Introduction

Characterized by clinical deficits in attention, hyperactivity, and impulsivity, attention deficit hyperactivity disorder (ADHD) is a highly prevalent neurodevelopmental condition with an estimated global prevalence of 3.4% of children worldwide [1]. Presentation types (i.e. subtypes) of ADHD are categorized as either predominantly inattentive (ADHD-I), predominantly hyperactive-impulsive (ADHD-HI) or combined (ADHD-C) [2]. Cognitive and behavioral differences between clinical presentations have been well documented across the neuropsychological literature [3]. Furthermore, the knowledge gap remains in advancing the neurobiological framework in ADHD. Consequently, neuroimaging studies have investigated structural, functional, and more recently connectome features in ADHD which may inform the underlying pathophysiology of ADHD presentations. Moreover, growing evidence from these studies may help better reconceptualize ADHD by linking brain-based features with improved clinical care models and treatment outcomes [4].

Research from task and resting-state functional connectivity studies reveal neural differences between the ADHD-C and ADHD-I types with atypical patterns of increased and reduced activation in the cingulo frontal parietal attentional (CFP) and default mode (DMN) networks [5]. The regions involved in these networks are of significance to the underlying pathophysiology in ADHD as they are concordant with the clinical deficits associated with the ADHD-I and ADHD-C types, respectively. However, evidence of neuroanatomical differences between the ADHD presentation types remains conflicted, with multiple studies reporting no grey matter volumetric differences [6–10], yet others reporting decreased volumes of the caudate and anterior cingulate cortex (ACC) in ADHD-C relative to ADHD-I type [11], regions associated with nodes of the DMN in ADHD-I than ADHD-C [12] and also in the hippocampus in ADHD-C relative to ADHD-I type [13]. A small number of studies investigating microstructural white matter (WM) properties between the ADHD-C and ADHD-I subtypes, using diffusion tensor imaging (DTI), also yield equivocal findings. DTI is an MRI technique which can provide information on WM microstructure by observing the directionality and coherence of water diffusion [14]. This is typically quantified using scalar metrics such as fractional anisotropy (FA), a marker of microstructural architecture; radial diffusivity (RD) to assess axonal myelination; axial diffusivity (AD) as a variable of axonal maturation; and mean diffusivity (MD) as an average of molecular diffusion, independent of directionality [15]. Using these scalar measures to evaluate the ADHD subtypes, Svatkova, Nestrasil [16] reported increased RD in the forceps minor in ADHD-I relative to ADHD-C [16]. Other studies have also reported increased RD bilaterally and AD in mostly left-lateralized fronto-striato-cerebellar regions [17], increased FA and RD in the right thalamus, increased AD in the left postcentral gyrus and right caudate, and increased RD in the left postcentral gyrus and supplementary motor area [18] for ADHD-C relative to ADHD-I. Additionally, studies comparing the ADHD-C type only to controls have observed lower FA values in ADHD-C in regions surrounding the basal ganglia [19,20]. Considering the dearth of studies available in this field, the goal of this study, using both voxelwise analyses of scalar measures and connectomic analyses of tractography measures, was to investigate whether differences in white matter microstructural properties and connectivity, distinguish the two most common ADHD types, ADHD-C and ADHD-I. In this same cohort, we have previously demonstrated differences in graph properties of structural volume covariance networks between the two ADHD types [6]. Based on this our goal was to test whether these ADHD types also differed in organization of their brain white matter.

The evolving framework of ADHD pathophysiology is shaped by the application of neuroimaging measures which include the more recent connectomic approach in assessing possible disruptions in brain network connectivity. An example of this paradigm shift can be seen in the proposal of a neurocircuitry based model in ADHD, which incorporates knowledge on the role of inter-regional network organization involving frontal, temporal and parietal regions, from the historical view of dopaminergic regulation and frontal-striatal circuitry deficit [5]. Network-based Statistics (NBS) and graph theoretical analysis can be applied to assess structural connectivity, which is represented by anatomical connections formed by WM axonal fiber tracts to understand whether these connections underpin functional network connections. Network topology can be characterized by the application of graph modeling of the structural or functional links characterizing the interregional neuronal connections [21]. Only one study to date has compared DTI-based structural connectomic measures between the two ADHD subtypes and found decreased white matter connectivity in ADHD-C compared to ADHD-I, involving mostly right-hemispheric frontal, cingulate and supplementary motor regions [22].

