Sleep Architecture Relates to Daytime Affect and Somatic Complaints in Clinically-Anxious but not Healthy Children

Abstract

Objective:

It is increasingly clear that seminal sleep-affective relationships begin to take root in childhood yet studies exploring how nighttime sleep characteristics relate to daytime affective symptoms, both in clinical and healthy populations of children, are lacking. The current study sought to explore these relationships by investigating whether trait-like and/or daily reports of affective and somatic symptoms of children with generalized anxiety disorder and matched controls relate to sleep architecture.

Method:

Sixty-six children (ages 7–11; 54.4% female; 56.1% Caucasian, 18.2% biracial, 6.1% African-American, 3% Asian; 16.7% Hispanic) participated including 29 with primary generalized anxiety disorder (without comorbid depression) and 37 healthy controls matched on age and race/ethnicity. Participants underwent structured diagnostic assessments including child-report measures and subsequently reported on their negative affect and somatic symptoms over the course of one week. Children also completed one night of polysomnography.

Results:

Among children with generalized anxiety disorder only, greater amounts of slow wave sleep corresponded with less negative affect, and greater amounts of rapid eye movement sleep was related to more somatic complaints across the week. Similarly, for trait-like measures, more rapid eye movement sleep and shorter latency to rapid eye movement sleep were related to greater depressive symptoms in the anxious group only.

Conclusions:

The current findings suggest that physiologic sleep characteristics may contribute in direct ways to the symptom profiles of clinically-anxious children. The functional relevance of such findings (e.g., how specific sleep characteristics serve to either increase or reduce long-term risk) is a vital direction for future research.

Although sleep complaints are prevalent across various types of child internalizing disorders (Alfano & Gamble, 2009) differences in sleep architecture (i.e., the basic structural organization of sleep) as compared to typically-developing children are far less commonly observed. For example, during the pre-pubertal years, youth with anxiety disorders and major depressive disorder exhibit objective sleep patterns largely similar to those of healthy youth (for reviews see Ivanenko, Crabtree, & Gozal, 2005; McMakin & Alfano, 2015). These findings stand in contrast to reliable evidence of sleep abnormalities in adult patients, particularly those with depression (Palagini, Baglioni, Ciapparelli, Gemignani, & Riemann, 2013). Notwithstanding the fact that child-based studies are fewer overall and often comprised of broad age ranges of youth, lack of sleep-based differences in child patients may well be attributable to maturational differences (e.g., stronger homeostatic sleep pressure) that override sleep-based expression of affective illness (Carskadon, 2002; Dahl, 1996). Still, as the vast majority of children with anxiety disorders and depression complain of poor sleep (Alfano, Ginsburg, & Kingery, 2007; Alfano, Patriquin, & De Los Reyes, 2015; Chorney, Detweiler, Morris, & Kuhn, 2008) and early sleep problems reliably presage the later onset of these disorders (Gregory et al., 2005; Kelly & El-Sheikh, 2014; Ong, Wickramaratne, Tang, & Weissman, 2006), it is clear that critical sleep-affective relationships begin to take root in childhood. Thus, there is need for research to move beyond group-based comparisons of sleep in favor of exploring how nighttime sleep characteristics relate to daytime emotional functioning in both clinical and healthy populations of children.

Sleep and affective functioning are also recognized to share bidirectional associations (see Alvaro, Roberts, & Harris, 2013) yet differential relationships are evident in across clinical and non-clinical samples. Sleep deprivation, particularly loss of rapid eye movement (REM) sleep, produces reliable, albeit transient improvements in mood in a majority of clinically-depressed patients whereas a worsening of mood is typically observed in non-depressed samples (see Giedke & Schwärzler, 2002). Other research has found affective responses to sleep deprivation to be predicted by degree of diurnal mood variation more specifically (Haug, 1992). Sleep deprivation studies are infrequent in child samples but other research highlights the bidirectionality of nighttime sleep and daytime affect in youth both with and without internalizing psychopathology. For the most part, this growing body of research reveals stronger pathways from sleep to affect/mood than the reverse relationship (McMakin & Alfano, 2015). For example, using three waves of data over a five year period (from 8 to 13 years) Kelly and El-Sheikh (2014) showed shorter sleep duration and worse sleep quality to predict increased anxiety and depressive symptoms whereas less robust associations were identified in the opposite direction. Similarly, in a study among youth ages 9–16, subjective sleep complaints predicted a later diagnosis of generalized anxiety disorder (GAD; Shanahan, Copeland, Angold, Bondy, & Costello, 2014). In one of the only studies to include youth with anxiety and depressive disorders as well as healthy children, one week of actigraphy and daily affect monitoring revealed more time spent asleep at night to predict greater positive affect the next day in the clinical groups only (Cousins et al., 2011). These findings in particular highlight the presence of differential relationships between nighttime sleep patterns and daytime affect in clinical versus healthy children.

