Meta-analysis of structural MRI studies in children and adults with attention deficit hyperactivity disorder indicates treatment effects

Authors

  • T. Frodl,

    1. Department of Psychiatry, Trinity College Dublin, Dublin
    2. St. James’s, Adelaide and Meath Hospitals Incorporating the National Children’s Hospital, Psychiatric Services, Dublin
    Search for more papers by this author
  • N. Skokauskas

    1. Department of Psychiatry, Trinity College Dublin, Dublin
    2. Lindara Child and Adolescent Mental Health Service, Dublin, Ireland
    Search for more papers by this author

Thomas Frodl, MD, Department of Psychiatry, Trinity College Dublin, Dublin, Ireland.
E-mail: thomas.frodl@tcd.ie

Abstract

Frodl T, Skokauskas N. Meta-analysis of structural MRI studies in children and adults with attention deficit hyperactivity disorder indicates treatment effects.

Objective:  About 50–80% of ADHD cases have been found to persist into adulthood, but ADHD symptoms change with age. The aim of this study was to perform a meta-analysis of MRI voxel-based morphometry (VBM) and manual tracing studies to identify the differences between adults and children with ADHD as well as between treated and untreated individuals.

Method:  Several databases were searched using keywords ‘attention-deficit and MRI’, ‘ADHD and MRI’. Gray matter volumes from VBM studies and caudate volumes from tracing studies of patients and controls were analyzed using signed differential mapping.

Results:  Meta-analyses detected reduced right globus pallidus and putamen volumes in VBM studies as well as decreased caudate volumes in manual tracing studies in children with ADHD. Adult patients with ADHD showed volume reduction in the anterior cingulate cortex (ACC). A higher percentage of treated participants were associated with less changes.

Conclusion:  Basal ganglia regions like the right globus pallidus, the right putamen, and the nucleus caudatus are structurally affected in children with ADHD. These changes and alterations in limbic regions like ACC and amygdala are more pronounced in non-treated populations and seem to diminish over time from child to adulthood. Treatment seems to have positive effects on brain structure.

Summations

  •  Children with attention deficit hyperactivity disorder (ADHD) have reduced right globus pallidus and putamen volumes as well as decreased bilateral caudate volumes.
  •  With treatment and time, these changes seem to diminish from childhood to adulthood, whereby adults with persistent ADHD symptoms seem to be still characterized by anterior cingulate cortex (ACC) volume reductions.
  •  The meta-analytic data also suggest that untreated children would have additional structural changes in limbic regions like amygdala and ACC.

Considerations

  •  Cross-sectional studies indicate in the meta-analysis that treatment results in the recovery of structural deficits; however, longitudinal studies are necessary to prove this finding.
  •  Gender differences need to be kept with caution as in children most participants were male, whereby adult studies did include equally distributed both males and females.
  •  Voxel-based morphometry studies do not detect caudate nucleus changes, whereby manual tracing studies indicate reduced caudate volumes in children with ADHD, an inconsistency possibly due to heterogeneity in studies.

Introduction

Attention deficit hyperactivity disorder (ADHD) is one of the most prevalent mental disorders of childhood with prevalence rates varying from 5% to 10% (1). Contrary to earlier assumptions, ADHD is no longer considered a disorder exclusive to childhood, as about 50–80% of cases have been found to persist into adulthood, and thus, 3–4% of adults suffer from ADHD (2). In a 11-year longitudinal study, only 20% of the sample either attained functional remission without the aid of psychopharmacological treatment or had an absence of major psychopathology, while the remaining 80% maintained some form of persistence of psychopathology or required continued treatment. Persistent symptoms of ADHD in adults can significantly impair activities of daily living, such as academic, social, occupational, and family functioning (3), which over time can negatively affect the quality of life, especially in the absence of adequate coping skills (4). In adults with ADHD, there seems to be a shift of symptoms with more prominent problems in emotion regulation and disorganization (5). Impulsive aggression like hot temper or short fuse is characteristic for adults with ADHD, and it was proposed that these features should be included into the diagnostic criteria (6, 7). This does not mean that adults with ADHD are not hyperactive any more, because symptoms of hyperactivity can still objectively be measured in adults with ADHD (8). ADHD in adults is associated with significant psychiatric comorbidity, including mood and anxiety disorders, irritability, substance use disorders, and personality disorder (9, 10). Kessler et al. (11) reported that up to 50% of adult patients with ADHD had a complicating mood or anxiety disorder. Adults with ADHD are at a higher risk to develop mood disorders, probably not only because of the psychosocial problems that accompany ADHD, but also because of neurobiological changes that may be predictive of or shared with depression. However so far despite extensive research in experimental and in vivo studies, neither the exact underlying neurobiological substrates of ADHD nor the underlying common biological underpinnings of ADHD and mood disorders have been identified.