This study utilized DTI data to assess whether structural WM microstructural properties and connectivity may distinguish the ADHD- C and ADHD-I types, and comparatively to neurotypical controls. Firstly, we applied tract-based spatial statistics (TBSS), using DTI derived scalar measures [23] of fractional anisotropy (FA), mean (MD), radial (RD), and axial (AD) diffusivity as indices of water diffusion properties in WM tracts. Secondly, we applied tractography analyses linking 84 brain regions distributed throughout the brain and used network-based Statistics (NBS) to explore connectomic differences and brain network organization, and graph theoretical analysis to capture topographic properties, which may underpin these subtypes.

Methods

Results

Demographic and clinical characteristics for the ADHD-C type, ADHD-I type, and control participants are summarized in Table 1. No significant differences were present between the three groups in terms of age and gender. Medication treatment history, comorbid disorders, or the ADHD-RS-IV (inattentive symptom items) sum of items 1–9 did not significantly differ between subtypes. Characteristically, ADHD-C type significantly differed from ADHD-I on the sum of items 10–18 (hyperactive/impulsive symptom items) and total item scores on the ADHD-RS-IV, in concordance with combined type criteria and severity (p < .05). Of the ADHD-I type participants, one participant qualified for seven items out of 9 hyperactive/impulsive subscale items, less than four 4 items (n = 4), 3 items (n = 2), 2 items (n = 1) and ≤1 item (n = 11).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Participant demographic and clinical characteristics.

https://doi.org/10.1371/journal.pone.0245028.t001

Tract-based statistical analyses and atlas based region of interest analysis

No significant differences were found between the subtypes or relative to controls for any of the DTI scalar measures FA, AD, RD, or MD using voxel-wise analysis. No significant differences were observed for the FA average values for the 46 defined WM regions, however findings from the ROI analyses which were significant at the uncorrected level is shown in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Group comparison of FA values using atlas-based region of interest analysis between the combined (ADHD-C) and inattentive (ADHD-I) subtypes, and controls.

https://doi.org/10.1371/journal.pone.0245028.g001

Forrest plots showing FA values for 46 defined White Matter Regions using region of interest analysis comparing the combined and inattentive ADHD types (Fig 1A), Combined and controls (Fig 1B), and inattentive relative to controls (Fig 1C). The dotted line represents the .05 p-value threshold, and the solid line represents the Bonferroni corrected p-value threshold.

Graph theoretical analysis

Neither of the global topological measures: clustering coefficient or global characteristic path length, showed significant differences between the ADHD-C and ADHD-I groups or compared to controls. Group comparison of the regional nodal degree revealed a single significant finding between the ADHD-C and control group, (p < .05 corrected for multiple comparisons) in the right insula (corrected p = .029; ADHD-C < Controls). There were no significant differences found between the two ADHD subtypes, or the ADHD-I type compared to controls. Regional nodal degree measures which were significant but did not survive after correcting for multiple comparisons are illustrated in Fig 2A–2C.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Group comparison of regional nodal degree measures using graph theoretical analysis between the combined (ADHD-C) and inattentive (ADHD-I) subtypes, and controls.

https://doi.org/10.1371/journal.pone.0245028.g002

Forrest plots showing p-values for 84 brain regions comparing the combined and inattentive ADHD types (Fig 2A), Combined and controls (Fig 2B), and inattentive relative to controls (Fig 2C). The dotted line represents the .05 p-value threshold, and the solid line represents the Bonferroni corrected p-value threshold.