Building on this burgeoning area of research, the current study was interested in whether, irrespective of quantitative differences in sleep, the daytime symptoms of youth with anxiety disorders might show distinct relationships with sleep architecture. Our study focuses on children with GAD, specifically, for several reasons. Most prominently, the vast majority (up to 90%) of children with GAD and/or their parents report sleep to be problematic (Alfano, Beidel, Turner, & Lewin, 2006; Alfano et al., 2007; Alfano, Pina, Zerr, & Villalta, 2010). Second, defining features of the disorder, including chronic worry and hypervigilance for threat, are inherently counterproductive to achieving and maintaining sleep (Dahl, 1996). A large proportion of children with GAD also eventually develop depressive disorders (Brown & Barlow, 1992; Kessler et al., 1994; Pine, Cohen, & Brook, 2001), vulnerability for which is marked by sleep-related disruptions (Palagini et al., 2013). Third, examinations of sleep architecture in this population which are notably limited have produced equivocal findings. In a sleep lab setting, children with GAD evidenced prolonged sleep onset and reduced latency to rapid eye movement (REM) sleep compared to controls (Alfano, Reynolds, Scott, Dahl, & Mellman, 2013) but results were not replicated using home-based polysomnography (PSG; Patriquin, Mellman, Glaze, & Alfano, 2014). Understanding whether and how the underlying structure of sleep maps onto the daytime symptoms of children with GAD may provide greater insight into these discrepancies, early sleep-affective relationships in general, and the role of sleep in the symptom profiles of anxious youth.

We begin with an overview of normal sleep architecture including seminal relationships between specific sleep stages and aspects of affect and psychopathology as a basis for the current investigation which includes children with a primary GAD diagnosis and a healthy control sample.

Sleep Architecture and Affective Links in Childhood

Polysomnography (PSG), which includes electroencephalography (EEG), electrooculography (EOG), and electromyography (EMG), is considered the ‘gold standard’ for assessing sleep architecture. In general, sleep consists of two main types – rapid eye movement sleep (REM) and non-rapid eye movement sleep (NREM) – each with unique neurobiological and physiological characteristics. NREM sleep is further divided into three stages (N1 – N3) corresponding with increasing depth of sleep as determined by progressive dominance of high-voltage, low-frequency EEG activity. NREM sleep stages are also marked by parasympathetic dominance, including decreased muscle tone, reductions in blood pressure/heart rate, and slowed and rhythmic respiration. N1, sometimes referred to as ‘relaxed wakefulness’, characterizes the transition from wake to sleep and occupies the smallest percentage of total sleep time. In contrast, N2 occupies approximately 50% of the nighttime sleep period and is characterized as light sleep. N3 is the deepest and most restorative of all sleep stages, marked by slow, delta brain waves, and thus, is referred to as slow wave sleep (SWS).

As compared to other stages of sleep, dramatic changes in N3 are observed in late childhood as children transition into adolescence, including an approximate 40% decrease in SWS (Campbell & Feinberg, 2009; Carskadon & Dement, 2011; Colrain & Baker, 2011). This significant decline primarily accounts for observable reductions in total sleep time during adolescence compared to the childhood years (Karacan, Anch, Thornby, Okawa, & Williams, 1975). Studies experimentally disrupting SWS find increased daytime sleep propensity (Dijk, Groeger, Stanley, & Deacon, 2006) and subsequent increases in SWS (Dijk, Beersma, Daan, Bloem, & van den Hoofdakker, 1987; Ferrara, De Gennaro, & Bertini, 1999), highlighting its functional significance. SWS enhances various daytime functions and is believed to be important for learning and memory consolidation (Walker, 2009). During childhood, SWS is theorized to play a critical role in brain maturation and development, including increased synaptic density (Campbell & Feinberg, 2009; Kurth et al., 2010).

A multitude of evidence reveals the presence of decreased N3 sleep in adults with depression as compared to healthy individuals (Berger, van Calker, & Riemann, 2003; Palagini et al., 2013; Riemann, Berger, & Voderholzer, 2001; Tsuno, Besset, & Ritchie, 2005). This robust finding is also evident in the first degree relatives of those who are depressed and observed even after a depressive episode has remitted (Lauer, Schreiber, Holsboer, & Krieg, 1995; Pillai, Kalmbach, & Ciesla, 2011). In childhood, N3 sleep may be distinctly protective against the later onset of depression in at-risk children. In a longitudinal study of non-depressed, pre-pubertal children (ages 6–11) at genetic risk for depression, participants completed diagnostic assessments annually through early adulthood (18–29 years) in order to examine risk and protective factors (Silk et al., 2007). Participants who did not develop depressive disorders in adulthood (i.e., resilient at-risk participants) evidenced greater amounts of SWS in childhood.

In contrast to NREM sleep, REM sleep is often referred to as “paradoxical sleep” due to the presence of low-amplitude mixed-frequency EEG activity (e.g., theta and alpha waves) that resembles brain activity during wakefulness. Increased eye movements, irregular respiration, and rapid changes in heart rate and blood pressure are also seen during REM sleep. This is the stage during which most dreaming activity occurs, with concomitant muscle atonia inhibiting motor activity so cortical signals during dreaming do not create bodily responses. REM sleep accounts for approximately 25% of nighttime sleep (Anders, Sadeh, & Appareddy, 1995) with mild decreases observed after the transition to adolescence. Unlike N3 sleep however, REM sleep percentage does not change significantly in relation to total sleep time (Ohayon, Carskadon, Guilleminault, & Vitiello, 2004). Also, whereas N3 sleep dominates the first half of the night, REM sleep is dominant during the second half, with REM periods becoming progressively longer and denser across the night.

REM sleep shows intimate relations with emotional learning and memory (Goldstein & Walker, 2014; Walker & van der Helm, 2009). At a basic neurobiological level, increased activity in emotion-related brain regions are observed during REM sleep, including the amygdala, striatum, hippocampus, insula, and the medial prefrontal cortex, along with reductions in levels of various neurotransmitters (e.g., noradrenaline; Hobson & Pace-Schott, 2002; Walker & van der Helm, 2009). During REM, the activation of these emotion-related brain structures in the absence of noradrenergic changes is believed to provide the backdrop for consolidating emotional events experienced during the day into memory while attenuating the intensity of associated emotions (Walker & van der Helm, 2009). For example, in a study that selectively deprived healthy adults of REM sleep, subsequent increases in reactivity to emotional stimuli were found in comparison to a NREM sleep deprived group (Rosales-Lagarde et al., 2012).