As highlighted earlier, ADHD symptoms change with age, and thus, we hypothesized that certain structural brain abnormalities are more prominent in children, whereby others are more prominent in adults with ADHD. Over the past few years, a number of neuroimaging studies were carried out that allowed us to test this hypothesis.

Aim of the study

The aim was to perform a meta-analysis of MRI voxel-based morphometry and manual tracing studies to identify the differences between adults and children with attention deficit hyperactivity disorder as well as between treated and untreated individuals.

Material and methods

Literature search

PubMed, Science Direct, Web of Knowledge, and Scopus were searched from databases inception through January 2011 using the following key words: ‘attention-deficit and MRI’, ‘ADHD and MRI’, ‘attention-deficit and voxel-based morphometry’, ‘attention-deficit and VBM’, ‘ADHD and voxel-based’, ‘ADHD and VBM’, and ‘ADHD and caudate’. The search was confined to English-language articles. Selected articles, as a criterion for inclusion, had to describe an original study. Only those voxel-based morphometry (VBM) MRI studies, which used a whole-brain correction methods like false discovery rate (FDR), familywise error (FWE), or cluster enhanced thresholding, were included. Moreover, another inclusion criteria were that studies used structural MRI. Studies containing duplicated datasets (i.e. reporting the same data in different manuscripts) and studies with <10 patients were excluded. Studies that employed individuals with ADHD and comorbid psychopathic traits and studies that used only screening instruments to confirm the diagnosis of ADHD were excluded too.

Comparison of regional gray matter volumes

Regional differences in gray matter volumes between patients and controls were analyzed using signed differential mapping (SDM; version 1.31, http://www.sdmproject.com), a novel voxel-based meta-analytic approach that improves upon other existing methods (12) and has been used in previous meta-analyses of imaging studies (13–15). The main advantage of SDM is that it uses the reported peak coordinates to recreate maps of the signed (i.e. positive and negative) volume difference between patients and controls, rather than just assessing the probability or likelihood of a peak. This unique feature makes SDM an optimal method for comparing patients with controls without biasing the results toward those brain regions with more interstudy heterogeneity (16).

Signed differential mapping converts VBM coordinates to Talairach space with cluster peaks from VBM studies being represented on an SDM or MRIcroN brain map, highlighting areas of the brain where volume alterations reach significant values, with positive and negative volume changes being represented by different colors. Peak coordinates of volume differences between patients and controls are extracted from each dataset. Peaks that are not statistically significant at the whole-brain level are excluded from these maps. This is carried out in order to ensure that the same statistical threshold throughout the brain is used within each study. Therefore, biases toward liberally thresholded brain regions are avoided, as it is not uncommon in neuroimaging studies that the statistical threshold for some ROIs is more liberal than for the rest of the brain. Then, a standard Talairach map of the differences in brain volume is recreated separately for each study by means of a Gaussian kernel that assigns higher values to the voxels closer to peaks. This includes limiting voxel values to a maximum to avoid biases toward studies reporting various coordinates in close proximity, and reconstructing both increases and decreases in brain volume in the same map. The statistical maps were obtained by calculating the corresponding statistics from the study maps, weighted by the squared root of the sample size of each study, so that studies with large sample sizes have more weight. An omnibus test (Q statistic) was performed to determine whether there were differences in gray matter volume between adult with ADHD and child/adolescent with ADHD. Percentage of treatment and gender were considered as possible confounders and included as cofactors in the analysis. Standard SDM meta-analyses were conducted separately in adult with and child/adolescent with ADHD to determine the differences in gray matter between patients and healthy controls. Percentage of treatment, age, and gender were considered as cofactors in these analyses.

Moreover, meta-regression was carried out in order to determine the association between volumetric differences in patients compared with controls and percentage of treatment, gender, and age in the samples. Statistical significance was determined using standard randomization tests, thus creating null distributions from which P values can be obtained directly (13). We focus on results with P < 0.001 for significance in the group differences and P < 0.0002 for the meta-regression analysis.

Jackknife analysis was also carried out on the included studies to ensure that one study was not significantly affecting our results and that the volumes obtained were highly replicable throughout all of the studies. Moreover, descriptive analyses of quartiles were used to find the actual proportion of studies reporting results in a particular brain region. Maximum location of difference is given in Talaraich coordinates.

