Movement Distribution: A New Measure of Sleep Fragmentation in Children with Upper Airway Obstruction

Abstract

Study Objectives:

To develop a measure of sleep fragmentation in children with upper airway obstruction based on survival curve analysis of sleep continuity.

Design:

Prospective repeated measures.

Setting:

Hospital sleep laboratory.

Participants:

92 children aged 3.0 to 12.9 years undergoing 2 overnight polysomnographic (PSG) sleep studies, 6 months apart. Subjects were divided into 3 groups based on their obstructive apnea and hypopnea index (OAHI) and other upper airway obstruction (UAO) symptoms: primary snorers (PS; n = 24, OAHI < 1), those with obstructive sleep apnea syndrome (OSAS; n = 20, OAHI ≥ 1) and non-snoring controls (C; n = 48, OAHI < 1).

Interventions:

Subjects in the PS and OSAS groups underwent tonsillectomy and adenoidectomy between PSG assessments.

Measurements and Results:

Post hoc measures of movement and contiguous sleep epochs were exported and analyzed using Kaplan-Meier estimates of survival to generate survival curves for the 3 groups. Statistically significant differences were found between these group curves for sleep continuity (P < 0.05) when using movement events as the sleep fragmenting event, but not if stage 1 NREM sleep or awakenings were used.

Conclusion:

Using conventional indices of sleep fragmentation in survival curve analysis of sleep continuity does not provide a useful measure of sleep fragmentation in children with upper airway obstruction. However, when sleep continuity is defined as the time between gross body movements, a potentially useful clinical measure is produced.

Citation:

Coussens S, Baumert M, Kohler M, Martin J, Kennedy D, Lushington K, Saint D, Pamula Y. Movement distribution: a new measure of sleep fragmentation in children with upper airway obstruction. SLEEP 2014;37(12):2025-2034.

INTRODUCTION

Body movements are a normal feature of sleep, show distinct ontogenetic changes, and are altered in various medical and sleep disorders.^1–4 Too many or even too few nocturnal body movements have been shown to reduce sleep quality, and in the case of excessive body movements can significantly fragment sleep.^5,6 Furthermore, adverse physiological or psychosocial events during infancy or childhood have been found to be associated with an abnormal pattern of body movements during sleep in adulthood.⁷ Given these findings, it is surprising that comparatively little research has focused on nocturnal body movements as a marker of sleep disturbance, despite the advent of more sophisticated polysomnographic monitoring techniques including digital video recording.

Restless sleep is a common presenting complaint in children with sleep disorders. Sleep-related body movements may be a more specific marker of sleep disturbance in children compared to adults due to developmental and age-specific differences in both sleep structure and arousal processes.^8–10 Studies have demonstrated that abnormalities in nocturnal body movements can occur with respect to their frequency, timing, and sleep stage distribution in association with various diseases impacting sleep.^11,12 These findings suggest that altered body movements during sleep may be an indirect marker of central nervous system (CNS) or arousal dysfunction.¹³

Sleep-related upper airway obstruction (UAO) is a common disorder during childhood with a spectrum of severity ranging from primary snoring (PS) to obstructive sleep apnea syndrome (OSAS). Primary snoring, characterized by frequent snoring in the absence of gas exchange abnormalities, is estimated to occur in 5% to 10% of children, while the prevalence of OSAS, which causes intermittent hypoxia and sleep fragmentation, is between 1% and 4%.¹⁴ It is now well documented that OSAS in children is associated with a range of neurocognitive and behavioral deficits including problems with attention, memory, learning, executive function, hyperactivity, and aggression.^15,16 During childhood, peak presentation of OSAS is observed between 2-5 years of age, which is an important period of cognitive development.¹⁷ Initial efforts to understand the pathophysiological mechanisms underlying these deficits focused largely on the role of hypoxia resulting from the apneas and hypopneas characteristic of OSAS. However, emerging evidence now suggests that children with PS or with disturbed sleep arising from medical conditions not associated with hypoxia, such as eczema and juvenile arthritis, also demonstrate reduced neuro-cognitive performance.^18,19 This has focused debate on the relative contribution of hypoxia versus sleep fragmentation to the observed neurocognitive sequelae of UAO.²⁰ However, while sleep fragmentation is a cardinal feature of many sleep disorders, it has proven difficult to accurately measure and quantify.