REM sleep is also heavily implicated in depressive symptomology. Among the most reliable sleep alterations observed in depressed patients are decreased latency to the first REM sleep period and increased time spent in REM (Palagini et al., 2013), which have been replicated in some samples of youth (Arana-Lechuga et al., 2008; Dahl et al., 1991; Emslie, Rush, Weinberg, Rintelmann, & Roffwarg, 1990; Lahmeyer, Poznanski, Bellur, 1983). Several studies among those at risk for depression have found alterations in REM sleep to be present even before onset of the disorder (Giles, Roffwarg, & Rush, 1987). REM sleep abnormalities are therefore considered among the most robust biomarkers for depressive illness (Palagini et al., 2013).

To a lesser extent, REM sleep has been implicated in somatic/physiological hyperarousal processes. Sleep disturbance generally is associated with somatic complaints in youth (e.g., Lewandowski, Ward, & Palermo, 2011; Miller, Palermo, Powers, Scher, & Hershey, 2003), but research among depressed patients suggests relationships between REM sleep specifically and inflammatory markers/disease (Motivala, Sarfatti, Olmos, Irwin, 2005). For example, greater amounts of REM sleep have been observed in patients with irritable bowel syndrome (Kumar et al., 1992). Also, dreams (typically occurring during REM sleep) have been found to precipitate migraines in adults (especially when containing negative content) and relate to somatic distress (Heather-Greener, Comstock, & Joyce, 1996; Levin, Lantz, Fireman, & Spendlove, 2009). Our research group has also shown percentage of REM sleep to relate positively to pre-sleep somatic arousal in children with GAD but not controls (Patriquin et al., 2014).

The Current Study

Overall, relationships between sleep and daytime affective functioning are evident throughout the sleep literature but few studies have focused on children. Studies among clinical populations of youth are especially lacking. The current study examined these relationships among youth with GAD and healthy controls. We focused on a school-aged sample (ages 7 to 11 years) in order to examine sleep-affective relationships in children predominantly likely to be pre-pubescent, as puberty results in significant neurobiological, hormonal, and social changes that impact both affect and sleep-wake regulation (e.g., Carskadon, Acebo, & Jenni, 2004; Darchia & Cervena, 2014). Better understanding of these relationships earlier in development may also serve to alter high-risk trajectories. In the current study, we examined sleep architecture during a night of PSG in relation to trait-based measures of psychopathology symptoms (anxiety, depression, and hyperarousal) as well as daily reports of negative affect and somatic symptoms across one week. Based on previous research, we expected less N3 sleep, greater REM sleep, and reduced latency to REM sleep to be related to overall depressive and anxiety symptoms in children with GAD. We also expected less N3 and greater REM sleep to be associated with more daytime affective problems in this group. Conversely, we hypothesized positive relationships between total REM sleep and both overall and daytime somatic/hyperarousal symptoms in the anxious group.

Method

Participants

Participants were 66 children aged 7–11 (M = 8.79, SD = 1.36) including youth with primary GAD (n = 29) and healthy controls (n = 37) recruited from two large metropolitan cities in the U.S. (Washington, D.C. and Houston, TX). Approximately half of the participants were female (n = 36, 54.5%), most were Caucasian (56.1%; African-American, 6.1%; Asian-American, 3%; biracial, 18.2%) and 16.7% were Hispanic. About half (51.5%) reported household incomes above $100,000, 25.7% between $60–100,000, 15.2% between $20–60,000, and 6% below $20,000. Most children had at least one parent with a college (40.9%) or advanced degree (48.5%). There were no differences in demographic characteristics by group (see Table 1 for more information). Pubertal status was assessed using the parent-reported pubertal development scale (PDS; Carskadon & Acebo, 1993), and all but one child was considered pre-pubertal.¹ The original study sample from which participants were drawn included a total of 83 children, but 17 were excluded from the current set of analyses because of missing PSG data. Specifically, 7 families opted not to complete the PSG, 4 were unable to complete the PSG due to scheduling issues, and 6 participants’ data was lost due to technical errors. Comparison of children with and without available PSG data did not reveal significant differences in terms of demographic, clinical, or sleep characteristics.

Table 1.

Demographic Characteristics of Children with GAD and Controls

	GAD (n = 29)	Control (n = 37)	t or X2 statistic (df)	p value
Age: M (SD)	8.76 (1.46)	8.81 (1.31)	.15 (64)	ns
Female: n (%)	17 (58.6%)	19 (51.4%)	.35 (1)	ns
Race/Ethnicity n (%)			1.86 (4)	ns
Caucasian	19 (65.5%)	24 (64.9%)
African-American	1 (3.4%)	3 (8.1%)
Asian-American	1 (3.4%)	1 (2.7%)
Biracial/Other	4 (13.8%)	2 (5.4%)
Hispanic/Latino	7 (24.1%)	10 (27%)
Parental Education: n (%)			.04 (1)	ns
No college or some college	4 (13.8%)	6 (16.2%)
At least one parent with college degree	25 (86.2%)	31 (83.7%)
Household Income: n (%)			−.97 (63)	ns
< $20K	2 (6.9%)	2 (5.6%)
$20–40 K	1 (3.4%)	4 (11.1%)
$40–60 K	3 (10.3%)	2 (5.6%)
$60–80 K	2 (6.9%)	6 (16.7%)
$80–100K	3 (10.3%)	6 (16.7%)
> $100 K	18 (62.1%)	16 (44.4%)

Note. GAD = generalized anxiety disorder.

