Longitudinal Stability of Neural Correlates of Pediatric Attention Deficit Hyperactivity Disorder: A Pilot Study of Event Related Potentials and Electroencephalography

Abstract

Objective.

Stability and developmental effects of electroencephalography (EEG) and event related potential (ERP) correlates of ADHD is understudied. This pilot study examined stability and developmental changes in ERP and EEG metrics of interest.

Methods.

Thirty-seven 7–11-year-old children with ADHD and 15 typically developing (TD) children completed EEG twice, 11–36 months apart. A series of mixed effects linear models were run to examine stability and developmental effects of EEG and ERP metrics.

Results.

Stability and developmental effects of EEG and ERP correlates of ADHD varied considerably across metrics. P3 amplitude was stable over time and showed diverging developmental trajectories across groups. Developmental differences were apparent in error related ERPs and resting aperiodic exponent. Theta-beta ratio was stable over time among all children.

Conclusions.

Developmental trajectories of EEG and ERP correlates of ADHD are candidate diagnostic markers. Replication with larger samples is needed.

Attention deficit hyperactivity disorder (ADHD) is a common neurodevelopmental disorder characterized by developmentally-inappropriate and impairing levels of inattention and/or hyperactivity and impulsivity (American Psychological Association, 2013). ADHD symptoms occur on a continuum, with a wide range of possible ADHD and co-existing psychiatric symptom constellations, which adds complexity to differential diagnosis and treatment of affected individuals (Arnett, Pennington, et al., 2013; Luo, Weibman, Halperin, & Li, 2019). Moreover, because diagnoses rely heavily on subjective reports of behavioral symptoms, children with ADHD often go undiagnosed until they demonstrate functional impairments in academic, behavioral, and/or social domains (Wehmeier, Schacht, & Barkley, 2010; Wilens et al., 2011). In fact, current behaviorally-based assessment tools are not particularly sensitive to ADHD symptoms in young children (Halperin et al., 2020). As such, there has been an increased effort to identify biomarkers of ADHD using electroencephalography (EEG) and event related potential (ERP) methods (Jeste, Frohlich, & Loo, 2015; Lenartowicz & Loo, 2014). Biomarkers for ADHD could not only support diagnostic processes, but could also advance the precision medicine approach to clinical care of ADHD by differentiating among children with ADHD who are likely to experience particular constellations of symptom development, treatment response, psychiatric comorbidity, and functional impairment; and who may have distinct etiological profiles for their symptoms. This prospective clinical knowledge has potential to inform provision of preventative and augmentative interventions for affected children and families. As described by the task force of the World Federation of ADHD, a biomarker is an objective measure of a biological or pathological process, such as brain development. Critically, a fundamental criterion of a biomarker relates to its reliability (or longitudinal stability) across time (Thome et al., 2012).

Regulatory deficit models suggest that vulnerabilities in information processing and cognitive-behavioral regulation account for ADHD symptoms (Shiels & Hawk Jr, 2010). Accordingly, strong neurocognitive biomarker candidates for pediatric ADHD are metrics that reflect the brain’s self-regulation processes, including attention allocation, inhibitory control, and error monitoring. ERP components reflect time-locked neural firing in response to external and internal stimuli. They capture discrete stages of information processing, including attention orienting, stimulus evaluation, response preparation, and self-monitoring – all of which are relevant to cognitive-behavioral dysregulation associated with ADHD (Luck, 2014).

A number of ERP components have been identified as correlates of pediatric ADHD, including the P300 (P3), error related negativity (ERN), and error positivity (Pe). We also consider three EEG-derived indices of cognitive and behavioral regulation: aperiodic spectral slope, individual alpha peak frequency (iAPF), and theta-beta ratio (TBR). To date, the stability of ERP components and EEG indices over time in children with ADHD has not been thoroughly investigated. This study aims to address this knowledge gap with results of a pilot investigation.

Executive processes: P3

The P3 component is associated with novelty detection and attention switching (Friedman, Cycowicz, & Gaeta, 2001); it is typically measured in the central-parietal scalp region and is believed to originate from frontal-temporal-parietal circuitry (Polich, 2007). Among pediatric ADHD samples, the P3 amplitude is consistently attenuated relative to typically developing (TD) children (Johnstone, Barry, & Clarke, 2013; Kaiser et al., 2020). This difference can be corrected with administration of psychostimulant medication (Peisch, Rutter, Wilkinson, & Arnett, 2021), supporting evidence that the neural activation underlying the P3 is supported by availability of dopamine and other catecholamines (Arnett, Rutter, & Stein, 2022; Polich, 2007).

A number of studies have examined developmental effects of the P3 component in TD samples, with findings being somewhat mixed. Stige and colleagues (Stige, Fjell, Smith, Lindgren, & Walhovd, 2007) reported that visual P3 amplitudes were greater in younger as compared to older children. However, a review by van Dinteren and colleagues found that amplitude of the P3 complex increases during childhood (van Dinteren, Arns, Jongsma, & Kessels, 2014). These inconsistencies in the extant literature are likely due to variations in study design, stimulus modality, and participant age ranges. Test-retest reliability of P3 components has likewise been investigated, largely in neurotypical populations, with stability of P3 amplitudes being moderate to strong over spans of one week (Lin, Davies, Stephens, & Gavin, 2020) to 18 months (Brunner et al., 2013) in these samples. A few studies have examined longitudinal stability of P3 components in ADHD samples. A small study of children with ADHD found the P3 amplitude to be moderately reliable over the span of 30 minutes (Kompatsiari, Candrian, & Mueller, 2016). Doehnert and colleagues (Doehnert, Brandeis, Schneider, Drechsler, & Steinhausen, 2013) reported that although ADHD diagnosis did not affect the stability of P3b amplitude from childhood into adulthood, reduced amplitude in the ADHD group was only statistically significantly in childhood, suggesting this component is a less valid correlate of ADHD symptoms in adults.

