|Home | About | Journals | Submit | Contact Us | Français|
To highlight recent advances in the conceptualization of Attention-Deficit/Hyperactivity Disorder (ADHD) emerging from neuroimaging and endophenotypic approaches.
Selective review of the recent published literature on the phenomena of resting state functional connectivity, intra-subject variability, and diffusion tensor imaging pertaining to ADHD.
Recent advances based on the novel approach of resting state functional connectivity appear to be highly promising and likely to link to studies of intra-subject variability.
Endophenotypic fractionation may offer a means of addressing the complex heterogeneity of ADHD on the path to testable models of pathophysiology. Such models focusing on intra-subject variability, intrinsic brain activity, and reward-related processing are progressing rapidly.
Attention-Deficit/Hyperactivity Disorder (ADHD),1 a highly prevalent condition in childhood, is increasingly recognized as an important contributor to lifelong impairment,2,3 although it has been infrequently studied in adults until recently.4–20 Certainly, a substantial literature unequivocally demonstrates that ADHD in children is associated with globally decreased total brain and cerebellar volumes and striatal volumetric abnormalities that are age-dependent (see reviews21–23). Increasingly sophisticated methods examining cortical topography,24 cortical thickness12,25,26 or voxel-based morphometry27–29 are demonstrating more delimited anatomic abnormalities. However, the overall pattern of results is not satisfyingly convergent, which suggests that the multiple types of heterogeneity that occur in ADHD must be addressed before we can develop neurobiological methods with adequate explanatory power.
Recently, functional magnetic resonance imaging (fMRI) studies of ADHD have been increasing, as recently reviewed qualitatively by Bush et al.,30 who concluded that the evidence supports dysfunction of frontal-striatal circuitry, particularly lateral prefrontal cortex (PFC), dorsal anterior cingulate cortex (ACC), and striatum.30 These conclusions were recently extended using a quantitative meta-analytic method, Activation Likelihood Estimation,31,32 which was applied to 17 published functional imaging studies. Results showed significant frontal hypoactivity in patients with ADHD relative to controls in ACC, dorsolateral and inferior PFC, basal ganglia, thalamus, and portions of parietal cortex.33 Dickstein et al. noted that the widely distributed results of the meta-analysis did not provide strong support implicating a single frontal subregion in ADHD.33
In brain imaging studies, low- or no-demand attentional conditions have been used primarily as control tasks to provide a baseline against which task-specific activations could be estimated, rather than as a worthy topic of study in its own right.34 This changed recently with the confirmation that the brains of individuals apparently at rest (i.e., where no externally imposed cognitive task is being performed), display strikingly similar patterns of spontaneous intrinsic activity synchronized across multiple dissociable neural networks.34–48
Raichle and colleagues used the absolute quantitative values yielded by positron emission tomography to suggest the existence of an organized ‘default-mode’ of brain function which serves as a physiological baseline44,49 that is suppressed during specific goal-directed behaviors.34 This phenomenon has been confirmed with a variety of methods39,40,44,50–53 and extended by other investigators35,37,38,41–43,47,54–62 and has focused attention on “investigating the features of the baseline state of the brain” (p. 13848).37 Initial analyses of the default-mode focused on regions showing patterns of deactivation during task performance – also described as the “task-negative network” because of the lack of association between this network with cognitive tasks.41,43 Recently, investigators have noted that the default-mode network is negatively correlated across time (anti-correlated) with a “task-positive network.”39,42 These two anti-correlated anti-phase networks are each robustly and consistently related to a distinctive set of widely distributed brain regions related to specific cognitive functions.39 The task-positive elements of the network involve specific regions of the prefrontal (especially the dorsolateral PFC), the supplementary motor areas and regions of the parietal cortex. The task-negative components include the medial and lateral parietal and medial prefrontal cortex and the precuneus/posterior cingulate cortex (retrosplenial complex). The striking degree of anti-correlation of the task-positive and task-negative networks39,41,42 suggests reciprocal relationships.39 These networks are remarkably replicable across labs using a variety of methods,36 and the test-retest stability of these networks has also been documented.37,63