All children were required to live with a primary caretaker and be enrolled in regular educational classes. Children were excluded if they were taking any medications known to impact anxiety or sleep (e.g., selective serotonin reuptake inhibitors, melatonin), had a chronic medical condition, or a known or suspected sleep disorder (e.g., sleep-disordered breathing). Also, because the goal of the larger study was to examine sleep problems and patterns associated with childhood GAD specifically, youth could not have a current or lifetime history of depressive disorders. Psychotic, pervasive developmental, or bipolar disorders were also exclusionary. Other exclusion criteria included an IQ of less than 80, a body mass index (BMI) of more than 25, or a non-English speaking parent or child. In addition, control children could not meet diagnostic criteria for any psychiatric or sleep disorder.

Procedure

Children were recruited via community flyers, mailing to local schools and pediatrician offices, and advertising in local publications and at local events for a study about “behavior and emotions.” The study protocol was approved by an Institutional Review Board at both locations. Participants were first consented/assented and then completed an initial assessment that included a diagnostic interview to determine clinical diagnoses, abbreviated IQ testing (Wechsler, 1999), and parent and child questionnaires.

Within one week of this initial assessment, participants completed seven days of daily reports. Each evening, a member of the study staff called the participant on the phone and asked them to provide reports of their negative affective symptoms and somatic complaints during that day. Some participants (34.8%, n = 23) missed at least one phone call resulting in a significant amount of missing data on these indicators (9.5%). Specifically, 11 children were missing one night, 7 children were missing two nights, 3 children were missing three nights, and 2 children were missing five nights of data. We examined if children who missed at least one phone call differed from children who completed all reports, and there were no differences on any demographic, clinical, or sleep characteristics. However, children at the Houston location were significantly more likely to have missed at least one phone call during week (Χ² = 13.68, p < .001) than those from the Washington, D.C. location.

On the final night of the daily reports (a Friday or Saturday) participants completed one night of PSG monitoring. Participants from Washington, D.C. completed PSG monitoring in a sleep laboratory in a pediatric hospital (n = 29) and participants from Houston completed unattended PSG monitoring at home (n = 37).² Equal proportions of anxious and control children completed at-home and lab-based PSGs. Children wore a wrist actigraph (Ambulatory Monitoring, Inc.) during the week prior to the PSG to ensure they were not sleep deprived on the PSG night. During this week, children were not given any specific sleep instructions other than to follow their normal sleep routines. A majority of children participated in the study during the school year (n = 54, 81.8%). Families who completed the entire study were compensated $160 for their time and effort, and were provided clinical referrals if needed.

Questionnaires and Sleep Measures

Clinical Interview.

The Anxiety Disorders Interview Schedule for the DSM-IV for Children and Parents (ADIS-C/P; Silverman & Albano, 1996) was administered at the initial assessment. Separate interviews were conducted with the child and parent by a Ph.D. level psychologist or trained doctoral level graduate student, and all cases were reviewed with a licensed clinical psychologist prior to assigning final diagnoses. Clinician severity ratings (CSR; range 0–8) are used to determine the severity of each disorder in order to differentiate primary (most severe) from secondary diagnoses. The mean CSR for the current sample of children with GAD was 5.66 (SD = 1.26, range = 4–8). The ADIS C/P is considered the gold standard for assessing child anxiety, and past studies demonstrate strong inter-rater reliability, test-rest reliability, and concurrent validity (Lyneham, Abbott, & Rapee, 2007; Silverman, Saavedra, & Pina, 2001). Reliability for a GAD diagnoses in the current study was excellent (Kappa = 1.0). Some children with GAD (n = 15) had secondary diagnoses, including separation anxiety disorder (n = 3), social anxiety disorder (n = 11), specific phobia (n = 6), attention deficit/hyperactivity disorder (n = 3), and oppositional defiant disorder (n = 1).

Depressive Symptoms.

General depressive symptoms were measured using the 27-item Children’s Depression Inventory (CDI; Kovacs, 1992) at the initial assessment. For each item, children are presented with 3 statements and asked to choose which statement most describes how they have been feeling over the last two weeks (e.g., I have fun in many things, I have fun in some things, or nothing is fun at all). Each item is scored from 0 (not indicating depression) to 2 (indicating higher depression). The CDI is a reliable, well-validated and widely used assessment to assess depressive symptomology in youth (Kovacs, 1992). Raw scores were converted to T-scores adjusted for age and gender, and items were totaled so that higher values indicate greater depressive symptoms. Cronbach’s alpha indicated good reliability in the current sample (α = .86).

Anxiety Symptoms.

At the initial assessment, anxiety symptoms were measured using the child-report version of the Screen for Child Anxiety Related Emotional Disorders (SCARED-C; Birmaher et al., 1999), which has demonstrated good internal consistency, validity, and test-retest reliability in child populations (Birmaher et al., 1997). The SCARED-C is a 41-item questionnaire that asks children to indicate how much each statement (e.g., “I worry about other people liking me”) describes how they have felt for the last 3 months from 0 (not true or hardly ever true) to 2 (very true or often true). The SCARED-C offers subscale scores for panic disorder/somatic symptoms, generalized anxiety, separation anxiety, social anxiety, and school avoidance. For the current investigation, summed total scores of the anxiety-related subscales were used to represent overall anxiety (α = .93).