Error monitoring: ERN & Pe

Behavioral research has shown that children diagnosed with ADHD differ from TD children in error recognition (Shiels & Hawk Jr, 2010) and error rates (van Meel, Heslenfeld, Oosterlaan, & Sergeant, 2007). Accordingly, EEG research has focused on two ERP components that are related to error monitoring: the ERN and Pe (Balogh et al., 2017). The ERN is typically observed 20–100ms following an incorrect response (e.g., omission or commission), regardless of whether the individual is consciously aware of having made a mistake (Michael Falkenstein, Jörg Hoormann, Stefan Christ, & Joachim Hohnsbein, 2000). The ERN is likely generated by the anterior cingulate cortex (ACC) (Holroyd & Coles, 2002). A developmental effect on the ERN has been described, such that the amplitude of the ERN increases across childhood and adolescence (Lo, 2018). The Pe represents a positive deflection that occurs 200–500ms following a consciously-perceived error (Balogh et al., 2017) and is also likely generated by the ACC (Herrmann, Römmler, Ehlis, Heidrich, & Fallgatter, 2004). Similar to the ERN, the amplitude of the Pe appears to increase with age (Moser, 2017). Although research on these error monitoring components is less consistent compared to the body of research on the P3, findings generally suggest reduced ERN and Pe amplitudes in children (Johnstone et al., 2013) and adults (Ehlis, Deppermann, & Fallgatter, 2018a) diagnosed with ADHD compared to non-ADHD control participants. Such attenuated ERP component amplitudes of the error monitoring system could help account for differences at the behavioral level, such as reduced behavioral monitoring and behavioral adjustment (Ehlis, Deppermann, & Fallgatter, 2018b).

Developmental effects on the ERN have been reported such that the amplitude of this component increases with age (Moser, 2017; Tamnes, Walhovd, Torstveit, Sells, & Fjell, 2013), possibly driven by increased consistency of the timing of ERN deflections with increasing age (DuPuis et al., 2015). The effect of age on Pe amplitude is mixed, and may suggest greater developmental change in younger as opposed to older children (Grammer, Carrasco, Gehring, & Morrison, 2014; Wiersema, van der Meere, & Roeyers, 2007). Relatedly, stability reliability of ERN and Pe amplitudes may be lower in children as compared to adults (Davies & Gavin, 2009). However, a study of ERN and Pe stability across the brief span of one week found that ERN and Pe amplitudes were moderately reliable among both neurotypical children (ages 8–12) and adults (Lin et al., 2020). Similarly, a small study of adult men found good short-term test-retest reliability for ERN amplitudes (Segalowitz, Santesso, Murphy, et al., 2010). With respect to long-term reliability, the ERN and Pe were found to be moderately reliable in young children from kindergarten to first grade (Grammer, Gehring, & Morrison, 2018). Moreover, longitudinal change in the ERN was associated with simultaneous change in attention control in this study. We are not aware of any studies that have investigated ERN/Pe stability among ADHD samples.

Aperiodic Spectral Slope

Aperiodic spectral slope, an EEG-derived measure of the 1/f spectral density power distribution (Gao, Peterson, & Voytek, 2017), has been suggested as an endophenotype for cognitive-behavioral control (Gao et al., 2017). Aperiodic spectral slope is thought to reflect the brain’s relative emphasis on spontaneous slower oscillations that support broad spatial and temporal integration, versus spontaneous faster oscillations that permit segregation and local processing of environmental information. Several studies have suggested that the aperiodic slope is atypical in individuals diagnosed with ADHD. For example, Robertson and colleagues (Robertson et al., 2019) reported steeper aperiodic slopes among preschool-age children with ADHD compared to control children. In contrast, others have found flatter slopes among school-age and adolescent ADHD samples (Ostlund, Alperin, Drew, & Karalunas, 2021; Pertermann, Bluschke, Roessner, & Beste, 2019). Developmentally, there appears to be an effect such that the aperiodic slope flattens with increasing age (Voytek et al., 2015). In a previously published study of a small sample that partially overlaps with that of the current analyses, our group (Arnett, Rutter, et al., 2022) found that children classified as methylphenidate medication non-responders showed atypically flat aperiodic slope compared to control children; this same group difference was not found between medication responders and control children. Such results suggest that aperiodic spectral slope may aid in treatment planning, in line with the precision medicine model of care. As such, aperiodic spectral slope, and its potential stability across time, deserves further investigation.

Individual Alpha Peak Frequency

Individual alpha peak frequency (iAPF) is a trait-like EEG-derived index thought to reflect individual differences in cortical development and cognitive performance (Berger, 1934). More specifically, iAPF has been associated with working memory (Clark et al., 2004), processing speed (Klimesch, Doppelmayr, Schimke, & Pachinger, 1996), and reaction times (Surwillo, 1961). iAPF is typically derived from parieto-occipital regions when individuals are awake and at rest (Adrian & Matthews, 1934). With regards to psychometric properties, strong intra-individual test-retest values have been documented (Näpflin, Wildi, & Sarnthein, 2007), suggesting potential utility of iAPF as a biomarker of cortical development.

Developmentally, iAPF increases across development in neurotypical individuals up until adolescence (Chiang, Rennie, Robinson, Van Albada, & Kerr, 2011), ranging from 3–5 hz at birth to 10–12 hz during early adulthood (Marshall, Bar-Haim, & Fox, 2002). iAPF is considered to be both trait-like and heritable (Grandy et al., 2013; Posthuma, Neale, Boomsma, & De Geus, 2001). Although children with ADHD do not consistently differ from TD peers in average iAPF (Lansbergen, Arns, van Dongen-Boomsma, Spronk, & Buitelaar, 2011), individual differences in this metric might reflect clinically relevant neurobiology. Specifically, a series of studies by Arns and colleagues suggests that low iAPF predicts pharmacological treatment resistance in male youth with ADHD (Arns, 2012; Arns, Gunkelman, Breteler, & Spronk, 2008; Voetterl et al., 2023).

Theta Beta Ratio

TBR was first described by Lubar (Lubar, 1991) and reflects the ratio of power between slow-wave theta and fast-wave beta oscillations. It is typically measured at the midline of the scalp during resting conditions (Loo & Arns, 2015). Although the functional significance is still being debated, TBR is believed to reflect cortical activation (Loo & Arns, 2015). A negative association between TBR and ADHD-related behaviors has been documented, including self-reported trait and state attentional capacity (Putman, van Peer, Maimari, & van der Werff, 2010; Putman, Verkuil, Arias-Garcia, Pantazi, & van Schie, 2014; van Son, Angelidis, Hagenaars, van der Does, & Putman, 2018).