Restless sleep is frequently reported in children with UAO, but overnight polysomnography often demonstrates that the macrostructure of sleep is relatively preserved, with relatively fewer sleep stage changes and awakenings compared to adults with UAO.^21–25 Furthermore, in children comparatively fewer apneas and hypopneas are terminated by cortical arousal.^26,27 While frequent movement during sleep has been reported in children with OSAS,^22,28,29 conventional sleep scoring rules have tended to disregard these events. Recently, researchers have noted that many physiological processes can be described by exponential models,³⁰ including the length of continuous sleep bouts. Norman et al., using an exponential model of sleep continuity, utilized survival curve analysis (SCA) to measure sleep continuity in adults with OSAS.³¹ This novel approach to analysing sleep fragmentation demonstrated that individuals with OSAS had less stable sleep than controls, even though conventional sleep architecture measures did not differ between the groups. Furthermore, SCA was sensitive enough to show dose-dependent differences in sleep continuity between controls, those with mild UAO, and the moderate to severe OSAS group.³¹

This aim of this study was to use SCA to measure sleep continuity in children with UAO before and after adenotonsillectomy. However, while Norman used the occurrence of wake and stage 1 sleep as their marker of sleep fragmentation, we used the occurrence of body movements as the indication of a sleep fragmenting event and to define the end of a sleep period. Several studies have demonstrated increased NREM stage 1 sleep and decreased stage 2 NREM sleep bout duration in children with UAO.^32–34 However, only two small studies, with results reported in conference proceedings have directly investigated using stage 1 NREM sleep as a marker of sleep fragmentation in children with UAO, and both failed to distinguish between disease-based groups.^35,36 We hypothesized that Norman's findings using wake and stage 1 NREM sleep as the marker of sleep fragmentation would not be replicated in children with UAO, as children show relatively less wake and stage 1 NREM sleep than adults with UAO. Instead, we hypothesized that children with UAO would have an altered distribution of body movements during sleep compared to normal controls. Lastly, we hypothesized that following treatment of children with UAO by adenotonsillectomy, the pattern of body movements post-surgery would approximate that of the control children.

METHODS

Participants

As part of a broad-ranging study looking at neurocognitive deficits in children with UAO that used a prospective repeated measures design, overnight polysomnography (PSG) was performed in children with UAO awaiting adenotonsillectomy at baseline and 6 months following surgery.¹⁶ Non-snoring control children matched for age and gender underwent PSG at the same time points. Children with UAO were those with a history of frequent snoring who were scheduled for adenotonsillectomy for suspected obstructive sleep apnea as diagnosed by an experienced pediatric otorhinolaryngologist at the Adelaide Women's and Children's Hospital prior to being recruited in this study. It is routine clinical practice to diagnose OSAS in children based on clinical history and examination findings in these circumstances. The decision to operate was made prior to recruitment to the current study on factors other than objective sleep measures. The PSG results played no role in the decision to operate on the participants, with PSG results not available prior to scheduled adenotonsillectomy in most cases.

Children were aged between 3 and 12.9 years of age at baseline. Children were excluded from participating in the study if they had undergone previous ear, nose, or throat, or craniofacial surgery, had a medical condition (other than UAO) associated with hypoxia or sleep fragmentation, or were taking medication known to affect sleep or cardiorespiratory physiology. Control children were recruited through the recommendation of parents of the participating UAO children and from advertisements in local newspapers and schools. The same exclusion criteria were applied to controls with the additional requirement that they did not snore on more than two nights per week as determined from parental report. Parental consent and child assent was obtained from all participants. This study was approved by the Women's and Children's Hospital Human Ethics Committee.

