Evidence of a Maturational Disruption in Non-Rapid Eye Movement Sleep Slow Wave Activity in Youth with Attention-Deficit/Hyperactivity, Learning and Internalizing Disorders

Abstract

Background:

Sleep slow wave activity (SWA) peaks during childhood and declines in the transition to adolescence during typical development (TD). It remains unknown whether this trajectory differs in youth with neuropsychiatric disorders.

Methods:

We analyzed sleep EEGs of 664 subjects 6 to 21y (449 TD, 123 unmedicated, 92 medicated) and 114 subjects 7–12y (median 10.5y) followed-up at 18–22y (median 19y). SWA (0.4–4 Hz) power was calculated during non-rapid eye movement sleep.

Results:

TD and unmedicated youth showed cubic central and frontal SWA trajectories from 6 to 21y (p-cubic<0.05), with TD youth showing peaks in central SWA at 6.8y and frontal at 8.2y. Unmedicated attention-deficit/hyperactivity (ADHD) and/or learning disorders (LD) showed peak central SWA 2y later (at 9.6y, coinciding with peak frontal SWA) than TD, followed by a 67% steeper slope by 19y. Frontal SWA peak and slope in unmedicated ADHD/LD, and that of central and frontal in internalizing disorders (ID), were similar to TD. Unmedicated ADHD/LD did not differ in the longitudinal SWA percent change by 18–22y; unmedicated ID showed a lower longitudinal change in frontal SWA than TD. Medicated youth showed a linear decline in central and frontal SWA from 6 to 21y (p-linear<0.05).

Conclusions:

ADHD/LD youth show a maturational delay and potential topographical disruption in SWA during childhood and steeper decline throughout adolescence, suggesting faster synaptic pruning. Youth with ID experience less changes in frontal SWA by late adolescence. Psychotropic medications may impact the maturational trajectory of SWA, but not the magnitude of developmental decline by late adolescence.

1. INTRODUCTION

The brain undergoes developmental changes characterized by increasing synaptic density during childhood, followed by synaptic pruning and myelination in adolescence, which streamline neural circuits and mature cortical regions.^1–4 Researchers have hypothesized that neuropsychiatric disorders result from altered neurodevelopment.^5,6 Sleep electroencephalography (EEG) activity has emerged as a potential biomarker of brain maturation that, if altered, may have implications in the development of psychopathology, including attention-deficit/hyperactivity (ADHD), learning (LD) and internalizing (ID) disorders.^4,7 However, the association of clinical disorders with specific sleep oscillations at different developmental stages remains elusive.

Slow wave activity (SWA) in the delta-frequency (0.4–4 Hz) range during non-rapid eye movement (NREM) sleep is a marker of synaptic plasticity and sleep intensity that declines in typically developing (TD) children as they transition to adolescence.^4,8,9 SWA undergoes a posterior-to-anterior shift in predominance as it declines in the transition to early adolescence (age 14).^10,11 This age-related trajectory of SWA mirrors that of brain maturation biomarkers, including synaptic density.^12–14 It remains unknown whether the maturational trajectory of SWA differs between TD youth and those with neuropsychiatric disorders and their pharmacotherapy.

Previous studies examining mean differences in SWA levels in youth with diverse conditions have produced mixed results.^15–26 While one study found increased central SWA relative to other brain regions in medicated youth with ADHD (n=18, 9–14y),¹⁷ another found no significant differences in central SWA in children with ADHD (n=21, 10–13y),¹⁸ while a more recent study found lower overall SWA across all cortical regions in unmedicated youth with ADHD compared to controls (n=136, 8–16y).¹⁹ Such lower SWA was hypothesized to reflect delayed cortical maturation in children with ADHD.¹⁹ Furthermore, a recent meta-analysis of 11 cross-sectional studies found an inversion point at age 10 in SWA mean differences between youth with ADHD and TD youth, a finding that the investigators interpreted as indirect evidence of a potentially delayed cortical maturation and excessive synaptic pruning in the transition to adolescence.²⁰ When examining youth with LD, one study showed evidence of increased SWA in youth with dyslexia (n=27, 7–16y),²¹ while another found no significant differences in central, frontal or occipital SWA between youth with dyslexia and TD youth (n=52, 8–13y).²² As it pertains to ID,^23–27 one study found increased frontal SWA in adolescents with major depressive disorder, comorbid with ADHD or on psychotropic medications, leading the authors to suggest it may be indicative of altered synaptic pruning in the frontal cortex (n=30, 12–16y)²⁷; however, the potential impact of psychotropic medication use could not entirely be ruled out. Overall, the conflicting results in the literature may be due to studies focusing on mean differences in SWA in age groups with potentially overlapping developmental stages. This is important because, as reported above, SWA has been shown to undergo a distinct age-related developmental trajectory with an increase during early childhood and decline with the onset of adolescence in TD youth.^8–11 Therefore, it is important to determine whether the age-related trajectory of SWA differs among youth with neuropsychiatric disorders compared to that observed in TD youth.

