Contemporary theories of the pathogenesis of attention deficit hyperactivity disorder (ADHD) conceptualize symptoms of inattention, hyperactivity, and impulsivity as consequences of dysfunction in the cortico-striato-thalamo-cortical (CSTC) loops that subserve, among other processes, executive function and effortful control (1—3). Abnormal activation in these networks is thought to produce hypoarousal and inadequate suppression of both irrelevant sensory inputs and premature behavioral responses, which in turn manifest as the cardinal symptoms of inattention, hyperactivity, and impulsivity that define ADHD (3—6). Animal studies have shown that CSTC loops convey information from the cortex to the basal ganglia, then to thalamic nuclei, and then back to the cortex (7—9). In principle, disturbances in any of the structures along these CSTC pathways could produce abnormal information processing and ultimately the symptoms of ADHD (6, 10). Most anatomical imaging studies in youths with ADHD, however, have focused on the basal ganglia and cortex (11—17) and have largely neglected the thalamus, a key component of the larger CSTC network that relays basal ganglia output to the cortex and mediates information flow between cortical circuits (4, 18—22).
Neuropsychological and neuroanatomical imaging studies generally implicate the thalamus in the pathogenesis of ADHD. Trauma to the thalamus and basal ganglia, for example, can produce new-onset ADHD symptoms in children and adolescents (23). Functional imaging studies have shown that activation of distributed portions of CSTC circuits, including the thalamus, is abnormal in individuals with ADHD (6). Stimulant medications administered to adolescents with ADHD have been reported to alter portions of the EEG that originate in the thalamic reticular nucleus (24). These reports together provide compelling preliminary evidence for the presence of anatomical and functional abnormalities of the thalamus in persons with ADHD. To our knowledge, however, no studies have examined thalamic morphology in youths with ADHD. We present findings from the first high-resolution anatomical MRI study that tests the hypothesis that morphological features of the thalamus in youths with ADHD differ significantly from those of healthy comparison subjects.
We studied a cohort of 105 children and adolescents, 8—18 years of age, including 46 youths who met DSM-IV criteria (25) for ADHD, combined type, and 59 comparison youths who were recruited randomly from a telemarketing list and matched to the patient group by postal code. Written informed consent was obtained from all parents, and all participants provided written assent. Participants' demographic characteristics are summarized in Table 1.
Diagnoses of ADHD were established using the Schedule for Affective Disorders and Schizophrenia for School-Age Children—Present and Lifetime Version (26) and a best-estimate consensus procedure that considered all available clinical and diagnostic information (27). Additional instruments used included the ADHD Rating Scale (28), the Revised Children's Manifest Anxiety Scale (29), and the Children's Depression Inventory (30). Exclusion criteria for participants with ADHD were premature birth (gestation ≤36 weeks) and lifetime diagnosis of any bipolar, psychotic, obsessive-compulsive, or tic disorder. Exclusion criteria for comparison subjects were a lifetime or a current DSM-IV axis I disorder. Additional exclusion criteria for both groups were epilepsy, head trauma with loss of consciousness, lifetime substance abuse, developmental delay, or IQ below 80, as measured by the WISC-III (31), WAIS (32), or Kaufmann Brief Intelligence Test (33). Socioeconomic status was estimated using the Hollingshead Four-Factor Index of Social Status (34).
+
MRI Scanning and Image Analysis
High-resolution MR images were obtained using a single 1.5-T scanner (GE Signa, Milwaukee). Head position was standardized using canthomeatal landmarks. T1-weighted sagittal three-dimensional volume images were acquired using a spoiled gradient echo pulse sequence (repetition time=24 msec, echo time=5 msec, flip angle=45º, matrix=256×192, field of view=30 cm, excitations=2, slice thickness=1.2 mm, contiguous slices=124).
Image processing was performed on SunUltra10 workstations with the ANALYZE 7.5 software program (Biomedical Imaging Resource, Mayo Foundation, Rochester, Minn.). Operators were blind to participant characteristics and hemisphere (images were randomly flipped left to right prior to analysis). Large-scale variations in image intensity were removed (35), and images were reformatted to standardize head positioning prior to region definition (36). Axial slices were oriented parallel to both the anterior and posterior commissures, and sagittal slices were oriented parallel to standard midline landmarks (36).
An isointensity contour function was used in conjunction with manual editing to isolate the cerebrum. This whole-brain volume measure included gray and white matter, ventricular CSF, cisterns, fissures, and cortical sulci. CSF was included using a connected components analysis. Whole-brain volume was used as a covariate in statistical analyses to control for scaling effects (37).