Overnight Polysomnography

Overnight PSG was conducted without sedation or reported sleep deprivation and began close to each child's usual bedtime with a parent present throughout the procedure. Polysomnography was only performed if children reported as well on the night and free of any recent illness including respiratory infection. The following parameters were measured and recorded continuously using a commercially available computerized PSG system (Compumedics S-Series Sleep System, Melbourne, Australia): electroencephalogram (EEG; C3-A2 or C4-A1), left and right electroocculogram (EOG), submental and diaphragmatic electromyogram (EMG) with skin surface electrodes; leg movements by piezoelectric motion detection; heart rate by electrocardiogram (ECG); oro-nasal airflow by thermistor and nasal pressure; respiratory movements of the chest and abdominal wall using uncalibrated respiratory inductive plethysmography (RIP); arterial oxygen saturation (SpO₂) by pulse oximetry (Nellcor N-595; 2-3 second averaging time); and transcutaneous CO₂ (TcCO₂) using a heated (43°C) trans-cutaneous electrode (TINA, Radiometer Pacific). All data were digitized and stored for subsequent analysis. Each child was monitored continuously overnight via infrared camera by a pediatric sleep technician who also documented observations of sleep behavior including the presence or absence of snoring. Height and weight were measured on the night of PSG, and established growth charts corrected for age and gender were used to determine body mass index (BMI) z-scores.³⁷

Sleep stages were scored visually in 30-s epochs by a single trained technician (SC) according to the standardized EEG, EOG, and EMG criteria of Rechtschaffen and Kales.³⁸ Stage 3 and 4 NREM sleep were combined as slow wave sleep (SWS). Epochs were scored as movement if the EEG and EOG signals were obscured for ≥ 50% of the epoch by muscle tension or artifact associated with movement of the subject.³⁸ Movement time was scored as a separate category and was not included in either sleep or wake time. Wake after sleep onset (WASO) refers to time spent awake during the recording period after initial sleep onset and before the final awakening. Sleep efficiency (SE) was calculated as the percentage of time spent in sleep of the time available for sleep (i.e., from lights out to the final epoch of sleep).

Respiratory variables were scored according to standard guidelines recommended for pediatric sleep studies.³⁹ The obstructive apnea-hypopnea index (OAHI) was calculated as the total number of obstructive apneas, mixed apneas, and obstructive hypopneas divided by the total sleep time (TST) and expressed as the number of events per hour of sleep. An OAHI ≥ 1 was considered indicative of UAO. The central apnea hypopnea index (CAHI) was calculated as the total number of central apneas and central hypopneas divided by the TST and expressed as the number of events per hour of sleep. The total apnea and hypopnea index (AHI) was calculated as the total number of all respiratory events divided by the TST and expressed as the number of events per hour of sleep.

Cortical arousals were scored according to the criteria of the American Sleep Disorders Task Force.⁴⁰ The total arousal index (AI) represents all arousals combined (excluding arousals caused by external stimuli) expressed as the number of arousal per hour of sleep. The spontaneous arousal index (SAI) represents the total number of spontaneous arousals per hour of sleep. The respiratory arousal index (RAI) represents the total number of respiratory arousals per hour of sleep. Periodic limb movements (PLM) were scored using standard criteria.⁴¹ The PLM index (PLMI) was calculated as the number of PLMS per hour of sleep.

Movement Measures

Two different measures of body movement were analysed in this study. The first was movement time (MT) as defined by Rechtschaffen and Kales, which represents sleep epochs where the PSG signals are obscured or distorted by gross body movements.³⁸ An epoch was scored as MT when the EEG and EOG signals were obscured > 50% of the epoch by muscle tension or other artifact associated with movement of the subject.³⁸ Epochs were scored as MT only if the preceding and subsequent epochs were scored as sleep. Movement time is reported as the percentage of the TST.

Secondly, we utilized a new measure of movement during sleep termed movement events (ME). Movement events were defined as discrete body movements > 3 sec duration and were scored when there was evidence of movement on ≥ 2 independent PSG channels, but requiring at least one EMG channel (chin, diaphragmatic, or legs) to show evidence of movement. Evidence of movement included but was not limited to distortion or artifact in a PSG signal, an increase in EMG activity, generation of a cortical arousal, or a visually discernible increase in heart rate or respiratory rate above baseline levels. An ME was scored for an indefinite period until evidence of movement ceased and the subject remained asleep. Movement events separated by < 0.5 sec were combined into one event. Our definition of ME is modified from 2 existing measures of body movements: (1) movement arousals defined by Rechtschaffen and Kales³⁸ and (2) the movement/arousal definition of Mograss et al.²⁹ We modified these 2 previous definitions to improve inter-scorer reliability and to incorporate additional signals that are now routinely collected during PSG, which provide improved evidence of body movement. Movement event number and duration were recorded. Movement events included movements of relatively short duration and intensity to gross body movements accompanying postural shifts. Movement events are reported as the number of events per hour of sleep for each subject, rectified for that subject's TST (movement event index, MEI) and a mean value then generated for each of the 3 experimental groups.