The complex relationship between NREM SWA and developmental psychopathology requires additional attention to clarify its maturational trajectory and shed light on this potential biomarker. We aimed to determine whether the age-related trajectories of SWA from age 6 to 21 differ between unmedicated youth with neuropsychiatric disorders (n=123) and youth on psychotropic medications (n=92), including those with ADHD, LD or ID, compared to TD (n=449). Additionally, exploratory analyses examined the proportion of change in SWA in a subset of subjects (n=114) that had longitudinal data in the transition from childhood to the observed age at nadir (18y), including those taking stimulants or antidepressants/anxiolytics.

2. MATERIALS AND METHODS

2.1. Penn State Child Cohort (PSCC)

The PSCC is a randomly-recruited sample of 700 children (47.7% female, 23.7% racial/ethnic minority) from the general population who underwent an in-lab study at ages 5–12 (median 9y).^28–32 Out of the 700 children, 421 returned 6–13 years later (median 7.3y) for a follow-up study at ages 12–23 (median 16y, 46.1% female, 21.9% racial/ethnic minority).^9,28–36 The 279 subjects (median 9y, 47.8% female, 29.3% racial/ethnic minority) that did not return for the follow-up study were not significantly different from the 421 who were followed-up in their demographic characteristics at ages 5–12.^28–36 All subjects or parents/legal guardians provided informed written consent for the study protocol, which was approved by Penn State’s Institutional Review Board.

2.2. Demographic and Clinical Measures

All in-lab sleep studies consisted of a clinical history and physical examination, neurobehavioral assessments, and one-night-9-hour polysomnography (PSG) performed by registered PSG technicians (RPSGTs). Sex, race/ethnicity, and date of birth were reported during the clinical history. Height and weight were measured during the physical examination and body mass index (BMI) percentile was calculated. A 3-hour standardized neurobehavioral assessment was administered by a trained psychometrist and was performed to obtain measures of cognitive and behavioral functioning.^36–46 Please see the Supplement for a detailed description of the neurobehavioral assessments performed.

A structured clinical history schedule was used to gather past and current history of medical and psychiatric conditions, with specific modifications for a child cohort (e.g., focus on ADHD, learning and other behavioral disorders). The clinical history and physical examination was performed by trained clinical research technologists, medical students or residents, and supervised by an experienced scientist-clinician (physician or psychologist). During the clinical history, subjects or their parent were asked whether there was a presence of a psychiatric or behavioral disorder (“Has your child/have you ever been treated for a psychiatric/behavioral disorder?”).^36–38 This question was followed-up with the option to specify whether the disorder diagnosed was ADHD, LD or other disorder; for the latter, the diagnosis was specified with an open-ended question by which the presence of ID (e.g., depressive, generalized anxiety, post-traumatic stress, obsessive-compulsive disorders), externalizing disorders (e.g., oppositional defiant disorder) or autism spectrum disorder was determined.^29,36–38 Please see Table S1 for comorbidity across these disorders.

Current medication use was ascertained during the clinical history and on an evening pre-PSG questionnaire.^36–38 Medications were classified by an experienced registered pediatric nurse.²⁹ Psychotropic medication use was defined by a report of currently taking stimulants, other psychoactive medications (e.g., antidepressants, anxiolytics, antipsychotics) and/or sleep (e.g., melatonin, trazodone) medications.³⁷ Data on non-pharmacological (e.g., behavioral, school-based) treatments were not available. Please see Table S2 for co-existence across psychotropic medications and Table S3 for a list of the psychotropic drug in each category.