Comparison of caudate volumes

Signed differential mapping was further used to carry out a comparison of global caudate volume differences between patients with ADHD and age- as well as gender-matched healthy controls. Statistical significance was obtained by performing a randomization test, and age of study populations was included as covariates.

Results

Included Studies for meta-analysis of voxel-based morphometry studies

Using the literature search terms described in the methods, 890 articles showed up in the search libraries, among which there were 34 VBM studies. Fifteen studies had to be excluded, because they did not investigate the patients with ADHD. Other methods than structural MRI were used in six studies (one diffusion tensor imaging (DTI), one fMRI, one spectroscopy, and two perfusion measurements). From the remaining 13 studies, two had to be excluded, because they did not correct for multiple comparisons or used surface-based measurements. The remaining 11 studies were included in the meta-analysis: four studies were conducted on adults with ADHD and seven on children with ADHD, comprising 320 individuals with ADHD (145 adults and 175 children and adolescents) and 288 healthy controls (115 adults and 173 children and adolescents) (Table 1). The cutoff for age in these studies on children and adolescents was 18 years. The percentage of treated subjects (child: 45% and adult: 34%) did not differ significantly between the two age groups (P = 0.66, Mann–Whitney U-test). However, in children studies, there were significantly more males in comparison with adult studies (P = 0.02, Mann–Whitney U-test). Thus, gender was considered as a covariate in the analysis.

Table 1.   Demographic characteristics of included voxel-based morphometry studies. Depicted is also information on whether studies included depressed individuals or those with an anxiety disorder. Total number of patients with attention deficit hyperactivity disorder (ADHD) = 320. Total number of healthy controls = 288
StudyADHDInstrumentControlsAdults = 1, child = 0Mean age of ADHD (years)Mean age of controls (years)Males (%) in ADHDMales (%) in controlsMethod
FSL = 1
SPM = 2
others = 3
Major depressive disorder
yes = 1
Anxiety
yes = 1
Per cent treatedPer cent (%) affective/anxiety
  1. *Structured Clinical Interview for DSM-IV supplemented with modules from the Schedule for Affective Disorders and Schizophrenia for School-Age Children Epidemiological Version to cover ADHD and other childhood disorders.

  2. †Schedule for affective disorders and schizophrenia for school-age children – epidemiologic version.

  3. ‡Diagnostisches Interview Kiddie Sads-Present and Lifetime Version.

Seidman et al., 2011 (34)74Structured Clinical Interview for DSM-IV, (SCID)*54137.334.35146110328
Amico et al., 2011 (60)20SCID, Conners Adult ADHD scale (CAARS), the self-rate CAARS, the Wender Utah Rating Scale20133.634.775752103030
Almeida Montes et al., 2010 (61)20Clinical judgment; DSM-IV criteria MINI-plus (Spanish version)20128.9527.57505020000
Depue et al., 2010 (62)31A structured interview regarding DSM-IV ADHD2112019.361.148.910077.40
Kobel et al., 2010 (63)14Clinical judgment; DSM-IV-TR criteria, Conners Scale12010.410.910010010110028.7
Sasayama et al., 2010 (64)18Clinical judgment; DSM-IV-TR criteria, ADHD Rating Scale-IV1709.91072.270.6200NA0
Yang et al., 2008 (65)57Chinese version of the K-SADS-E†57011.111.761.459.620087.70
Brieber et al., 2007 (66)15K-SADS-PL‡15013.1313.310010020066.60
McAlonan et al., 2007 (19)28Chinese Diagnostic Interview Schedule for Children for DSMIV3109.69.91001003001000
Carmona et al., 2005 (20)25Diagnosed with ADHD according to the DSM-IV TR criteria, Conners scale25010.811.2848421110064
Overmeyer et al., 2001 (18)18Diagnosed with ADHD according to the DSM-IV, ICD criteria, Conners scale16010.410.383.393.430088.90
 320 288          

Included studies for global differences of caudate volumes

With respect to region of interest (ROI) analysis of structural imaging, 89 studies appeared with search terms ADHD and caudate. Of these, 14 were identified of having used a manual tracing of the caudate. Three of these studies used overlapping samples, and thus, two of these were excluded from the meta-analysis. One study was excluded, because patients with ADHD had psychopathic traits in addition to the ADHD diagnosis. Another study was excluded because of missing data for the meta-analysis and because it did not indicate left and right caudate volumes. And finally, one study was excluded because only the caudate area – not the volume – was measured and reported (17). Two studies were published 20 years ago and were not available online. We contacted corresponding authors of both article; unfortunately, authors did receive our request or did not respond. The remaining seven studies resulted in eight samples, as one study included a group of treated and a group of untreated patients with ADHD compared with healthy controls that were included in the meta-analysis. All studies were conducted in children. The demographic and clinical characteristics of the seven included studies comprising 218 individuals with ADHD and 228 healthy controls are presented in Table 2. Studies on other regions were reviewed too, but not included in the meta-analysis, because there were only <5 studies available.

