Introduction
One of the most fascinating aspects of problems with motor coordination or clumsiness is its frequent overlap with problems with inattention, hyperactivity and impulsivity. Here we examine this overlap, taking both dimensional and categorical approaches. A dimensional approach considers how motor coordination skills, treated as continuous variables, are associated with the severity of symptoms of inattention, hyperactivity and impulsivity, also treated continuously. This dimensional approach is appropriate given that the tool most widely used to assess motor coordination returns values that are normally distributed (Brown & Lalor, Reference Brown and Lalor2009). Similarly, epidemiological and neuropsychological evidence suggests that attention-deficit/hyperactivity disorder (ADHD) can be considered dimensionally, lying at the extreme end of a continuous distribution of symptoms (Polderman et al. Reference Polderman, Derks, Hudziak, Verhulst, Posthuma and Boomsma2007; Lubke et al. Reference Lubke, Hudziak, Derks, Van, jsterveldt and Boomsma2009). In general, normally distributed quantitative traits contain more information about between-individual variability than a dichotomous diagnostic category. Thus treating motor coordination skills as continuous variables is a powerful strategy when mapping their neuroanatomic substrate. A similar approach can be adopted to determine whether the severity of symptoms of inattention, hyperactivity and impulsivity (considered dimensionally) moderate the relationship between motor coordination skills and the brain.
The overlap can also be examined using Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) diagnostic categories. Developmental coordination disorder (DCD) represents the impairing end of an underlying dimension of motor coordination skills and ADHD reflects the presence of clinically significant symptoms of inattention, hyperactivity and impulsivity. Epidemiological studies show that around 50% of children with DCD also have ADHD and vice versa (Kadesjo & Gillberg, Reference Kadesjo and Gillberg1999; Kaiser et al. Reference Kaiser, Schoemaker, Albaret and Geuze2015). This co-morbidity is so marked that it has been argued that it delineates a distinct nosological entity: a ‘Disorder of Attention, Movement and Perception’ (Gillberg et al. Reference Gillberg, Rasmussen, Carlström, Svenson and Waldenström1982; Gillberg, Reference Gillberg1987). Support for this proposition comes from natural history, as children with the combination of ADHD and impairing motor dyscoordination have a worse adult outcome than those with ADHD alone. A twin study suggests that the combination of ADHD and DCD has genetic etiology which is distinct from the genetic factors contributing to either ADHD or DCD alone (Martin et al. Reference Martin, Piek and Hay2006). The combination of ADHD and DCD clusters within families, also consonant with the argument that this combination is a distinct subtype of ADHD rather than just an expression of increased severity of the disorder (Fliers et al. Reference Fliers, Vermeulen, Rijsdijk, Altink, Buschgens, Rommelse, Faraone, Sergeant, Buitelaar and Franke2009). Given the clinical interest in the overlap between ADHD and DCD, we conduct categorical analyses. Specifically, we ask whether the brain regions associated with motor coordination skills differ between four groups: DCD alone, ADHD alone, combined DCD/ADHD, and completely unaffected. The most pertinent question is whether the motor coordination regions differ between those with DCD alone compared with those with DCD and co-morbid ADHD.
What do we already know of the neural circuits underpinning motor coordination? A wealth of functional neuroimaging studies implicates circuitry encompassing the cerebral cortex (motor/premotor/anterior cingulate), basal ganglia (motor putamen, globus pallidus), thalamus (ventrolateral and ventromedial) and cerebellum (lobules VIIA and VIIB) (Graff-Radford et al. Reference Graff-Radford, Eslinger, Damasio and Yamada1984; Picard & Strick, Reference Picard and Strick1996; Karussis et al. Reference Karussis, Leker and Abramsky2000; Paus, Reference Paus2001; Stoodley & Schmahmann, Reference Stoodley and Schmahmann2010; Brown-Lum & Zwicker, Reference Brown-Lum and Zwicker2015). A consistent finding is that the different aspects of motor coordination have partly distinct neural substrates (Park et al. Reference Park, Lee, Han, Lee, Lee and Park2009, Reference Park, Lee, Han, Lee, Lee and Park2011, Reference Park, Lee, Kim, Park, Won, Jung and Yoon2012; Pangelinan et al. Reference Pangelinan, Zhang, Vanmeter, Clark, Hatfield and Haufler2011; Kühn et al. Reference