Reut Gruber, PhD, Department of Psychiatry, McGill University, Douglas Mental Health University Institute, 6875 LaSalle Blvd, Verdun (Quebec), Canada H4H 1R3. Tel.: (514) 761-6131 ext. 3476; fax: (514) 762-3858, e-mail: email@example.com
The present study assessed the association between habitual sleep patterns and one night of PSG measured sleep with daytime sleepiness in children with ADHD and typically developing children. Eighty-two children (26 ADHD, 56 typically developing children), between 7 and 11 years, had nighttime sleep recorded using actigraphy over five nights (habitual sleep patterns) and polysomnography during one night (immediate sleep patterns), both within their home environments. Daytime sleepiness was examined using the multiple sleep latency test within a controlled laboratory setting the following day. Using Spearman correlations, the relationships between mean sleep latencies on the multiple sleep latency test and scores on a modified Epworth Sleepiness Scale with polysomnographic measures of sleep quality and architecture and with actigraphic sleep quality measures were examined. Longer sleep latency, measured using polysomnography and actigraphy, was related to longer mean sleep latencies on the multiple sleep latency test in typically developing participants, whereas actigraphic measures of sleep restlessness (time awake and activity during the night), as well as time in slow-wave sleep, were positively related to mean sleep latency on the multiple sleep latency test in children with ADHD. These results show a differential relationship for children with ADHD and typically developing children between habitual and immediate sleep patterns with daytime sleepiness and suggest that problems initiating and maintaining sleep may be present both in nighttime and daytime sleep.
Recent evidence suggests children with ADHD experience hypoarousal and that hyperactive, impulsive symptoms may be compensatory behaviors for fatigue (Cortese et al., 2008). Waking electroencephalography studies support this by showing enhanced theta activity, theta/alpha and theta/beta ratios and decreased alpha and beta activity in children with ADHD compared to typically developing children (TDC) (Barry et al., 2003). Similar changes in activity occur following sleep deprivation, suggesting daytime sleepiness increases hypoarousal (Cajochen et al., 1995).
Despite the prevalence of sleepiness in children and the increased vulnerability of individuals with ADHD, little research has examined whether nocturnal factors are related to daytime sleepiness in these populations. To our knowledge, only two studies have examined the association between nocturnal sleep and objectively measured daytime sleepiness in children with ADHD (Golan et al., 2004; Lecendreux et al., 2000). In Golan et al. (2004), children with ADHD were more likely than controls to present primary sleep disturbances, such as sleep disordered breathing (SDB), likely contributing to increased daytime sleepiness on the multiple sleep latency test (MSLT). Controlling for SDB, Lecendreux et al. (2000) found that, following one night, no relationship existed between nocturnal sleep and mean sleep latency on the MSLT. Consistent with Golan et al. (2004), however, children with ADHD had greater daytime sleepiness, suggesting further need to investigate this phenomenon.
As children with ADHD have considerable night-to-night variability in sleep patterns (Gruber et al., 2000), one night of sleep may not be as representative of habitual sleep patterns in these children. Additionally, sleep in a laboratory is often not representative of normal sleep patterns, due to the ‘first-night-effect’ (Agnew et al., 1966), with children with ADHD being particularly sensitive to the effect (Palm et al., 1992). The present study proposes to overcome these obstacles by examining sleep using actigraphy over five nights and by examining one night of PSG, both within a home environment.
The objective of the present study was to determine the relationship for both habitual sleep patterns and one night of PSG sleep patterns with daytime sleepiness in children with ADHD and TDC. We hypothesize that (i) higher levels of habitual activity during sleep and time awake during the sleep period will be related to increased daytime sleepiness and (ii) that, due to increased hypoarousal, this relationship will be more robust in children with ADHD.
Eighty-seven children (27 ADHD, 60 TDC) between the ages of 7 and 11 years (mean = 8.8, SD = 1.3) participated in the study. The Diagnostic Interview Schedule for Children, version IV [DISC-IV; (Shaffer et al., 2000)], which produces DSM-IV based diagnoses, was administered to parents to determine ADHD diagnosis. Diagnoses were further confirmed through responses to the Conners’ Parent Rating Scale – Revised [CPRS-R; (Conners et al., 1998)]. When diagnoses between measures were in disagreement, partici-pants were excluded from the analyses. One (1.1%) child with ADHD and 4 (4.6%) TDC were excluded from analyses on this basis.
