In a prospective 33-year longitudinal follow-up of 59 probands (mean age 41 years) with established ADHD in childhood and 80 prospectively enrolled non-ADHD comparisons, we found an overall significant reduction in mean cortical thickness in probands. Beyond this global difference, the greatest cortical thinning associated with childhood ADHD was located in bilateral parietal lobes, temporal poles, insula, precentral gyri, frontal poles, and right precuneus. No cortical region was significantly thicker in probands than comparisons. Although less sensitive,48
VBM also revealed significantly decreased GM in probands versus comparisons in right precentral, bilateral parietal, left temporal, and right cuneus. Additionally, VBM detected decreased GM in probands in caudate, thalamus and cerebellar hemispheres.
With respect to current adult diagnosis, probands with persistent ADHD differed most from non-ADHD comparisons in the same cortical regions identified in our primary analyses, as well as in additional clusters in left medial occipital cortex and subgenual ACC. Probands with remitted ADHD did not differ significantly from persistents when analyses were corrected for full-brain comparisons. In exploratory uncorrected analyses, probands with persistent ADHD exhibited reduced cortical thickness relative to remitters in bilateral medial occipital lobes, temporal lobes extending to insula, and left parahippocampus.
Our results extend prior volumetric and cortical thickness findings in ADHD. First, consistent with decreased total cerebral volume in ADHD,2–4
our observation of reduced global cortical thickness in probands with ADHD confirms prior reports.13, 14, 20
Furthermore, although we found less frontal and prefrontal cortical thinning in ADHD than others,12–15, 20, 49
we confirmed thinner cortical mantle in occipito-parietal,12, 13, 20
temporal cortex and precentral regions13, 14
in ADHD. In subcortical analyses, we also confirmed anatomic abnormalities in caudate,3, 50, 51
Studies of cortical thickness in adults with ADHD have focused on specific regions associated with executive function and attentional control.54, 55
Makris et al.9
selected nine parcellation units (from 48) per hemisphere and found thinner cortex related to ADHD in prefrontal and cingulate cortex and inferior parietal lobe, albeit without correcting for multiple comparisons.9
A cross-sectional study of children, adolescents and adults found that individuals with ADHD, regardless of age, had significantly thinner right superior frontal cortex than controls.15
In the adults with ADHD, the specific reduction, with correction for multiple comparisons limited to the frontal lobe, was localized to BA9. In contrast, we did not find group differences in much of prefrontal cortex but found widespread cortical thinning in bilateral parietal-temporal cortex. We found similar results in analyses that included all participants as well as in those limited to probands with persistent ADHD versus non-ADHD comparisons. The latter contrasts are comparable to studies in adults that define group membership by current diagnostic status.15, 20
Studies of cortical thickness in children with ADHD are more numerous than those in adults,12–14, 33, 47, 56, 57
and typically have examined the entire cerebrum, although nearly all (except14
) report results uncorrected for multiple comparisons. Thinner cortex has been reported in children with ADHD in prefrontal and precentral regions12, 14
parietal and temporal lobes12, 13
and inferior frontal gyrus bilaterally.58
In our main analyses, we applied FDR full-brain correction for multiple comparisons, and observed significant differences whether groups were defined by initial childhood history or by current adult diagnoses. We speculate that the robustness of our results reflects having established the diagnosis of ADHD in childhood as well as our medium to large sample sizes.
Broadly, our results implicate disruptions in large-scale neural systems involved in the regulation of both attention and emotion in adults with childhood ADHD. We found convincing converging anatomic evidence implicating the dorsal attentional network55
and distributed regions within limbic circuits that were thinner in ADHD probands than in comparisons. Similar findings were obtained when we contrasted probands with persistent ADHD versus comparisons without ADHD. However, we failed to observe hypothesized group differences in prefrontal regions.1, 3
Below we discuss our main findings and non-findings in turn.