Kühn, Romanowski, Schilling, Banaschewski, Barbot, Barker, Brühl, Büchel, Conrod, Czech, Dalley, Flor, Garavan, Häke, Ittermann, Ivanov, Mann, Lathrop, Loth, Lüdemann, Mallik, Martinot, Palafox, Poline, Reuter, Rietschel, Robbins, Smolka, Nees, Walaszek, Schumann, Heinz and Gallinat2012; Paola et al. Reference Paola, Caltagirone and Petrosini2013; Baer et al. Reference Baer, Park, Bailey, Chakravarty, Li and Penhune2015). However, there are several gaps in our knowledge. First, it is unclear to what extent brain structure follows function in motor coordination. Most neuroanatomic studies have focused on just one region of interest (mainly the cerebellum; Doya, Reference Doya1999; Hutchinson et al. Reference Hutchinson, Lee, Gaab and Schlaug2003; Park et al. Reference Park, Lee, Han, Lee, Lee and Park2009, Reference Park, Lee, Han, Lee, Lee and Park2011; Paola et al. Reference Paola, Caltagirone and Petrosini2013; Baer et al. Reference Baer, Park, Bailey, Chakravarty, Li and Penhune2015), have assessed only one domain of motor coordination (Pangelinan et al. Reference Pangelinan, Zhang, Vanmeter, Clark, Hatfield and Haufler2011; Kühn et al. Reference Kühn, Romanowski, Schilling, Banaschewski, Barbot, Barker, Brühl, Büchel, Conrod, Czech, Dalley, Flor, Garavan, Häke, Ittermann, Ivanov, Mann, Lathrop, Loth, Lüdemann, Mallik, Martinot, Palafox, Poline, Reuter, Rietschel, Robbins, Smolka, Nees, Walaszek, Schumann, Heinz and Gallinat2012; Baer et al. Reference Baer, Park, Bailey, Chakravarty, Li and Penhune2015) or considered individuals with exceptional skill in motor coordination (Park et al. Reference Park, Lee, Han, Lee, Lee and Park2009, Reference Park, Lee, Kim, Park, Won, Jung and Yoon2012; Abdul-Kareem et al. Reference Abdul-Kareem, Stancak, Parkes and Sluming2011; Di et al. Reference Di, Zhu, Jin, Wang, Ye, Zhou, Zhuo and Rao2012). The literature thus needs a definition of the neural substrate of motor coordination that encompasses all of the pertinent brain regions (cerebrum, cerebellum, thalamus and basal ganglia) and uses a comprehensive standardized assessment of all facets of motor coordination.
We have three aims. First, we define the neuroanatomic substrate for motor coordination in youth. Second, we ask if this motor behavior–brain link is moderated by symptoms of ADHD, treating both motor coordination and ADHD symptoms as continuous variables. Third, we ask if the neural substrate of motor-coordination substrate differs between four diagnostic groups: children with DCD and ADHD combined, children with either DCD or ADHD alone, and completely unaffected children.
Method
Participants
In all, 226 children participated. The general inclusion criteria were: (1) age between 4 and 16.9 years; (2) intelligence quotient (IQ) of 80 or greater, defined using age-appropriate version of the Wechsler intelligence scales. For the ADHD group, we used the DMS-5 criteria for inclusion. Specifically, parents or guardians of all participants were interviewed by one of two clinicians (W.S. or P.S.) using the Diagnostic Interview for Children and Adolescents to assess for symptoms of inattention, impulsivity and hyperactivity (Reich, Reference Reich2000). A symptom was defined as present if it had an adverse impact on functioning at school, home or with peers. The inclusion criteria for the probable DCD group were ascertained using the Developmental Coordination Disorder Questionnaire (http://www.dcdq.ca). This is a parent-completed screening tool, which is widely used to aid in the diagnosis of probable DCD. It determines if a child shows impairment relative to peers in three domains, namely, control during movement, fine motor/handwriting skills and general coordination. Age-appropriate cut-offs are applied to determine if the scores indicate or raise or suspicion of DCD. The questionnaire has a sensitivity of 84.6% and specificity of 70.8% for the DSM-based diagnosis of DCD (Wilson et al. Reference Wilson, Crawford, Green, Roberts, Aylott and Kaplan2009). As we used a questionnaire, the term ‘probable DCD’ is used to refer to this group. General exclusion criteria were: (1) neurological disorders affecting movement; (2) gross anatomic anomalies on magnetic resonance imaging (MRI); (3) psychiatric disorders other than ADHD, oppositional defiant disorder or conduct disorder; (4) psychotropic medications other than psychostimulants and guanfacine (standard treatments for ADHD).
Medication histories were obtained from parents. In the ADHD group, 66 (61%) were medication naive, 30 (27%) were treated with methylphenidate preparations, 11 (10%) amphetamines and two (2%) guanfacine. Participants were recruited through advertisement and community contacts, and were predominantly from areas surrounding the study center. The procedures were approved by the institutional review board. Parents or guardians gave informed consent and children gave assent.