After excluding non-brain tissue, an anisotropic diffusion filter was applied to the remaining brain tissue to improve the discrimination of the lateral surface of the thalamus from the white matter of the internal capsule. The procedures for defining the thalamus were identical to those previously published (38, 39). The thalamus was segmented by sampling gray-scale values of the thalamus and internal capsule throughout the entire three-dimensional extent of each structure and then averaging the peaks for white and gray matter. An isointensity contour function applied at the calculated threshold and grown from a seed within the thalamus provided an initial definition of the structure that was then edited manually. The intra- and interrater intraclass correlation coefficients were 0.95. The thalamus was distinguished from the hypothalamus by a line defining the hypothalamic sulcus on sagittal views, which excluded portions of the geniculate nuclei from the analysis.
We calculated the distance from a voxel-sized point on the surface of each participant's thalamus to the corresponding point on the surface of the thalamus in a template brain. This previously validated method of surface analysis (40) was customized to accommodate independent analysis of right and left hemi-thalami. Briefly, a rigid-body similarity transformation with global scaling was used to register the entire brain of each participant with the template brain, thereby eliminating the need to further adjust for differences in overall brain volume. The thalamus was then rigidly coregistered to the template thalamus. This second transformation created a "refined registration" in which isolated thalami could be compared. Each hemi-thalamus was warped to the corresponding anatomy of the template thalamus using a high-dimensional nonrigid warping based on fluid-flow dynamics that permitted point-to-point matching of homologous tissue between the two thalami. The warped images were then unwarped to the refined registration while maintaining these point-to-point correspondences, which permitted calculation of the distance of each point on the surface of the test thalamus from the corresponding point on the surface of the template thalamus.
We applied a rigorous two-step procedure to select a thalamic template that was as close as possible (in the sense of least-squares mean) to the average shape of the thalamus for all our healthy comparison subjects. First, a preliminary reference was selected as the thalamus of the comparison subject who demographically was as representative as possible of all healthy comparison subjects. The thalami for all other comparison subjects were then normalized to this preliminary reference. The point correspondences on the thalamic surfaces were determined, and we computed the distance from the template surface for each of the corresponding points on the surfaces of the thalami for all other participants. Second, the thalamus for which all points across its surface were closest, in terms of least-squares mean, to the average of the computed distances was selected as the final template. As we have done previously, we used a single representative brain as a template rather than an averaged brain from many youths, because use of a single brain, which has sharp borders at the CSF-gray matter or gray matter-white matter interface, improves the accuracy of registration (36, 41, 42). Averaging images for a template blurs these boundaries and increases registration errors that are subtle but important when distinguishing subtle effects across populations. In addition, precise surface morphometry requires a brain with smooth gray and white matter surfaces that are devoid of topological defects, which cannot be reconstructed by averaging brains from many youths.
Statistical analyses were performed in SAS, version 9.0 (SAS Institute, Inc., Cary, N.C.). We tested our a priori hypothesis that conventional measures of overall thalamic volume would differ across diagnostic groups by assessing the main effect of diagnosis in a mixed-models analysis with repeated measures over a spatial domain (volumes of each hemi-thalamus). The model included the within-subjects factor "hemisphere" with two levels (left and right), the between-subjects factor of diagnosis (ADHD=1 and comparison subjects=0), and the covariates of age, sex, and whole-brain volume. In addition to these independent variables, we considered all two- and three-way interactions of diagnosis, sex, hemisphere, and age, as well as the two-way interactions of whole-brain volume with hemisphere. Other variables included medication, IQ, and lifetime diagnoses of depression, oppositional-defiant disorder, and anxiety disorders. Statistically nonsignificant terms were eliminated via backward stepwise regression, with the constraint that the model at each step had to be hierarchically well formulated (i.e., all possible lower-order component terms of an interaction were included in the model, regardless of statistical significance). We considered p values <0.05 statistically significant. All p values were two-sided.
The distances between points on the surface of the thalamus for each participant and the corresponding points on the template thalamus were compared statistically between groups using linear regression while covarying for age and sex. We color-coded p values at each voxel and displayed them across the surface of the template thalamus. We used the theory of Gaussian random fields (GRFs) to correct p values appropriately for the multiple comparisons performed across the thalamic surface (42).