In considering the neural circuitry underlying the regulation of behavioral performance, researchers investigating increases in intra-subject variability in populations such as traumatic brain injury patients and the elderly have implicated PFC and ACC. Given the vast literature implicating these regions in top-down control processes, their involvement in maintaining consistent behavioral performance is not surprising. With respect to ADHD, such findings are intriguing given consistent reports of ADHD-related hypofunction in some of the same frontal regions, thus suggesting a potential locus of dysfunction capable of explaining ADHD-related increases in intra-subject variability.64
However, the specific mechanism(s) by which frontal regions modulate response time (RT) variability remained poorly defined until recently.54 In an event-related fMRI examination of an attentional task, Weissman et al.54 demonstrated that the episodic expression of longer RT was associated with failure to suppress activity in the default-mode network. Decreased cue-related activation in three prefrontal regions (right middle frontal gyrus, right inferior frontal gyrus and right dorsal ACC) predicted slower RT and decreased default-mode network suppression, suggesting their involvement in default-mode network regulation. This study provided the first demonstration of an association between activation in the default-mode network and behavioral performance, and supported the conclusion that increased RT variability reflects weaker suppression of activity in key default-mode network regions.
Accordingly, Sonuga-Barke and Castellanos65 enunciated the “default-mode hypothesis of ADHD” which states that these fronto-default-mode interactions may represent a novel locus of dysfunction in ADHD, potentially accounting for ADHD-related increases in intra-subject variability. Increased intrasubject variability in ADHD has been well documented. Such findings have been demonstrated in ADHD across tests of sustained attention,66–68,68–73 attentional control,64 motor sequencing and control,74–78 inhibitory control,10,79–81 time perception,82 and choice RT tasks.68,83,84
While consistent findings of ADHD-related hypofunction in ACC may suggest a mechanism by which the hypothesized impairments in fronto-default-mode network interactions may occur, morphometric studies suggest possible abnormalities within the default-mode network itself. Although not typically examined as an explicit region of interest, full-brain morphometric examinations of ADHD have reported volumetric abnormalities in posterior portions of the default-mode network, most notably posterior cingulate cortex and precuneus. Cortical thickness in adults with ADHD was found to be significantly decreased in the retrosplenial complex.12 Similarly studies using whole-brain voxel-based morphometry also identified gray matter volume reductions in precuneus and posterior cingulate cortex.27,28
Task-based functional imaging studies of ADHD have typically neglected deactivations which are characteristic of default-mode network involvement (with a recent exception85). In the past year, intrasubject RT variability has been examined in healthy children with a Go/No-go task,86 and in adolescents with ADHD and healthy controls during an oddball task.73 Consistent with the overarching default-mode hypothesis,65 during the Go/No-go task, higher default-mode network activity in medial PFC (Brodmann area (BA) 9/10) predicted higher RT variability, and higher pre-supplementary motor area (BA 6) predicted lower RT variability.86 Similarly, on an oddball task, adolescents with ADHD were significantly more variable on RT and had significantly less activation in ACC than controls.73 One other group has examined resting state functional connectivity in adolescents with ADHD. In a series of intriguing papers describing alternative analytical methods applied to a single dataset of 13 boys with ADHD and 12 controls,87–90 Wang et al. described between-group differences in default-mode network properties during resting state, finding decreases in the amplitude of the spontaneous low frequency fluctuations that form the basis of default-mode network connectivity, though not commenting on any analyses focusing on default-mode network connectivity specifically.89 Interestingly the same group reported findings of increased functional connectivity between a region-of-interest consisting of dorsal ACC (not otherwise defined) and widespread regions including thalamus, cerebellum, insula and pons.88 Such findings are seemingly inconsistent with both the default mode hypothesis and the findings from preliminary study by Castellanos et al.,14 which provided evidence of ADHD-related decreases in ACC functional connectivity with the retrosplenial complex. There are several possible explanations for these divergent findings. First, ACC is functionally and structurally complex.91,92 The seed region of interest used by Tian et al. comprised the entire dorsal ACC, unlike the highly focused seed used by Castellanos et al., which was identified as a source of top-down cognitive control by Weissman et al.54 Second, Tian et al. did not differentiate the anti-phase relationships (“anti-correlations”)39,41 that accounted for the Castellanos et al. findings, potentially confounding their interpretation.
