In other populations, indices of sleep dysfunction, such as sleepiness, sleep quality, and sleep duration, are correlated with a child’s school performance and daytime functioning [13], and behavioral regulation and neurobehavioral function [14]. Documenting the prevalence and severity of sleep dysfunction is a critical step towards the development of future clinical trials targeting control of circadian rhythms and sleep/wake cycles among individuals with RTT. Further, abnormal sleep patterns may reflect underlying physiological dysfunction and may therefore serve as a non-invasive marker that could be used to evaluate the effects of pharmacological and/or environmental manipulations designed to improve physiological outcomes in this population. To date, two clinical trials of therapeutics targeting core symptoms of RTT included parent-reported sleep questionnaires as outcome measures, including the Children’s Sleep Habit Questionnaire (CSHQ). The CSHQ is a questionnaire designed to screen children to identify those who need further evaluation for a pediatric sleep disorder [15] and has been used to characterize sleep in other developmental disability populations, such as autism spectrum disorders (ASD) [16], Angelman syndrome [17], and Williams syndrome [18]. While it is being used as an outcome measure in a clinical trial for individuals with RTT, there is currently no published literature on its use in this population, and thus, further investigation is needed.

Investigations into sleep dysfunction in RTT clinical samples have yielded inconsistent results. Some studies have documented increased total sleep time compared to typically developing norms when grouped by age [7], while others have found less total sleep time (TST; nighttime + daytime sleep) [19] or less total nighttime sleep (TNS) with increased daytime sleep compared to comparison samples [20, 21]. Ellaway et al. [7] found that younger girls with RTT slept less than developmental norms, but older teens with RTT slept more. Frequent night wakings, increased sleep onset latencies, and low sleep efficiency have also been reported [19, 21]. Preliminary evidence suggests that differences in total daytime sleep may be related to seizure status, as Ellaway and colleagues also found that girls with RTT who had a seizure disorder had more daytime sleep and more of their total sleep time was during the day than those without a seizure disorder [7]. Polysomnography (PSG) studies have had conflicting results, with some indicating relatively normal sleep architecture [22, 23], while others reported increases [23] or decreases [24] in duration of rapid eye movement (REM) sleep.

There are many potential explanations for the divergent findings across the studies conducted to date. It should be noted that many of the studies were completed prior to the seminal work by Amir and colleagues [3] implicating MECP2 in the development of RTT. Therefore, it is plausible that some of the participants in the earlier studies had distinct, but clinically similar syndromes. In addition, the studies vary considerably with respect to sample sizes, recruitment methods, and data collection methods. Data collection methods have included caregiver questionnaires [6, 8, 25], direct observations [20, 21], caregiver-completed sleep diaries [7], PSG [22‐24], and actigraphy [19]. Each of these methods is associated with its own set of advantages and disadvantages. Survey data are based on subjective caregiver reporting that may require parents to recall sleep behaviors, sometimes over long time scales. This method has the advantage of including the caregivers’ full, long-term knowledge of their children and their sleep patterns, but may lead to over-reporting of memorable events due to psychological biases, making it unclear whether overall results reflect the child’s current sleep patterns. Issues can also arise due to differences in interpretation (e.g., what is considered a sleep problem by a parent), which may differ across caregivers or between caregivers, researchers, and clinicians [12]. Direct observational methods allow in-home recording of sleep behaviors, as well as more sensitivity to analyze a participant’s typical sleeping pattern, but may not be sensitive to brief awakenings that occur, as behavior is not recorded continuously but intermittently across intervals throughout the night [20, 21]. Finally, PSG offers insight into the specific stages of sleep and its cortical physiology, but often requires participants to spend the night in a medical setting attached to equipment, removing the participant from their home environment and usual bedtime routine. The across-study differences and limited replication within collection methods may contribute to the apparent inconsistencies.

One approach that preserves the “home environment” for recording while allowing continuous recording is actigraphy, a method of data collection increasingly applied in sleep research because of its sensitivity and ability to record in-home for extended periods of time [26]. When worn day and night, a watch-like actigraph transduces movement which can be used to make inferences about sleep/wake periods. Actigraphy has been used to characterize sleep in typically developing populations as well as populations with developmental disabilities [26], including individuals with motor impairments associated with neurodevelopmental disorders such as cerebral palsy [27] and Angelman syndrome [28].

