The x, y, and z are Montreal Neurologic Institute coordinates. Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) showed significant influences of diagnosis. Statistically significant clusters converged in the corpus callosum, especially the anterior portions. FWE indicates familywise error; R, right; and TFCE, threshold-free cluster enhancement. The numbers in the Venn diagram denote the number of voxels in the clusters that were significant for each diffusion tensor imaging metric alone or combined with any of the others. Secondary post hoc group comparisons showed that these differences were driven by autism spectrum diagnosis (eFigure 6 in the Supplement).
The x, y, and z are Montreal Neurologic Institute coordinates. The dimensional approach identified clusters in which fractional anisotropy (FA), radial diffusivity (RD), and mean diffusivity (MD) were associated with SRS-P total T scores. Clusters identified by these 3 diffusion tensor imaging metrics converged in the corpus callosum from its anterior to posterior regions. The numbers in the Venn diagram denote the number of voxels in the clusters that were significant for each diffusion tensor imaging metric alone or combined with any of the others. In the right column, scatterplots show the relationship across all participants between each diffusion tensor imaging metric and SRS-P total T scores (residuals accounting for the nuisance covariates included in the model are plotted). ADHD indicates attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; FWE, familywise error; R, right; TDC, typically developing children; and TFCE, threshold-free cluster enhancement.
eAppendix. Supplemental Appendix
eTable 1. Characteristics of the Study Clinical Sample
eTable 2. Comparisons Between Data Passing DTI QA and Those Who Had Not
eTable 3. Comparisons Between ADHD and TDC Kids Included in Final Sample and Those Excluded
eTable 4. Clusters Results of Categorical TBSS Analyses
eTable 5. Clusters Results of Dimensional TBSS Analyses With SRS-P Covarying for CPRS-R:LV DSM-IV Total T Score
eTable 6. Clusters Results of Dimensional TBSS With CPRS-R:LV DSM-IV Inattentive T Score Covarying for CPRS-R:LV Hyperactive/Impulsive
eTable 7. Clusters Results of Dimensional TBSS With CPRS-R:LV DSM-IV Inattentive T Score Covarying for SRS-P and CPRS-R:LV Hyperactive/Impulsive
eFigure 1. Scatterplots of the Social Responsiveness Scale by Parents (SRS-P) and Conners’ Parent Rating Scales-Revised: Long Version (CPRS-R:LV) DSM-IV Scores by Diagnostic Group
eFigure 2. TBSS Fractional Anisotropy (FA) Analyses Based on Skeletons Generated With Different FA Thresholds
eFigure 3. Raw Images of Categorical Analyses of Each DTI Metric
eFigure 4. Raw Images of the Dimensional Analyses of the Social Responsiveness Scale by Parents (SRS-P)
eFigure 5. Raw Images of Dimensional Analyses of Conners’ Parent Rating Scales-Revised: Long Version (CPRS-R:LV) DSM-IV Inattentive T Scores
eFigure 6. Post Hoc Analyses of Categorical Approaches in Data With and Without DTIPrep Preprocessing
eFigure 7. Results of Voxelwise Dimensional Model Conducted in the Primary Analyses Adding Autism Spectrum Disorder (ASD) Membership as a Covariate
eFigure 8. Scatterplots of the Children’s Communication Checklist-2 (CCC-2) Subscores Against Extracted DTI Metrics
eFigure 9. Results of Dimensional Analyses of the Conners’ Parent Rating Scales-Revised: Long Version (CPRS-R:LV) DSM-IV Inattentive T Scores
eFigure 10. Results of Cluster Level Dimensional Analyses Using Data With DTIPrep
eFigure 11. Results of Cluster Level Dimensional Analyses of Subdomain of Attention-Deficit/Hyperactivity Disorder Symptoms Using Data With DTIPrep
Customize your JAMA Network experience by selecting one or more topics from the list below.
Aoki Y, Yoncheva YN, Chen B, et al. Association of White Matter Structure With Autism Spectrum Disorder and Attention-Deficit/Hyperactivity Disorder. JAMA Psychiatry. 2017;74(11):1120–1128. doi:10.1001/jamapsychiatry.2017.2573
Do the neural correlates of autistic traits extend across diagnostic boundaries among children with autism spectrum disorder and children with attention-deficit/hyperactivity disorder?