Physiological Hyperarousal.

During the initial assessment, physiological arousal was examined using the Physiological Hyperarousal Scale for Children (PHS-C; Laurent, Catanzaro, & Joiner, 1995; Laurent, Catanzaro, & Joiner, 2004), a reliable and valid measure of anxiety-related hyperarousal in clinical and non-clinical populations (Laurent, Joiner, & Catanzaro, 2011). The PH-C is a 18-item self-report scale assessing physiological symptoms over the last 2 weeks (e.g., “How much did you feel like your heart was pounding during the past 2 weeks?”) on a 5-point scale from 1 (very slightly or not at all) to 5 (extremely). Items were summed to create a total score, with higher scores indicating greater hyperarousal (α = .85).

Daily Reports of Negative Affect and Somatic Complaints.

After the initial assessment, participants completed 7 days of daily phone calls. During each phone call, participants provided responses to 3 questions assessing negative affect and 3 questions assessing somatic complaints that day on a 4-point scale from 0 (none) to 3 (a lot). For negative affect (α = .77), children reported how much they felt 1) nervous or anxious, 2) sad or upset, and 3) irritable or angry that day. For somatic symptoms (α = .60), children reported how much they experienced 1) muscle aches or tight muscles, 2) headaches, and 3) stomachaches. To account for missed phone calls, we opted to calculate average symptoms across the week as opposed to total scores. Specially, the three items on each scale were averaged across each day and the averages of each day were averaged across the week, resulting in a total score with higher values indicating greater average symptoms during the one week period. The affective and somatic items were developed for the purpose of this study, and previous research suggests that the use of daily diary is feasible to assess symptoms in anxious children (Beidel, Neal, & Lederer, 1991).

Polysomnography (PSG).

For all participants, PSG monitoring (on a Friday or Saturday night following the 7 nightly phone calls and one week of actigraphy monitoring) included 6 channels of EEG (frontal, central, and occipital regions), right and left EOG, submental and right/left tibial EMG, electrocardiogram (ECG), respiratory inductance plethysmography, and pulse oximetry. For ambulatory PSGs (conducted using NicoletOne ambulatory equipment) participants were prepared at a pediatric sleep center by a registered technologist and followed their typical bedtime and sleep routine at home that night. PSGs conducted in a sleep-lab in pediatric hospital were performed with Medcare amplifiers and Rembrandt 9.0 Sleep Acquisition Software. For all studies, registered PSG technicians with pediatric experience scored sleep stages in 30 second epochs according to AASM scoring rules (Iber, Ancoli-Israel, Chesson, & Quan, 2007) and studies were reviewed by a certified sleep medicine specialist before generating a final report. All scoring technicians were masked to child diagnostic status. For the current study, the following sleep architecture variables were of interest: percentage of total sleep time spent in N3 sleep (N3 %), percentage of total sleep time spent in REM sleep (REM %), and the number of minutes from the first epoch of sleep to the first epoch of REM sleep (REM Latency).

Results

Preliminary Analyses

We examined if primary study variables differed as a function of group (GAD vs. control), type of PSG (ambulatory vs. lab-based), or whether children participated during the school year versus the summer using a series of t-tests and ANOVAs. There were no differences in sleep architecture or psychopathology symptoms based on the type of PSG participants completed. Across the entire sample, children reported marginally more physical symptoms in the summer compared to the school year [t(64) = −2.00, p = .05].

Means and standard deviations for primary study variables and other PSG characteristics for both groups and the total sample are reported in Table 2. As expected, children with GAD reported significantly greater anxiety symptoms, physiological hyperarousal, and depressive symptoms than controls. Across the one week assessment, children with GAD reported marginally greater somatic complaints and significantly greater negative affect. There were no differences in sleep architecture variables based on group. We also examined descriptive sleep information from each group from the week prior to the PSG using actigraphy-measured variables. This information is presented in Table 3. A series of independent samples t-tests indicated that there were no group differences in any of the actigraphy-measured sleep indicators.

Table 2.

Descriptive Statistics for Primary Study Variables and Additional PSG Characteristics

	GAD		Control		Total
	M	SD	M	SD	M	SD	t statistic (df)	95% CI
PSG TST (minutes)	496.28	45.64	484.33	47.66	489.58	46.81	−1.03(64)	−35.13, 11.22
N2 minutes (%)	246.83(49.66)	38.53(5.42)	242.20(50.29)	40.37(8.29)	244.23(50.01)	39.34(7.13)	−.47(64)	−24.23, 12.98
N3 minutes (%)	125.40(25.40)	26.11(5.36)	124.39(25.74)	29.10(5.79)	124.83(25.59)	27.62(5.57)	−.15(64)	−14.80, 12.79
REM minutes (%)	113.38(22.81)	21.10(3.48)	109.20(22.16)	32.27(5.26)	111.04(22.45)	27.80(4.54)	.55(64)	−18.02, 9.67
REM Latency (minutes)	134.71	58.33	148.97	54.35	142.70	56.15	1.03(64)	−13.54, 42.08
Daily Negative Affect	.33	.40	.14	.17	.22	.31	−2.56(36.13)*	−.36, −.04
Daily Somatic Complaints	.23	.33	.10	.14	.15	.25	−2.02 (35.89)+	−.27, .00
CDI	50.84	10.57	42.60	6.47	46.03	9.29	−3.46 (36.72)**	−13.06, −3.42
SCARED-C	27.36	14.44	13.77	9.56	19.81	13.68	−4.47 (61)***	−19.65, −7.52
PHS-C	30.90	11.49	25.59	6.04	28.11	9.35	−2.22 (41.41)*	−10.12, −.49

Note. Significant differences between groups are marked with

⁺p = .05

^*p < .05

^**p < .01

^***p < .001.