Participants presenting with psychiatric or medical diagnoses which may interfere with sleep [e.g. depression, anxiety, SDB or periodic limb movement disorder (PLMD)], were excluded from the study. We chose to exclude primary sleep disorders (SDB and PLMD) that are more commonly reported in children with ADHD so as to help us better understand what may be due to the aetiology of ADHD versus what might just be a consequence of these disorders. SDB was assessed at screening by the Sleep Disordered Breathing subscale of the Pediatric Sleep Questionnaire (Chervin et al., 2000) and both SDB and PLMD were more thoroughly screened using PSG measures during the study. Six (6.9%) children with ADHD and 10 (11.5%) TDC were excluded due to the presence of periodic limb movement disorder, such that a total of 20 children with ADHD and 46 TDC were included for analyses.
All participants were asked to refrain from taking medication and caffeine containing products (e.g. cola or chocolate) for at least 48 h prior to sleep and sleepiness evaluation. Participants were recruited from the greater Montreal region from local schools, psychologists, and clinics for helping children with ADHD. All study procedures were approved by the Research Ethics Board of the Douglas Hospital Research Center. Parents and children were informed about the study and written consent was obtained from parents prior to participation.
For both TDC and children with ADHD, the majority of participants were male (60.9 and 65.0%, respectively) and/or Caucasian (60.9 and 75.0%, respectively). Most children came from households where the mother had a college or university education (80.4% of TDC and 90.0% of children with ADHD). Of the 20 children with ADHD, 4 (20.0%) had combined subtype, 13 (65.0%) had inattentive subtype, and 3 (15.0%) were hyperactive/impulsive subtype. Four children with ADHD were taking medication for ADHD symptoms prior to the study. The Conners’ Parent Rating Scale – Revised (Conners et al., 1998) and the Child Behavioral Checklist (Achenbach et al., 1987) were used to examine clinical characteristics of the children. Full clinical and demographic characteristics are presented in Table 1.
Table 1. Demographic and clinical characteristics of children with ADHD and typically developing children
Participant’s sleep was evaluated for five consecutive nights using actigraphy (AW) and for one night with polysomno-graphy (PSG) within the child’s natural home environment. Children were asked to maintain their regular sleep-wake patterns throughout the study. Parents filled out a modified version of the Epworth Sleepiness Scale [ESS; (Melendres et al., 2004)] the day of the multiple sleep latency test (MSLT) to evaluate children’s level of sleepiness. The day following nocturnal PSG sleep evaluation, children underwent the MSLT at the laboratory as a means of objectively measuring sleepiness.
Polysomnography: Polysomnographic data was collected using an ambulatory digitized recording device (Vitaport-3 System; TEMEC Instruments, Kerkrade, Netherlands). EEG signals were low-passed filtered at 60 Hz and digitized at a sampling rate of 256 Hz. Electrodes were placed on bilateral ocular sites (EOGs), bipolar submental muscular sites (EMGs), F3, F4, C3, C4, P3, P4, O1 and O2 with linked ears reference, according to the 10–20 system.
Two respiratory belts were fitted on the chest and abdomen to detect hypopneas and central apneas and electrodes were placed on the anterior tibialis of the legs to detect periodic leg movements during sleep. Hypopneas were considered to be present with a reduction of ≥50% in a chest or abdominal belt signal, while a complete halt in respiratory signal was defined as central apnea. A hypopnea/central apnea index (AHI) >2 h−1 of sleep was considered means for exclusion and for referral for complete assessment of sleep breathing disorders. This threshold was chosen as previous studies on the relationship between sleep and ADHD have used AHI values from 2 (Golan et al., 2004) to 5 (Sangal et al., 2005). Since an AHI >2 is consistent with the clinical practice of several leading Canadian sleep centers and is at the low end of the thresholds set in previous works, we retained it as the cut-off for our study. The presence of 5 leg movements or more per hour of sleep was taken to indicate the possible presence of PLMD and lead to exclusion from study participation.