The above data informed the definition of our diagnostic groups. The TD group included youth who were never treated for or diagnosed with any psychiatric/learning disorder and were not taking any psychotropic medications. The unmedicated group included youth who had a diagnosis of a psychiatric/learning disorder and were not being treated with psychotropic medications. The medicated group included youth who had a diagnosis of a psychiatric/learning disorder and were being treated with psychotropic medications. In addition, specific groups of youth with ADHD and/or LD (ADHD/LD), given their high comorbidity and similar diagnostic characteristics, treated or untreated with stimulants and absent of any other comorbid disorder as well as youth with ID treated or untreated with antidepressants/anxiolytics, were also identified. Please see Table S1 for comorbidity across diagnostic groups and detailed data in the Supplement regarding the reliability and validity of the ADHD/LD and ID groups.

2.3. Polysomnography

Sleep data were recorded for 9-hours of time-in-bed from the time of “lights out” (21:00–23:00) until “lights on” (06:00–08:00) using Grass PSG equipment (Grass-Telefactor, West Warwick, RI) and included EEG, electrooculography, electromyography, electrocardiography (ECG) and respiratory measures. PSG recordings were visually scored by RPSGTs in 30-second epochs following standard criteria.^47,48 Out of the 700 recordings at ages 5–12, 646 were analyzable (48 were on paper PSG, 6 were digital but corrupted). All 421 recordings at ages 12–23 were digital and analyzable. There were unavoidable PSG system updates during this long-term study that collected baseline data across 4 years and follow-up data 6–13 years later across 3 years, with differences in the number of EEG channels, referencing method, filter settings, and sampling rates. All of these were accounted for in spectral analyses (Supplement) and controlled for in statistical analyses.⁹ All 1,067 digital PSG records were converted into European data format (EDF).

2.4. NREM SWA

All EDFs were analyzed in a blind manner using sleepFFT (Biosoft Studio, Pennsylvania State University, Hershey, PA).^9,31,49 All EDFs underwent thorough procedures for rejecting epochs with movement artifacts, correcting ECG interference intruding into EEG channels, sorting spectral data according to 30-second visually-scored sleep-staged epochs, and calculation of EEG power during sleep/wake states using the fast Fourier transform (FFT) with correction for rejected epochs (Figure S1). All-night spectral analysis of PSG data was automatically processed with sleepFFT by a graduate assistant blind of subject’s characteristics. Common EEG frequencies range from 0.3–30.0 Hz and activities beyond this range were removed for consistent data processing across all records. SleepFFT used 8-orders of Butterworth band pass filter with a high pass filter at 0.3 Hz and a low pass filter at 30.0 Hz. Each 30-second epoch was applied with 22 overlapping Hann windows lasting 2.56 seconds, with overlaps between windows by approximately half. The FFT was performed on each overlapping window to generate power density data with 0.39 Hz resolution. The resulting data were averaged across these 22 windows as the power spectral data for the epoch. Absolute power within the 0.39–3.91 Hz range was computed by summing the power density data (including the lower and upper limits of the frequency band) and expressed as microvolts squared (μV2). Absolute SWA power was computed for all artifact-free NREM (stages 2, 3 and 4) sleep epochs, adjusted for rejected epochs, and averaged at central (C3/C4) and fronto-occipital (F3-O1/F4-O2) derivations. Please see the Supplement and our prior publications^9,50 for detailed information regarding differences in referencing methods and their reliability and validity.