GAD = generalized anxiety disorder, TST = total sleep time, N2 = stage 2 sleep, N3 = stage 3 sleep, REM = rapid eye movement sleep, CDI = Children’s Depression Inventory, SCARED-C = Screen for Child Anxiety Related Emotional Disorders, PHS-C = Physiological Hyperarousal Scale for Children.

Table 3.

Actigraphy Measured Sleep Variables Prior to the PSG Recording

	GAD		Control		Total
	M	SD	M	SD	M	SD	t statistic (df)	95% CI
TST (minutes)	510.32	32.31	501.10	43.42	505.15	39.78	−.93(64)	−28.95, 10.51
TIB (minutes)	593.46	47.92	574.24	51.03	582.57	50.21	−1.48(58)	−45.15, 6.70
SE (%)	86.75	5.79	87.36	5.01	87.10	5.32	.44(59)	−2.17, 3.38
WASO (minutes)	49.29	29.07	44.61	24.58	46.67	26.54	−.71(64)	−17.88, 8.52
SL (minutes)	23.43	13.94	21.01	11.13	22.04	12.35	−.75(59)	−8.84, 4.01

Note. Values are averaged across the 7 day pre-PSG period, GAD = generalized anxiety disorder, TST = total sleep time, TIB = time in bed, SE = sleep efficiency, WASO = wake after sleep onset, SL = sleep latency. Actigraphy variables did not significantly differ between groups.

Bivariate correlations among the primary variables of interest are reported in Table 4. As would be expected, longer latency to REM was related to less time spent in REM. Daily reports of negative affect and somatic symptoms, and global physiological hyperarousal, depressive and anxiety symptoms were all positively related to one another. We also examined if age was related to any variables of interest, given typical developmental changes in sleep architecture across childhood. Age was unrelated to sleep architecture, daily symptoms, or psychopathology symptoms.

Table 4.

Bivariate Correlations for Primary Study Variables in the Full Sample

	1	2	3	4	5	6	7	8
1. Age
2. N3 %	−.11
3. REM %	.13	−.03
4. REM Latency	−.10	−.02	−.30*
5. Daily Negative Affect	.07	−.16	.18	−.24+
6. Daily Somatic Complaints	.01	.003	.11	−.09	.61***
7. CDI	−.13	.15	−.01	−.17	.48***	.46***
8. SCARED-C	.09	.07	.13	−.13	.43***	.54***	.73***
9. PHS-C	−.08	.21	−.04	−.08	.44***	.64***	.60***	.67***

Note.

⁺p = .05

^*p < .05

^***p < .001.

N3 = stage 3 sleep, REM = rapid eye movement sleep, CDI = Children’s Depression Inventory, SCARED-C = Screen for Child Anxiety Related Emotional Disorders, PHS-C = Physiological Hyperarousal Scale for Children.

Sleep Architecture and Daily Affective and Somatic Reports

To examine if sleep architecture was related to daily reports of negative affect and somatic complaints we conducted a series of hierarchical regression analyses with each (continuous) sleep architecture variable, diagnostic status, and their interaction term as predictors, and daily negative affect and somatic complaints as outcome variables. We included age, PSG type, and school year/summer participation as covariates. Results are reported in Table 5. The interaction between group and N3% was significant, revealing greater N3% was related to less negative affect over the course of the week in the GAD (simple slope = −.27, p = .01) but not control group (simple slope = .18, p = .49). A significant interaction also emerged between group and REM%, indicating greater REM sleep was related to more somatic complaints, but again, for children with GAD (simple slope = .37, p = .02) and not controls (simple slope = −.08, p = .55). These interactions are displayed in Figures 1a and 1b.

Table 5.

Sleep Architecture and Daily Reports

	Negative Affect		Somatic Complaints
Variable	Beta	95% CI	Beta	95% CI
Age	−.01	−.05, .05	−.04	−.05, .04
School/Summer	.36***	.12, .45	.25*	.01, .32
PSG type	.06	−.10, .18	.16	−.05, .21
Diagnostic Status	.29**	.05, .31	.23	−.01, .23
N3%	.10	−.05, .11	.05	−.06, .09
REM%	.03	−.07, .09	−.08	−.09, .06
Latency to REM	−.04	−.11, .09	.06	−.08, .11
Group X N3%	−.37*	−.28, −.01	.05	−.10, .14
Group X REM%	−.24	−.25, .04	.36*	.03, .32
Group X REM Latency	.13	−.08, .23	−.21	−.20, .06
Adjusted R2	.30***		.13

Note.

^***p ≤ .001

^**p < .01

^*p < .05.

Diagnostic status is coded as GAD = 1 and control = 0. Summer is coded as during summer = 1 and during school year = 0. Location is coded as lab-based = 1 and ambulatory = 2. The pattern of significant findings remain unchanged when removing covariates from each model.

Figure 1a–d.

Interactions between group and sleep architecture predicting daily and trait-like psychopathology symptoms. High and low values represent +/− 1 SD from the mean.