Developmentally, TBR decreases with increasing age in neurotypical individuals (Clarke, Barry, McCarthy, & Selikowitz, 2001a). A meta-analysis suggested that children with ADHD showed increased absolute power in the theta band compared to control children (Arns, Conners, & Kraemer, 2013). Similarly, in a meta-analysis of nine studies, Snyder and Hall (Snyder & Hall, 2006) suggested that individuals with ADHD have high TBR compared to controls, although some authors have proposed that this could be accounted for by the overlap between traditional theta frequency ranges and low iAPF in some youth with ADHD (Lansbergen et al., 2011; Saad, Kohn, Clarke, Lagopoulos, & Hermens, 2018). Studies of EEG profiles in youth suggest that only a subset of children with ADHD have high TBR, and this group may have distinct cognitive behavioral profiles (Arnett & Flaherty, 2022; Clarke, Barry, McCarthy, & Selikowitz, 2001b). Further, significant associations between ADHD symptoms and theta power have been documented (Ogrim, Kropotov, & Hestad, 2012). A notable gap in the literature pertains to TBR stability, especially among children with a diagnosis of ADHD.

The Current Study

We examine the extent to which ERP and EEG indices are stable across time in children with and without ADHD. We also examine developmental effects of these neural indices. The ultimate goal is to evaluate the potential utility of these neural correlates to serve as biomarkers for pediatric ADHD.

Methods

Participants

Participants were recruited from a sample of 107 children with a prior ADHD diagnosis and 34 TD children who participated in a larger study investigating neurocognitive correlates of ADHD (Anne B Arnett, M Fearey, V Peisch, & A R Levin, 2022b). Exclusion criteria for the larger study were diagnoses of autism spectrum disorder or intellectual disability, perinatal complications requiring a NICU stay, prenatal exposure to alcohol or drugs, seizure disorder, and color blindness. TD children did not have a diagnosis of ADHD or a reported family history of ADHD. A subset of 42 confirmed ADHD and 16 TD children completed a second laboratory visit 11–36 months after their initial visit (M = 23.20, SD = 5.44). Comparisons between participants who completed the second time point versus those who did not revealed no differences in age, IQ as measured by the 2-subtest Wechsler Abbreviated Scales of Intelligence, Second Edition (Wechsler, 2011), proportions of non-white racial identity, or proportion of females (p’s > 0.394).

ADHD diagnoses were confirmed via caregiver report on the Kiddie Schedule for Affective Disorders and Schizophrenia Computerized Version (K-SADS-Comp) (Townsend et al., 2019) as well as brief clinical interview conducted by a licensed child psychologist when necessary to clarify K-SADS-Comp responses. Five ADHD participants and one TD participant were excluded from the current analyses for the following reasons: identification of epileptiform activity during the EEG (n = 1), change in child or immediate relative’s diagnostic status since the first visit (n = 3), or excessive artifact during the EEG (n = 1. The final sample included 52 participants (ADHD n = 37; TD n = 15). Participant information can be found in Table 1.

Table 1.

Sample Description (at baseline)

	ADHD	TD	difference
N	37	15	-
T1 Age in years M (SD)	9.51 (1.42)	9.22 (1.24)	p = .487
T2 2 Age in years M (SD)	11.48 (1.51)	11.03 (1.50)	p = .334
ΔT1T2 in months M (SD)	23.74 (5.15)	21.85 (6.07)	p = .258
Female	30%	33%	p = .999
Hispanic	8%	20%	p = .461
African American/Black	3%	0%	p = .140
American Indian/Alaskan Native	0%	0%
Asian/Pacific Islander	0%	13%
Multiple Races	16%	13%
T1 Easy Task Accuracy M (SD)	89% (9%)	92% (11%)	p = .370
T2 Easy Task Accuracy M (SD)	89% (9%)	94% (7%)	p = .002
T1 Hard Task Accuracy M (SD)	77% (12%)	85% (11%)	p = .038
T2 Hard Task Accuracy M (SD)	80% (14%)	87% (7%)	p = .059
Full-Scale IQ M (SD)	107.11 (10.06)	117.33 (10.34)	p = .002
Average ADHD Symptom Severity M (SD)	1.30(0.65)	−0.67 (0.63)	p < .001

Note: M = mean; SD = standard deviation. Average ADHD symptom severity was calculated as the mean severity of parent ratings of 18 ADHD symptoms on the SWAN, with a range of −3 to 3; higher scores indicate greater ADHD symptom severity. ΔT1T2 = number of months between first and second time point. Difference tests were independent samples t-tests for continuous variables (age, accuracy, novel trials, IQ); and chi-square tests for categorical variables (female, non-white). T1 = first time point; T2 = second time point. Bolded p-values indicate significance at p <.05. Four ADHD participants performed below the accuracy cutoff (50%) on the Hard ERP task at either the first or second time point; their ERP data for that task was not included in the analyses.

Procedures

All study procedures were approved by the IRB at Boston Children’s Hospital. Informed consent and assent were obtained from participating caregivers and children, respectively. At Time 1 (“T1”), children visited the university laboratory where they completed a 45-minute EEG that included baseline as well as ERP paradigms. During this initial visit, participants also completed 90 minutes of neuropsychological testing, while the child’s caregiver completed behavioral questionnaires and a brief health history. At Time 2 (“T2”), children completed a 90-minute visit to the same laboratory for EEG only. Baseline EEG and the two visual ERP tasks were repeated in the same order at both time points. When applicable, participants abstained from taking stimulant medications for at least 48 hours prior to the visit; and non-stimulant medications for the duration necessary to wash out the drug from their system (determined via consultation with their provider).

EEG Acquisition

Continuous EEG was measured using a 128-channel Magstim-EGI Hydrocel geodesic sensor net and Netstation Acquisition software version 4.5.6 (T1) or 5.4.2 (T2), integrated with a 400-series high impedance amplifier (Magstim-EGI; Plymouth, MN). Electrode impedances were reduced to below 50 kOhms at the start of the session and monitored throughout. EEG signals were referenced to the vertex electrode, analog filtered (0.1 Hz high-pass, 100 Hz elliptical low-pass), amplified and digitized with a sampling rate of 1000 Hz. At Time 1, amplifier offsets were corrected offline during processing; at T2, amplifier offsets were corrected during acquisition using the Anti-Alias filter. ERP tasks were presented with Eprime 2.0 (T1) or Eprime 3.0 (T2). Timing of the presentation of the visual stimuli was recorded using a Cedrus Stimtracker (Cedrus Corporation, San Pedro, CA) at both time points. Participant behavior was monitored by the experimenter via camera and trials in which the child was not attending to the task or moving excessively were coded during acquisition for exclusion.

