Background
Attention-Deficit/Hyperactivity Disorder (ADHD) is an impairing neurodevelopmental disorder that remains inadequately understood. Along with the observable behavioral symptoms of inattention and hyperactivity/impulsivity, there is robust evidence of structural, functional, and neurochemical brain differences in ADHD [
1‐
3] particularly in regions involved in vital executive functions (EFs) that regulate the ability to identify, extract, and interpret what is relevant for executing the correct response, as well as monitoring, inhibiting, and changing the prepotent response as needed [
4,
5]. The pathophysiology of ADHD remains unclear, although converging evidence suggests that alterations in brain structure, function, and physiology likely arise from an interaction of genetic and environmental causes and experience [
5‐
8]. For example, structurally, prominent volumetric decreases are evident in the posterior-inferior lobules of the cerebellar vermis in both male and female children with ADHD [
9‐
12]. There are decreases in prefrontal volume, particularly the right prefrontal cortex [
9,
13]. Also reported are regional differences in cerebral blood flow in the cerebellum, striatum [
14] and prefrontal cortex (PFC) [
15]. Moreover, differences in baseline oscillatory activity between those with ADHD and controls have been observed in frontal regions, particularly the PFC [
16,
17]. Consistent with the neuroimaging findings, psychological research indicates clearly that subtle but impairing problems in EFs are correlates of ADHD, regardless of gender or age [
18].
While the majority of ADHD research focuses on deficits in EF, it is apparent that not all individuals with ADHD have EF deficits [
18,
19] and that not all neuropsychological difficulties can be explained by EF theory alone [
20]. Moreover, EF tasks in which individuals with ADHD do show deficits often include processing and responding to simple sensory stimuli that vary in predictability. This suggests that deficits in anticipatory or perceptual processing of simple stimuli could also contribute to impairments on tasks that assess higher-order functions. Accordingly, an important goal of ADHD research is to address not only the concept of multiple
forms of impairment but also of multiple
sources of impairment. Emerging evidence not only shows abnormalities in neural regions governing higher order function but also in regions governing basic function such as somatosensory cortex [
21‐
24], motor cortex [
21,
25,
26] and visual cortex [
27]. Although people with ADHD have shown behavioural deficits in responding to simple stimuli during sensorimotor tasks [
28‐
30], methodological shortcomings in the limited studies available have precluded an adequate understanding of the role of neural networks in processing predictable and non-predictable stimuli in ADHD. Specifically, existing studies have relied almost exclusively on behavioural measures (i.e., accuracy, reaction time), which cannot assess moment-by-moment activities that are driving these processes on the order of milliseconds.
Our aim was to examine basic sensory processing of predictable and non-predictable stimuli in those with ADHD using magnetoencephalography (MEG), a non-invasive functional neuroimaging technique that records neural activity on the order of a millisecond. This high temporal resolution combined with novel source reconstruction techniques capable of mm spatial resolution makes MEG an optimal technique for capturing spatial and temporal information during sensory processing for which the time scale is on the order of milliseconds. MEG studies of the human somatosensory system using median nerve stimulation have shown that only the contralateral primary somatosensory cortex (SI) responds to unilateral tactile information whereas bilateral secondary somatosensory cortices (SII) show activity in response to unilateral stimulation [
31,
32]. The earliest somatosensory activity occurs at approximately 20 ms post-stimulation in SI just caudal to the central sulcus in the corresponding topographical location. Subsequent somatosensory activation occurs in the bilateral parietal opercula located in the dorsal regions of the lateral sulci [
32‐
34]. Source activity in SI and SII, following median nerve stimulation, is composed of both alpha and beta cortical rhythms [
35]. In association with MEG, median nerve stimulation has been used to examine evoked responses to somatosensory stimuli in order to examine somatosensory cortical function [
31,
36,
37] and ascending pathways from the peripheral receptors to the spinal cord, brainstem, thalamus, and cortex [
38]. This technique has also been used to examine physical and cognitive impairments in individuals with Alzheimer's [
39], stroke patients [
40], and infantile autism [
41], for example. Using MEG, we investigated the oscillatory changes during somatosensory activation in adults with and without ADHD.