+
Cytoarchitectonically Defined Thalamic Atlas
We superimposed a cytoarchitectonic atlas for the thalamus (43, 44) on our thalamic template to identify precisely the regions of significant differences in regional volumes between groups of subjects. The cytoarchitectonic atlas was coregistered to our thalamic template (40) using the same registration and warping procedures as employed for surface morphometry (above).
The main effect of diagnosis indicated no significant differences across groups in overall conventional volumes of the thalamus.
Significant covariates in the model included whole-brain volume (F=24.32; df=1, 99, p<0.001), indicating the presence of significant scaling effects in the thalamus; hemisphere (F=9.90; df=1, 104, p<0.002), reflecting significantly larger volumes in the left hemisphere for all participants; and sex (F=7.55; df=1, 99, p<0.007), demonstrating sex-specific differences in thalamic volumes across the whole sample. The main effect for age was not significant, and the absence of a diagnosis-by-age interaction indicated the stability of findings across the age range of the youths studied (Table 2). The effects of stimulant medication, evaluated by including stimulant treatment as an independent variable in the linear regression, revealed that patients who were receiving stimulants at the time of scanning had larger thalamic volumes than did untreated patients, although this difference fell short of statistical significance (F=3.84, df=1, 98, p=0.053).
+
Morphological Features of the Thalamic Surface
Relative to comparison subjects, youths with ADHD exhibited significantly smaller regional volumes corresponding to the anterior, posterior, and ventral lateral surfaces of the right hemi-thalamus and the posterior surface of the left hemi-thalamus. Using the thalamic atlas template, these morphological differences between the groups were localized in the right anterior and lateral posterior nuclei and the lateral portion of the pulvinar bilaterally. GRF-corrected analyses showed that the differences in regional morphology were concentrated predominantly in the lateral pulvinar bilaterally (Figure 1).
Youths with ADHD who were receiving stimulant medications at the time of scanning had significantly larger regional volumes over the posterior dorsal, ventral, and anterior dorsal thalamic surfaces bilaterally compared to untreated patients. The thalamic atlas template further localized these regions to the anterior nucleus and pulvinar bilaterally. GRF-corrected maps revealed larger regional volumes in stimulant-treated youths predominantly in the pulvinar, closely overlapping previously identified locations of smaller regional volumes in the ADHD group relative to the comparison group. These findings did not change when we compared youths with ADHD on and off stimulants after excluding the five participants with ADHD who were taking both stimulant and nonstimulant medications and one participant who had taken stimulants in the past but had been off them for 19 months before the scan. Additionally, GRF-corrected analyses demonstrated smaller regional volumes of the right anterior nucleus and the pulvinar bilaterally in untreated patients with ADHD relative to comparison subjects (Figure 2). Thus, we obtained the finding of smaller regional volumes in both the posterior and anterior thalamic surfaces in the ADHD group relative to the comparison group when we compared healthy youths either to all patients with ADHD or to untreated patients only.
Surface maps for the subgroup of youths with ADHD for whom we collected sufficient data on duration of treatment with stimulants (N=17) revealed that a longer duration of treatment was associated with smaller regional volumes in the right lateral and medial posterior thalamic surfaces (Figure 3). GRF-corrected analyses localized these correlations predominantly to the right lateral and medial pulvinar, a location that did not overlap with the main effects of diagnosis or stimulant treatment.
+
Correlations with symptom severity
More severe hyperactivity accompanied smaller regional volumes in the lateral and posterior thalamic regions bilaterally, which were localized in the ventral anterior, ventral lateral, ventral posterior, lateral posterior, and pulvinar nuclei. GRF-corrected maps showed that these differences were concentrated on the ventral lateral nucleus bilaterally and the left lateral posterior nucleus. In contrast, greater inattention scores accompanied larger regional volumes in the right posterior and medial thalamus and in the left anterior and lateral thalamic regions, corresponding to the right pulvinar, medial dorsal, left central medial, ventral lateral, and lateral posterior nuclei. GRF-corrected maps localized these differences to the right pulvinar and the medial dorsal nuclei (Figure 4), extending beyond the locations of the main effects near the pulvinar, where regional volumes were smaller in the ADHD group.
In analyses of conventional volumes and surface morphology, we did not discern appreciable effects of age, comorbid disorders (see Table 2, as well as Figures S1 and S2 in the data supplement that accompanies the online edition of this article), duration of stimulant treatment (not shown), or IQ (not shown), which suggests that these variables did not unduly influence our findings.