Measures of Sleep Continuity

Sleep continuity was calculated by measuring the duration of uninterrupted periods of sleep (contiguous sleep epochs) throughout the night, each period being termed a “sleep run.” Two different definitions of sleep runs were used to assess sleep continuity. In the first method, a sleep run began with a change from wake to any stage of sleep and was terminated by the appearance of either an epoch of stage 1 NREM sleep or an epoch of wake as per Norman et al.³¹ In the second method, a sleep run began with either a change from wake to any stage of sleep or following an ME termination in sleep and was terminated by the appearance of either an ME or an epoch of wake. The occur-rence of stage 1 NREM sleep was not used to terminate a sleep run in the second definition as we wanted to utilize movement rather than sleep stage lightening as a measure of sleep fragmentation, particularly given that children do not commonly exhibit much stage 1 sleep compared to adults.

Survival Curve Analysis

For each individual the epoch based sleep staging and all scored movement measures were exported to MATLAB (Math-works, USA) for further analysis. Using custom written software sleep runs throughout the night were calculated (for an example of this process see Figure 1, diamonds), twice for each individual using the 2 definitions described above. Survival curves representing sleep continuity were then calculated for each individual using the Kaplan-Meier method.⁴²

Figure 1.

The distribution of sleep runs terminated by movement events (ME) or wake epochs for the baseline PSG (pre) for a single control subject. The diamonds represent individual sleep runs in seconds and the proportion of those runs from the single subject's population of total sleep runs for that night. The solid black line is the approximated survival curve for the subject using the Kaplan-Meier method.⁴²

Survival curves of sleep runs between sleep disrupting events (e.g., ME) were computed for each participant based on empirical cumulative distribution functions and exponential regression functions fitted (see Figure 1 for an example, solid black line). Assumptions of normality were not valid for survival curve exponent values and a square root transform [$\sqrt{x}$] was used to correct for skewness when performing some statistical tests. For a more detailed explanation of the methods involved in performing survival curve analysis see Norman et al.³¹ For each subject, sleep continuity is modelled on the equation y = m^{− θt} where m is a constant and t is sleep run length; thus each subject has a theta (θ, survival curve exponent) value that characterizes the distribution of sleep disrupting events.

Statistics

For analysis between groups a one way analysis of variance (ANOVA) test was utilized. For correlation calculations between the various sleep fragmentation measures, Pearson r method was employed. Kolmogorov-Smirnov statistic, with a Lilliefors significance level, was utilized in testing the normality of variables.⁴³ Assumptions of normality were valid for all PSG variables with the exception of PLMI, RAI, frequency of SpO₂ desaturations ≥ 3%/h TST, percentage of sleep time with SpO₂ < 95%, TcCO₂ > 50 mm Hg, OAHI, and AHI. Inverse transformation [1/(x + 1)] was required for these variables to correct skewness. Post hoc testing was conducted using Tukey Honestly Significant Difference (HSD) method for multiple comparisons. Within group changes were tested using a general linear models multiple measures test and paired samples t-test. All P values reported are 2-tailed, with statistical significance determined at α = 0.05. Correlations between normal variables was performed using the Pearson r method and for non-normal variables via the Spearman rho test. Data are presented as mean ± standard deviation unless stated otherwise.

RESULTS

Subjects

Forty four children with sleep-related UAO and 48 control children underwent both baseline and follow-up PSG. Children with UAO were divided into 2 groups based on their OAHI: PS (OAHI < 1, N = 24) and those with OSAS (OAHI ≥ 1, N = 20). Demographic characteristics for the 92 children are presented in Table 1. Children with OSAS had a significantly greater BMI compared to both PS and control so BMI was entered as a covariate where appropriate. There were no other significant differences in demographic variables between the 3 groups or within the groups between baseline and follow-up PSG.