Table 2.   Demographic characteristics of included manual caudate nucleus tracing studies. Depicted is also information on whether studies included depressed individuals or those with an anxiety disorder. Total number of patients with attention deficit hyperactivity disorder (ADHD) = 218. Total number of healthy controls = 227
StudyADHDInstrumentControlsMean age of ADHDMean age of controlsMales (%) in ADHDMales (%) in controlsMajor depressive disorder, yes = 1Anxiety, yes = 1Per cent treated
  1. + = 15 girls and 36 boys in the sample, numbers in controls and patients not indicated.

  2. *Child Behaviour Checklist (CBCL).

  3. †Diagnostic Interview Schedule for Children–IV–Parent Version.

  4. ‡Schedule for affective disorders and schizophrenia for school-age children – epidemiologic version.

Tremols et al., 2008 (67)39Clinical judgment, DSM-IV TR, Conners’ Scale, CBCL*, Edelbrock Scale3910.911.789.769.200100
Garrett et al., 2008 (68)24(DISC†, DSM-IV criteria, Conners Adult ADHD Diagnostic Interview2216.917.1758011100
Castellanos et al., 2001 (69)50Diagnostic Interview for Children and Adolescents, DSM-IV criteria, Conners scale,509.710000070
Semrud-Clikeman et al., 2000 (70)10Criteria for DSM-III-R on K-SADS‡1112.915.110010000100
Castellanos et al., 1994 (71)50Criteria for DSM-III-R based on Diagnostic Interview for Children and Adolescents4812.312.11001000078
Filipek et al., 1997 (72)15Criteria for DSM-III-R, CBCL*, K-SADS‡, ADHD module1512.412.510010000100
Semrud-Clikeman et al., 2006 (52)16DISC†, DSM-IV, Conners’ Scale2112.7513.2++00100
Semrud-Clikeman et al., 2006 (52)14DISC†, DSM-IV, Conners’ Scale2112.513.2++000
 218 227       

Meta-analysis of voxel-based morphometry studies

Coordinates for the SDM analyses were obtained from all 11 studies. A significant main difference between ADHD and controls was detected for the right globus pallidus and right putamen. There was a main interactive effect of the age group (adult vs. child) on the gray matter volumes in the anterior cingulate cortex (ACC) bilaterally, the right putamen and the right globus pallidus. Adults with ADHD showed decreased gray matter in the ACC, whereas children with ADHD showed decreased right globus pallidus and right putamen volumes (Table 3, Fig. 1). Heterogeneity was not seen on the putamen and ACC but was present in the globus pallidus and in the frontal, temporal, and parietal lobes that did not show significant differences. The main effects remained unchanged when the rate of treatment in per cent was introduced as a covariate. When gender was introduced as a covariate, only the right globus pallidus was significantly different across the age groups (adult vs. child) (122 voxels, maximum at (22, −8, 6), SDM = 1.0, P = 0.000015).

Table 3.   Regional differences in gray matter volume: main effects from the omnibus test
RegionTalairach coordinates, maximum at x, y, zP valueNumber of voxelsAdult with ADHD vs. controlsChild with ADHD vs. controls
  1. ADHD, attention deficit hyperactivity disorder; ACC, anterior cingulate cortex.

Right putamen20, 14, 0<0.00190 0.285
Right globus pallidus22, −8, 6<0.001126 0.277
Right ACC, Brodman 2416, −4, 44<0.001220.193 
Left ACC Brodman 32−2, 18, 30<0.001380.189 
Figure 1.

 Neuroimaging changes derived from voxel-based morphometry studies. (a) At the top shows changes in children with attention deficit hyperactivity disorder (ADHD) compared with healthy controls in the right globus pallidus and the right putamen. (b) Represents the changes in adults with ADHD compared with healthy controls in the right and left ACC.