Sleep Architecture and Global Symptom Measures

Similar regression models were used to examine relationships between sleep architecture and trait-like measures of psychopathology symptoms (see Table 6). For the model predicting depressive symptoms, a significant interaction between group and REM latency showed that, for children with GAD only, decreased latency to REM sleep was related to greater depressive symptoms (simple slope = −.45, p = .002). A similar interaction emerged between REM% and group. Increased REM sleep was related to greater depressive symptoms in children with GAD (simple slope = .33, p = .02) but not controls (simple slope = −.33, p = .14). These interactions are displayed in Figures 1c and 1d. Sleep architecture and the interaction between sleep architecture and group in the models predicting anxiety symptoms and physiological hyperarousal were not significant.

Table 6.

Sleep Architecture and Psychopathology Symptoms

	Anxiety Symptoms		Depressive Symptoms		Physiological Hyperarousal
	Beta	95% CI	Beta	95% CI	Beta	95% CI
Age	.08	−1.66, 3.39	−.05	−1.93, 1.30	−.07	−2.37, 1.43
School/Summer	−.02	−8.42, 7.31	.04	−4.19, 5.83	.07	−4.44, 7.65
PSG type	.20	−1.18, 12.14	.23	−.16, 8.47	.22	−1.05, 9.27
Diagnostic Status	.47***	6.63, 19.05	.38***	2.96, 11.13	.25	−.25, 9.27
N3%	.16	−1.81, 6.28	.23	−.45, 4.85	.24	−1.02, 5.51
REM%	−.06	−4.58, 3.06	−.21	−4.29, .62	−.23	−5.13, 1.00
Latency to REM	−.08	−5.75, 3.64	.12	−1.87, 4.03	−.13	−4.71, 2.42
Group X N3%	−.10	−8.60, 4.25	−.01	−4.58, 4.27	−.01	−5.14, 4.78
Group X REM%	.24	−1.13, 13.97	.39**	2.13, 11.85	.25	−1.25, 10.14
Group X REM Latency	−.02	−7.01, 6.29	−.43**	−9.78, −1.37	.04	−4.57, 5.52
Adjusted R2	.23**		.34***		.05

Note.

^***p ≤ .001

^**p ≤ .01

^*p < .05.

Diagnostic status is coded as GAD = 1 and control = 0. Summer is coded as during summer = 1 and during school year = 0. Location is coded as lab-based = 1 and ambulatory = 2. The pattern of significant findings remain unchanged when removing covariates from each model.

Discussion

The current study investigated interactions between aspects of sleep architecture and affective symptoms in both clinically-anxious and healthy children toward developing a better understanding of the functional relevance of these relationships for emotional functioning in childhood. We specifically examined both general measures of psychopathology symptoms as well as negative affect and somatic complaints across a one week period. First, as expected, children with GAD reported greater anxiety symptoms, depressive symptoms, and physiological hyperarousal overall, and experienced greater negative affect and somatic complaints across the seven day period. Moreover, these group differences were moderated by physiologic sleep characteristics whereby significant relationships were observed for the anxious children only. The implications of these novel findings for developmental models of affective disorders are considered below.

We found a greater percentage of N3 (SWS) sleep to relate significantly to less negative affect across the week among children with GAD but not controls. This finding aligns with emerging experimental research suggesting SWS attenuates the effects of distressing daytime emotional experiences (e.g.,Talamini, Bringmann, de Boer, & Hofman, 2013) as well as evidence for the protective role of SWS in at-risk children. Specifically, in their longitudinal study, Silk and colleagues (2007) found greater that SWS during childhood was protective against the later development of depression in genetically at-risk youth. Importantly, our anxious group was somewhat unique in that children in the current study did not have comorbid depressive disorders. In light of robust genetic and phenomenological overlap between GAD and depression (Kendler, Neale, Kessler, Heath, & Eaves, 1992; Middeldorp, Cath, Van Dyck, & Boomsma, 2005), these children might be conceived of as an at-risk but resilient anxious subgroup. Although we found no differences in the amount of N3 sleep between our two groups, it may be the case that intensity rather than quantity of SWS, as indexed by slow wave activity (SWA; EEG power density between 0.5 and 4 Hz) is more germane for understanding this relationship. Specifically, among depressed and at-risk youth, greater negative emotional experiences during the day might accelerate SWA at night as a compensatory response aimed at reprocessing amplified negative emotions (Talamini et al., 2013). The typical daytime emotional experiences of healthy children, comparatively, may not trigger the same accelerations in SWA. Another possibility is that spectral power of SWS may be compromised in those who report sleep problems not corroborated by macro-level differences in sleep architecture (Krystal, Edinger, Wohlgemuth, & Marsh, 2002).This is a particularly intriguing question for anxious youth who often report sleep problems in the absence of objective sleep alterations (McMakin & Alfano, 2015).

Results from the current study also revealed a positive relationship between percentage of REM sleep and the daytime somatic complaints of anxious children. The mechanisms underlying this relationship require further investigation but several specific explanations appear worthy of exploration. First, research in adults has linked dream content, and negatively-valenced dream mentation specifically with migraines and somatic distress (Heather-Greener et al., 1996; Levin et al., 2009). Anxious/depressed adults and children report more emotionally-negative dream content than controls (McNamara, Auerbach, Johnson, Harris, & Doros, 2010; Nielsen et al., 2000) but direct relationships with daytime somatic complaints have yet to be examined. Second, some sleep-related behaviors including bruxism (i.e., involuntary teeth grinding or clenching) have been shown to contribute to daytime somatic symptoms (Shang, Gau, & Soong, 2006) and may occur more frequently during REM sleep in childhood (Herrera et al., 2006; Mahowald & Rosen, 1990). Bruxism is more common in anxious individuals (Ohayon, Li, & Guilleminault, 2001) as well as patients with frequent tension headaches (De Luca Canto, Singh, Bigal, Major, & Flores-Mir, 2014). However, because bruxism occurs outside of conscious awareness, and is often imperceptible by others research utilizing PSG is needed to explore these ostensive links in anxious youth.