Table 1.

Demographic characteristics for control (C), primary snoring (PS), and obstructive sleep apnea syndrome (OSAS) children at baseline and follow up polysomnography (PSG).

Polysomnography

Baseline PSG data are presented in Table 2. There were no significant differences between the 3 groups with respect to TST, the amount of time spent in all sleep stages, REM sleep latency, spontaneous arousals and PLM index. As expected, children with OSAS had higher OAHI, higher AHI, higher RAI, greater number of oxygen desaturations ≥ 3%, and lower nadir oxygen saturation compared to both the PS and control groups. Standard PSG measures of sleep disturbance (shaded variables in Table 2) including sleep efficiency, number of awakenings and frequency of sleep stage shifts were not significantly different between the three groups of children during baseline measurement.

Table 2.

Baseline polysomnography results for control (C), primary snoring (PS), and obstructive sleep apnea syndrome (OSAS) children.

The mean time between baseline and follow-up polysomnography was 29.4 ± 5.9 weeks. The mean time between adenotonsillectomy and follow-up polysomnography for children with UAO was 27.5 ± 6.0 weeks. In the follow-up PSG (Table 3), there were no significant differences between the 3 groups with respect to TST, the amount of time spent in all sleep stages, REM sleep latency, spontaneous arousals, or PLMI. Following adenotonsillectomy the OSAS children still had a higher mean OAHI and AHI and a greater number of oxygen desaturations ≥ 3% than both the PS and control groups and a lower nadir oxygen desaturation compared to the PS group suggesting the presence of residual OSA. As in the first PSG standard measures of sleep disturbance including sleep efficiency, number of awakenings, and frequency of sleep stage shifts were not significantly different between the 3 groups of children. However, at the second PSG, the OSAS children spent more time awake after sleep onset than controls.

Table 3.

Follow-up polysomnography results for control (C), primary snoring (PS), and obstructive sleep apnea syndrome (OSAS) children.

Movement Measures

Descriptive statistics for the 2 movement measures are summarized in Table 4. There was no effect of age, gender, BMI, or PLM index on any of the movement measures. Movement time (epochs obscured by gross body movements) was not significantly different between controls and children with UAO at the baseline or follow-up PSG.

Table 4.

Summary of movement measures for controls (C), primary snorers (PS), and children with clinically defined obstructive sleep apnea syndrome (OSAS) at baseline (1) and follow up (2) polysomnography (PSG).

During the first PSG a total of 7,152 ME from all subjects met the inclusion criteria for analysis. These data demonstrated that the total number of body movements per hour of sleep (MEI) was significantly higher in children with OSAS than controls for TST (TST), NREM sleep, and REM sleep (P < 0.01, Table 4). The MEI was also significantly higher in children with OSAS than PS during REM sleep (P < 0.01, Table 4). The MEI did not differ between PS and controls.

During the follow-up PSG, a total of 6,590 ME met inclusion criteria for analysis. Following adenotonsillectomy in the children with UAO, there were no longer any significant differences between the groups with respect to the frequency of body movements (Table 4). The mean duration of ME did not differ between groups during either the first or the follow-up PSG (Table 4).

Survival Curve Analysis

The survival curve exponent, θ, represents the cumulative distribution of the length of sleep runs for the night; the higher the value of θ, the more sharply the survival curve drops, indicating a higher frequency of shorter sleep runs. There was no effect of age, gender, BMI, or PLM index on θ values in any subject group.

Stage 1 NREM Sleep and Awakening as Markers of Sleep Fragmentation

When using a shift in sleep stage to stage 1 NREM sleep or an awakening as the marker of sleep fragmentation there was no significant difference (P > 0.05) between groups at baseline for θ, the survival curve exponent value (Figure 2). The same result was observed in the follow-up PSG (θ = 0.97 ± 0.2 for controls, 0.98 ± 0.1 for PS and 0.96 ± 0.1 for OSAS, P > 0.05).