Neuropsychological testing was conducted by research fellows supervised by a clinical neuropsychologist. Motor coordination was assessed by age-appropriate versions of the Movement Assessment Battery for Children (MABC; Henderson et al. Reference Henderson, Sugden and Barnett2007). This assesses three domains of motor coordination: aiming/catching, manual dexterity and balance. Scores were standardized using population norms. This tool has become the ‘gold standard’ for the assessment of motor coordination (Chow et al. Reference Chow, Hsu, Henderson, Barnett and Lo2006). We did not assess inter-rated reliability for the MABC. If subjects were taking psychostimulant medication this was withdrawn at least 24 h before testing as the medication may have an impact on performance on neuropsychological tests (Brossard-Racine et al. Reference Brossard-Racine, Shevell, Snider, Bélanger and Majnemer2012).
A high-resolution (1.07 × 1.07 × 1.2 mm) T1 weighted volumetric structural image was obtained using a magnetization prepared rapid gradient echo sequence (with ASSET preparation; 124 slices, 1.2 mm slice thickness, 224 × 224 acquisition metric, flip angle = 6°, field of view = 24 cm2) on a 3 T General Electric Signa scanner (USA) using an eight-channel head coil. We examined the links between motor coordination and four brain ‘compartments’ (cerebral cortex, cerebellum, basal ganglia and thalamus) [see online Supplementary Fig. S1 for an overview of the processing pipeline, conducted on the National Institutes of Health High Performance Computer Cluster (Biowulf). Volumes for the striatum, globus pallidus, and the cerebellum and its lobules were estimated using the MAGeT Brain technique (Chakravarty et al. Reference Chakravarty, Steadman, Van, de, Calcott, Gu, Shaw, Raznahan, Collins and Lerch2013; Park et al. Reference Park, Pipitone, Baer, Winterburn, Shah, Chavez, Schira, Lobaugh, Lerch and Voineskos2014). To summarize this technique, single atlases for the striatum, the globus pallidus and the thalamus that were defined using a three-dimensional reconstruction of serial histological data (Chakravarty et al. Reference Chakravarty, Bertrand, Hodge, Sadikot and Collins2006a ) were warped to an MRI-based template. The atlas for the cerebellum was defined through expert manual labeling of five cerebella. The atlases were then customized to a subset of the dataset (21 randomly selected subjects) using a non-linear transformation estimated in a region of interest defined around the subcortical structures and cerebellum (Chakravarty et al. Reference Chakravarty, Bertrand, Hodge, Sadikot and Collins2006a , Reference Chakravarty, Sadikot, Germann, Bertrand and Collins2008, Reference Chakravarty, Steadman, Van, de, Calcott, Gu, Shaw, Raznahan, Collins and Lerch2013). This set of subjects now acts as a set of templates and all other subjects are now warped to these templates. This provides 21 candidate segmentations for each subject's basal ganglia, thalamus and cerebellum. The final segmentation is decided upon using a voxel-wise majority vote, that is, the label occurring most frequently at a specific location is retained (Collins & Pruessner, Reference Collins and Pruessner2010). Volumes of each structure were determined from these segmentations. These methods are reliable in comparisons against ‘gold standard’ manual definitions of the striatum and thalamus (κ = 0.86) and the cerebellum (κ > 0.7) (Chakravarty et al. Reference Chakravarty, Steadman, Van, de, Calcott, Gu, Shaw, Raznahan, Collins and Lerch2013). We used this subcortical segmentation tool as it affords a finer level of subregional volumetric definition than is available in most other tools. It provides 33 cerebellar subregions (11 lobular regions in each hemisphere and 11 vermal subregions); 22 thalamic subregions and 12 basal ganglia regions. All image processing was performed using the MINC suite of image processing tools (http://www.bic.mni.mcgill.ca/ServicesSoftware/MINC) and all non-linear registrations were performed using a version of the ANTs algorithm (that was adapted to work with the MINC tools; https://github.com/vfonov/ANTs). The pipeline and atlases are all publicly available (http://cobralab.ca/software/; http://cobralab.ca/atlases/). We have validated the localization of the thalamic and basal ganglia subregions though comparison with expert manual segmentation (Chakravarty et al. Reference Chakravarty, Sadikot, Germann, Bertrand and Collins2008, Reference Chakravarty, Steadman, Van, de, Calcott, Gu, Shaw, Raznahan, Collins and Lerch2013), intra-operative electrophysiological recordings and stimulations (Chakravarty et al. Reference Chakravarty, Sadikot, Germann, Bertrand and Collins2008), post-operative localization for deep brain stimulation electrodes (Chakravarty et al. Reference Chakravarty, Sadikot, Mongia, Bertrand and Collins2006 b), activations of the sensory network as recorded using functional MRI (Chakravarty et al. Reference Chakravarty, Broadbent, Rosa-Neto, Lambert and Collins2009a , Reference Chakravarty, Rosa-Neto, Broadbent, Evans and Collins b ), and diffusion-tensor tractography based connectivity (Leh et al. Reference Leh, Ptito, Chakravarty and Strafella2007, Reference Leh, Chakravarty and Ptito2008).