2.5. Statistical Analyses

Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). To examine age-related trajectories of NREM central and fronto-occipital SWA, we derived a cross-sectional sample spanning from ages 5 to 23 by aggregating independent subjects who contributed with data at ages 5–12 (n=279) and at ages 12–23 (n=421) with no subject represented twice. After excluding participants who were recorded on paper PSG (n=27), whose EDF file was corrupted (n=3), or who had missing spectral data or were outliers (n=6), 664 subjects were included in the analyses, 449 of whom were TD, 123 unmedicated, and 92 medicated. To examine the dependence between age and SWA within each group, multivariable-adjusted linear models regressed subjects’ age against SWA at central and fronto-occipital derivations. SWA was normally distributed. Age was treated as a continuous variable truncated at age 6 and age 21 as only 2 subjects were 5 years old and only 8 were 22 years or older. Given previous study findings of quadratic or cubic SWA trajectories, depending on the age range examined,^4,8,9 we tested these non-linear associations between age and SWA by including those terms in the models. The highest-ordered significant (p<0.05) age term was used as the final model. Population-level means of SWA between age 6 and 21, based on the final model, were plotted to estimate the cross-sectional age-related trajectories, from which, both TD and unmedicated groups showed cubic trajectories. Given prior evidence of SWA increasing since birth, peaking in childhood, and declining during adolescence,^4,8,9 the cubic trajectories can be analyzed as part of a cosine function determined by three parameters: overall mean, difference between maximum and mean values, and time, in our case age, at which the peak is reached. This approach allowed us to reliably estimate the multivariable-adjusted peak age with its 95% confidence interval (95%CI) through a non-linear regression model based on a cosine function. Additionally, piece-wise linear regression analyses were performed to obtain regression coefficients and their standard error (SE) of the age-related SWA slope between peak and nadir age within TD, unmedicated and medicated youth. To explore the impact of each diagnosis on the age-related distribution of SWA, we estimated the age at which peak and nadir SWA was reached among youth with ADHD/LD and ID. Covariates adjusted for in these models included sex, race/ethnicity, BMI, apnea hypopnea index (AHI), periodic limb movement index (PLMI), and PSG system (coded as B1=0, B2=1, F=2 and treated as a nominal factor).

Exploratory analyses examined the magnitude of longitudinal change in central and fronto-occipital SWA from baseline to the observed nadir (age 18) across diagnostic groups. We leveraged a subsample of 114 subjects (51.8% female, 21.9% racial/ethnic minority) who had analyzable longitudinal data from ages 7–12 (median 10.5y) to ages 18–22 (median 19.0y). Of these subjects, 59 were TD, 30 were unmedicated and 25 were medicated. The within-subjects change in SWA by age 18–22 was the dependent variable in these longitudinal analyses and was calculated as a percent change with the formula [(follow-up value–baseline value)/baseline value]*100. General linear models estimated the percent change in SWA by age 18–22 while adjusting for sex, race/ethnicity, BMI, AHI, PLMI, baseline PSG system (coded as B1=0, B2=1 and treated as a binary factor), baseline SWA, and length of follow-up. Differences in mean percent change in SWA were also explored between specific unmedicated and medicated ADHD/LD and ID groups. These longitudinal results were reported as multivariable-adjusted means (95%CI).

3. RESULTS

3.1. Sample Characteristics

Table 1 shows the characteristics of the sample. Overall, unmedicated and medicated youth had worse behavioral and neurocognitive scores than TD youth, with medicated youth reporting worse attention (p=0.033) and thought (p=0.023) problems than unmedicated youth. ADHD, LD and ID were the most prevalent disorders among unmedicated and medicated youth (Table 1). Out of the 162 youth with ADHD/LD, 38.9% were taking stimulants. Out of the 71 youth with ID, 47.9% were taking antidepressants/anxiolytics. PSG parameters were commensurate with the age range of the sample whose total sleep time was 448 minutes, 81.0% of which was spent in NREM sleep and 27.9% in slow wave sleep (SWS). Unmedicated (p=0.050) and medicated (p=0.003) youth both had more awakenings than TD youth, while unmedicated youth spent less time in stage 1 than TD (p=0.025) or medicated youth (p=0.033). Medicated youth had a higher PLMI than TD (p<0.001) and unmedicated youth (p=0.056). There were no significant differences between groups in the proportion of SWS.

Table 1.