Figure 2.

The exponential coefficient (theta, θ) of the distribution of sleep runs terminated by NREM stage 1 sleep (stage 1) or wake for the baseline PSG (pre) in the control (C), primary snoring (PS), and obstructive sleep apnea syndrome (OSAS) groups. Box ends represent the 25^th and 75^th percentile points, the center line represents the mean value for each group and the whiskers represent the range of the data. No significant differences between groups were found.

Movement Time as a Marker of Sleep Fragmentation

Utilizing movement time as a marker of sleep fragmentation showed no significant difference (P > 0.05) between groups in the survival curve exponent values (Figure 3). The same result was observed during the follow-up PSG (θ = 1.03 ± 0.3 for controls, 0.96 ± 0.1 for PS, and 0.97 ± 0.1 for OSAS, P > 0.05).

Figure 3.

The exponential coefficient (θ, theta) of the distribution of sleep runs terminated by movement epochs for the baseline PSG (pre) in the control (C), primary snoring (PS), and obstructive sleep apnea syndrome (OSAS) groups. Box ends represent the 25^th and 75^th percentile points, the center line represents the mean value for each group and the whiskers represent the range of the data. No significant differences between groups were found.

Movement Events as a Marker of Sleep Fragmentation

When using body movement as a sleep fragmentation event, significant differences between groups in the mean θ value were observed. At baseline, children with OSAS had a significantly higher exponent value (θ = 1.00 ± 0.12) than both normal controls (θ = 0.92 ± 0.06, P < 0.001) and the PS group (θ = 0.93 ± 0.09, P < 0.05, Figure 4). There was no significant difference in the distribution of sleep run duration between controls and children with PS (P > 0.05). Control children had the longest average sleep run duration (median length of 105 sec, range 4 to 5,483 sec), followed by the primary snorers (median of 69.4 sec, range 3 to 4,819 sec), while the OSA group had the shortest average sleep runs (median of 57.9 sec, range 3 to 5,912.4 sec).

Figure 4.

The exponential coefficient (θ, theta) of the distribution of sleep runs terminated by movement events (ME) for the baseline PSG (pre) in the control (C), primary snoring (PS) and obstructive sleep apnea syndrome (OSAS) groups. Box ends represent the 25^th and 75^th percentile points, the center line represents the mean value for each group, and the whiskers represent the range of the data. *** P < 0.0001, * P < 0.05.

Following adenotonsillectomy, children in the OSAS group continued to have a higher survival curve exponent value (θ = 0.99 ± 0.07) than the PS group (θ = 0.94 ± 0.04, P < 0.01) and normal controls (θ = 0.95 ± 0.04, P < 0.05), indicating that despite surgical treatment they still had an altered distribution of sleep duration, which included a higher number of shorter sleep runs. No difference was seen between the PS and control groups (P > 0.05). Once again the control group had the longest average sleep runs (median 96 sec, range 3 to 4,819 sec) compared to primary snorers (median 73 sec and range 4 to 4,910 sec) and the OSAS children (median 62 sec, range 3 to 5,120 sec). There were no significant differences in θ for each subject group between the first and second PSG (P > 0.05).

Sleep Fragmentation and AHI

The relationship between the degree of sleep fragmentation and disease severity (AHI) was evaluated for the OSAS group only as the AHI in the primary snorers and controls was less than one. In the OSAS children (N = 20), the survival curve coefficient was significantly correlated with the apnea-hypopnea index (r = 0.62, P < 0.01; Figure 5), indicating that as the AHI increased, sleep run duration decreased. Following adenotonsillectomy this association was no longer significant (r = 0.11, P > 0.05). In addition, the amount of change in the survival curve coefficient from baseline to follow-up PSG was positively correlated with the respective change in AHI (r = 0.58; P < 0.01; Figure 6). Thus the children who showed the greatest reduction in sleep-related obstruction following adenotonsillectomy also demonstrated bigger improvements in sleep consolidation.

Figure 5.

Correlation between the exponential coefficient (θ, theta) of movement events (ME) and apnea-hypopnea index (AHI) for the children in the OSAS group at baseline PSG. (AHI is presented on a logarithmic scale axis as the data was log transformed to correct for a skewed distribution in calculations). r² = 0.349, P < 0.01.