Cerebral cortical reconstruction and volumetric segmentation were performed with the Freesurfer image analysis suite version 5.3.0 (http://surfer.nmr.mgh.harvard.edu/). Technical procedures of this widely used method are described in the weblink above. All raw T1 images were visually inspected, and those with moderate or severe motion or other artifacts were excluded (about 20%). The subcortical, cerebellar and cerebral cortical segmentations were also visually checked for segmentation quality, resulting in the further exclusion of around 10% of the segmentations.
Analyses
First, we defined the relationship between motor coordination skills and the brain. Volumes within each of the brain's major compartment – the cerebral cortex, basal ganglia, thalamus and cerebellum – were highly correlated, raising issues of multicollinearity. We thus extracted the principal components for each of the four brain compartments. Varimax rotation was used and components were retained with eigenvalues greater than 1. These components were then used to determine four latent variables (the cerebral cortex, basal ganglia, thalamus and cerebellum). These latent variables were then mapped onto each measure of motor coordination using partial least squares pathway modeling (implemented in R, using the program PLSPM; Sanchez, Reference Sanchez2013). The technique maximizes the explained variance of all dependent variables based on how they relate to their neighboring constructs with a predictive purpose (Tenenhaus et al. Reference Tenenhaus, Vinzi, Chatelin and Lauro2005). This approach was chosen as the latent variables are determined by the multiple morphometric components (rather than reflecting these components). It also does not impose any distributional assumptions and is well suited to medium sample sizes (Sanchez, Reference Sanchez2013).
Partial least squares pathway models are comprised of an outer model which reflects the relationships between a latent variable and its block of manifest variables. Each latent variable is calculated as a weighted sum of its manifest variables. In our model, the manifest variables are the principal components of the four different brain regions. Within each region, these are uncorrelated as they are principal components extracted using a varimax rotation which is critical for a model in which the latent variable is considered to be formed by the manifest components.
The inner model considers the relationships between the different latent neuroanatomic and outcome variables. Path coefficients are calculated between each of the four neuroanatomic latent variables and the three measures of motor coordination separately. Effect sizes and corresponding significance values for the paths between the latent variables are determined and bootstrapping (with 10 000 resamples) is used to evaluate the precision of the parameter estimates. We modeled each of the three aspects of motor coordination separately, as they are thought to be underlined by partly distinct neural substrates (Park et al. Reference Park, Lee, Han, Lee, Lee and Park2009, Reference Park, Lee, Han, Lee, Lee and Park2011, Reference Park, Lee, Kim, Park, Won, Jung and Yoon2012; Pangelinan et al. Reference Pangelinan, Zhang, Vanmeter, Clark, Hatfield and Haufler2011; Kühn et al. Reference Kühn, Romanowski, Schilling, Banaschewski, Barbot, Barker, Brühl, Büchel, Conrod, Czech, Dalley, Flor, Garavan, Häke, Ittermann, Ivanov, Mann, Lathrop, Loth, Lüdemann, Mallik, Martinot, Palafox, Poline, Reuter, Rietschel, Robbins, Smolka, Nees, Walaszek, Schumann, Heinz and Gallinat2012; Baer et al. Reference Baer, Park, Bailey, Chakravarty, Li and Penhune2015), and they were only modestly correlated (aiming/catching and manual dexterity, r = 0.29; aiming/catching and balance, r = 0.37; balance and manual dexterity, r = 0.34). This analysis thus mapped paths between the brain and motor coordination behavior for the entire sample. We treated motor coordination as a continuous variable as it was normally distributed in this cohort.
To address our second aim, we determined whether the paths between motor coordination skill and the brain were moderated by symptoms of ADHD. We created two additional latent variables; the first for total ADHD symptom count and the second reflecting the interaction between motor coordination and ADHD symptom count. We tested whether the ‘interaction’ latent variable mapped significantly to the neuroanatomic latent variables. A significant path coefficient would indicate that the relationship between motor coordination and the brain is moderated by ADHD symptom severity.