Regional volumes were significantly reduced over the pulvinar bilaterally in youths with ADHD, despite the presence of normal overall thalamic volumes. These morphological abnormalities were not associated with the effects of naturalistic stimulant medication treatment or the presence of comorbid conditions in the ADHD group. In addition, conventional thalamic volumes in youths with ADHD treated naturalistically with stimulants tended to be larger than in untreated individuals (p=0.053), and regional volumes in those treated with stimulants were significantly larger in the pulvinar bilaterally. Finally, smaller regional thalamic volumes in the ventral lateral nucleus bilaterally and the left lateral posterior nucleus were associated with more hyperactivity, and larger regional volumes in the right pulvinar and medial dorsal nuclei were associated with more inattention.
The absence of a significant main effect for diagnosis in analyses of conventional thalamic volumes likely derives from the insensitivity of our ability to detect abnormalities in conventional volumes of a heterogeneous brain region such as the thalamus when the anatomical abnormalities in that region are restricted to a small number of substructures. Several studies, however, have reported significant group differences in conventional volumes of the hippocampus (36), the putamen (45, 46), the globus pallidus (47), the frontal cortex (47, 48), and the cerebellum (47, 49) in youths with ADHD relative to comparison subjects. Thus, examination of both conventional thalamic volumes and morphological features of the thalamic surface is appropriate in the study of ADHD.
+
The Role of the Thalamus in the Pathogenesis of ADHD
We used the cytoarchitectonically defined thalamic atlas (43) to determine precisely the regional morphological features on the surface of the thalamus overlying specific thalamic nuclei and to interpret morphological alterations in the context of the known thalamic connectivity derived from studies in humans. This atlas allowed us to map anatomical findings in youths with ADHD onto larger anatomical circuits that subserve specific cognitive processes. These morphological findings suggest that the pulvinar and ventral lateral nuclei are involved in the pathophysiology of ADHD.
The pulvinar projects to the frontal and parietal association cortices (50), two thalamocortical pathways that support attentional processes. These circuits are important for stimulus seeking and for distinguishing between contextually relevant and irrelevant somatosensory stimuli (51). Within this overall thalamocortical network, the pulvinar integrates and coordinates responses to auditory and visual stimuli (52—54). Morphological disturbances in the pulvinar in youths with ADHD therefore may contribute to difficulties in allocating and directing attentional resources toward salient stimuli.
The pulvinar also connects with limbic structures, including the amygdala (55—59). We previously reported morphological disturbances in the basolateral complex of the amygdala (36), and others have reported abnormal amygdala activation (60, 61) in individuals with ADHD. The basolateral complex of the amygdala supports emotional regulation, fear conditioning, and the attribution of affective valence to sensory stimuli (62—64). Therefore, disturbances of the thalamolimbic network could interfere with emotional learning and the regulation of affect, thereby contributing to the emotional dysregulation and affective illnesses that affect many children with ADHD (65, 66).
+
The ventral lateral nucleus
The ventral lateral nucleus receives projections from the cerebellum and connects reciprocally with the basal ganglia, motor cortex, and premotor cortices to form a sensorimotor network that drives the acquisition and implementation of learned motor behaviors (67—69). Smaller regional volumes in the ventral lateral nucleus could therefore contribute to the gross and fine motor disturbances and the slower processing speed that have been documented in children with ADHD (70—74).
+
Effects of Stimulant Medications on Thalamic Morphology
Surface maps show that the pulvinar was significantly larger bilaterally in treated compared with untreated youths with ADHD. These areas closely overlapped regions where volumes were smaller in youths with ADHD, which suggests that stimulant treatment may help attenuate morphological abnormalities in the thalamus in individuals with ADHD. If that is true, then the absence of statistically significant differences in conventional volumes of the thalamus in youths with ADHD relative to comparison subjects may be partially explained by these hypothesized effects of stimulant treatment. Alternatively, youths with ADHD who had larger thalami may for unknown reasons have been preferentially treated with stimulants. The effects of stimulants on local morphology should be more definitively determined in a longitudinal randomized controlled trial that would obviate any possible selection biases that may have confounded treatment effects in this cross-sectional, naturalistic study.