Discussion

This study investigated whether structural WM microstructural properties and inter-regional white matter connectivity distinguished the combined (ADHD-C) and inattentive (ADHD-I) DSM-V presentations of ADHD. TBSS analysis of FA, AD, MD, and RD values, ROI analysis of FA values for 46 white matter tracts defined by the JHU WM atlas and NBS analyses of inter-regional connectivity of 84 cortical-subcortical brain regions revealed no significant differences between the subtypes or compared to controls. Of all the white matter measures, only graph properties of nodal degree for the insula were found to significantly differ in the ADHD-C type relative to controls after controlling for multiple comparisons. Studies of WM microstructural properties between the subtypes of ADHD are limited. In line with the only two previous TBSS studies [16,17] available in the literature, our findings did not find differences in FA, between the two ADHD subtypes. However, both the Svatkova [16] and Ercan [17] studies did observe differences in RD, MD, and AD between the two ADHD types. An earlier study [18] demonstrated FA, RD, and AD differences between the two subtypes, although the analytical approach employed in that study is no longer accepted in current DTI studies which favor TBSS, as a sophisticated model for measures of DTI scalar values [34]. While we did not find any significant white matter tractography differences between the combined and inattentive subtypes of ADHD, only one other study exists that has reported on white matter connectivity differences in the frontal, cingulate and supplementary motor regions between patients with ADHD- C and ADHD-I type using NBS [22]. As such, further research is warranted to determine differences in WM structural integrity between the hyperactive/impulsive subtype relative to the other subtypes.

Another possible reason for null findings in our study may be due to a number of treatment-experienced participants in our study compared to the previous subtype-based studies, which recruited participants who were treatment-naive or with treatment history < 6 months [17,18,22]. Svatkova, Nestrasil [16] had mostly treatment-experienced participants (n = 21, 63%) without washout prior to scanning. Participants in our study who were treatment-experienced (n = 17, 46%) commenced scanning procedures after entering a washout period of 7 days. Interestingly, previous subtype inclusive studies which compared ADHD participants to controls and found significant differences also had a majority of treatment naïve participants [35–38]. However, it is important to note that these previous studies are not comparable with our study in terms of methodological approaches. Furthermore, a recent DTI study investigated a medication naïve ADHD group of children and adults and found no differences to controls in the child group but reported significant findings in the treatment naïve adult group [39]. Nevertheless, it may be likely that white matter differences in ADHD may be influenced by stimulant treatment effects but may also be due to neurodevelopmental periods in line with the proposed maturational lag theory in ADHD [40].

Consistent with the two previous subtype comparison studies utilizing TBSS, we found no differences in FA values [16,17] or AD and MD values, distinguishing the two types [16]. Notably, while both studies employed cohorts that were similar to our study in terms of clinical characteristics and sample size, the Ercan, Suren [17] study included a treatment naïve sample. Interestingly, increased FA values were found relative to controls, in the bilateral cingulum bundle in ADHD-C, and the anterior thalamic radiation, inferior longitudinal fasciculus and left corticospinal tract involving regions linked to fronto-striatal-thalamic circuits in ADHD-I [16]. Increased RD in the forceps minor in ADHD- I relative to ADHD-C [16] and increased RD bilaterally and increased AD in brain regions mostly on the left side linked to fronto-striato-cerebellar regions in ADHD- C relative to ADHD-I [17] have also been reported. In contrast, also using a TBSS approach, Rossi et al. [41] measured FA values for 20 fiber tracts in ADHD-I children compared to controls and found no significant difference between the groups.

Additionally, we performed a region of interest (ROI) analysis and did not find significant differences in FA values between the two types. Despite the absence of significant differences between groups at the corrected thresholds, increased FA values in ADHD-C compared to ADHD-I and controls were observed at uncorrected thresholds. That is, the bilateral posterior corona radiata characterized the ADHD-C type from ADHD-I and controls, the left posterior limb of internal capsule and left anterior limb of internal capsule in ADHD-C relative to ADHD-I, and the left uncinate fasciculus in ADHD-C compared to controls. Interestingly, similar findings from a previous study comparing an ADHD group reported reduced FA values in the ADHD groups in the posterior corona radiata, the uncinate fasciculus with reduced MD in the posterior limb of internal capsule, relative to controls [42].