ERP Tasks

Both ERP tasks were adapted from Jonkman et al. (Jonkman et al., 2000). In the Easy ERP experiment, task-related visual stimuli (targets) were presented alternately with task-irrelevant passive visual oddball stimuli, using a design adapted from experiments previously described by Jonkman and colleagues (Jonkman et al., 1997). Target stimuli were rectangles of four colors (red, blue, green or orange). Oddball task stimuli included a white bracket presented 60% of the time (standard); an identical bracket oriented in the opposite direction presented 20% of the time (deviant); and non-repeated white line drawings of animals and vehicles, presented 20% of the time (novel). The task lasted approximately eight minutes and included up to three practice sets of 10 trials, followed by 140 target and 140 oddball task stimuli, presented with a stimulus duration of 300 ms and interstimulus interval of 0.8–1.4 seconds. Instructions for the Easy task were consistent with a forced-choice discrimination task; participants were told to respond with a right-hand button press to blue rectangles (50%) and a left-hand button press to all other targets. Participants were instructed not to respond to the oddball task stimuli which were presented between each target stimulus. See Figure 1.

Figure 1.

Schematic of the dual-task ERP paradigms.

The Hard ERP experiment was identical to the Easy Task, except that participants were given instructions for a standard 1-back working memory task. Specifically, they were told to press the right button in response to two consecutively presented identical targets (50%), and the left button in response to two incongruent consecutive targets. As with the Easy task, they were instructed not to press any button in response to oddball task stimuli.

EEG Task

All participants completed two resting EEG tasks: “lights on” and “lights off.” Visual stimuli for the lights on condition were presented using E-Prime 2.0 and consisted of six 30-second, silent, abstract color videos (Webb et al., 2018). During the lights off condition, the participant room was reduced to near-total darkness for two minutes. Participants were instructed to sit quietly with their eyes open during both experiments. The lights off condition was selected over an “eyes-closed” paradigm due to young children’s difficulty keeping their eyes shut for extended periods of time, and the muscle artifact that is often introduced by squinting.

EEG Processing

EEG data were processed in MATLAB R2019b using EEGLAB 2021.1 and ERPLAB 8.10 packages. Continuous EEG was downsampled to 250 hz and bandpass filtered at 0.3–80 hz. Electrical line noise in the range of 55–65 hz was removed using the EEGLAB Cleanline plugin. To reduce artifact, eye electrodes and 14 rim channels were excluded from analyses. Additional noisy channels with were automatically detected and subsequently interpolated back into the dataset prior to average referencing, following methods outlined in Gabard-Durnam and colleagues (Gabard-Durnam, Mendez Leal, Wilkinson, & Levin, 2018). To remove remaining artifact (e.g., eye blinks, muscle movement or cardiac signal), we used extended independent component analysis (ICA) with primary component analysis dimension reduction consistent with our prior published research and established pipelines (Arnett, Fearey, et al., 2022b; Levin, Méndez Leal, Gabard-Durnam, & O’Leary, 2018).

ERP Component Extraction.

ERP component time windows and electrode clusters were selected based on previous literature as well as visual examination of ERP waveforms and topographic plots (Figure 2). Mean P3 amplitude was extracted from two scalp regions: parietal (Pz, P3, P4) and occipital (Oz, O1, O2) for task-relevant target and novel oddball stimulus conditions at both time points. Only trials with correct responses were included in the P3 analyses. P3 component time windows differed for each stimulus condition (novel: 300–600 ms; target: 400–600 ms). Mean amplitudes of the response-locked ERN and Pe components were each extracted from a central electrode cluster (Cz, C3, C4) at 0–150 ms and 200–550 ms, respectively. ERN/Pe data included trials for which the participant provided an incorrect response (commission errors) and did not include trials for which no response was provided (omission errors).

Figure 2.

Grand average event related potential (ERP) waveforms by diagnostic group, stimulus condition, experiment, and time point, measured over four midline electrode clusters (Frontal: Fz, F3, F4; Central: Cz, C3, C4; Parietal: Pz, P3, P4; Occipital: Oz, O1, O2). A) Stimulus-locked ERP following presentation of task- irrelevant novel stimuli. B) Stimulus-locked ERP following task-relevant target stimuli. C) Response-locked ERP on trials in which the participant responded incorrectly. T1 = Baseline; T2 = Second time point. ADHD = attention deficit hyperactivity disorder.

Aperiodic Exponent.

The Fitting Oscillations and One-Over-f (FOOOF) MATLAB toolbox (Donoghue, Dominguez, & Voytek, 2020; Donoghue, Haller, et al., 2020) was used to compute the aperiodic exponent across a frequency range of 1–50 hz, at each electrode for each individual. Parameters were selected consistent with our prior work (Anne B Arnett, Margaret Fearey, Virginia Peisch, & April R Levin, 2022a) and the aperiodic (1/f) exponent was extracted for analysis and averaged over anterior frontal (Afz, Af3, Af4), frontal (Fz, F3, F4), central (Cz, C3, C4), parietal (Pz, P3, P4), and occipital (Oz, O3, O4) electrode clusters; values ranged from 0.22 – 2.30.

Peak Alpha Frequency.

Using Welch’s method, spectral power was calculated at each electrode in 0.10 hz increments from 7.0 to 30.0 hz, with a one second hamming window with 50% overlap. Individual alpha peak frequencies were then extracted within the range of 7.0 to 13.0 hz for each individual at midline electrodes (Fz, Cz, Pz, and Oz), separately for lights on and lights off resting experiments. Alpha peaks were defined as the highest amplitude in the alpha range for which the amplitudes at the previous and subsequent two data intervals (i.e., 0.2 hz) were lower. This assured that the alpha peak would reflect a true peak in the spectral distribution, rather than a local maximum.

Theta Beta Ratio.

Welch’s method was used to perform fast Fourier transformation (FFT) on continuous EEG data from 1–55 hz with a 50% overlap and 1-second Hamming window. Theta amplitude was defined as the average power at 4–8 hz and beta as the average amplitude from 12–21 hz. TBR was calculated using absolute power at frontal (Fz, F3, F4) and central (Cz, C3, C4) regions of interest.

Data Analytic Plan.

One T1 ERN and two T1 Pe outlying data points were winsorized to three standard deviations (SDs) from the mean. After this transformation, all ERP and EEG variables demonstrated a normal distribution with skew and kurtosis values < |3|. Participants with < 50% accuracy on the easy or hard task were excluded from those analyses (ADHD exclusions: T1 hard task n = 1; T2 hard task n = 3. There were no TD or easy task exclusions). For error-related analyses, participants with fewer than six error trials were also excluded. See Table 2 for a summary of data points available for each analysis.

Table 2.