The general assumption of cortical oscillations is that populations of neurons exist in varying states of synchrony as they respond to externally or internally generated events. Event-related desynchrony (ERD) and event-related synchrony (ERS) phenomena are thought to represent decreases and increases, respectively, in synchronization within a specific frequency range in relation to an event [
42]. Previous MEG studies of cortical activity following median nerve stimulation in healthy adults report brief suppression of mu (an alpha wave variant oscillating at approximately 10 Hz) and beta (15–30 Hz) cortical activity in primary and secondary somatosensory cortex (ERD) followed by a marked increase in beta band activity above baseline (late-ERS, known as
beta rebound) [
42]. Basic or complex sensory processing requires a dynamic interaction between groups of neurons oscillating at particular frequencies and differing degrees of coupling. Oscillations in the alpha and beta bands are of particular interest in ADHD research as these frequencies are thought to mediate perception [
43,
44] and attention [
45‐
47]. To our knowledge, MEG has not yet been used to investigate changes in SI alpha or beta oscillations in individuals with ADHD. Accordingly, our aim was to characterize ERD and ERS in the alpha and beta bands in SI and SII in response to randomly and predictably presented electrical stimulation of the median nerve in adults with and without ADHD. Comparison of random versus predicted median nerve stimulation is a novel approach to determine whether basic somatosensory processing differs between those with ADHD and healthy controls and if stimulus predictability differentially influences somatosensory processing in those with ADHD compared to controls.
The neural basis of predictive responding to the absence of a stimulus in both SI and SII will be described in a subsequent report.
Methods
Participants
We studied nine adults (4 females/5 males) with a diagnosis of ADHD (mean age: 34.6 +/- 3.28 years) and ten healthy age-matched controls (4 females/6 males; mean age of 34.13 +/- 4.6 years). All were right-handed. Adults with ADHD were recruited from an outpatient neuropsychiatry clinic in a mental health centre in a large metropolitan city. All had completed the same comprehensive clinical diagnostic assessment including: a clinical diagnostic interview and various self-report rating scales including the Conners Scales [
48]; Wender Utah Rating Scale [
49], Brown Attention Deficit Disorder Scales (Brown, 1996); and Adult Self-Report Scale [
50]. Healthy adult volunteers were recruited by means of advertisements placed in the same institution and in other community organizations. All participants completed a telephone-based Intake Screening Questionnaire (screens for psychopathology and education level) and the Adult ADHD Self-Report Scale [
51] at the time of participation to estimate current levels of ADHD symptomatology. Participants were excluded if they wore orthodontic braces, had any non-removable metal, or had a diagnosis of psychosis, neurological disorder, or uncorrected sensory impairments.
MEG Recordings
A whole-head 151 channel MEG system (VSM MedTech Ltd, Vancouver, Canada) was used to measure somatosensory evoked fields. Participants lay in a supine position in a magnetically shielded room with their head resting in the MEG helmet. The MEG signals were bandpass filtered at 0.3 – 300 Hz and recorded at a 1250 Hz sampling rate. Head position in relation to the MEG sensors was determined by measuring the magnetic field generated by 3 fiducial reference coils just before and after each experimental session. T1-weighted structural magnetic resonance images (MRI) (axial 3D spoiled gradient echo sequence) were obtained for each participant using a 1.5 Tesla Signal Advantage system (GE Medical Systems, Milwaukee, USA). During MRI data acquisition, 3 radiographic markers were positioned on the same anatomical landmarks as the fiduciary coils to allow coregistration accuracy of the MEG and MRI data. Single equivalent current dipole (ECD) models were also fit to the N20m median nerve responses in order to confirm coregistration accuracy.
Experimental Paradigm
Individuals were asked to lie comfortably on a bed in the MEG room and relax. Stimuli were non-painful, current pulses of 0.2 ms duration, presentation rate of 0.5 Hz (ISI: 2000 ms between onset of each stimulus), just above motor threshold (eliciting a small, passive, thumb twitch) applied cutaneously to the right median nerve. Somatosensory stimuli were presented in two counterbalanced conditions: a) Predicted Stimulus Pattern and b) Random Stimulus Pattern. In the Predicted Pattern, stimuli occurred in trains of four followed by a long break before the next train (4000 ms) giving 332 stimulus events (83 trains) and 83 breaks in between trains. In the Random Pattern, stimuli and long breaks (4000 ms) were randomly dispersed throughout a 415 event trial (totaling 332 stimuli and 83 breaks). Each condition was 12 minutes in length. Participants were naïve as to the specific patterns that were presented. Upon completion of both stimulus conditions, each participant was asked if they recognized a presentation pattern or not in each of the paradigms.
This research was conducted in compliance with the Helsinki Declaration and approved by the Research Ethics Board at The Hospital for Sick Children, Toronto, Canada, File Number 1000010728.