Sleep stages were scored visually using sleep software (Stellate Harmonie, Montreal, Canada) according to the standards of Rechtschaffen and Kales (1968), with 20 second epochs. The night polysomnographic measures utilized for analyses were sleep duration (total time asleep from Lights Off to Lights On), sleep efficiency (percentage of time asleep during the sleep period) and sleep onset latency (time to fall asleep from Lights Off), as well as total time in stage 2, slow wave sleep (SWS) and REM sleep.
Actigraphy: Nighttime sleep was also monitored using actigraphy, which uses a wristwatch-like device (AW-64 series; Mini-Mitter, Sunriver, OR, USA) to evaluate sleep through the measurement of ambulatory movement. Actiwatches were worn on the non-dominant wrist from shortly before bedtime until shortly after awakening. Actigraphy produces indices on sleep duration (total time asleep from sleep onset to wake time), sleep onset latency (time to fall asleep from bedtime), total time awake (time awake from bedtime to wake time), and activity (amount of movement during the total sleep period). Measures were averaged over five nights in order to examine children’s habitual sleep patterns. One-minute epochs were used to analyze actigraphic sleep data and Actiware Sleep 3.4 (Mini-Mitter) was used as the sleep scoring software. Actigraphy has been shown to be a reliable method of sleep evaluation and the Actiware Sleep algorithm for scoring sleep indices has been validated previously, with high correspondence with PSG measures (Kushida et al., 2001). Reported bedtime and wake time (provided by sleep logs) were used as the start and end times for the analyses. The total sum of activity counts was computed for each 1-min epoch and if they exceeded a threshold sensitivity value of the mean score during the active period/45, then the epoch was considered waking. Otherwise, the epoch was considered sleep.
Sleep log: Bedtime and wake time were determined by a parent-filled sleep log. Any exceptional circumstances which may have taken place during the night were also noted in this log.
Modified Epworth Sleepiness Scale (ESS): Sleepiness was evaluated subjectively using a modified ESS (Melendres et al., 2004), which consists of 8 items examining the propensity for a child to fall asleep in various situations, with higher scores indicating a higher level of sleepiness. The sleepiness scale was filled out by parents on the evening of the PSG.
Multiple Sleep Latency Test (MSLT): The day following night-time PSG sleep evaluation, sleepiness was objectively assessed in the laboratory using the MSLT. The MSLT uses polysomnographic measures (as described previously) to measure sleep latency time at different points during the day, starting 1.5–3 h after termination of sleep recording. Consistent with guidelines from the American Academy of Sleep Medicine (Littner et al., 2005), the shorter four nap opportunity, with the time points 10 am, 12 pm, 2 pm, and 4 pm, was used for the MSLT. During these times, children are asked to lie down on a bed in a quiet room without resisting sleep and are given a 20 min opportunity to try and fall asleep. Children were determined to have fallen asleep after 1 min of continuous Stage 1 sleep, or by the appearance of a deeper stage of sleep. Sleep latency was measured from Lights Off until sleep began. If children did not fall asleep during the 20 min time period, a value of 20 min sleep latency was assigned. Average sleep latency on the MSLT was taken as the variable of interest.
Clinical and demographic characteristics were compared between ADHD and TDC using analyses of variance (anovas) or Chi square analyses, depending on the nature of the data. Spearman rank correlations were used to analyze the sleep and sleepiness data and were conducted separately for children with ADHD and TDC. To compare the pattern of associations between nocturnal sleep variables and sleepiness measures for children with ADHD and TDC, canonical correlations were conducted separately for each group. Canonical correlation is a multivariate analysis that examines correlations of sets of variables – one set is considered the independent variable set and one is considered the dependent variable set. A significant canonical correlation means that a pattern of associations is present between the variable sets. As in factor analysis, several latent, or canonical, variables are produced and, thus, multiple canonical correlations result. The number of canonical variables and correlations can be no larger than the number of variables in the smallest variable set.