Data Points by Task, Component, and Time Point

	TD (n = 15)		ADHD (n = 37)
	Easy Task	Hard Task	Easy Task	Hard Task
T1 P3	15	15	37	36
T2 P3	15	15	36	33
T1 ERN/Pe	8	8	27	26
T2 ERN/Pe	6	7	27	23

To test for stability across T1 and T2 and developmental effects on ERP amplitudes and EEG metrics we ran a series of linear mixed models in R Studio version 2022.12. ERP models were run separately for target and novel oddball stimuli; EEG models were run separately for lights on and lights off experiments. A random intercept for individual was included in all analyses.

In stability analyses, linear mixed models were estimated for each EEG/ERP outcome. T2 measurement was modeled as the dependent variable, with the following independent variables: T1 measurement, diagnostic group, and their interaction; time between T1 and T2 (ΔT1T2), the interaction between T1 measurement and ΔT1T2; and main effects of additional covariates relevant to the EEG/ERP metric of interest (e.g., scalp location). To measure developmental trajectories, linear mixed models were estimated with EEG/ERP as the dependent variable and age (grand mean centered), diagnostic group, and their interaction as independent variables. Additional covariates relevant to the EEG/ERP metric were also included.

Power analyses were conducted with 2000 simulations, using the simr library in R. Our most complicated model, with four main and two interaction fixed effects, one random effect, and the full sample of 52 participants, had 94% power to detect a small continuous main effect (β = 0.20) and 84% power to detect a medium categorical-continuous interaction effect ( β = 0.32). Our model with the fewest data points (error-related ERPs; n=30) had 80% power to detect a large main effect (β = 0.65) and 92% power to detect a large continuous-categorical interaction effect (β = 0.50). Given these latter analyses were clearly underpowered, we report 95% confidence intervals rather than p values for error-related ERP results.

We did not have sufficient statistical power to additionally examine potential main and moderating effects of sex on stability and development of the EEG/ERP metrics. However, given previous literature indicating differences in patterns of cortical development among males and females {Kaczkurkin, 2019 #2050}, we conducted exploratory analyses with a series of independent-sample t-tests to examine sex differences in target variables. We limited our analyses to the T2 time point and the central scalp region (when applicable) to minimize the number of tests. P values were adjusted for multiple comparisons using the Bonferroni method.

Results

Stability

P3 Amplitude.

Target Stimuli.

We tested a model with target stimulus P3 amplitude at T2 as the dependent variable. Independent variables were target stimulus P3 amplitude at T1, diagnostic group, group by T1 interaction, ΔT1T2, ΔT1T2 x T1, scalp region (parietal versus occipital), and experiment (lights on versus lights off). T1 target P3 amplitude predicted the same at T2 (B = 1.14, SE = 0.31, p = .0003), indicating longitudinal stability. The main effect of diagnostic group on T2 amplitude was not statistically significant (B = 0.54, SE = 0.77, p = .487). However, there was an interaction such that the association between T1 and T2 P3 amplitude was stronger for the TD compared to the ADHD group (B = 0.32, SE = 0.15, p = .031). T2 target P3 amplitude was greater in the parietal region relative to the occipital region (B = 0.79, SE = 0.28, p = .0064) and with longer ΔT1T2 (B = 0.20, SE = 0.07, p = .0047). Additionally, the interaction between T1 P3 amplitude and ΔT1T2 indicated reduced longitudinal stability with greater time between time points (B = −0.03, SE = 0.01, p = .0064). There was no effect of experiment on T2 target P3 (B = −0.16, SE = 0.29, p = .5788).

Novel Stimuli.

The main association between T1 and T2 novel P3 amplitude did not reach statistical significance (B = −0.33, SE = 0.18, p = .0725), nor did the main effect of diagnostic group (B = −0.61, SE = 0.98, p = .535). However, as with target P3 amplitude, the association between novel P3 amplitudes across time points was stronger in the TD group (B = 0.43, SE = 0.10, p < .0001). There was no main effect of ΔT1T2 on T2 novel P3 amplitude, (B = −0.14, SE = 0.08, p = .0660). Surprisingly, there was an interaction effect indicating greater stability of P3 amplitude across T1 and T2 with longer ΔT1T2 (B = 0.02, SE = 0.01, p =.001). T2 novel P3 amplitude was smaller in the parietal relative to occipital region (B = −1.05, SE = 0.38, p = .0071), and during the hard experiment (B = −0.83, SE = 0.23, p = .0111).

ERN Amplitude.

To test for ERN amplitude stability across time points, T2 ERN was modeled as the independent variable, with T1 ERN, diagnostic group, T1 x group, ΔT1T2, T1 x ΔT1T2, and experiment as independent variables. The main association between T1 and T2 ERN amplitudes was positive but close to zero (B = 0.06, SE = 1.05, 95% CI: −1.87 – 2.00). Unexpectedly, T2 ERN amplitude was attenuated in the TD compared to ADHD group (B = 0.81, SE = 0.68, 95% CI: −0.45 – 2.08). ERN amplitude was attenuated during the hard relative to easy task (B = 0.63, SE = 0.36, 95% CI: −0.11 – 1.29). The main effect of ΔT1T2 and the interaction effects were all close to zero (range: −0.04 – 0.004).

Pe Amplitude.

To test for Pe amplitude stability across time points, T2 Pe was modeled as the independent variable, with T1 Pe, diagnostic group, T1 x group, ΔT1T2, T1 x ΔT1T2, and experiment as independent variables. The association between T1 and T2 Pe was positive but small (B = 0.04, SE = 0.49, 95% CI: −0.86 – 0.94). The TD group had greater T2 Pe amplitude than the ADHD group (B = 1.07, SE = 0.58, 95% CI: 0.01 – 2.15). Additionally, T2 Pe amplitude was attenuated during the hard experiment (B = −1.33, SE = 0.34, 95% CI: −1.96 - −0.65), and enhanced with longer time between measurements (B = 0.14, SE = 0.05, 95% CI: 0.06 – 0.23). The T1 x group and T1 x ΔT1T2 interaction terms were close to zero (−0.05 and 0.009, respectively).

Aperiodic Exponent.