Characteristics of the sample

	TD	Unmedicated	Medicated	Pa
N	449	123	92
Female sex	(49.4%)	(40.7%)	(42.4%)	0.146
Racial/ethnic minority	(26.5%)	(23.6%)	(17.4%)	0.173
Age	13.1 (4.5)	14.5 (4.1)	14.3 (3.6)	0.015
BMI percentile	64.2 (29.0)	67.1 (27.6)	62.3 (29.7)	0.581
Disorders
ADHD		(52.0%)	(69.6%)	n/a
LD		(35.8%)	(19.6%)
Other		(34.1%)	(47.8%)
ID		(26.0%)	(42.4%)
ED		(6.5%)	(0.0%)
ASD		(2.4%)	(2.2%)
Medications
Stimulants			(68.5%)	n/a
Other psychoactive			(46.7%)
Sleep			(9.8%)
Behavioral outcomes
Attention problems	53.6 (5.6)	59.0 (8.2)	61.1 (9.8)	<0.001
Thought problems	53.6 (5.1)	57.5 (7.3)	59.5 (8.5)	<0.001
Internalizing symptoms	48.8 (9.5)	55.0 (11.1)	55.8 (12.0)	<0.001
Externalizing behaviors	46.4 (9.4)	53.9 (10.1)	54.8 (11.0)	<0.001
Neurocognitive outcomes
Verbal IQ	104.2 (10.9)	98.4 (12.6)	99.9 (11.0)	<0.001
Non-verbal IQ	106.7 (12.8)	101.1 (13.6)	102.7 (12.6)	<0.001
Coding	10.3 (2.7)	8.8 (2.6)	8.3 (2.9)	<0.001
Digit span backward	6.1 (2.6)	5.5 (2.5)	5.8 (2.4)	0.080
Stroop interference	54.3 (7.1)	53.0 (6.5)	52.4 (6.9)	0.067
Reading achievement	107.5 (10.8)	100.5 (12.7)	102.2 (11.9)	<0.001
Math achievement	104.2 (13.2)	93.3 (13.7)	97.2 (14.4)	<0.001
Polysomnography
Sleep onset latency, min	28.2 (23.1)	24.5 (18.0)	29.4 (34.5)	0.262
Awakenings, #	25.3 (15.6)	28.6 (14.9)	30.8 (20.0)	0.005
Wake after sleep onset, min	60.2 (40.9)	66.5 (41.0)	64.1 (50.1)	0.304
Total sleep time, min	448.3 (52.2)	449.6 (47.9)	447.5 (70.0)	0.955
Sleep efficiency, %	83.7 (9.4)	83.4 (8.7)	83.0 (13.1)	0.832
Stage 1, %	2.1 (2.9)	1.4 (1.7)	2.3 (3.7)	0.049
Stage 2, %	50.8 (11.2)	51.9 (9.7)	51.5 (12.0)	0.596
Slow wave sleep, %	28.1 (10.3)	27.2 (9.6)	27.6 (10.3)	0.670
Rapid eye movement sleep, %	18.9 (5.2)	19.5 (5.2)	19.0 (6.8)	0.617
Apnea/hypopnea index, #/hour	1.7 (4.8)	2.5 (4.5)	2.2 (4.7)	0.216
PLMI, #/hour	2.3 (4.7)	3.0 (6.0)	4.3 (5.8)	0.002

Data are means (standard deviation) and number of cases (percentage) for continuous and categorical/ordinal variables, respectively.

^aBold p-values from Pearson chi-square test for categorical variables and between-subjects analysis of variance for continuous variables are statistically significant (p<0.05).

ADHD=attention-deficit/hyperactivity disorder; ASD=autism spectrum disorder; BMI=body mass index; ED=externalizing disorder; ID=internalizing disorder; IQ=intelligence quotient; LD=learning disorder; PLMI=periodic limb movement index; TD=typically developing.

3.2. Age-related Central SWA Trajectories

The age-related trajectories of central SWA in both TD and unmedicated youth were best fit by cubic terms (Figure 1.A). Unmedicated youth, regardless of diagnosis, reached peak central SWA at age 9.1, 2.3 years later than TD youth (peak at age 6.8, p=0.041), yet the nadir in central SWA occurred at age 18.6 compared to age 19.9 in TD youth (Table 2); thus, unmedicated youth experienced a 37.3% steeper peak-to-nadir decreasing slope in central SWA than TD youth (p=0.043; Figure 1.A). This delay in peak central SWA was found in unmedicated youth with ADHD/LD but not in those with ID, with peaks at ages 9.6 and 6.9 and nadirs at ages 18.8 and 18.9, respectively (Table 2). Consequently, unmedicated youth with ADHD/LD showed a 66.9% steeper peak-to-nadir decreasing slope in central SWA [−27502.3 (4355.2)/y] than TD youth (p=0.029), while that of youth with ID was similar to TD youth (p=0.245). Despite the steeper slope, unmedicated youth regardless of diagnosis [−62.5% (3.4), p=0.308], and those with ADHD/LD [−63.6% (3.8), p=0.329] or ID [−62.6% (3.9), p=0.480], showed a similar proportion of longitudinal change in central SWA by ages 18–22 as TD youth [−58.8% (2.5)].