First, we found widespread thinner cortex and decreased GM density in bilateral parietal and precentral regions, overlapping areas of the dorsal attentional network. The bilateral dorsal network, which mediates goal-directed, top-down executive control processes, interacts with a right-sided ventral system (stimulus-driven, bottom-up) during attentional functioning,1, 55
particularly in redirecting attention. The core areas constituting the dorsal attentional network include the intraparietal sulcus and the conjunction of the precentral and superior frontal sulcus (frontal eye fields)55
which were particularly affected in the ADHD probands. Strikingly, we also observed significantly thinner cortex in precuneus and superior parietal lobe, which along with the dorsal network core regions are implicated in top-down processing of shifting of attention.59
These findings are consistent with studies of ADHD that report abnormal patterns of activation in parietal regions52
during working memory,60–62
or response inhibition tasks.66, 67
We also found occipital cortical thinning in probands with persistent ADHD versus non-ADHD comparisons. Occipital cortex has been recently found to interact with the dorsal network in maintaining attention59
and in suppressing responses to irrelevant stimuli.68, 69
Individuals with ADHD are easily distracted when required to ignore extraneous signals.70, 71
Top-down control deficits when responding to irrelevant stimuli are associated with impaired working memory.72, 73
Abnormal activation of occipital cortex has been found in youth74
with ADHD during working memory tasks. Similarly, in a meta-analysis of functional imaging studies, children and adolescents with ADHD showed activation decreases in left middle occipital gyrus (BA19) compared to controls.52
Additionally, a recent VBM study in adults with ADHD found significant bilateral reduction of GM volume only in early visual cortex.78
Our VBM analysis revealed cerebellar, thalamic and striatal GM deficits in ADHD. Cerebellar involvement in ADHD is well-established, with findings in children reported mostly in the vermis,1–4, 79
and in the hemispheres in adults, as in this sample.60, 80, 81
Early anatomical studies of ADHD did not specifically examine thalamic nuclei, although thalamic hypoactivation emerged in an unbiased meta-analysis.52
Recently, several studies have identified thalamic abnormalities in children/adolescents53, 82
and adults with ADHD.83, 84
Second, our analyses revealed thinner cortex in probands, and particularly those with persistent ADHD, across multiple limbic regions such as temporal poles (BA38), insula (BA13) and subgenual ACC (BA25). The insula and ACC play important roles in sensorimotor, emotional and cognitive function.85, 86
Specifically, subgenual ACC is implicated in emotional processing and pain perception.87
In humans, subgenual ACC is functionally connected with multiple limbic regions including temporal poles88
In turn, the insula, along with participating in performance of demanding tasks,90
is clearly also related to affective processing.91
Abnormal activations in insula and subgenual ACC were reported in a meta-analysis of ADHD functional imaging.52
Cortical thickness studies in ADHD have downplayed findings in the temporal pole, which have been reported but not discussed.12–14
The temporal pole (BA38) is classified as a paralimbic region, based on its interconnections with both amygdala and orbitofrontal cortex, and is implicated in social and emotional processes.92
Altered activation in temporal pole is associated with deficits in face recognition93–100
and mentalizing, i.e., theory of mind.101–104
The temporal poles have been proposed as a channel for the integration of emotion and perception, playing an important role in both emotional and social functions.92
Our findings are consistent with pathophysiological models of ADHD highlighting not only cognitive executive functions (“cool” processes) but also emotion/motivational deficits (“hot” processes).105
Anatomic “spiraling” circuits begin with emotion/motivation pathways which influence “cool” cognitive processes, which in turn control motor responses.106
We observed thinner cortex in regions subserving both emotional regulation (temporal pole, insula, parahippocampus and subgenual ACC) and top-down attentional regulation (dorsal attentional network and medial occipital cortex). Further, our exploratory analyses suggest that thinner cortex and diminished gray matter in the dorsal attentional network and limbic relay regions is related to the trait of having had ADHD in childhood, regardless of current diagnostic status.
Third, the lack of proband-comparison differences in prefrontal cortex or ACC was unexpected.8, 9, 17, 20, 21
To better understand possible differences between persistents and remitters, we performed uncorrected exploratory analyses. In regions in which we found suggestive differences, we observed remarkable congruence between remitters and controls in left superior temporal gyrus, ACC, parahippocampus, and occipital cortical thickness as well as in thalamus and cerebellum gray matter density. We cannot rule out that remitters may have differed from persistents in these regions since childhood, but the most parsimonious explanation is offered by the hypothesis that remission entails compensatory processes12, 107
underpinned by prefrontal cortical maturation. While we found supporting evidence for ACC and orbitofrontal involvement in diagnostic remission of ADHD, our data also suggest superior temporal, medial occipital and thalamo-cerebellar involvement in remission.