Figure 6.

Correlation between changes in movement event (ME) exponential coefficient (Θ, theta) between studies and changes in AHI between studies in the OSAS group (The AHI change is presented on a logarithmic scale axis as the data was log transformed to correct for a skewed distribution in calculations). r² = 0.376, P < 0.01.

DISCUSSION

This study has demonstrated significantly increased sleep fragmentation in children with OSAS compared to primary snorers and healthy controls when ME, as defined by this research protocol, are used as the marker of sleep disturbance. Furthermore, utilizing survival curve analysis to assess sleep continuity (and hence fragmentation), as indicated by the exponent of movement event distribution (θ), we found that θ was significantly correlated with the severity of UAO. In fact, 38% of the variance in θ was explained by the AHI. Additionally the extent to which sleep fragmentation “improved” (as indicated by decreases in θ) following adenotonsillectomy was correlated with the level of improvement in the AHI. However, the use of conventional measures of sleep fragmentation (movement time, stage 1 NREM sleep, or awakenings) in SCA failed to distinguish children with UAO from controls. These latter results are consistent with previous findings that children with OSAS show relatively preserved sleep when examining traditional measures of sleep fragmentation such as the number of cortical arousals or sleep stage shifts. Our findings demonstrate three important points: (1) traditional measures of sleep fragmentation do not adequately describe the degree of sleep disturbance in children with OSAS; (2) children with OSAS have an altered arousal response compared to normal, healthy controls; and (3) movement distribution rather than movement frequency may be a more precise measure of sleep fragmentation and its impact.

Frequent body movements are commonly observed in children with OSAS but these often normalize following adenotonsillectomy. In a video-recording study of sleep in children with OSAS, Stradling reported that 65% of the children spent more than 8% of sleep time moving compared to only 4% postoperatively.²⁸ In our study there were no differences in the percentage of sleep epochs obscured by movement or in the mean duration of ME between control children and those with PS or OSAS. However the OSAS children had a significantly higher number of movements compared to controls in NREM sleep and primary snorers in REM sleep. In addition, when using movement as a marker of sleep disruption, children with OSAS showed an altered distribution of sleep periods, demonstrating a higher frequency of short sleep runs. Following adenotonsillectomy for the children with UAO, all three subject groups showed a similar number of movements during sleep. However, in spite of this normalization in movement frequency, the OSAS children still demonstrated a higher θ value than the control and PS children, indicating an ongoing alteration in the distribution of their movements during sleep. This result is in keeping with the finding that this group of children still showed evidence of residual OSAS following surgery as demonstrated by a mildly elevated OAHI. It is noteworthy that these children also continued to show significant neurocognitive deficits six months post adenotonsillectomy.¹⁶

A recent actigraphy study by Suratt et al. examined the relationship between nocturnal body movements and both reaction time and cognitive performance in children with suspected OSAS.⁴⁴ They found that a higher frequency of body movements during the sleep period was significantly correlated with slower reaction times, while children whose body movements occurred in close clusters had reduced performance on vocabulary and memory tasks. Furthermore, it appeared that the distribution of movements during the night had a significantly greater impact on cognitive performance than movement frequency. In our study, the OSAS children had a similar number of movements during sleep post- adenotonsillectomy as the controls but a higher number of shorter sleep runs. This means that the OSAS children must have had a combination of both greater consolidation of movements (i.e., greater clustering of ME) and correspondingly longer, uninterrupted periods of sleep. This is a similar pattern of movement distribution as found by Suratt,⁴⁴ and thus both of these studies suggest that it is not just movement per se that may be an important marker of sleep fragmentation but also how movement is distributed throughout the sleep period.