In the present analyses, variables were separated according to nocturnal sleep duration and continuity (duration/continuity) measures which were assessed over multiple nights (AW sleep duration, AW time awake and AW activity) and nocturnal sleep duration/continuity measures from the night preceding sleepiness measurement (PSG sleep efficiency, PSG sleep duration, S2, REM and SWS). These variable sets were created, as both the duration and continuity of sleep have been found to be important variables related to levels of sleepiness (Gillberg, 1995). Sleep latency was assessed in separate canonical analyses, as it can also be thought of as a variable representing sleepiness levels and we did not want to confound our findings by including it within the variable sets. Significant canonical correlations were compared between the children with ADHD and the TDC using Fisher’s z-transformation. All analyses were conducted using SPSS 15.0 for Windows (SPSS Inc., Chicago, IL, USA) and a cut-off of P <0.05 was considered statistically significant.
A one-way anova revealed that children with ADHD scored significantly higher than TDC on all clinical measures (Table 1). No other significant differences were found between groups.
Group differences in sleep and sleepiness
Mean sleep and sleepiness values are presented in Table 2.
Table 2. Means and standard deviations for sleep measures
aBedtime and Wake time refer to the night before the MSLT testing.
*P <0.05. NS, non-significant.
F1,53 = 0.53
21 : 26 (0 : 49)
21 : 29 (0 : 40)
F1,53 = 0.20
7 : 04 (0 : 40)
7 : 04 (0 : 39)
F1,53 = 0.00
Sleep duration (min)
F1,53 = 0.57
Sleep latency (min)
F1,53 = 2.70
Wake time (min)
F1,53 = 0.02
F1,53 = 1.10
Sleep duration (min)
F1,53 = 0.01
Sleep latency (min)
F1,53 = 0.51
F1,53 = 0.07
Total Stage 1 (min)
F1,53 = 0.64
Total Stage 2 (min)
F1,53 = 0.28
Total SWS (min)
F1,53 = 0.27
Total REM (min)
F1,53 = 0.11
MSLT 1 latency (min)
F1,53 = 0.81
MSLT 2 latency (min)
F1,53 = 0.62
MSLT 3 latency (min)
F1,53 = 0.17
MSLT 4 latency (min)
F1,53 = 2.43
MSLT average latency (min)
F1,53 = 1.46
The distribution of children for number of nap opportunities in which they fell asleep was not significantly different between groups. Of the children with ADHD, 12 (60.0%) fell asleep for at least one of the nap opportunities and 3 (15.0%) fell asleep for two or more of the nap opportunities. Children with ADHD fell asleep most often for the 2 pm (30.0%) and the 4 pm (30.0%) naps. In the typically developing group, 26 (56.5%) of the children fell asleep during at least one nap opportunity, while 17 (37.0%) fell asleep during two or more of the sleep opportunities. TDC fell asleep most often at the 12 pm (34.8%) and 2 pm (41.3%) naps, although 30.4% still fell asleep at the 4 pm nap. Mean sleep latencies over the total sample were towards the higher end of the time limit (mean = 17.7 min, SD = 3.1) and were not significantly different between groups (P >0.05).
No differences were found between groups on the modified ESS.
No differences in sleep between children with ADHD and TDC were apparent for either the actigraphic or PSG data.
Relationship between nocturnal sleep and daytime sleepiness
Children with ADHD: Actigraphic sleep measures were not correlated with ESS scores for children with ADHD (Table 3). Longer PSG nocturnal sleep onset latency and higher PSG sleep efficiency, however, were related to lower scores on the ESS (r = −0.56, P =0.01 and r = −0.41, P =0.08, respectively).
TDC: Higher habitual activity scores in TDC were related to lower sleepiness scores on the ESS (r = −0.30, P <0.05).
Children with ADHD: Longer time spent in SWS was related to longer mean sleep latencies on the MSLT for children with ADHD (r =0.46, P <0.05). Additionally, longer AW time awake and more AW activity were associated with longer mean latency on the MSLT (r =0.48, P <0.05 for both). A trend for an association between longer AW sleep latency and longer MSLT latency was also present (r =0.42, P =0.08). These results show that children with ADHD who had higher mean scores on these measures, also took longer to fall asleep when given the opportunity during the day.