Our findings must be interpreted in light of several limitations. First, despite our prospective longitudinal design, we examined brain imaging data only cross-sectionally in middle adulthood. Nevertheless, this is the largest sample of children with ADHD followed into adulthood, obviating the unreliability of retrospective recall of childhood symptoms. Additionally, we report on the largest sample to date of adults with confirmed childhood ADHD who had remitted. We were able to analyze imaging data from only 28% of initially diagnosed probands with ADHD and 45% of comparison subjects. However, these probands and comparisons did not differ from the original sample, and the probands studied did not differ significantly from those excluded on nearly all clinical and demographic variables, except for significantly higher rates of substance use disorders at 18FU in scanned probands. Nevertheless, we did not observe significant relationships between brain anatomic measures and substance use disorders. Finally, as is generally the case, our probands had significantly lower IQ than comparisons both in childhood/adolescence and adulthood. The issue of whether to covary for IQ in disorders such as ADHD is not settled.108
As shown in eFigure 1
and Author e-Table 7
, our principal findings of persistent differences in brain anatomy survived covarying for IQ even with conservative full-brain correction.
We were surprised by the rate of ADHD-NOS diagnosed in comparisons, which was comparable to the rate in probands. We speculate that secular changes in the general public’s awareness of ADHD may have contributed. While we cannot rule out instrument-related error (see Author e-Instrument), using similar approaches did not yield high rates of ADHD symptoms in comparisons in two previous “blind” assessments.24, 26
Nevertheless, analyses excluding ADHD-NOS did not alter results appreciably.
Subjects were limited to Caucasian males, since the number of originally diagnosed females with ADHD was too small for meaningful statistical comparisons. Thus our results may not generalize to ADHD in women or to other racial or ethnic groups. However, this constraint avoided potential confounds from possible sex, ethnic, or socioeconomic differences. Exclusion of conduct disorder comorbidity (see Author e-Text) in childhood also averted confusion as to the origin of the deficits found in cortical thickness or GM density.
We cannot comment on cortical thickness or GM density in ADHD in the absence of medication treatment, as all but four of the scanned probands were treated with methylphenidate as children. We also did not detect significant effects of childhood treatment with stimulants or thioridazine in cortical thickness or VBM analyses. Medication treatment has been reported to affect cortical thickness47
although the durability of such effects is unknown, and treatment had been discontinued for all subjects for several decades.
For logistical reasons, we used two scanners. Fortunately, scans were approximately counterbalanced across probands and comparisons, and there were no significant main effects or interactions related to scanner type. Secondary analyses (see eFigure 2
) also showed that we obtained comparable results when we examined only the 98 scans obtained on the Allegra scanner. Finally, the analyses presented here were limited to cortical thickness and VBM; ongoing analyses will examine white matter structure using diffusion tensor imaging.
In conclusion, in this first study of childhood ADHD prospectively examined in adulthood, we found thinner overall cortex in probands with childhood ADHD that was even more pronounced in those with persistent ADHD. Beyond this global effect, we also detected significant reductions in cortex thickness in parietal, temporal and posterior frontal regions corresponding to the dorsal attentional network and limbic areas. These findings were largely echoed by VBM, which additionally highlighted decreased GM in caudate. These regions underpin top-down control of attention and the regulation of emotion and motivation and were comparably diminished in probands with remitted ADHD or persistent ADHD. Thus these differences seem to primarily reflect the childhood diagnosis of ADHD. By contrast, remitters tended to differ from persistents in medial occipital cortex, temporal pole, insula, orbitofrontal cortex, parahippocampus, frontal pole, and subcortically in cerebellum and thalamus. This supports the suggestion that symptom amelioration and diagnostic remission may result in part from compensatory maturation of frontal thalamic cerebellar circuits.107, 109