The emphasis of the default-mode network hypothesis on ADHD-related decreases in the integrity of neural circuits, rather than the responsiveness of a single region or set of regions to task demands presents a novel challenge. Recently emerging approaches to studying functional connectivity during resting state appear to provide an appropriate solution for this challenge. Specifically, applying correlational analyses to resting state fMRI data enables characterization of task-independent patterns of functional connectivity.93 Studies using this analytic approach have demonstrated that functionally relevant patterns of activity, commonly observed during task performance, are also intrinsically represented in spontaneous brain activity.35,37,38,41,43 Fox et al. demonstrated the utility of resting state approaches in mapping neural systems, successfully differentiating the dorsal and ventral attentional systems,40 two functionally related but distinct networks. The utility of resting state approaches in mapping neural systems has been recently demonstrated in neurotypical controls, by mapping intricate patterns of functionally connectivity throughout the anterior cingulate cortex,92 basal ganglia,94 and amygdala.95
Castellanos et al.14 obtained blood oxygenation level dependent (BOLD) fMRI scans on a 3.0 Tesla scanner from 20 adults with ADHD (mean age: 34.9±9.9) and 20 age- and sex-matched neurotypical controls NC (mean age 31.2±9.0; p>.22) during rest. All subjects were strongly right-handed and underwent clinical assessments to confirm current presence of ADHD (patients), and absence of axis I psychiatric disorders in controls and absence of medical disorders in both groups. The objective was to test the hypothesis that fronto-default mode interactions may represent a possible locus of dysfunction in ADHD. Specifically, given reports that fronto-default mode interactions are intrinsically represented in spontaneous activity at rest,39,41 the authors tested for the presence of ADHD-related differences in functional connectivity between each of the three previously identified frontal foci (dorsal ACC; right inferior frontal gyrus; right medial frontal gyrus) and the default-mode network.
Although both right inferior frontal gyrus and right middle frontal gyrus ROIs were significantly negatively related to precuneus and posterior cingulate cortex (collectively retrosplenial complex) in both groups, between-group differences did not exceed the conservative whole brain Gaussian random field threshold set for statistical significance. By contrast, FC analyses of the dorsal ACC ROI (see Figure 1; significantly positively predicted voxels shown for entire sample in red; negatively predicted voxels in blue; p<.05, corrected) demonstrated significantly less negatively correlated activity in the retrosplenial complex in subjects with ADHD (p<.0004, corrected) [shown in yellow on the Z-maps]. The plot shows the regression parameter estimates for the negatively correlated relationship between dorsal ACC and retrosplenial complex in the two groups further documenting the extent of this group difference (means ±SD −0.23 ± 0.12 and −0.01 ± 0.15 for NC and ADHD groups, respectively; t(38)=5.10, p=9.6 × 10−6; d=1.61 SD, 95% CI 0.79 – 2.43). In the context of increasing awareness of the complex role of precuneus and posterior cingulate cortex in “high-level integration between posterior association processes and anterior executive functions” 96, p. 578, Castellanos et al. interpreted their findings as suggesting that structural and functional circuits linking the dorsal ACC to the retrosplenial complex may represent ‘small-world network’ long-range connections 97 that should be considered as a candidate locus of dysfunction in ADHD. This intriguing finding of a circuit-based diagnostic difference raises questions regarding structural connectivity that are amenable to examination through approaches such as diffusion tensor imaging.