Figure 1. Cross-sectional trajectories of central (A) and fronto-occipital (B) slow wave activity (SWA) in 664 subjects.

Data points are multivariable-adjusted means and lines are multivariable-adjusted regression curves. Values above the cubic curves are estimated peak age (95%CI). Slopes are estimated from peak to nadir.

Table 2.

Summary of peak and nadir ages for central and fronto-occipital slow wave activity (SWA) observed in the cubic trajectories of typically developing and unmedicated youth, including those with attention/learning disorders and with mood/anxiety disorders

	Central SWA		Fronto-Occipital SWA
	Peak	Nadir	Peak	Nadir
TD	6.8	19.9	8.7	19.2
Unmedicated	9.1	18.6	9.4	18.8
ADHD/LD	9.6	18.8	9.7	19.7
ID	6.9	18.9	8.0	18.9

ADHD=attention-deficit/hyperactivity disorder; ID=internalizing disorder; LD=learning disorder; SWA=slow wave activity; TD=typically developing.

Overall, regardless of medication type (e.g., stimulants, antidepressants/anxiolytics), medicated youth showed a linearly decreasing central SWA trajectory from age 6.0 (highest) to age 21.0 (lowest), with a similar (p=0.734) slope to TD youth (Figure 1.A). Medicated youth [−58.4% (3.6), p=0.930], and those with ADHD/LD taking stimulants [−55.5% (4.7), p=0.389] or ID taking antidepressants/anxiolytics [−61.7% (4.9), p=0.681], showed a similar proportion of longitudinal change in central SWA by ages 18–22 as TD youth.

3.3. Age-related Fronto-occipital SWA Trajectories

The age-related trajectory of fronto-occipital SWA was also best fit by cubic terms in TD and unmedicated youth (Figure 1.B). TD youth reached peak fronto-occipital SWA 1.9 years later than peak central SWA (age 6.8 in Figure 1.A vs. age 8.7 in Figure 1.B). Unmedicated youth reached peak fronto-occipital SWA at a similar age as TD youth (age 9.4 in unmedicated youth vs. age 8.7 in TD youth, p=0.488), with nadirs occurring at ages 18.8 and 19.2, respectively (Table 2); thus, unmedicated youth experienced a similar peak-to-nadir decreasing slope in fronto-occipital SWA (p=0.412) compared to TD youth (Figure 1.B). Peak fronto-occipital SWA in unmedicated youth with ADHD/LD or ID occurred at ages 9.7 and 8.0, with nadirs at ages 19.7 and 18.9, respectively (Table 2). Unmedicated youth regardless of diagnosis [−36.5% (6.5), p=0.196], as well as those with ADHD/LD [−44.4% (5.8), p=0.688] or ID [−37.4% (7.8), p=0.253], showed a similar proportion of longitudinal decline in fronto-occipital SWA by ages 18–22 (nadir) as TD youth [−45.5% (4.8)]. However, when examining unmedicated youth with persistent or incident ID (i.e., excluding those in whom childhood ID had remitted by ages 18–22), they showed a lower longitudinal decline [−27.4% (8.3)] in fronto-occipital SWA than TD youth (p=0.023) or youth with ID taking antidepressants/anxiolytics (p=0.007) by ages 18–22.

Medicated youth, regardless of medication type, showed a linear age-related trajectory in fronto-occipital SWA from age 6.0 (highest) to 21.0 (lowest), with a similar (p=0.525) slope to TD (Figure 1.B). Medicated youth [−48.4% (6.7), p=0.680], as well as those with ADHD/LD taking stimulants [−45.3% (6.8), p=0.826] or ID taking antidepressants/anxiolytics [−51.3% (9.6), p=0.667], showed a similar proportion of longitudinal decline in fronto-occipital SWA by ages 18–22 as TD youth.