Children

Children with ADHD showed decreased volumes in the right putamen [40 voxels, maximum at (20, 14, 0), SDM = 0.285, P = 0.0001] and in the right globus pallidus [46 voxels, maximum at (22, −8, 6), SDM = 0.277, P = 0.0003]. When percentage of treatment in the studies was used as a covariate, significant differences were also seen in the right ACC, indicating the studies with more untreated children were more likely to show changes in the ACC too [26 voxels, maximum at (16, −4, 44), SDM = 0.699, P = 0.00008, Brodmann area 24]. Age and gender were not associated with any differences within the children samples.

Using treatment as factor in the meta-regression analysis, children with ADHD showed an inverse association between treatment and left amygdala/uncus volumes [73 voxels, maximum at (−20, −2, −20), SDM = 0.470, P = 0.00005], indicating that experience of treatment might be associated with smaller changes in the area.

Whole-brain jackknife sensitivity analysis revealed that the gray matter decrease in putamen and globus pallidus was highly replicable, as this finding was observed in five of seven combinations of studies/analyses. Interestingly when the study by Overmeyer et al. (18) was excluded from the analysis, there was no significant difference observed. When the study by McAlonan et al. (19) was excluded in addition to the right putamen, the left uncus was found to be significantly reduced in patients with ADHD compared with controls [15 voxels, maximum at (−34, −18, −30), SDM = 0.264, P = 0.0005]. When the study by Carmona et al. (20) was excluded, the right globus pallidus, but not the right putamen, was found to be significantly reduced in patients with ADHD compared with the controls.

In the analysis of quartiles, decreases in gray matter volumes in the putamen and globus pallidus were detected in the third quartile [maximum at (24, 14, −2) and (22, −8, 6], meaning that at least 25% but <50% of studies had found some degree of decreased gray matter in these regions.

Adults

Adults with ADHD had decreased gray matter volumes in the left ACC [22 voxels, maximum at (−2, 18, 30), SDM = 0.189, P = 0.00013, Brodmann area 32] and right ACC [17 voxels, maximum at (16, −4, 44), SDM = 0.193, P = 0.00005, Brodmann area 24].

The small number of studies limits meta-regression analysis between age, gender, treatment, and gray matter differences. Nevertheless, again effects of treatment on left ACC [five voxels, maximum at (−2, 18, 28), SDM = 0.116, P = 0.00004, Brodmann area 24] and on right ACC [six voxels, maximum at (16, −4, 44), SDM = 0.232, P = 0.000009, Brodmann area 24] point toward larger effects of smaller ACC volumes changes in studies that included more patients without previous treatment.

In the meta-regression analysis, age was associated with changes in the left ACC [five voxels, maximum at (−2, 18, 28), SDM = 0.264, P = 0.00004, Brodmann area 24] and the right ACC [six voxels, maximum at (16, −4, 44), SDM = 0.276, P = 0.000009, Brodmann area 24]. Gender was significantly associated with the left ACC [five voxels, maximum at (−2, 18, 28), SDM = −0.098, P = 0.000036, Brodmann area 24] and right ACC [six voxels, maximum at (16, −4, 44), SDM = 0.327, P = 0.000009, Brodmann area 24].

Whole-brain jackknife sensitivity analysis revealed that the gray matter decrease in the ACC was replicable in all four of four combinations of studies. When the study from Amico et al. 2010 was excluded, only the right ACC was significant.

In the analysis of quartiles, decreases in gray matter volumes left and right were detected in the third quartile [maximum at Talairach (6, 6, 36) and (−2, 2, 38)], meaning that at least 25% but <50% of studies had found some degree of decreased gray matter in the same regions.

Results of global caudate volumes derived from manual tracing studies

The meta-analysis of caudate volume studies revealed significant smaller right (d = 0.57, z = 2.5, P = 0.01, upper CI = 1.0, lower CI = 0.13) and significant smaller left (d = 0.50, z = 2.9, P = 0.004, upper CI = 0.85, lower CI = 0.15) caudate volumes in ADHD compared with healthy controls (Fig. 2). The mean volume was 7% smaller for the left and 8% smaller for the right caudate in ADHD compared with healthy controls. All of these studies were carried out in children, and no difference between adult or children population thus can be calculated. Age and gender did not significantly influence the results. The heterogeneity index was significant for the right (τ = 0.29, P < 0.001) and the left (τ = 0.15, P = 0.012) caudate.

Figure 2.