To address our third aim, we conducted categorical analyses, comparing the four groups (unaffected, probable DCD alone, ADHD alone, combined DCD/ADHD). A one-way analysis of variance (ANOVA) tested for group difference in the regions that our primary analyses linked with motor coordination skill. A planned pairwise contrast tested for differences between the DCD groups with and without ADHD.
Finally, we tested for sex differences in the path coefficients, using permutation tests. To test for age effects we compared the three age group versions of the motor coordination tool (4 up to 6 years, 11 months; 7 up to 10 years, 11 months; 11 up to 16 years, 11 months).
Results
Behavioral
Demographic data for the cohort with motor coordination skill assessment (n = 226) are shown in Table 1. Motor coordination scores were negatively correlated with symptoms of inattention (total score r = −0.24, p = 0.0003; for aiming/catching r = −0.15, p = 0.03; balance r = −0.21, p = 0.002; manual dexterity r = −0.17, p = 0.009) and with hyperactivity/impulsivity (total score r = −0.26, p = 0.001) and aiming/catching (r = −0.22). IQ was inversely correlated with total motor coordination score (r = 0.16, p = 0.02) and with manual dexterity (r = 0.28, p < 0.0001) and balance (r = −0.14, p = 0.05) but not aiming/catching (r = −0.04, p = 0.62). Results from a repeated-measures ANOVA indicated no main effect of sex on motor coordination (F 1,224 = 0.35, p = 0.55). However, there was a significant interaction between sex and the motor coordination subtest (F 2,448 = 25.6, p < 0.0001). Post-hoc analyses showed that females outperformed males in manual dexterity [females mean score 9.7 (s.d. = 3.2), males 8.4 (s.d. = 3.0), t = 3.1, p = 0.002], males outperformed females in aiming and catching [males 10.3 (s.d. = 3.4), females 9.0 (s.d. = 2.9), t = 2.7, p = 0.007], and there was no sex difference in balance [males 10.3 (s.d. = 2.8), females 10.8 (s.d. = 3.1), t = 1.24, p = 0.22]. The interaction effect between sex and diagnosis was not significant for motor coordination subtests (F 2,444 = 0.74, p = 0.48).
Table 1. Demographic, neuropsychological and clinical detailsa
Data are given as mean (standard deviation) unless otherwise indicated.
DSM-5, Diagnostic and Statistical Manual of Mental Disorders, fifth edition; ADHD, attention-deficit/hyperactivity disorder; DCD, developmental coordination disorder.
aThere were 226 participants with motor coordination data; 170 with data on DCD. Missing data are indicated by the degrees of freedom.
b n = 170.
Developmental Coordination Disorder Questionnaire data were available on 170 (75%) of the participants. The diagnostic groups did not differ in age, sex or IQ (see Table 1). As would be expected, the ADHD groups had more symptoms of inattention, hyperactivity/impulsivity; and the DCD groups had lower motor coordination scores.
Brain–behavior relationships
We first estimated the path coefficients from the anatomic latent variables to each subtest of motor coordination separately. These paths were significant for links between the aiming/catching test and the cerebral cortical latent variable [t = 4.31, p < 0.0001, effect size 0.29, 95% confidence interval (CI) 0.2–0.48] and between aiming/catching and the cerebellar latent variable (t = 2.31, p = 0.02, effect size 0.18, 95% CI 0.06–0.31) (Fig. 1). Path coefficients linking aiming/catching with the thalamic and basal ganglia latent variables did not reach significance. Overall, the latent variables explained a moderate amount of the variance in the aiming/catching scores (R 2 = 0.22, 95% CI 0.2–0.48). Next, we identified the brain regions with the strongest contributions to these links. This is provided by the weights of each principal component onto the latent constructs (see online Supplementary Table S1 for all external weights). For the cerebral cortical construct, two components weighted significantly onto the latent construct (Fig. 2). The first component (weight effect size of 0.49, 95% CI 0.13–0.64) reflected the volumes of the bilateral precentral, superior frontal and caudal middle frontal cortex. A second cortical component (weight 0.34, 95% CI 0.007–0.53) reflected the volumes of the inferior temporal, fusiform and lateral occipital cortex. For the cerebellar latent variable, the component that weighed most heavily (weight 0.66, 95% CI 0.31–0.82) reflected the volumes of lobules VIIA, VIIB and Crus II bilaterally. A second significant cerebellar component (weight effect size of 0.52, 95% CI 0.13–0.78) reflected the volumes of lobules and vermal regions V, VI and X. There were no significant path coefficients between the anatomic latent variables and balance or manual dexterity.