While structural imaging studies have long appreciated gross differences in total brain, cerebellar and striatal volumes, as well as specific decreases in grey matter,21 it is only recently that potential ADHD-related differences in white matter (WM) have come into focus. A recent meta-analysis of region-of-interest-based morphometric studies of ADHD highlighted a number of specific volumetric reductions in WM abnormalities in fronto-striatal and fronto-cerebellar circuits, with the largest effect sizes in deep frontal regions and the splenium.98
With the advent of diffusion tensor imaging (DTI), WM abnormalities can be examined at the voxel level. DTI provides information on the directionality and coherence of the self-diffusion of water.99 In DTI, diffusion is measured in multiple directions to form the diffusion tensor that summarizes the three dimensional diffusion properties.99,100 The magnitude of diffusion in the three orthogonal directions (principal eigenvectors) can be computed from the diffusion tensor.101,102 These eigenvectors and their corresponding eigenvalues yield quantitative measures such as the trace and fractional anisotropy. The trace is the average diffusion coefficient over the three orthogonal directions. Fractional anisotropy is a measure of the degree of anisotropy within a voxel with a value of 0 corresponding to completely isotropic diffusion and 1 corresponding to free diffusion in one direction only. White matter abnormalities are expected to increase the trace but decrease fractional anisotropy due to axonal disorganization. Measures of the self-diffusion of water have implications for the structural anatomical organization of neurons (e.g., axons, cell membranes, and related myelin). These structural elements are responsible for WM integrity. Given the importance of WM, DTI measures have begun to be applied to ADHD.
The first DTI examination of ADHD-related compromises in WM connectivity was published in 2005.103 Voxel-wise fractional anisotropy measures in 18 children with ADHD, ages 7–11, and 15 age- and gender-matched controls revealed ADHD-related fractional anisotropy decreases predominantly in the motor circuit (e.g., right premotor, right striatal, right cerebral peduncle, left middle cerebellar peduncle, and left cerebellum). A 2007 comparison of adolescents born at very low birth weight found decreased fractional anisotropy in several frontal tracts in the eight subjects who had ADHD symptoms or diagnosis.104 Twenty parent-child dyads with ADHD and 10 dyads without ADHD provided both fMRI data from a Go/No-Go task and DTI fractional anisotropy.105 Prefrontal fractional anisotropy correlated significantly within ADHD parent-child dyads but not in control dyads. Prefrontal fractional anisotropy in parents and youth with ADHD correlated significantly with performance and BOLD activation.105 Of interest, fractional anisotropy in cingulum bundle and superior longitudinal fasciculus II was significantly reduced in 12 adults with childhood ADHD compared to 17 unaffected controls.16 Thus, DTI is also beginning to contribute although this literature is still in its early stages, particularly in terms of sample size.
ADHD is increasingly recognized as a syndrome associated with lifelong impairment, which merits study in affected adults, not just in children and adolescents. At the same time, the incontrovertible heterogeneity of ADHD106,107 calls for an approach that can be called endophenotypic fractionation.108 One promising approach seems to be to focus on RT variability both in neuropsychological assessments64,109 and in brain imaging studies.86,110 Besides RT variability, reward-related circuitry is also being productively examined because the neural substrates are relatively well defined.18,19,111,112 By establishing a finite set of such dimensions that are grounded in neuroscience, investigators will increasingly be able to formulate appropriate models of the pathophysiology of ADHD and related conditions – which remains the next necessary bridgehead.
This work was supported in part by grants provided to FXC by the Stavros S. Niarchos Foundation, NIMH (5R21MH066393 and 5T32MH067763), the Leon Lowenstein Foundation, NARSAD and gifts from Linda and Richard Schaps, Jill and Bob Smith and the Taubman Foundation and the endowment established by Phyllis Green and Randolph Cōwen.
The authors report no financial conflicts of interest.