 Effect sizes and standard errors of manual tracing studies for the caudate nucleus. Included is also the effect size and low and higher confidence interval. Mean in the last row represents that mean effect size across studies. (a) Presents the left caudate nucleus; that before mean effect size needs to be deleted. Lower (b), the right caudate nucleus.

When the number of treated children was included as a covariate, there was a significant effect of treatment for the right caudatus (d = 1.34, z = 2.4, P = 0.017, upper CI = 2.5, lower CI = 0.24) and the left caudatus (d = 0.97, z = 2.6, P = 0.009, upper CI = 1.7, lower CI = 0.24), indicating that structural differences were smaller in studies, where more children were treated.

Discussion

To our knowledge, this is the first meta-analysis of VBM MRI studies in subjects with ADHD analyzing the differences between adults and children as well as between treated and untreated patients with regard to the structural neuroimaging abnormalities. The main findings of the meta-analysis of VBM studies are that right globus pallidus and right putamen volumes are reduced in children with ADHD. These changes in the basal ganglia seem to diminish over time from child to adulthood; however, adults with ADHD are characterized by changes in the ACC, in particular when they had not been treated for ADHD. Moreover, meta-analysis of studies using manual tracing of the caudate volume found reduced caudate volumes in children with ADHD most pronounced in those samples that were less treated. Meta-regression indicated that studies in untreated children might have shown also changes in the ACC and the amygdala region.

Reduced right globus pallidus volumes in patients with ADHD compared with age- and gender-matched control subjects are in line with two of three volumetric MRI studies that manually assessed the globus pallidus and found reduced volumes in children with ADHD compared with healthy children (21, 22), whereas a study of 24 youths with ADHD vs. healthy controls found no differences in the globus pallidus (23). Therefore, both VBM and manual tracing studies provide strong evidence that the globus pallidus is volumetrically reduced in subjects with ADHD.

Smaller right putamen volumes in ADHD compared with healthy controls did not survive significance when gender was considered as covariate, indicating that gender plays an important role. As ADHD is an early-onset disorder of childhood affecting males four times more frequently than females (24) the studies in children had mainly included males and only small numbers of females. Thus, the smaller numbers of females might have resulted in smaller power for gender differences in these studies. A possible gender effect here should be not overinterpreted and needs to be investigated in future studies. Interestingly, manual tracing studies were inconsistent with respect to the putamen supporting a higher variability between study populations. A study of 24 youths with ADHD vs. healthy controls found reduced volumes of the mean (left and right) caudate and the putamen in the ADHD group, but no differences in the globus pallidus were reported (23). The anterior putamen also was found to be smaller in a study on 47 children with ADHD compared with 66 healthy controls, whereby the posterior putamen showed increased volumes in ADHD (22). A smaller study of 12 ADHD males and 12 healthy age- and gender-matched controls did not detect significant differences in the putamen (25). In another study of 57 males with ADHD and 55 healthy matched controls, no significant differences were found for the putamen volume (21). Differences in the putamen seem to be less consistent compared with those in the globus pallidus.

Based on the meta-analysis, VBM studies did not provide evidence for changes in the caudate nucleus, whereas the meta-analysis of studies assessing manually the volume of the right and left caudate nucleus showed reductions in children with ADHD vs. control children. This result shows that there is variability between studies and methods with respect to the caudate nucleus, and this might also be indicated by the significant heterogeneity index. The significant heterogeneity index suggests that there is more variation between studies than it would be expected by chance alone. Because age and gender did not modify the results and all studies were carried out in children, this heterogeneity might go back to different study populations in terms of ADHD symptoms. Thus, more research will be necessary with larger samples to further explore this question. For a complete picture, it has also to be acknowledged that there are reports of structural changes in other regions like the hippocampus, amygdala, dorsolateral prefrontal cortex, corpus callosum, and cerebellum: a study on the hippocampal and amygdala volumes in 51 children with ADHD and 63 healthy controls, all aged from 6 to 18 years, found larger hippocampal head volumes, which correlated negatively with clinical ADHD symptoms. The amygdala total volume was not significantly altered, whereas surface analyses indicated reduced size bilaterally over the area of the basolateral complex (26). A study on 27 patients with ADHD and 27 healthy controls did not find significant differences in hippocampus and amygdala volumes between the two groups (27), whereby smaller amygdala volumes, but no changes in hippocampal volumes, were detected in another study of 20 adults with ADHD compared with 20 healthy controls (28). Thus, changes in the amygdala have not consistently been reported in children and adults with ADHD. The hippocampus seems not to be volumetrically altered in ADHD. Abnormalities in frontal lobe regions were detected in children with ADHD (19, 20, 29–31). One study in eight adult patients with ADHD found smaller volumes in the orbitofrontal cortex for the ADHD group (31). Moreover, it also has to be mentioned that the cerebellum, which is involved in a range of cognitive and affective processes (32), was found also to be reduced in volume in some studies (33–35). In a ROI study of the posterior vermis, its inferior posterior lobe was found to be decreased in male children with ADHD compared with comparison children (36). A detailed review exists for these structural changes in children with ADHD (37). However, these changes in the frontal cortices, hippocampus, amygdala, and cerebellum did not show up in our meta-analysis on the basis of VBM, suggesting that they are not so consistent across samples compared with the changes in the basal ganglia. Changes in smaller regions like the amygdala and hippocampus might be more difficult to detect when large cluster threshold corrections for the whole brain are used. The prefrontal cortex and the cerebellum and in particular their subregions are difficult to trace manually, so that this could be another reason why these structures are less frequently investigated and reported results are more inconsistent.