Data analyses
Neural activities during the experiments were analyzed with respect to brain location, latency, and frequency to determine spatiotemporal profiles of event-related activity time-locked to stimulus presentation. Initial spatial analyses were performed using a novel application of a minimum-variance beamformer algorithm (synthetic aperture magnetometry: SAM) [
52‐
54]. In order to map the median nerve initial response we created SAM differential images by subtracting control periods (-200 to 0 ms prior to stimulus or gap onset) from active periods (0 to 200 ms after stimulus or gap onset) and filtering the data from 3 – 50 Hz. This resulted in high resolution (2 mm) three-dimensional differential images which were time-locked to median nerve stimulation and averaged over time to identify peak activation sites in the brain during the active period relative to baseline. Grand averaged localizations of regional activity peaks for each group were determined by warping SAM images to a template brain and averaging across subjects using Statistical Parametric Mapping software [
55]. Source activity was overlaid on the template brain and imaged using mri3dX software [
56].
We then computed the single trial output of the spatial filters ('virtual sensors') for peak locations of source activity in the SAM images displaying millisecond changes in source power. Time-frequency response (TFR) plots were constructed from the virtual sensor data using a wavelet-based technique which demonstrates both phase-locked and non-phase-locked changes in power at different frequencies over time relative to the baseline period (-100 ms to 0 ms prior to stimulus onset).
Following TFR results we wished to examine group differences for specific frequency sets. Selected bandwidths were averaged across subjects, demonstrating the time course of averaged group response amplitude for a chosen frequency set. Regions on the line graphs were highlighted wherever standard errors did not overlap between controls and those with ADHD in order to exemplify bandwidths where the two groups diverged significantly over time. To determine statistical differences between groups and conditions for each separate time-frequency value we used a permutation program that extracted the normalized, source power value for each time-frequency bin. Individual data were subject to 1000 permutations and then collapsed across participants within a specific group and experimental condition to derive a mean value which could then be statistically compared between groups or conditions. The group mean difference for each pixel was plotted (i.e. – control group data minus ADHD group data for SI random condition) and subsequently thresholded so that only statistically significant differences remained, being expressed as a P-Value plot. Multiple comparison corrections (such as a Bonferroni correction) were not made to the data as each TFR point was not independent.
Discussion
This study used MEG with median nerve stimulation to determine whether somatosensory processing was altered in adult ADHD. We measured frequency specific changes in evoked spatiotemporal patterns of neural activation in response to non-painful electrical stimuli in adults with and without ADHD. Major findings included a marked reduction in the duration of beta rebound in the ADHD group compared to controls in both SI and SII. Beta rebound is a post-stimulus beta phenomenon which commences approximately 400 – 600 ms after median nerve stimulation. Additionally, the ADHD group showed a substantial decrease in SII alpha and beta power during ERD (decreases in power of cortical oscillations below baseline) and ERS (increases in power of cortical oscillations above baseline).
When the stimuli were randomly presented, the ADHD group showed reduced SI ERS power during the immediate N20m response and a significantly shorter SI beta rebound than the controls. This suggests that incoming somatosensory information is less well-characterized at a basic neural level in those with ADHD. Irrespective of whether stimuli were randomly
or predictably presented, the ADHD group showed substantive power decreases in SII alpha and beta ERD and SII beta rebound ERS relative to controls as well as a significantly shorter SII beta rebound. From SI, somatosensory information is thought to project to SII, where stimulus information is integrated and contextualized [
62,
63]. Without sufficient consolidation at SI the deficit may become even more profound as the information is volleyed to the higher processing region of SII. This would explain the marked reduction in SII ERD and ERS in the ADHD group.
To our knowledge this is the first demonstration of reduced duration of somatosensory evoked beta rebound in a clinical population. Little is known regarding the functional significance of the beta rebound response. Historically, beta rebound was thought to be an epiphenomenon that originated in the motor cortex in response to volitional movement [
64]. More recent MEG recordings show that beta rebound also occurs in somatosensory cortex and can be initiated by a tactile stimulus, in the absence of volitional movement [
47]. Moreover, attending to a stimulus can suppress beta rebound relative to that occurring when the stimulus is intentionally ignored [
47]. Both movement imagery [
65] and observation [
66] have been found to suppress the rebound effect. Collectively, these findings suggest that beta rebound can be associated with cognitive state.