Fig. 1. Pathway model between the neuroanatomic latent variables and the aiming/catching subtest of motor coordination. The pathways coefficients between the cerebral, cerebellar latent variables and aiming/catching were significant. The brain regions which made the strongest contributions to these links are listed, along with the weights of the brain regions onto the corresponding neuroanatomic latent construct. Fig. 2 shows the location of these brain regions.
Fig. 2. Brain regions most robustly associated with aiming/catching skills. The weight of each component onto the latent neuroanatomic variables is shown in Fig. 1.
Next, we tested whether these path coefficients differed in the presence of ADHD symptoms (treated as a continuous variable). We created two latent variables – one for ADHD symptoms and the second for its interaction with aiming/catching. The path from the ‘interaction’ latent variable to the neuroanatomic variables was not significant (cerebellar path coefficient, t = 0.87, p = 0.37, effect size = −0.07, 95% CI −0.21 to 0.09; cortical path, t = 1.11, p = 0.27, effect size = 0.08, 95% CI −0.07 to 0.24; basal ganglia path, t = 1.06, p = 0.29, effect size = −0.1, 95% CI −0.27 to 0.07 ; thalamic path, t = 0.14, p = 0.89, effect size = 0.01, 95% CI −0.18 to 0.21). Thus, the relationship between motor skill and the brain was not moderated by the severity of ADHD symptoms.
Group contrasts
Next, we tested for differences between the diagnostic groups (unaffected, probable DCD alone, ADHD alone, combined DCD/ADHD). We examined the regions that our primary analyses associated with motor coordination skill. We summed the two cerebellar regions into one variable, and the two cerebral cortical regions into another. There was a significant group difference in the cortical region (F 3,166 = 2.85, p = 0.04) and a trend difference in the cerebellar region (F 3,166 = 2.4, p = 0.07) (Fig. 3). There were no differences in the planned contrast of the group with probable DCD only against the group with DCD/ADHD (cerebellar region, t 61 = 0.37, p = 0.72; cortical region t 61 = 0.69, p = 0.49). Likewise, there were no differences between the ADHD group only compared with the DCD/ADHD group (cerebellar region t 81 = 0.1, p = 0.92; cortical region t 81 = 1.31, p = 0.19). There were, however, significant pairwise differences between the unaffected group and the clinical groups (see Fig. 3). Thus, brain regions associated with motor coordination skill were atypical in those with probable DCD, but did not differ further with the presence or absence of co-morbid ADHD.
Fig. 3. Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) group differences in regions associated with motor coordination skill. Values are means, with 95% confidence intervals represented by vertical bars. The horizontal lines indicate pairwise group differences (p < 0.05). DCD, Probable developmental coordination disorder, ADHD, attention-deficit/hyperactivity disorder.
Medication effects
We repeated the pathway analysis removing the 43 participants who were on psychostimulant medication, leaving 183 subjects with motor coordination data. The path from the cerebellar latent variable that mapped to aiming/catching was essentially unchanged (effect size 0.22, 95% CI 0.08–0.34). The path from the cerebral cortical latent variable had a similar central estimate but the CI now crossed zero (effect size 0.3, 95% CI −0.31 to 0.50). We repeated the categorical analyses removing those on medication. This left 22 in the ADHD-only group and 28 in the ADHD/DCD combined group. The overall group difference remained significant for the cerebellar regions associated with motor coordination (F 3,133 = 3.48, p = 0.02) and the cerebral cortical region showed a trend (F 1,133 = 2.52, p = 0.06). As before, the unaffected group differed significantly from the clinical groups, which did not differ significantly from one another.
We next tested for age effects. There were no significant pairwise differences in the path coefficients between the aiming/catching variable and the cerebral cortical and cerebellar variables for the youngest v. middle; youngest v. eldest; and middle v. eldest groups (all p > 0.05). Finally, there were no significant sex differences in the path coefficients between the latent variables (all p > 0.05).