In our study, we also used tractography to evaluate white matter connectomic differences. Our study did not find WM connectivity differences between the ADHD-C and ADHD-I subtypes or compared to controls. However, previous findings from Hong, Zalesky [22], the only other study evaluating connectome differences between the subtypes, observed decreased WM connectivity differences in ADHD-C relative to ADHD-I connecting 17 right hemispheric brain regions including the superior frontal gyrus, anterior cingulate gyrus, and supplementary motor areas. Additionally, they found significant negative correlations between FA values and continuous performance task scores to differentiate ADHD-C and ADHD-I. Also, we did not find any significant differences between the two ADHD subtypes for global topological characteristics or regional nodal graph measures. However, a single significant finding was observed which showed reduced nodal degree involving the right insula in the ADHD- C group when compared to controls. Whilst speculative, similar findings involving the insula have been noted in previous structural [43–45] and functional connectivity [46,47] studies in ADHD and warrant further research. The insula is considered part of the salience network which interacts with the default mode network. Reduced degree is this region may account for attentional processing and goal-directed action deficits and tied to mechanisms of executive functioning and sensorimotor dysregulation associated with ADHD-C type [48]. Furthermore, this finding may have implications for attentional and inhibitory emotional processing in ADHD [45]. At the uncorrected thresholds, regional network measures revealed some interesting results; with increased nodal degree in regions involving the occipital temporal, and parietal lobes in ADHD-I relative to ADHD-C and controls and increased nodal degree in regions of the frontal lobe and the basal ganglia in ADHD-C compared to ADHD-I and controls. These findings are consistent with the Hong, Zalesky [22] study that also revealed decreased connectivity in ADHD-C, relative to ADHD-I, in frontal regions, which are involved in regulating inhibitory response and motor activity, both of which are linked to the ADHD-C type.

Although the availability of DTI studies which specifically examine subtype differences are sparse, significant findings have been reported, albeit equivocal. Undoubtedly, further studies are warranted to establish structural connectivity alterations in ADHD types, however, emerging evidence from DTI studies examining an all-inclusive ADHD group with controls have reported significant findings. For example, studies investigating basal ganglia and thalamic connectivity have found lower FA values in ADHD-C relative to controls [19,20]. Fall and colleagues [49] indirectly support these findings reporting mean reaction times correlated with MD values in the striatum and thalamus in ADHD-C compared to controls. Interestingly, these results are consistent with studies using structural gray matter and resting state functional connectivity, reported in a recent review [50] which have shown disruptions in the motor network in ADHD-C. Whether these structural connectivity disruptions or the absence of WM abnormalities attribute to functional deficits associated with the subtypes of ADHD, remain unconfirmed. And so, while the present study focused on whether hard-wired networks may shed insight that distinguishes the subtypes, further exploration of the functional dynamics of brain network organization may potentially provide greater insight toward the underlying neurobiology of the ADHD subtypes. Importantly, the rationale for the DSM-V based diagnostic criteria adopting ‘presentation type’ in place of ‘subtypes’ was to try to convey the somewhat transient nature to these classifications and how these clinical symptoms evolve throughout development [50,51]. Perhaps this is another potential reason for the null findings as ADHD presentation types are not stable across time, and thus they may be better linked with functional measures.

Conclusion

In summation, we did not find white matter microstructural properties or network connectivity to differentiate the two subtypes from each other or relative to controls. However, we did find the ADHD-C group to differ in regional nodal degree in the insula relative to controls. While the overall results of this study may be inconsistent with the findings of previous ADHD subtype DTI studies, the absence of evidence for differences in WM microstructural properties relative to FA values between the two subtypes is in line with previous research. To the best of our knowledge, this is the first study to utilize both voxel-wise analysis of scalar white matter measures in addition to tractography incorporating network and graph theoretical measures to examine differences between these two ADHD subtypes.

Supporting information

Acknowledgments

We thank Dr. Lavier Gomes and Ms. Sheryl Foster and the Department of Radiology at Westmead Hospital for their substantial contributions to magnetic resonance imaging data acquisition. We thank Tracey Tsang who served as the iSPOT-A trial coordinator, along with Sariah Hobby, Yennie Hyunh and Jodie Logan who assisted in data acquisition. We also thank the individuals who gave their time to participate in the study.