For conclusion, the basal ganglia regions globus pallidus and putamen are found to be reduced in volume in children with ADHD compared with controls. Based on our meta-analysis, volume reduction in the globus pallidus and putamen is the most prominent finding, most pronounced at the right hemisphere and also most consistent across different methods, which are manual tracing vs. VBM. The globus pallidus is known to be involved in controlling subconscious voluntary movement and transmitting information from the putamen and caudate to the thalamus (38). Changes in the basal ganglia are in line with current models that describe a dysfunction of fronto-striatal pathways, which might be related to the imbalances in dopaminergic and noradrenergic systems observed in ADHD (39). Evidence for functional changes in the basal ganglia also derives from positron emission tomography (PET) and single photon emission computer tomography (SPECT) studies that have reported higher than normal dopamine transporter (DAT) density in the striatum of participants with ADHD (40–43). However, some studies also found no differences (44, 45) or decreases (46, 47) in DAT bindings. These studies show that functional changes in the basal ganglia exist; however, the discrepancies between studies have never been elucidated.

Interestingly, these basal ganglia changes mainly were found in children populations but are not detected in adults with ADHD compared with controls. Manual tracing studies of the basal ganglia regions in adults with ADHD had not been performed to date, and this might be an interesting area of research. In adults with ADHD compared with healthy controls, ACC volume reductions were found based on the VBM studies. Moreover, studies that included more previous treated patients reported less change in the ACC, indicating that treatment might influence and protect brain structure. Interestingly, genetic variations in the dopamine system have been found to be important determinants in ADHD. In particular, polymorphism in the 3′ untranslated region (UTR) of the DAT gene SLC6A3, associated with ADHD, might contribute to the regulation of the dorsal ACC function during multisource interference tasks in subjects with ADHD (48). The role of the ACC goes far beyond the often cited function in the processing and regulation of emotional information. Reciprocal connections between the ACC and the lateral prefrontal cortex support its role in cognition. Dense projections from the ACC to the motor cortex and spinal cord seem to implicate this region in motor control. Moreover, extensive afferents from the midline thalamus and the brainstem nuclei point to the importance of arousal/drive state to the function of the ACC (49). Moreover, recent manual structural MRI studies showed significantly smaller ACC in a sample of 24 adults with ADHD when compared with 18 age-matched healthy controls (50). A 21% volume reduction in the left ACC of 13 treatment-naïve adults with ADHD relative to 22 controls was further detected, whereby this study also found a reduction in the right ACC in 13 treated adult ADHD patients compared with the 22 controls indicating a persistence of ACC deficits (51).

Voxel-based morphometry studies in children did not detect ACC changes; however, interestingly, the majority of these studies were carried out in children who received treatment for ADHD and the majority were males. When treatment was considered as a covariate, we detected that in the case of more untreated children, we might also have detected ACC changes. This is in line with one ROI study demonstrating that treatment-naïve children had smaller ACC and caudate volumes compared with controls, whereby this was not the case for treated children (52). In the present meta-analysis, there were less changes in the amygdala/uncus regions when more children were treated in the studies, suggesting that in drug-naïve populations, changes in other brain regions like ACC or amygdala might indeed be present. This hypothesis is supported by a study showing that adolescents receiving pharmacological treatment differed from those not taking medications within a 4-year interval in the rate of change in the cortical thickness in the left middle/inferior frontal gyrus, the medial and inferolateral right precentral gyrus, and the right parieto-occipital region. In this study, an analysis of cortical thickness was undertaken, which showed more cortical thinning in the group of ADHD children not treated compared with the group taking psychostimulants (53). These findings would direct to a protective effect against neural changes when treated for ADHD. Treatment seems to have a normalizing effect on structural deficits and potentially on the developing brain. However, better-designed studies with larger samples using appropriate measures are still needed.