4. DISCUSSION

This study provides population-level evidence for a maturational delay and potential topographical disruption of SWA during childhood, a biomarker of reorganization of brain systems, in unmedicated youth with ADHD/LD, but not in those with ID. These unmedicated youth with ADHD/LD also experienced a steeper declining slope in SWA from age 9.6 (peak) to age 18.8 (nadir), yet the proportion of longitudinal change in SWA by age 18 was similar to that of TD youth, suggesting unmedicated youth with ADHD/LD may experience faster, but not necessarily greater, synaptic pruning, as indexed by SWA, during adolescence. In addition, unmedicated youth with ID experienced less longitudinal changes in frontal SWA in the transition to late adolescence, suggesting there may be altered connectivity in the developing frontal cortex. Finally, psychotropic medications appear to alter the age-related trajectories of SWA without impacting its magnitude of decline by the nadir in late adolescence.

It is worth first discussing the age-related trajectory of SWA in TD youth, as it was commensurate with that of brain maturation biomarkers, including synaptic density and cortical gray matter volume, characterized by an increase in childhood followed by a decline with the onset of puberty.^1–4,8–14 We replicated prior study findings, including the topographical shift in SWA predominance^10,11 and decline across adolescence⁸ in a population-based sample. TD children showed a posterior-to-anterior SWA topography^10,11 with a central peak at age 6.8 and a frontal peak 2 years later at age 8.7. They also showed a smooth decline in the transition to adolescence, reaching their nadir by late adolescence with a 45–60% longitudinal loss in frontal and central SWA, commensurate with prior studies.^8,13 Thus, the present study further delineates the age at which peak central and frontal SWA occurs during TD and provides further support for SWA, and its local distribution, signaling brain maturation.⁴

As compared to TD youth, this study provides evidence of an altered maturational trajectory in SWA from childhood through late adolescence in unmedicated youth with neuropsychiatric disorders. Data showed that unmedicated youth, specifically those with ADHD/LD, experienced a peak in central SWA 2–3 years later than TD youth, which coincided with their peak in frontal SWA, suggesting not only a maturational delay in central SWA, but also a potentially altered topographical shift in SWA predominance compared to TD youth. While these findings may be interpreted as a sign of a topographical disruption in the posterior-to-anterior developmental gradient, it is important to note that a higher spatial resolution is needed to confirm and compare these findings to prior high-density EEG (hdEEG) studies.^10,11 Additionally, these unmedicated youth with ADHD/LD showed a 67% steeper decreasing slope in central SWA from their peak (age 9.6) to their nadir (age 18.8) compared to TD youth. These findings are consistent with previous research identifying a delay in cortical development or an inversion point at age 10 in unmedicated youth with ADHD/LD.^20,51,52 Our findings, thus, support that neuro-maturational processes taking place during childhood, including an increase in synaptic density and gray matter volume, may be developmentally altered in youth with ADHD/LD, while those taking place with the onset of puberty, such as synaptic pruning and myelination, may occur at a faster rate during adolescence. This accelerated synaptic pruning may be due to altered cortical thickness and excitatory-inhibitory imbalance, which may be involved in neurocognitive deficits.⁵³ Importantly, while unmedicated youth with ADHD/LD showed a faster SWA decline by age 18, the proportion of longitudinal loss was similar to that of TD youth by the nadir in late adolescence. These data suggest that, although faster synaptic pruning may occur during adolescence, these youth may not experience a greater magnitude of pruning than TD youth. Together with previous studies supporting the SWA topographical shift in TD^10,11 and its purported role in sustaining brain plasticity,⁴ our data suggest a maturational delay and potential topographical disruption in SWA during childhood occurs in unmedicated youth with ADHD/LD, indicative of potentially delayed cortical maturation in these youth.¹⁹

In contrast to those with ADHD/LD, unmedicated youth with ID showed peaks in central (age 6.9) and frontal (age 8.0) SWA similar to those of TD youth. Thus, they did not show a maturational delay or altered topographical shift during childhood, nor a steeper slope in adolescence, in the age-related trajectory of SWA. However, we did find that unmedicated youth with chronic or newly-developed ID experienced less longitudinal changes in frontal SWA in the transition from childhood to late adolescence. While we do not provide cross-sectional evidence of greater frontal SWA in these late adolescents with chronic or newly-developed unmedicated ID,²⁷ our finding of a lower longitudinal decline in frontal SWA is consistent with prior interpretations by Tesler et al.²⁷ based on hdEEG data and by Reynolds et al.⁵⁴ based on neuroimaging data, who suggested that dysfunctional neural connections and greater cortical thickness in the frontal cortex during development may be at play in adolescents with ID.