In the only published study of using actigraphy to measure sleep in RTT, McArthur and Budden [19] characterized the sleep patterns of nine girls before and during a melatonin treatment trial. Sleep was measured with actigraphy as well as with caregiver-completed sleep diaries for seven consecutive nights before treatment began and for 10 weeks during treatment. Although specific inclusion/exclusion criteria were not described specific to presence or absence of sleep disorders, all participants were enrolled in a sleep intervention trial. Given that parents are unlikely to consent to a treatment trial in the absence of a clinical problem, it is likely that enrolled participants had pre-existing sleep problems. For this reason, it is likely that the results may not reflect the average sleep patterns among a non-referred sample of individuals with RTT. Therefore, additional studies using actigraphy to examine sleep patterns among a broader sample of individuals with RTT are warranted.

The purpose of the present study was to demonstrate the feasibility of using actigraphy to characterize sleep in a sample of girls with clinically diagnosed RTT. To do so, we partially replicated the baseline phase of McArthur and Budden’s clinical sleep trial in which they used actigraphy as part of their outcome measures. In the current sample, girls were not participating in a sleep treatment trial. Further, we present data from three different measures of sleep to characterize sleep in the sample: actigraphy, parent report via sleep diary, and the Child’s Sleep Habits Questionnaire. It was hypothesized that actigraphy would be feasible and that our sample would have sleep characteristics that deviated from age-specific developmental norms. We were also able to explore several RTT relevant clinical variables, including parent-reported seizure status.

The purpose of this study was to extend the current literature on sleep patterns among individuals with RTT by demonstrating the feasibility of using in-home actigraphy, parent-reported sleep diaries, and a published sleep problem scale in a community sample of girls with RTT. Several summary-level points can be made. First, from a feasibility perspective, actigraphy was successfully used in-home by parents. There were no reported application problems or technical issues, despite the stereotypic hand movements associated with RTT. The night-by-night movement data in each participant’s report look comparable to data obtained from typically developing samples [37]. In comparison, there were more issues with the analysis of data from the sleep diaries and the CSHQ due to inconsistencies in completion and missing data. Second, from a normative perspective, the data show that most participants had appropriate onset latency, as derived by actigraphy. All but one participant, however, had low sleep efficiency (below 85%) [16], and several participants had TST outside of their recommended range according to both parent report and actigraphy measurement. Third, in relation to the indirectly measured RTT clinical characteristics, sleep efficiency was the only sleep variable significantly related to seizure status. Due to the number of participants taking sleep and seizure medications, it was not feasible to perform between-group analyses to examine differences in sleep characteristics by medication status. Visual analysis, however, did not show overlap of sleep characteristics between those prescribed and not prescribed sleep or seizure medications, and that pattern is consistent with McArthur and Budden [19], who did not find differences between those of and off anticonvulsant medication.

Aspects of the present findings that suggest poor sleep for some individuals in this sample (e.g., less TST than is recommended) align with the findings of literature reviews of sleep quality in individuals with and without intellectual disabilities that individuals with pain, genetic syndromes, and/or health conditions such as epilepsy, all of which have been documented in individuals with RTT [1, 38, 39] tend to have poorer sleep [40, 41]. The link between epilepsy and poor sleep may explain why individuals in this sample with uncontrolled seizures had lower sleep efficiency than individuals with controlled or no seizures, and differences in sleep variables by seizure status have been reported in other disability populations, such as in cerebral palsy and RTT [7, 42‐44]. Further, there is a significant intra-individual variability across some sleep variables in this sample, including some participants with standard deviations of nearly 2 h for TNS across the week-long collection. This variability is also consistent with other sleep studies in developmental disabilities. For example, Hoffman, Sweeney, Gilliam, Apodaca, Lopez-Wagner, and Castillo [45] found a relation between sleep values and phenotypic differences in individuals with ASD, and Annaz, Hill, Ashworth, Holley, and Karmiloff-Smith [18] found a relation between medial history (e.g., the presence of cardiac problems) and CSHQ scores in children with Williams syndrome. Individual ranges and/or standard deviations are often not reported in investigations of sleep, though, and thus, the comparison here is limited. Further research may illuminate the reasons for within-group variability in sleep parameters in developmental disability populations.

There were no age-related changes across any of the measured sleep behaviors. In typically developing samples, some sleep characteristics remain stable across the lifespan, whereas others show developmental changes [46]. Sleep efficiency and onset latency tend to be stable and independent of age [46], and the pattern of these variables in the current sample, as measured by actigraphy, is consistent with those of other populations in that respect. Typical samples, however, also show a decrease in TNS across the lifespan, and daytime sleep tends to end around 6 years of age [36]. The current sample deviated from typical patterns in these sleep characteristics regardless of measurement method, as there were no age-related changes in TNS measured by parents or actigraphy, and all participants napped at least once during the study period. The lack of age-related changes replicated patterns found in some previous studies of sleep in RTT [19], but is inconsistent with Piazza [21], who found that nighttime sleep decreased and daytime sleep increased with age. Due to the small sample size in the current study, as well as the cross-sectional nature of the study design, longitudinal studies of change in sleep patterns in RTT would elucidate whether the lack of age-related changes observed in the current sample is due simply to large inter-individual differences obscuring such patterns, or a true lack of developmental change.