Several studies have recently addressed how methylphenidate changes and normalizes the function in the brain using functional MRI: While treated with methylphenidate, children with ADHD exhibited increased activation of the right frontal cortex during an fMRI task on interference suppression supporting that methylphenidate facilitates cognitive processes (54). Methylphenidate normalized attention differences between children with ADHD and controls by both upregulation of dysfunctional fronto-striato-thalamo-cerebellar and parieto-temporal attention networks and downregulation of hypersensitive orbitofrontal activation for reward processing (55). Moreover, methylphenidate (OROS formulation) increased dorsal anterior middle cingulate cortex activation during a Multi-Source Interference Task in ADHD (56). These observations in fMRI are in line with clinical observation of improved attention and cognition under therapy. However, while methylphenidate had an effect on the verbal working memory performance behaviourally, this effect was not reflected by changes in fMRI (57). There is clearly a need for more research in the area of how treatment in ADHD would be effective on brain function and structure.

Typically developing children without ADHD reach peak cortical thickness in the frontal cortex around the age of 7–8; however, children with ADHD seem to have a slower brain development and reach this milestone later at the age of 10–11 (58). In typically developing individuals, the relative thickness of the right orbitofrontal and inferior frontal cortex increased with age, while the left occipital cortical regions also increased. In ADHD, the right prefrontal component of this brain development was lost, indicating changes in the right hemisphere (59). These findings would be in line with the structural changes that seem to be right prominent in patients with ADHD.

Thus, we may expect more pronounced structural changes in those suffering from ADHD for a longer time, namely those with adult ADHD. This group of patients might have prominent biological symptoms, because ADHD did not disappear. Adult samples might show more changes in the ACC or the amygdala (function of ACC and amygdala in emotion regulation) and not in the basal ganglia, which might be related to the change in symptomatology from hyperactivity to more disorganization and affective regulation problems in those adults or persistence of these symptoms in adults.

Some limitations of the meta-analysis need to be addressed. All studies in the meta-analysis were cross-sectional. While they showed differences between ADHD and controls, it is not entirely possible to conclude on the longitudinal effects from childhood to adulthood based on cross-sectional studies. Whether treatment has an effect on the brain structure in ADHD, e.g. prevents from volume decline in adolescent and adult years, can finally only be answered in follow-up investigations. Currently, no study addressed this issue. Here, we show evidence that a higher number of treated patients in studies are associated with smaller structural change in the basal ganglia, ACC, or amygdala region. Thus, treatment might have a positive effect on long-term changes in the brain, but follow-up studies are required to confirm this hypothesis. While only four studies were undertaken in adults with ADHD, the total sample (145 adults with ADHD and 115 controls) was reasonable and allowed us to make some conclusions about cross-sectional differences, with regard to treatment, gender, and age. The sensitivity test with the analysis of quartiles showed that only between 25% and 50% of the VBM studies found the same results indicating a high variability between VBM studies. When one child study by Overmeyer et al. (18) was excluded in the jackknife analysis, the difference in basal ganglia volumes disappeared in the children subsample. The manual tracing studies showed a high heterogeneity, indicating differences between studies related to the study sample and the methods used to analyze the data. Thus, while there seems to be a volumetric change in the basal ganglia regions, more research in larger samples and clearly longitudinal studies that track changes from childhood to adulthood are required to look at the different development of the brain with regard to ADHD symptomatology. Another general limitation of meta-analysis is that more positive studies might have been included, because positive studies get published more and easier. Although we searched the databases carefully, it might be that a publication bias exists.

For conclusion, changes in the basal ganglia, most prominent in the right globus pallidus, are common in children with ADHD. These changes in the basal ganglia seem to diminish over time from child to adulthood; however, adults with ADHD are characterized by changes in the ACC, mainly when they had not been treated for ADHD. Whether these ACC changes might go ahead with problems in emotional regulation needs further research. Finally, our meta-analysis suggests that treatment might have positive effects in preventing further structural changes during the illness course in regions like the ACC. As there is a considerable variability between studies, the results of the meta-analysis need to be confirmed in larger samples of children and adults with ADHD.

Declaration of interest

None.

Ancillary