Discussion
In children, greater skill in aspects of motor coordination was associated with increased cerebral cortical and cerebellar volumes, particularly in the motor/premotor cortex and the superior cerebellar lobules. These regions are implicated in motor coordination by functional neuroimaging studies, demonstrating how structure mirrors function (Toyokura et al. Reference Toyokura, Muro, Komiya and Obara1999; Debaere et al. Reference Debaere, Swinnen, Béatse, Sunaert, Van, cke and Duysens2001; Zhuang et al. Reference Zhuang, Laconte, Peltier, Zhang and Hu2005; Mayka et al. Reference Mayka, Corcos, Leurgans and Vaillancourt2006; Mostofsky et al. Reference Mostofsky, Rimrodt, Schafer, Boyce, Goldberg, Pekar and Denckla2006; Stoodley & Schmahmann, Reference Stoodley and Schmahmann2010; Brown-Lum & Zwicker, Reference Brown-Lum and Zwicker2015). These links between motor coordination skill and brain were not moderated by symptoms of inattention, hyperactivity and impulsivity. Using DSM-5-based categories, there was a significant reduction in the volumes of these motor coordination regions in those with probable DCD compared with unaffected controls. However, the volume did not further differ between those with DCD alone and those who had combined ADHD and DCD.
Our primary analyses mapped motor coordination to the brain. We treated both motor coordination skill and symptoms of ADHD as continuous variables. Such quantitative traits contain more information on inter-individual variability than their associated diagnostic categories. Links emerged between cerebellar and cortical neuroanatomic variables and aiming/catching but not balance or manual dexterity. This is in keeping with prior reports of partly independent neural substrates for different aspects of motor coordination (Park et al. Reference Park, Lee, Han, Lee, Lee and Park2009, Reference Park, Lee, Han, Lee, Lee and Park2011, Reference Park, Lee, Kim, Park, Won, Jung and Yoon2012; Pangelinan et al. Reference Pangelinan, Zhang, Vanmeter, Clark, Hatfield and Haufler2011; Kühn et al. Reference Kühn, Romanowski, Schilling, Banaschewski, Barbot, Barker, Brühl, Büchel, Conrod, Czech, Dalley, Flor, Garavan, Häke, Ittermann, Ivanov, Mann, Lathrop, Loth, Lüdemann, Mallik, Martinot, Palafox, Poline, Reuter, Rietschel, Robbins, Smolka, Nees, Walaszek, Schumann, Heinz and Gallinat2012; Paola et al. Reference Paola, Caltagirone and Petrosini2013; Baer et al. Reference Baer, Park, Bailey, Chakravarty, Li and Penhune2015). The strongest associations were between motor coordination and the dimensions of both the motor/premotor cortex and cerebellar lobules. Reciprocal projects between these regions are pivotal in motor coordination, and have been directly delineated in non-human primate tracer studies (Kelly & Strick, Reference Kelly and Strick2003). We also found associations with the dimensions of some inferior temporal/parietal cortical regions, perhaps reflecting the recruitment of visual attention in aiming and catching (Petersen & Posner, Reference Petersen and Posner2012). To further place these findings in context, we draw comparisons with other morphometric studies in children and adolescents. A study of young children found no morphometric correlates of a task of visuomotor coordination (the groove pegboard) (Pangelinan et al. Reference Pangelinan, Zhang, Vanmeter, Clark, Hatfield and Haufler2011). However, only one domain of motor coordination was assessed and volumes of entire structures were defined, thus possibly missing more subregional change. Another study of adolescents found an association between manual dexterity and cerebellar dimensions, specifically right lobule VI (Kühn et al. Reference Kühn, Romanowski, Schilling, Banaschewski, Barbot, Barker, Brühl, Büchel, Conrod, Czech, Dalley, Flor, Garavan, Häke, Ittermann, Ivanov, Mann, Lathrop, Loth, Lüdemann, Mallik, Martinot, Palafox, Poline, Reuter, Rietschel, Robbins, Smolka, Nees, Walaszek, Schumann, Heinz and Gallinat2012). We similarly link motor coordination with the cerebellum, albeit not with manual dexterity, a difference which may reflect the methods of assessment. A recent study found that better motor timing skills were most strongly associated with a smaller volume of right lobule VI, which goes against the more usual finding of increased skills being associated with increased volumes (Baer et al. Reference Baer, Park, Bailey, Chakravarty, Li and Penhune2015). This discrepancy may be due to the inclusion of musicians with early training in this study, a highly selected but not epidemiological representative sample. A final study reported that intensive experience with a video game was associated with an increase in the gray matter volume in the cerebellar lobules IV, V and VI (Kühn et al. Reference Kühn, Gleich, Lorenz, Lindenberger and Gallinat2014). In summary, when neuroanatomic correlates of motor coordination skills are apparent in children, they localize to the same regions of the cerebellum (predominately lobules IV to VI), as found in the current study.
Motor coordination skills were not associated with striatal dimensions. This is perhaps consonant with the different roles played by cortico-striatal and cortico-cerebellar circuitry. The former supports the learning (and perhaps retention) of novel motor sequences, whereas the latter supports adaptation to environmental changes, particularly of well-practised skills (Doyon et al. Reference Doyon, Penhune and Ungerleider2003). The aiming/catching subtests probe motor adaptation, as they require a child to refine a familiar skill (throwing) in the light of environmental feedback (hitting or missing the target).