This study is the first to report scores on the CSHQ specifically in a sample of girls with RTT. In the current sample, over half of the participants scored above the cut score of the CSHQ, even after removing the four non-applicable items, suggesting that the majority of this sample may require evaluation of a sleep disorder. Additionally, CSHQ total scores were positively related to actigraphy-derived daytime sleep, indicating that individuals who endorsed more items related to sleep dysfunction also tended to nap more. Further, individuals who were reported to nap more in general on the CSHQ (and were 6 years old or older, as napping occurs in less than 10% of children by age six [36]), had lower sleep efficiency. This may indicate that excessive napping after an appropriate age leads to lower sleep efficiency or that poor sleep efficiency leads to increased napping. Future research should further examine this relationship to determine the direction of the relation, as this may have implications for sleep interventions related to low sleep efficiency. As described above, however, there were several non-applicable CSHQ items, such as sleep-walking and talking (all girls were non-ambulatory and non-verbal), that were discarded. Other items that were applicable (e.g., sleeps alone in own bed) did not contribute any variance when examining internal consistency. Although the CSHQ did show strong internal consistency in the sample overall, further research and a larger sample size are needed to continue evaluating the appropriateness and validity of the CSHQ in RTT.

It should be recognized that the current study was based on a convenience sample, and it is worth pointing out the limitations created by such an approach. Generalizing to the RTT population is not warranted given no random sampling, and the results are best considered specific to the sample. The small sample size also had reduced statistical power and limited the complexity of the analyses that could be conducted. It is possible that some of the non-significant relationships between variables, including the lack of age-related changes, may be attributable to reduced power. Further, there was no normative or disability-matched comparison sample, which limits the specificity of the conclusions (i.e., it is difficult to make claims that the findings are RTT-specific and/or abnormal), although we were able to make comparisons against national normative age-based databases. Future studies should be designed to expand the current sample, as well as add a comparison sample of age-matched typically developing females.

Possible limitations of actigraphy should also be taken into consideration when considering the results of the current study in the decision to use actigraphy to characterize sleep in this population. Collecting and analyzing sleep patterns via actigraphy has not been validated in RTT, and there exists only one validation study of actigraphy in a sample of children with developmental disorders, according to a review by Meltzer and colleagues [29]. Validation studies of actigraphy in pediatric samples showed that accuracy was quite good regarding the detection of sleep intervals but less accurate regarding wake time after sleep onset [29]. Additionally, one polysomnographic study did document an increased prevalence of periodic limb movements during sleep in a sample of individuals with RTT compared to healthy controls [47]. Although we acknowledge that this might have had a small effect on WASO detected by actigraphy, these types of movements are brief and would be unlikely to cause large errors in the estimation of total nighttime sleep. Parents, though, may also be unreliable at reporting WASO, as they may not be aware of times their child is awake at night due to being asleep themselves, and thus, this limitation of actigraphy may still be present if studies and/or clinicians are relying on parent report [48].

In addition to limitations of actigraphy itself, there seems to be a paucity of reporting specific details in the current published literature of studies using actigraphy including the device settings, scoring algorithm settings, such as sensitivity and requirements for determining a sleep epoch, and sleep characteristics definitions, making direct comparison across studies challenging. The current literature ranges from 1 to 20 consecutive minutes of sleep to count as sleep onset, which influences variables such as sleep efficiency and WASO. Similarly, many time-based characteristics, including onset latency and WASO, are dependent on scoring rules, rules which are established presumably by the research team yet not always reported. For example, “medium” sensitivity is the default and most frequently used sensitivity setting in pediatric sleep studies using actigraphy [29], but “low” has been shown to be more aligned with PSG for this device [32]. Because of a general absence of reporting of scoring rules and the effect different sensitivity settings have on the calculation of sleep characteristics, caution should be taken when comparing actigraphy data across studies.

This study provides evidence on the feasibility of using actigraphy, an objective, in-home recording system, to characterize sleep in RTT. Overall, some participants had age-appropriate levels of TST and sleep onset within the recommended guidelines. On the other hand, the results indicate the presence of dysfunction for some sleep parameters in this RTT sample, specifically the continuance of daytime sleep across adolescence, low sleep efficiency, a lack of age-related changes in total night sleep, and clinically significant scores on the CSHQ. Future work should investigate the validity of using actigraphy to measure sleep in RTT, to establish an objective, in-home method to assess sleep in this population.