Stability of aperiodic exponent was evaluated separately within lights on and lights off experiments, with T2 exponent as the dependent variable, and T1 exponent, diagnostic group, T1 x group, ΔT1T2, T1 x ΔT1T2, and region (anterior frontal, frontal, central, parietal, occipital) as independent variables. Results were replicated across experiments. Specifically, the T2 aperiodic exponent was reduced in the TD versus ADHD group (lights off B = −0.71, SE = 0.27, p = .0102; lights on B = −0.87, SE = 0.26, p = .0009). The main effect of T1 on T2 aperiodic exponent was not statistically significant (lights off B = −0.30, SE = 0.34, p = .3785; lights on B = −0.29, SE = 0.32, p = .3768). There was an interaction between T1 exponent and diagnostic group wherein the stability of aperiodic exponent across time points was stronger among TDs (lights off B = 0.35, SE = 0.16, p = .026; lights on B = 0.45, SE = 0.15, p = .0029). The aperiodic exponent was lower (i.e., flatter slope) in the occipital relative to anterior frontal region (p’s <.003). There were no statistically significant main effect of ΔT1T2 (p’s > .1395), nor was there an interaction between T1 exponent and ΔT1T2 (p’s > .1446).

Individual Alpha Peak Frequency.

Stability of iAPF across time points was estimated using mixed linear models with electrode as a covariate, separately for lights on and lights off resting experiments. During the lights off experiment, none of the main or interaction effects were statistically significant (p’s > .0972). During the lights on experiment, T1 and T1 alpha peak frequency were significantly associated (B = 0.74, SE = 0.29, p = .0112), and the TD group had lower peak alpha frequency than the ADHD group at T2 (B = −3.85, SE = 1.52, p = .0124). There were interaction effects wherein the stability of alpha peak frequency across time points was stronger for the TD group (B = 0.40, SE = 0.16, p = .0131) and trended toward being slightly weaker with greater time between measurement (B = −0.02, se = 0.01, p = .0600). There were no effects of electrode (p’s > .2751).

Theta Beta Ratio.

Stability of TBR across time points was estimated separately for lights off and lights on experiments, with region as a covariate (central versus frontal). In both experiments, TBR at T1 predicted the same at T2 (lights on B = 0.48, SE = 0.22, p = .0317; lights off B = 0.87, SE = 0.47, p = .0666). The main and interaction effects of diagnostic group, ΔT1T2, and region were not statistically significant for either experiment (p’s > .1741).

Developmental Effects

P3 Amplitude.

Target Stimuli.

In the developmental mixed linear models, we tested for effects of age on ERP amplitude. On average, there was a small negative main effect of age on target P3 amplitude (B = −0.02, SE = 0.01, p = .0434). However, TDs showed a positive association between age and target P3 amplitude compared to the ADHD group (B = 0.06, SE = 0.02, p = .0019). Target P3 amplitude was lower during the hard relative to the easy experiment (B = −0.70, SE = 0.23, p = .0026). There were no main effects of diagnostic group (B = 0.35, SE = 0.79, p = .6602) or region (B = −0.003, SE = 0.23, p = .9886) on target P3 amplitude. The interaction effects indicated more positive linear growth in P3 over age in the hard as compared to the easy experiment (B = 0.02, SE = 0.01, p = .0260), and in the parietal as opposed to the occipital region (B = −0.03, SE = 0.01, p = .0071).

Novel Stimuli.

Across the entire sample, there was a negative effect of age on novel P3 amplitude (B = −0.04, SE = 0.02, p < .0241); however, as with the target P3, the effect of age on novel P3 amplitude was positive for the TD group (B = 0.06, SE = 0.03, p = .0133). The TD group had higher novel P3 amplitude than the ADHD group, on average (B = 2.19, SE = 0.91, p = .0207). Novel P3 amplitude was lower in the parietal region than the occipital region (B = −3.84, SE = 0.31, p < .00001) and for the hard as compared to the easy experiment (B = −1.66, SE = 0.31, p < .00001). Finally, the interaction effects indicated that novel P3 amplitude increased with age more in the parietal than occipital region (B = 0.05, SE = 0.01, p = .0009). There was no interaction effect between age and experiment on novel P3 amplitude (B = 0.02, SE = 0.01, p = .1878).

ERN Amplitude.

The main effect of age on ERN amplitude was negative, indicating greater ERN amplitude with development (B = −0.01, SE = 0.01, 95% CI: −0.03 – 0.002). The TD group had lower ERN amplitude than the ADHD group, although this effect was close to zero (B = −0.18, SE = 0.27, 95% CI: −0.70 – 0.34). The age x diagnostic group interaction indicated that the developmental effect on ERN was weaker in the TD relative to ADHD group (B = 0.06, SE = 0.02, 95% CI: 0.03 – 0.10). As expected, ERN amplitude was weaker (more positive) in the hard task (B = 0.68, SE = 0.20, 95% CI: 0.29 – 1.07). The interaction between task and age was close to zero (B = 0.004, SE = 0.01, 95% CI: −0.02 – 0.02).

Pe Amplitude.

Pe amplitude increased with age (B = 0.03, SE = 0.01, 95% CI: 0.006 – 0.06). The TD group had higher Pe amplitude on average (B = 3.55, SE = 3.56, 95% CI: −3.31 – 10.46). The interaction between group and age suggested developmental effects on the Pe were comparable across ADHD and TD children (B = −0.02, SE = 0.03, 95% CI: −0.08 – 0.03). Pe amplitude was lower during the hard experiment (B = −1.60, SE = 0.30, 95% CI: −2.18 – −1.01). The interaction between age and experiment was close to zero (B = −0.01, SE = 0.02, 95% CI: −0.04 – 0.02).

Aperiodic Spectral Slope.

The effect of age on aperiodic slope was not significant in either experiment. However, there was an interaction between diagnostic group and age indicating that the TD participants showed a greater decline in aperiodic exponent over age (lights off B = −0.005, SE = 0.001, p < .0001; lights on B = −0.005, SE = −0.001, p = .0005). During the lights on experiment, there was a main effect of diagnostic group wherein the TD group had a steeper average slope than the ADHD group (B = 0.53, SE = 0.17, p = .0024), but this was not true during the lights off experiment (B = −0.05, SE = 0.04, p = .2660). As in the previous analyses, the aperiodic slope was flatter in the occipital scalp region for both experiments (lights off B = −0.08, SE = 0.03, p = .0020; lights on B = −0.08, SE = 0.02, p = .000).

Individual Alpha Peak Frequency.

During both experiments, the effect of age on iAPF was non-significant, although there was a trend toward increased iAPF with age during the lights off experiment (lights on B = 0.003, SE = 0.004, p = .4988; lights off B = 0.1, SE = 0.003, p = .0846). The main effects of group were not significant in either experiment (p’s > .2867), nor were the interactions between age and group (p’s > .5043). There was variability in iAPF among electrodes during the lights on (p = .0201) but not lights off (p = .8945) experiment.

Theta Beta Ratio.