In addition to being a sensitive marker of sleep fragmentation, movement may also contribute to the adverse sequela observed in children with OSAS. Compared to isolated movements, clustered body movements represent a more sustained level of activation of the CNS.⁴⁴ Increasing evidence now supports the view that sleep plays a pivotal role in both brain development and in the formation and consolidation of memories, thereby underpinning a significant component of neuro-cognitive performance. Periods of sustained arousal during sleep may impact adversely on CNS activities that occur only during sleep or may also alter other physiological regulatory processes. Support for this argument comes from a study by Loredo et al., who found that in adults with OSAS, movement arousals (defined similarly to our ME) but not cortical arousals correlated with awake sympathetic nervous system activity as measured by plasma norepinephrine.⁴⁵ Such up-regulation of the systems controlling sleep could make sleep inherently less stable in the face of perturbations, as found by Bianchi et al. where adults with OSAS demonstrated an increased instability in sleep architecture dynamics.⁴⁶ The increased periods of reduced motor activity seen in the OSAS children in our study in addition to the greater number of shorter sleep runs may reflect such sleep instability.

A recent study has proposed that chronic sleep disruption can generate long-term changes in the neuroendocrine stress response system.⁴⁷ It has also been proposed that similar changes would be seen in children with sleep chronically disrupted by the common symptoms of UAO such as respiratory arousals, transient hypoxic episodes, obstructive respiratory events, snoring, and end-apneic snorts.⁴⁸ The main components of this stress response system are the autonomic nervous system (ANS) and the hypothalamic-pituitary-adrenal axis (HPAA). Activation of this stress system leads to elevated plasma levels of stress hormones such as adrenaline and cortisol. Sleep would normally suppress the stress systems, and so sleep disruption abnormally maintains the activity of these systems at a significantly higher level, with one study showing that children with sleep disordered breathing (SDB) had an odds ratio of 3.48 for excessive autonomic activation compared to children without SDB.⁴⁸ A study in sleep deprived rats showed a change to underresponsiveness in the slower acting HPAA and an altered, up-regulated ANS response.^47,49 Sleep deprivation and sleep fragmentation are believed to result in similar physiological responses and thus a similar adaptive change to that seen in the chronically sleep deprived rats could potentially explain the result seen in the children with UAO. The UAO subjects in this study showed an increased arousal threshold as indicated by the longer periods of uninterrupted sleep, which may reflect under-responsiveness of the HPAA. In addition they also demonstrated clustered periods of short sleep runs which may reflect up-regulation of the ANS resulting in an increased propensity for sleep-wake transitions. This conclusion if correct provides a plausible mechanism by which sleep disruption arising from UAO in childhood could lead to increased cardiovascular morbidity in adulthood, as the link between increased stress system reactivity (e.g. increased cortisol) and cardiovascular disease is well established.⁵⁰

CONCLUSION

Several studies in adults with UAO have found novel associations between sleep parameters and disease severity when using survival curve analysis to model sleep-wake transitions.^52–54 Using the appearance of stage 1 NREM sleep and awakenings as markers of sleep disruption Norman, et al. showed large differences in sleep continuity between groups of adults with varying levels of OSAS.³¹ Using this same approach, we were not able to replicate these findings in a group of children with mild to moderate OSAS. However when we used the occurrence of movement as the marker of sleep fragmentation, we found significant differences in sleep continuity in the children with OSAS. Even though there is some overlap of this measure between groups, the movement measure exponent (θ) still does have potential clinical utility. As can be seen from Figure 5, all subjects in the OSAS group with a θ of > 1.0 were abnormal with AHI values > 10 (considered in the severe range for children), and furthermore, no subject in the control group had a θ > 1.0. The results of our study therefore suggest that movement, when appropriately analyzed, could be an important indicator of sleep fragmentation in children with UAO. Movement metrics are sometimes reported in clinical sleep study results but are often considered of lower importance than other PSG-derived measures, such as the apnea-hypopnea index. Our study, among others over many years suggests that the evaluation of movement measures should be considered in the routine clinical analysis of PSG results.^29,55

DISCLOSURE STATEMENT

This was not an industry supported study. This study was supported by grants from the Australian Research Council and the National Health & Medical Research Council Australia. Dr. Baumert was supported by the Australian Research Council (grant #DP0663345). Dr. Kohler was supported by the National Health & Medical Research Council Australia (grant #453669). The other authors have indicated no financial conflicts of interest. The work was performed at the Women's and Children's Hospital Adelaide, Australia.