TDC: In contrast to the results of children with ADHD (Fig. 1), only habitual and PSG sleep onset latency were related to the mean sleep latency of the MSLT for TDC (r =0.39, P <0.05 and r =0.43, P <0.05, respectively), showing that TDC who took longer to fall asleep at night also took longer to fall asleep during the MSLT.
Children with ADHD compared to TDC: Canonical correlation analyses revealed a strong correlation between the nocturnal sleep duration/continuity measures assessed over multiple nights (AW sleep duration, AW time awake, and AW activity) and sleepiness measures for the children with ADHD (r =0.74, P =0.03), but not for TDC (r =0.29, P =0.70; Table 4). These correlations were significantly different (P =0.02), confirming different patterns of associations between the sleep variables and sleepiness measures. For the PSG variables, measured the night prior to sleepiness assessment, sleep latency was significantly related to the sleepiness set for children with ADHD (r =0.57, P =0.04) and approached significance for TDC (r =0.38, P =0.052). These correlations were not statistically different (P =0.39).
Table 4. Canonical correlations and latent variable loadings between sleep quality and sleepiness measures
1st Canonical Variable
2nd Canonical Variable
1st Canonical Variable
2nd Canonical Variable
ADHD, attention deficit/hyperactivity disorder; TDC, typically developing children; AW, actigraphy; MSLT, mean multiple sleep latency test; ESS, modified Epworth Sleepiness Scale; PSG, polysomnography; S2, stage 2 sleep; REM, rapid eye movement sleep; SWS, total slow wave sleep.
Children with ADHD who fell asleep:Post-hoc analyses were conducted on the group of children with ADHD that fell asleep during the MSLT. This was done to ensure correlations were not purely driven by the proportion of children that did not fall asleep and to more closely examine the associations that were present. These analyses revealed that more habitual time awake during the night and higher levels of nighttime activity (r =0.65, P <0.05 for both; Fig. 2) were still related to mean sleep latency on the MSLT, although correlations with PSG measures changed. More specifically, time spent in SWS was no longer significant, although longer PSG nocturnal sleep onset latency was related to longer mean sleep latencies on the MSLT (r =0.60, P =0.051). These results suggest that those with more SWS during the night were less likely to fall asleep on the MSLT.
This study examined the relationship between habitual sleep patterns and daytime sleepiness in children with ADHD and TDC. Habitual sleep patterns correlated strongly with PSG daytime sleepiness in children with ADHD, while in TDC, both habitual and PSG sleep latency correlated with PSG daytime sleepiness. PSG sleep latency and efficiency were related to ESS scores in children with ADHD and habitual activity scores were related to ESS scores in TDC. Canonical correlation analyses confirmed the pattern of associations between habitual sleep duration/continuity and daytime sleepiness was present for children with ADHD, but not for TDC. They also confirmed our findings that PSG measures of sleep onset latency were related to daytime sleepiness in both children with ADHD and TDC.
Children with ADHD showing increased SWS the previous night took longer to fall asleep during the MSLT, suggesting better sleep quality was related to decreased sleepiness. However, more habitual nocturnal wake time and activity and longer habitual sleep latencies for children with ADHD, and longer nocturnal PSG sleep latencies for TDC and children with ADHD who fell asleep during the MSLT, were also related to longer MSLT sleep latencies. These findings suggest problems with initiating and/or maintaining sleep affect both nighttime and daytime sleep.
Although well-rested children with ADHD showed longer MSLT latencies, those that displayed restless sleep did not show shorter latencies. It is possible that children experiencing restless sleep are overtired, thereby further contributing to increased difficulties sleeping. In children, symptoms of sleepiness include increased activity (Fallone et al., 2002), likely compensating for increased fatigue. Thus, increased activity may be difficult to ‘shut off’ when given a sleep opportunity. Similar results are found in insomnia patients, where, despite an increased desire to sleep, patients have longer MSLT latencies (Bonnet and Arand, 2010). This finding has been explained in the context of increased physiological arousal (Bonnet and Arand, 2010), which is likely also increased in children who are compensating for fatigue through active behaviors. Although the mechanisms underlying these disorders are quite different, the explanation for longer sleep latencies may be similar. Our findings that increased nocturnal sleep latencies and habitually restless sleep were related to longer MSLT latencies are in support of this explanation, although further measures of daytime activity would be needed to examine this hypothesis.