There was no effect of age on TBR during either experiment (lights on B = −0.01, SE = 0.01, p = .2408; lights off B = −0.01, SE = 0.01, p = .1179). The main effects of group and region were likewise non-significant (p’s > .137), as were the interactions between group and age (p’s > .083).

Sex Differences.

After correction for multiple comparisons, sex differences were found in aperiodic exponent and TBR at T2. Specifically, males had steeper aperiodic exponent during the lights off (t[50] = 3.83, p = .004) and lights on (t[30] = 3.30, p = .024) resting experiments. Likewise, males had higher TBR at T2 compared to females during the lights on experiment (t[50] = 3.01, p = .038). All other comparisons did not reach statistical significance (p > .05).

Discussion

A core clinical feature of pediatric ADHD is cognitive-behavioral dysregulation (Shiels & Hawk Jr, 2010). This pilot study examined the stability of neurophysiological indices that have been previously associated with cognitive-behavioral dysregulation and pediatric ADHD. Such knowledge is critical to advancing etiological models and developing individualized clinical care for affected individuals. We examined the longitudinal stability of three ERP (P3, ERN, Pe) and EEG-derived indices (aperiodic spectral slope, individual alpha peak frequency, and theta-beta ratio) in a pilot sample of youth with and without ADHD. Specific study aims were to examine the 1) reliability (i.e., longitudinal stability) and 2) developmental effects of these EEG and ERP correlates of ADHD in a sample of school-age children.

ERP Indices

The P3 component is a cortical index related to executive processes (novelty detection; attention switching) (Friedman et al., 2001) and as such, has featured prominently in the pediatric ADHD literature. Prior studies have repeatedly found reduced P3 amplitudes during continuous performance and cognitive tasks (Barry, Johnstone, & Clarke, 2003; Johnstone et al., 2013; Moavero et al., 2020). However, our results suggested that this effect is moderated by age, with greater differences in later childhood. Additionally, we found that P3 amplitudes to task-relevant target stimuli were stable across 1–3 years among both ADHD and TD participants, although stability was stronger for the TD group. In contrast, P3 amplitude to novel stimuli was only stable among the TD group, after controlling for the effects of experiment, region, and time between EEG visits.

Reduced stability in P3 amplitudes among the ADHD group could be explained by greater trial-by-trial variability (i.e., “neural noise”) in P3 latency among children with ADHD (Arnett et al., 2023; Lazzaro et al., 1997). Additionally, Figure 2 shows significant heterogeneity in the linear associations between T1 and T2 novel P3 amplitudes, particularly among ADHD participants. Neurobiological heterogeneity in ADHD samples has been posited to relate to phenotypic heterogeneity in symptom presentation, cognitive profiles, and coexisting psychiatric symptoms (Arnett & Flaherty, 2022; Arnett, McGrath, Flaherty, Pennington, & Willcutt, 2021).

We found contrasting developmental effects of P3 amplitudes between TD and ADHD participants. Specifically, while the TD group showed greater target and novel P3 amplitudes with older age, the ADHD group showed the opposite trend (i.e., decreasing amplitude with age), across both experiments. Inspection of Figure 3 suggests that greater P3 amplitude in the TD group is not reliable until after age 8. This is consistent with prior research in which 5–7 year old children with ADHD did not show atypical P3 amplitudes on trials for which they provided a correct response (Spronk, Jonkman, & Kemner, 2008). This has important implications for consideration of P3 amplitude as a biomarker for ADHD. Given the neurobiological heterogeneity in this population, it may not be possible to establish age-based norms in P3 amplitude that would reliably distinguish a child with ADHD from a neurotypical child of the same age. Rather, developmental trajectories in brain and behavior (Arnett, MacDonald, & Pennington, 2013) may be more sensitive and specific objective markers for children with ADHD.

Figure 3.

Associations between T1 and T2 ERP amplitudes, by experiment, scalp region, and diagnostic group. A) Task-irrelevant novel stimulus-locked P3 amplitudes. B) Task-relevant target stimulus-locked P3 amplitudes. Note: these plots do not account for the effects of ΔT1T2.

An important consideration is that attenuated P3 amplitudes are not specific to ADHD; rather, this atypical ERP has been noted across multiple disorders, including schizophrenia, learning disorders, and depression (Jeon & Polich, 2003; Santopetro, Kallen, Threadgill, & Hajcak, 2020; Taroyan, Nicolson, & Fawcett, 2007). On the one hand, this positions the P3 component as a potential cross-diagnostic biomarker of neurodevelopmental risk. However, it will be useful to determine whether there are diagnostic group differences in the developmental trajectories of P3 amplitudes and the types of tasks eliciting reduced P3 amplitude, as this could provide important insight into diverging neurodevelopmental mechanisms across disorders.

Error related ERP analyses were underpowered and thus results are interpreted cautiously. In this pilot sample, the amplitudes of error-related ERPs (ERN, Pe) increased with age, and did not retain rank order stability over childhood. This finding was inconsistent with a meta-analytic review that found no significant association between age and Pe amplitude in childhood (Boen, Quintana, Ladouceur, & Tamnes, 2022). Unexpectedly, the developmental effect was enhanced in the ADHD group, particularly for ERN amplitudes. If replicated in a larger sample, this interaction would be consistent with the theory that ADHD reflects delayed development of self-regulatory cortical networks.

Notably, the ERN is influenced by several external and internal factors that could influence our results. First, the ERN amplitude is thought to depend on subconscious recognition of the error, and ERN amplitude is enhanced when errors are infrequent (Fischer, Klein, & Ullsperger, 2017). Although the ERN is expected to increase with age, inattention among children with ADHD may worsen, which would lead to reduced error monitoring (Larsson, Dilshad, Lichtenstein, & Barker, 2011). Additionally, anxiety is associated with greater ERN in school aged children (Moser, 2017), and we expect that our participants were more comfortable in the research setting on the second visit due to increased familiarity. Thus, there are competing influences on the ERN from the first to second time period that could contribute to low stability.

Unlike the ERN, the Pe has previously been linked to conscious awareness of errors (Balogh et al., 2017; M. Falkenstein, J. Hoormann, S. Christ, & J. Hohnsbein, 2000). This likely explains our finding that Pe amplitude was reduced during the task with greater cognitive load. The majority of prior studies on the Pe have utilized Go/No-Go tasks, where the errors are primarily commissions and thus more likely to be perceived at the conscious level. In contrast, a response was expected for every target stimulus in our tasks. Thus, errors may have been more difficult to recognize with conscious awareness.