This cross-sectional diffusion tensor imaging study analyzed data from 174 children, including 50 typically developing children and children with a primary diagnosis of autism spectrum disorder (n = 69) or attention-deficit/hyperactivity disorder (n = 55). While categorical comparisons detected a significant influence of autism spectrum disorder on multiple white matter metrics in the corpus callosum, dimensional analyses yielded an association with autism spectrum disorder symptoms and white matter metrics in a set of both callosal and other tracts, regardless of diagnosis.
The frequent co-occurrence of autism spectrum disorder and attention-deficit/hyperactivity disorder symptoms may reflect underlying neural mechanisms that transcend diagnostic boundaries.
Clinical overlap between autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) is increasingly appreciated, but the underlying brain mechanisms remain unknown to date.
To examine associations between white matter organization and 2 commonly co-occurring neurodevelopmental conditions, ASD and ADHD, through both categorical and dimensional approaches.
Design, Setting, and Participants
This investigation was a cross-sectional diffusion tensor imaging (DTI) study at an outpatient academic clinical and research center, the Department of Child and Adolescent Psychiatry at New York University Langone Medical Center. Participants were children with ASD, children with ADHD, or typically developing children. Data collection was ongoing from December 2008 to October 2015.
Main Outcomes and Measures
The primary measure was voxelwise fractional anisotropy (FA) analyzed via tract-based spatial statistics. Additional voxelwise DTI metrics included radial diffusivity (RD), mean diffusivity (MD), axial diffusivity (AD), and mode of anisotropy (MA).
This cross-sectional DTI study analyzed data from 174 children (age range, 6.0-12.9 years), selected from a larger sample after quality assurance to be group matched on age and sex. After quality control, the study analyzed data from 69 children with ASD (mean [SD] age, 8.9 [1.7] years; 62 male), 55 children with ADHD (mean [SD] age, 9.5 [1.5] years; 41 male), and 50 typically developing children (mean [SD] age, 9.4 [1.5] years; 38 male). Categorical analyses revealed a significant influence of ASD diagnosis on several DTI metrics (FA, MD, RD, and AD), primarily in the corpus callosum. For example, FA analyses identified a cluster of 4179 voxels (TFCE FEW corrected P < .05) in posterior portions of the corpus callosum. Dimensional analyses revealed associations between ASD severity and FA, RD, and MD in more extended portions of the corpus callosum and beyond (eg, corona radiata and inferior longitudinal fasciculus) across all individuals, regardless of diagnosis. For example, FA analyses revealed clusters overall encompassing 12121 voxels (TFCE FWE corrected P < .05) with a significant association with parent ratings in the social responsiveness scale. Similar results were evident using an independent measure of ASD traits (ie, children communication checklist, second edition). Total severity of ADHD-traits was not significantly related to DTI metrics but inattention scores were related to AD in corpus callosum in a cluster sized 716 voxels. All these findings were robust to algorithmic correction of motion artifacts with the DTIPrep software.
Conclusions and Relevance
Dimensional analyses provided a more complete picture of associations between ASD traits and inattention and indexes of white matter organization, particularly in the corpus callosum. This transdiagnostic approach can reveal dimensional relationships linking white matter structure to neurodevelopmental symptoms.
Shared clinical and biological traits across psychiatric diagnoses have challenged the usefulness of a categorical nosology of psychiatric disorders for biological research.1-3 The challenges associated with categorical perspectives of illness are exemplified by the frequent clinical overlap between autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD).4-6 Whether such shared clinical presentations reflect common underlying neural mechanisms remains unknown to date.
In both ASD and ADHD, abnormal large-scale networks have been consistently reported using a range of neuroimaging methods.3,7-13 Among these modalities, diffusion tensor imaging (DTI) can provide insight into the pathology of white matter organization.14 Diffusion tensor imaging studies have found atypical structural connectivity in individuals with ASD15-17 or ADHD18,19 compared with typical controls, but the findings were based on independent comparisons. These studies varied in regard to DTI metrics examined, specific spatial locations, and the nature of the DTI abnormalities found in ASD or ADHD. Nevertheless, in most cases, lower fractional anisotropy (FA) in different areas of the corpus callosum (CC) has been reported in ADHD19,20 or ASD15 relative to typical controls.