The duration of stimulant treatment was inversely correlated with regional volumes in the right lateral and medial posterior hemi-thalamus. These regions did not overlap those where the main effects of an ADHD diagnosis or where the main effects of stimulants were located. These findings suggest that stimulants may have differing short- and long-term morphological effects on differing thalamic nuclei. Treatment with stimulants for any duration was associated with larger volumes in the dorsal anterior, posterior, and ventral posterior surfaces bilaterally, whereas progressively longer treatment was associated with progressively smaller regional volumes in the right lateral and medial posterior surfaces. Alternatively, youths with ADHD who received stimulants for longer durations may have differed systematically and in unknown ways from those who received stimulants for shorter durations, producing only an apparent association of regional volumes with duration that was not actually caused by longer exposure to stimulant medication. Moreover, we must emphasize that the association of treatment duration with thalamic morphology should be interpreted with caution, as treatment duration is exceedingly difficult to estimate reliably. Treatment is often interrupted by medication holidays and lapses in compliance that are not documented or recalled, and pharmacological agents and their dosing often vary widely over time.
We do not know what neurobiological mechanisms produced the local morphological differences we observed between the ADHD and comparison groups and that seemed to have attenuated those differences in the stimulant-treated youths with ADHD. Previous reports of depleted levels of dopamine and norepinephrine suggest that disturbances in catecholamine transmission may contribute to the pathogenesis of ADHD (75). Stimulants are thought to attenuate ADHD symptoms by inhibiting presynaptic dopamine and norepinephrine transporters and thereby potentiating dopamine and norepinephrine neurotransmission. Both dopamine and norepinephrine have powerful influences on the physiology and cellular morphology of the prefrontal cortex, basal ganglia, and thalamic portions of CSTC circuits (76—79). In humans, multiple thalamic nuclei, including the pulvinar, receive dopamine and norepinephrine inputs (80—83), so theoretically stimulants should increase cathecholaminergic transmission within the thalamus, although this has not been shown definitively. Nevertheless, numerous preclinical studies in rodents have shown that repeated exposure to stimulants increases the length of dendrites and the density of dendritic spines within the basal ganglia, limbic system, and frontal cortices (84—86). We speculate that stimulants may have a similar effect in the thalamus. Alternatively, the stimulant-induced cellular and functional changes in the prefrontal cortex (79) or basal ganglia could in turn alter cellular activity and dendritic architecture within the thalamus. The effects of stimulants on catecholamines have also been shown to alter the number of astrocytes in the striatum (87), cells that regulate the energetics of neuronal excitation (88). Stimulants could have similar effects on thalamic astrocytes.
+
The Differential Association of Thalamic Nuclei With ADHD Symptom Domains
Smaller regional volumes in the ventral lateral nucleus bilaterally and the left lateral posterior nucleus were associated with more severe hyperactivity, whereas larger regional volumes in the right pulvinar and the medial dorsal nucleus were associated with more severe inattention. The localization of these differences is consistent with reports that the projections of the ventral lateral nucleus to the striatum participate in the control of motor actions (4, 89) and that the pulvinar, through its cortical connections, may be involved in alerting and in the allocation of attention (52—54). Smaller regional volumes in the lateral thalamus suggest the presence of disturbances in networks related to motor functions, possibly producing more severe symptoms of hyperactivity. Larger regional volumes in the pulvinar may represent alterations in systems that subserve attention, either exacerbating the symptoms of inattention or compensating for functional disturbances in attention-related networks located outside the thalamus.
Several limitations of this study should be mentioned. First, our study design did not allow us to determine the direction of causality in the associations of regional thalamic volumes with diagnosis or stimulant treatment. Second, the use of the thalamic atlas has inherent inaccuracies in the exact boundaries that it ascribes to thalamic subnuclei, and our analyses of the thalamic surface assign the source of morphological differences to the most superficial nucleus, when in fact it may derive instead from abnormalities in deeper underlying nuclei. Third, the imaging methods we used cannot alone identify the cellular and molecular bases for our morphological findings, which would require studies using either different imaging modalities or postmortem histological analyses. Finally, the multiple statistical tests in our exploratory analyses increased the likelihood of type I error, although we minimized false positive findings through the use of conservative statistical thresholds and GRF-based corrections of p values for the multiple comparisons performed (90).
Our findings provide new evidence that abnormalities at the level of the thalamus are involved in the pathogenesis of ADHD. Our findings additionally suggest that the therapeutic effects of stimulants may attenuate these morphological abnormalities, possibly via the local changes that they produce in dendritic architecture and cellular composition within CSTC circuits. Finally, we hypothesize that the hyperactivity and inattention symptoms of ADHD derive from abnormalities in different thalamic nuclei that may reflect the differential involvement of both pathogenic and compensatory processes.