It is important to point out that in addition to being a required criterion for the diagnosis of GAD (American Psychiatric Association, 2013), somatic symptoms in anxiety-disordered youth also correspond with a more severe form of anxiety and more impaired functioning (Ginsburg, Riddle, & Davies, 2006), which was evident in the current sample from the positive relationship between somatic symptoms and self-reported anxiety symptoms (a follow-up analysis indicated that this relationship was significant for children with GAD: r = .63, p < .001). In this way, a positive association between REM sleep and somatic complaints in our anxious group may be driven in part by symptom severity. Further, because high levels of somatization predict the later development of depression (Shanahan et al., 2015; Zwaigenbaum, Szatmari, Boyle, & Offord, 1999), prospective studies examining the potential prognostic value of REM sleep-somatic connections in youth at risk for depression are suggested as an important direction for research.

In terms of more general psychopathology measures, our results converge with a wealth of findings showing reduced latency to and a greater proportion of REM sleep to correspond with depression as well as depression risk (Mendlewicz, Sevy, & de Maertelaer, 1989; Palagini et al., 2013; Riemann et al., 2001; Tsuno et al., 2005). Comparatively, consistent and reliable sleep-based markers of anxious pathology have not been identified, supporting non-significant associations between sleep architecture and global anxiety symptoms found in our study. Still, anxiety and depression are highly comorbid across development (Brady & Kendall, 1992; Cummings, Caporino, & Kendall, 2014; Moffit et al., 2007; Pine et al., 2001) and childhood GAD in particular is characterized by hetereotypic more so than homotypic continuity as a majority of affected children lose this diagnosis as adolescents or adults (Bittner et al., 2007; Copeland, Shanahan, Costello, & Angold, 2009). These developmental trajectories are poorly understood at present but a better understanding of longitudinal relationships between nighttime sleep and daytime affect in clinical and at-risk populations of youth may hold relevance in this regard.

Limitations and Future Directions

Our findings should be viewed in light of several limitations. Only one night of PSG monitoring was conducted which has been associated with first-night effects (i.e., non-typical sleep patterns due to equipment and/or sleeping in an unfamiliar environment). Our sample also included youth who underwent different types of sleep monitoring procedures (at-home vs. sleep-lab based studies). Although PSG type was evenly distributed across our two groups and included as a covariate in all analyses, the use of uniform PSG procedures across multiple nights would clearly be more ideal. We are also unable to account for possible differences in sleeping arrangements and environments during ambulatory studies (e.g., co-sleeping) that may have affected sleep patterns (e.g., awakenings). Future research including multiple nights of PSG and using standardized procedures are therefore needed to replicate our results. Additionally, the somatic and negative affect questionnaires in the current study had lower than desirable reliability (though is it important to point out the range of symptoms assessed by both measures). Future studies utilizing validated measures (e.g., Positive and Negative Affect Schedule for Children; Laurent et al., 1999) and/or more distinct components of emotional and physical functioning (e.g., discrete types of negative affect emotion such as sadness or anger, or specific physical symptoms such as inflammation or muscle tension) are needed.

Although racially and ethnically diverse, participants in this sample came from educated and affluent families and results may therefore not generalize to other populations. The fact that children with primary GAD could not have secondary mood disorders, while providing a novel sample, further limits the generalizability of findings. Given the sample size, we did not explore gender differences in the current study but sleep by gender interactions have been reported in youth with affective disorders (Alfano et al., 2013; Armitage & Hoffmann, 2001). Finally, the directional nature of effects detected cannot be clarified based on these data. Experimental studies that specifically manipulate daytime affective experiences and/or nighttime sleep patterns (e.g., deprivation of REM sleep only) are necessary for deciphering causal relationships between sleep architecture and waking affective symptoms.

Clinical Significance and Implications

These findings contribute to existing knowledge in several ways. First, results build directly on data linking objective sleep characteristics with daytime symptoms in youth with affective disorders (Cousins et al., 2011; Silk et al., 2007) and suggest that for anxious children, aspects of sleep architecture may contribute to symptomatic profiles and possibly longer-term outcomes (e.g., depression). Given the sharp increases in depressive disorders in adolescence (e.g., Kessler, Avenevoli, & Merikangas, 2001), and the high rate of comorbidity between depression and anxiety (Brady & Kendall, 1992; Brown & Barlow, 1992), understanding these processes during childhood, before the onset of puberty, is critical for identifying potential targets for prevention and intervention. Furthermore, these findings highlight the potential importance of targeting problematic nighttime behaviors that serve to alter sleep architecture. As an example, late bedtimes during the week become more common and sometimes extreme during the transition to adolescence (Laberge et al., 2001), resulting not only in reduced overall sleep but in objective sleep alterations (Banks & Dinges, 2007; Banks, Van Dongen, Maislin, & Dinges, 2010). For at-risk youth in particular these sleep changes could produce or amplify affective risk both in the short and long–term. Future studies aimed at replicating and better delineating these relationships might therefore be harnessed to reduce illness severity and risk.