EEG Indices

The aperiodic exponent has been examined in relation to pediatric ADHD because it is thought to reflect cortical functioning. After statistically controlling for region and time between measurements, our results suggested that the aperiodic exponent was more stable for the TD group than for the ADHD group. There was also a developmental effect for the aperiodic exponent such that neurotypical children showed an increasingly flatter slope with age, which aligns with prior research (Voytek et al., 2015). The stability of the aperiodic exponent across both groups, in combination with stable group differences in the lights on experiment, support the potential utility of the aperiodic slope as a biomarker for pediatric ADHD. However, given the strong developmental effect on this metric, it will be important to standardize values according to age.

Individual alpha peak frequency (iAPF) is also relevant to pediatric ADHD as it is thought to capture individual differences in cortical development and cognitive performance (Berger, 1934; Segalowitz, Santesso, & Jetha, 2010). Regarding reliability, we found that the iAPF was stable during both experiments, consistent with prior research on adults which reported strong intra-individual test-retest values (Näpflin et al., 2007). As with the aperiodic exponent, iAPF was more stable for the TD group compared to the ADHD group during the lights on experiment, perhaps indicating intra-individual variability in cortical activity at rest among children with ADHD. In our sample, we were not able to document a clear developmental effect on iAPF. As was described above, prior research has suggested an increase in iAPF from infancy to early adulthood (Marshall et al., 2002). Given that the rate of change in iAPF slows down (i.e., asymptotes) in later childhood, for the developmental effect may be more pronounced in younger age groups.

The third EEG metric, TBR, is thought to index cortical activation (Loo & Arns, 2015), although its exact role in cognition is still being debated among scholars. At the behavioral level, there appears to be a negative association between TBR and attentional capacity (Putman et al., 2010; Putman et al., 2014; van Son et al., 2018). Our results suggested stability in the TBR across time points for both groups. Of note, the ADHD and TD groups did not differ significantly from one another with regards to TBR. This finding is consistent with a meta-analysis revealing significant heterogeneity in the TBR among ADHD as well as TD groups (Arns et al., 2013). In another paper by our group (Arnett & Flaherty, 2022), we found that elevated TBR may be found only in a subset of children with ADHD who have a particular cognitive profile. Regarding developmental effects, our pilot data did not support the notion that TBR changes with development, which is in contrast to prior research reporting a decrease in TBR with increasing age in neurotypical individuals (Clarke et al., 2001a).

Sex effects, though not examined in the full multilevel models, emerged specifically for aperiodic exponent and TBR. On both metrics, males showed immature values compared to females at T2, suggesting delayed or atypical development of cortical generators for these measures. Given comparable sex distributions within ADHD and TD groups, these findings are not likely to explain the diagnostic group effects on aperiodic exponent. Consistent with previous literature, our results indicate that sex differences{Kaczkurkin, 2019 #2050} may be specific to certain regions of the cortex rather than indicating globally immature cortical development among males.

Limitations & Future Directions

There are study limitations, which warrant attention. First, our sample size was relatively small and participant demographics were not fully representative of the overall ADHD population; thus, we consider this a pilot study. Relatedly, participants in this study made few errors on the easy task, limiting available data for this experiment. Task characteristics for the hard task, including the strong cognitive load due to the working memory component, may have altered error recognition. Prior research has successfully utilized the Go/No-Go and Flanker tasks in pediatric error monitoring research (Shiels & Hawk Jr, 2010) and the stability of error related ERPs should be examined in a pediatric ADHD sample using those tasks. Further, although we present longitudinal data in this study, each participant only contributed two data points; thus, developmental analyses are largely cross-sectional. Given that many of the ERP and EEG metrics appeared to undergo normative developmental change, our data were not able to fully capture nuances in the shape of this development at the individual level. This highlights the need to collect data across longer periods of time—ideally from early childhood through adolescence—to understand cortical maturation and establish age-based norms in a demographically representative sample.

Ideally, age-normative data for ERP amplitudes would eventually become available for use as clinical indicators of individual differences in ADHD etiology, treatment/intervention needs, symptom course, and risks and protective factors. This approach will require large and representative data sets. In this vein, Imburgio and colleagues (Imburgio et al., 2020) recently published norms for the ERN and Pe, elicited by the arrow Flanker task, to classify neural responses in young adults. Although some scholars have cautioned against the creation of normative data for ERP components (Clayborne, Varin, & Colman, 2019), normative data for neurotypical and clinical samples could refine diagnostic processes while also informing treatment planning. Such data collection should include multiple timepoints with consistent measurement intervals. As has been the case for other clinical populations (e.g., pediatric anxiety; Meyer, 2022), neural markers of psychopathology could eventually be tracked and targeted through clinical interventions. Longitudinal data collection could also help determine whether differences between ADHD and TD groups represent a developmental delay or a more lasting developmental deviation (Doehnert, Brandeis, Imhof, Drechsler, & Steinhausen, 2010; Doehnert et al., 2013).

Figure 4.

Linear associations between EEG measures at T1 and T2 in ADHD (red) and TD (black) children. A) aperiodic slope exponent, b) individual alpha peak frequency, and c) theta beta ratio. Scalp region abbreviations: AF=anterior frontal; FR=frontal; CE=central; PR=parietal; OC=occipital. Fz, Cz, Pz, and Oz are midline electrodes at frontal, central, parietal and occipital scalp regions, respectively.

Figure 5.

Developmental effects of age on a) P3 amplitude to novel stimuli, b) P3 amplitude to target stimuli, c) Pe, and d) ERN.

Figure 6.

Developmental change in a) aperiodic slope exponent, b) individual alpha peak frequency, and c) theta beta ratio among children with ADHD (red) and TD children (black). Scalp region abbreviations: AF=anterior frontal; FR=frontal; CE=central; PR=parietal; OC=occipital. Fz, Cz, Pz, and Oz are midline electrodes at frontal, central, parietal and occipital scalp regions, respectively.

Biographies

Virginia Peisch, Ph.D. is an instructor of pediatrics at Harvard Medical School and an attending clinical psychologist in the Division of Developmental Medicine at Boston Children’s Hospital. Her research interests include improving intervention approaches to ADHD and co-existing psychiatric symptoms.

Anne B. Arnett, Ph.D. is an assistant professor of pediatrics at Harvard Medical School and a PI in the Laboratories of Cognitive Neuroscience in the Division of Developmental Medicine at Boston Children’s Hospital. Her research aims to leverage the neurobiological, genetic and phenotypic heterogeneity in ADHD to develop precision medicine care guidelines for affected children and families.