We also produced important data on the age-related trajectories of SWA in youth taking psychotropic medications, in whom central and frontal SWA linearly decreased with age. Our exploratory longitudinal data showed youth with ADHD/LD taking stimulants and those with ID taking antidepressants/anxiolytics experienced a similar proportion of loss in central and frontal SWA as TD by late adolescence. Together, our cross-sectional and longitudinal data in medicated youth suggest that psychotropic medications may not alter the developmentally-appropriate SWA decline, neither in its speed (peak-to-nadir slope) nor proportion (percent loss). Future studies should examine whether stimulant medication for ADHD may be able to normalize the accelerated slope of decline in SWA observed in unmedicated youth, given prior findings by Furrer et al.¹⁹ In addition, the impact of psychotropic medications on standardized biomarkers of global/state control of NREM sleep, such as the odds ratio product (ORP), should also be further examined in clinical trials, as stimulants appear to be associated with normalized ORP levels in adolescents with ADHD/LD.³⁸

The present findings must be discussed in light of potential limitations. Sleep studies consisted of one-night PSG, which may be affected by the first-night-effect; however, spectral EEG measures have shown greater night-to-night stability than traditional sleep continuity and architecture parameters⁵⁵ and current sleep studies in routine care are conducted with one-night PSG, which assures translation to and replicability in clinical samples and cohorts. Although there were PSG system updates during the study, we accounted for each setting during spectral analysis and controlled for the PSG system used in the statistical analyses to mitigate their potential impact on SWA estimation. We did not have available frontal or occipital contralateral derivations (e.g., F3-M2) in all subjects and, thus, could not better estimate SWA topography, which optimally requires higher spatial resolution as measured by hdEEG in previous studies.^{10,11,19,27,56} Unfortunately, we did not have enough sample size and statistical power to conduct post-hoc analyses based on the observed age-related trajectories to examine mean differences in SWA between the 3–5 diagnostic groups across the 6 potential piece-wise age groups (i.e., 6–8y [pre-peak], 9–10y [at-peak], 11–12y [post-peak], 13–14y [early adolescence], 15–17y [mid-adolescence], 18–21y [late adolescence/early adulthood]). Our longitudinal analyses in 114 subjects had two time-points, which precluded growth curve analyses across multiple follow-ups and lacked specific developmental transitions (e.g., from peak at age 9 to early adolescence at age 14). Although neuropsychiatric diagnoses were based on a structured clinical history and physical examination, they could not be verified via medical records. Nevertheless, the behavioral and neurocognitive data in Table 1 and in the Supplement, as well as our prior reports,^36–38 support the reliability and validity of the diagnostic groups, including those with ADHD/LD and those with ID. Lastly, these longitudinal analyses may be underpowered due to the small sample sizes in the unmedicated and medicated ADHD/LD (n=24 and n=14, respectively) and ID (n=17 and n=11, respectively) groups, being deemed as exploratory.

In conclusion, we replicated previous study findings in TD youth showing age-related SWA cubic trajectories similar to brain maturation biomarkers, characterized by a posterior-to-anterior shift during childhood and a decline in the transition to adolescence. Furthermore, we provide the first population-level evidence that unmedicated youth with ADHD/LD experience a maturational delay and potential topographical disruption in SWA during childhood and a faster decline throughout adolescence, that yet does not lead to a greater proportion of SWA lost by late adolescence. In contrast, unmedicated youth with ID do not experience a maturational delay and instead show a preserved topographical shift in SWA predominance during childhood, but rather they experience less changes in frontal SWA by late adolescence, suggesting altered cortical connectivity in the maturing frontal regions. Lastly, while medicated youth show a linearly decreasing maturational trajectory for central and frontal SWA, there was no evidence of greater developmental loss in the proportion of SWA by late adolescence.

HIGHLIGHTS.

• Replicates, in a general population sample, the topographic shift in SWA power.

• Youth with attention/learning disorders show a maturational disruption in SWA.

• Adolescents with mood/anxiety disorders experience less changes in frontal SWA.

• Medicated youth show a linear decline in SWA from childhood to late adolescence.