The concept that the combination of ADHD and DCD represents a nosological entity, distinct from ADHD alone, has impressive support from natural history and genetic studies. However, our primary analyses found that the symptoms of ADHD did not moderate the relationship between motor coordination skills and the brain. Similarly, the volumes of motor coordination brain regions did not differ between those with probable DCD alone v. those with combined DCD and ADHD. This argues somewhat against seeing the combination of ADHD and DCD as defining a distinct entity, insofar as a distinct neuroanatomic basis would be predicted. This finding is consonant with a series of studies that contrasted the brain structure in children with ADHD alone, children with DCD alone and children with both ADHD and DCD (Langevin et al. Reference Langevin, Macmaster, Crawford, Lebel and Dewey2014, Reference Langevin, Macmaster and Dewey2015; McLeod et al. Reference Mcleod, Langevin, Goodyear and Dewey2014). A marker of corpus callosum microstructure – fractional anisotropy – was found to be reduced in anterior callosal regions in ADHD alone, in posterior callosal regions in DCD alone, and throughout the entire corpus callosum in those with both disorders (Langevin et al. Reference Langevin, Macmaster, Crawford, Lebel and Dewey2014). A similar finding emerged in cortical morphology. ADHD alone and DCD alone were characterized by highly regional thinning of distinct regions of the temporal cortex, while the combination of ADHD and DCD was linked with thinning through broad swathes of the temporal and prefrontal cortex (Langevin et al. Reference Langevin, Macmaster and Dewey2015). Thus at both the cerebral structural and white matter microstructural levels, the combination of ADHD and DCD represents an accentuation of the morphological anomalies associated with each disorder in isolation, rather than being linked with a unique profile. Nonetheless, it remains possible that a distinct neural substrate for the combination of ADHD and DCD might be found using probes of brain function or neurochemistry.
We also confirm previous reports of sex differences in motor coordination skills (Watson & Kimura, Reference Watson and Kimura1991; Ruff & Parker, Reference Ruff and Parker1993), but no interaction of sex with diagnosis. Additionally, we note that whatever the etiology of the sex differences in motor coordination skills, the neuroanatomic substrate of motor coordination did not differ between males and females. Finally, it is beyond the scope of this study to delineate the main effects of sex in the brain which has been explored elsewhere in larger cohorts (Lenroot & Giedd, Reference Lenroot and Giedd2010; Ingalhalikar et al. Reference Ingalhalikar, Smith, Parker, Satterthwaite, Elliott, Ruparel, Hakonarson, Gur, Gur and Verma2014).
There are four main limitations. First, 43 subjects were on psychostimulant treatment. However, the pattern of results held when these subjects were excluded, suggesting that the findings are unlikely to be due to psychostimulant medication effects. Second, the cohort was not recruited using a strategy to ensure it was epidemiologically representative. This may limit the generalizability of the findings. Third, we only assessed motor coordination skill and did not consider other motor features associated with ADHD such as some ‘soft’ neurological signs (Mostofsky et al. Reference Mostofsky, Newschaffer and Denckla2003; Cole et al. Reference Cole, Mostofsky, Larson, Denckla and Mahone2008). Finally, in this cross-sectional study we did not find differences between the age groups in the links between motor coordination skills and the brain. However, inferences about developmental processes can only be drawn with extreme caution from cross-sectional data (Kraemer et al. Reference Kraemer, Yesavage, Taylor and Kupfer2000). We plan to collect longitudinal data for more refined developmental mapping.
In summary, we map the neuroanatomic substrate of motor coordination to the motor/premotor cortex and the superior cerebellar lobules. We further show that these links are not moderated by the severity of symptoms of inattention, hyperactivity and impulsivity. At a diagnostic level, we find that the dimensions of these motor coordination regions do not differ significantly between those who had DCD alone compared with those with DCD and co-morbid ADHD.
Supplementary material
The supplementary material for this article can be found at http://dx.doi.org/10.1017/S0033291716000660
Acknowledgements
P.S. is funded by the intramural programs of the National Human Genome Research Institute and National Institute of Mental Health. M.M.C. is funded by the Canadian Institutes of Health Research, National Sciences and Engineering Research Council of Canada, Weston Brian Institute, Michael J. Fox Foundation for Parkinson's Research, and Alzheimer's Society. M.M.C. also receives salary and research support from the Fond de Recherches Santé Québec.
Declaration of Interest
None.