Neither AW nor PSG sleep characteristics differed between children with ADHD and TDC, consistent with previous research showing no differences in sleep quality or architecture between children with ADHD and TDC (Cortese et al., 2006; Golan et al., 2004; Lecendreux et al., 2000). The higher SDB scores for children with ADHD, although not clinically significant, are consistent with findings from previous studies, where SDB was higher in children with ADHD (Cortese et al., 2006).
Despite the lack of differences between AW and PSG sleep characteristics, habitual nocturnal wake time and activity were only related to daytime sleepiness in children with ADHD, suggesting that these children are more affected by restless sleep than TDC. This is further supported by findings that children with ADHD are more plagued by restless sleep and by increased movements during sleep than TDC (Cortese et al., 2006). As such, future research may wish to examine if tired children with ADHD produce more problematic behaviors and experience greater difficulties shutting off these symptoms when sleep is required.
Our findings are in contrast to those of Lecendreux et al., (2000) and Golan et al. (2004) who found no relationship between objective measures of nocturnal sleep quality and MSLT latencies for children with ADHD or controls. Two major differences in our study may account for this discrepancy: (i) PSG and habitual sleep evaluation took place within the participants’ homes; and (ii) the majority of relationships found were between measures of habitual sleep patterns, which were not measured in the previous studies, and MSLT latencies. By measuring sleep within the participants’ homes, more natural sleep patterns were likely attained, thereby giving more accurate results. It is also possible that cumulative effects of restless sleep, which were measured in our study, may contribute more to daytime sleepiness than the effects of one night of sleep, which was evaluated in previous studies. This is consistent with previous findings of progressively decreased sleep latencies on the MSLT across multiple nights of sleep restriction (Banks and Dinges, 2007).
Our finding that more restless sleep was related to longer MSLT latencies suggests a need to develop interventions focused on providing strategies to regulate arousal, such as engaging in quiet activities prior to sleeping. Additionally, the differential pattern of associations for children with ADHD and TDC emphasizes a need to provide interventions tailored to the individual.
Limitations and Future Directions
Several study limitations should be noted. Although our sample sizes were large enough to detect large effects, a greater sample size would provide more power in detecting true moderate sized relationships. Second, as pediatric norms are not well established for the MSLT (Hoban and Chervin, 2001), it is possible that longer times for children to fall asleep may prove more sensitive. This was so in the Golan et al. (2004) study, where children with ADHD took on average 21.9 min to fall asleep, while controls took an average of 27.9 min. Third, restless legs syndrome was not measured and SDB scores were higher for children with ADHD, so these cannot be ruled out as contributing to our findings. Finally, another limitation is the measurement of nocturnal sleep at home and daytime sleepiness in the laboratory.
Since children with PLMD were excluded from analyses, future studies may wish to compare results between children with ADHD with and without PLMD. As well, since wake time and activity, and longer nocturnal sleep latencies were related to longer latencies on the MSLT, future research should examine factors such as sleep onset insomnia or bedtime behavioral problems and how they influence both nighttime and daytime sleep.
This is the first study, to our knowledge, to assess the relationship between both habitual and immediate effects of sleep patterns on subjective and objective measures of sleepiness within both ADHD and typically developing pediatric populations. Relationships existed for both habitual nocturnal sleep and sleep the preceding night with measures of sleepiness, although the relationship was different for children with ADHD and TDC. These findings suggest that problems initiating and maintaining sleep may affect both nighttime and daytime sleep, albeit through different mechanisms for children with ADHD and TDC.
This study was supported by grants to Dr. Gruber from the Canadian Institutes of Health Research (CIHR; grant number 153139) and the Fonds de la recherche en santé (FRSQ; grant number 10091). The authors of this paper report no conflicts of interest with regards to any of the research material.