Further evidence supports the notion that beta rebound plays a significant role in cortical inhibition of neural regions unrelated to current task performance [
42]. For instance, Chen et al [
67] showed that the brain is less responsive to transcranial magnetic stimulation during the period of beta rebound following median nerve stimulation. If cortical inhibition is indexed by levels of beta activity then it might be argued that the lower levels of beta activity in individuals with ADHD reflect increases in cortical activity. Functional imaging studies indicate that individuals with ADHD activate more widespread brain regions than controls during task performance (review: [
68]).
To our knowledge, this is the first application of MEG to investigate changes in somatosensory alpha or beta power in individuals with ADHD. Here we demonstrate that adults with ADHD showed less changes in source power in the alpha and beta bands overall in response to a somatosensory stimulus. Correspondingly, reduced alpha and beta powers have consistently been associated with ADHD EEG profiles (review: [
69]). Intriguingly, when the adults in the ADHD group were able to predict the onset of an impending event, their SI response to a stimulus did not differ statistically from controls. It may be that the small sample size precluded our ability to detect an underlying effect, as the group mean time frequency plots for the SI Predicted condition in the ADHD and control groups appear different, however these differences did not reach statistical significance. Alternatively, it may be the case that, when a stimulus is predictable, individuals with ADHD are able to recruit additional resources to facilitate somatosensory processing, thereby concealing underlying primary deficits. A similar effect has been observed in individuals with obsessive-compulsive disorder whose behavioural performance was the same as controls in a visual working memory task [
70]. This occurred in spite of the fact that these patients had significantly weaker desynchrony in the alpha band in response to a visual stimulus during the task with a distracter present but not when the distracter was absent [
70].
Our findings support the notion that cortical oscillations are altered during somatosensory processing in those with ADHD. It is possible that impaired somatosensory processing may impede sensorimotor development, as has been found in a substantial proportion of children with ADHD [
71‐
74]. Our data may explain, in part, why individuals with ADHD perform poorly on tasks that require somatosensory feedback such as externally-paced finger-tapping tasks [
28,
30,
75,
76] especially when the tasks require that tactile information be integrated in higher processing regions. Alternatively, it is possible that deficits in attention or executive functions may exert top-down influences on somatosensory processing in the ADHD group.
Our study is limited by the fact that we were unable to investigate effects of gender, comorbidity, or treatment history as the sample of adults with ADHD used in this study was small and heterogeneous, with variation in age, comorbidity, and/or medication (although medication was stopped for at least 24 hours prior to the study). Additionally, right hemisphere SI activity was not investigated as median nerve stimulation was only delivered to the dominant arm. In spite of these limitations, several findings reach statistical significance, emphasizing the powerful nature of the differences in somatosensory processing between the two groups. Future studies will investigate the effects of gender, comorbidity, and medication, as well as the activity of the right SI in response to a contralateral stimulus. Future steps to examine the cortical activity in regions that are in communication with the somatosensory cortex are necessary goals to further elucidate differences in basic processing in individuals with ADHD.
In summary, this study revealed several novel observations regarding somatosensory activity in an ADHD population. It is the first to profile somatosensory ERS and ERD in ADHD and the first to show that beta rebound is not a uniform phenomenon but one that can be modified in the presence of a psychiatric disorder. Profiling impaired cortical rhythms in response to basic sensory processing in ADHD will provide a more in depth understanding of the breadth of deficits in individuals with ADHD and aid in reconstructing the conceptualization and clinical understanding of ADHD.
Acknowledgements
Many thanks to Travis Mills for developing the permutation program for statistical analyses of TFR data and to Sonya Bells for her help running participants in the MEG facility.
We greatly appreciated funding provided by a National Institute of Mental Health Operating Grant (FXC, RT), Hospital for Sick Children Psychiatry Endowment (RT and CD), Canada Research Chair Program (RT), a Post-Doctoral Fellowship from the Hospital for Sick Children Research Training Centre (CD), and operating grants from the Canadian Institutes of Health Research (DC).
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CD developed the design of the study, carried out MEG recordings (subject testing), statistical analyses, and drafted the manuscript. WG participated in the design of the study, contributed to the interpretation of the results. DC participated in the design of the study and contributed to the development of the source analysis methods and interpretation of the results. FW carried out MEG recordings, statistical analyses, and computer programming. FXC contributed to the interpretation of the results. RT participated in the overall conceptualization and supervision of project, including the design and interpretation of the results, All authors read, contributed to, and approved the final manuscript.