The only 2 studies21,22 that have directly contrasted white matter structure in individuals with ADHD, those having ASD, and typically developing children (TDC) yielded mixed results. One tractography study21 of 8 children with ASD, 20 children with ADHD, and 20 TDC revealed disorder-specific patterns of densely interconnected hubs (ie, rich clubs). Beyond concerns about sample size, that preliminary study did not assess the influence on white matter organization of co-occurring ASD or ADHD symptoms across diagnoses, thus missing a potential source of commonalities in DTI associations. A second DTI study22 applied tract-based spatial statistics (TBSS) to a larger sample (n = 200) of school-aged children with ASD, ADHD, or obsessive-compulsive disorder compared with TDC. Decreased FA within the splenium of the CC was common among the 3 groups.22 Brain-behavior relationships with symptom domains characteristic of each disorder, examined separately, did not yield significant findings.22 Therefore, the implications of shared CC abnormalities remain unclear. While functional MRI studies23,24 have shown the utility of stratifying ASD subgroups based on ADHD comorbidity, no imaging study to date has simultaneously examined both ASD and ADHD dimensionally in the same sample.
Accordingly, to identify specific or shared patterns of white matter organization, we analyzed DTI data from 174 school-aged children with ASD, those having ADHD, or TDC. We adopted both a categorical diagnostic approach (ie, comparisons of diagnostic groups) and dimensional analyses of ASD-related and ADHD-related traits across diagnostic groups. The Social Responsiveness Scale by Parents (SRS-P)25 and Conners’ Parent Rating Scales–Revised: Long Version (CPRS-R:LV)26 were used for dimensional analyses. To examine their unique contributions, these measures were included in the same model, hence removing shared variance. Fractional anisotropy was our primary measure of interest. Because other DTI metrics may provide distinct complementary information about white matter structure,27 although rarely investigated together in ASD or ADHD,20 secondary analyses also explored voxelwise mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD), and mode of anisotropy (MA). Data collection was ongoing from December 2008 to October 2015.
We analyzed data from 174 children (age range, 6.0-12.9 years), including 50 TDC and children with a primary diagnosis of either ASD (n = 69) or ADHD (n = 55), selected from a larger sample after quality assurance (see “Preprocessing” and the eAppendix in the Supplement) to be group matched on age and sex (Table and eTable 1 in the Supplement). Clinicians’ diagnoses of ASD and ADHD were based on DSM-IV-TR codes supported by parent interview, direct observation, available teacher forms, and prior records (eAppendix in the Supplement). Autism spectrum disorder diagnosis was supported by the Autism Diagnostic Observation Schedule (research reliable n = 68)28,29 and the Autism Diagnostic Interview–Revised (research reliable n = 65).30,31 Absence of Axis I diagnosis per the Schedule for Affective Disorders and Schizophrenia for School-Age Children–Present and Lifetime Version32 and absence of a history of psychotropic medication use were required for inclusion as TDC. Exclusion criteria for all participants were current use of antipsychotics, known genetic diseases, or below 80 on the estimated full-scale IQ.33,34
To discern brain-behavior relationships, we used parent ratings of ASD traits and ADHD traits indexed by SRS-P total T scores25 and CPRS-R:LV DSM-IV total T scores,26 respectively (eFigure 1 in the Supplement). Parent ratings of the Children’s Communication Checklist 2 (CCC-2)35 and the Child Behavior Checklist36 further characterized the sample. Parents also reported on race and socioeconomic status indexed by the Four-Factor Index of Socioeconomic Status by Hollingshead.37 Data from 29 TDC and 29 children with ADHD in our sample were included in a previous DTI study.38 The institutional review boards of the New York University and the New York University School of Medicine granted ethical approval of the study. Written parental consent and verbal assent were obtained for all participants; children older than 7 years also provided written informed assent.
Two DTI scans were acquired using a twice-refocused diffusion-weighted echoplanar imaging sequence (repetition time, 5200 milliseconds; echo time, 78 milliseconds; 50 sections; 64 × 64–pixel acquisition matrix; field of view, 192 mm; voxel size, 3 × 3 × 3 mm; 64 noncollinear diffusion directions, uniformly distributed around a unit sphere with B value of 1000 s/mm2; 1 image with no diffusion weighting) at the New York University Center for Brain Imaging using a 3.0-T imaging system (Allegra; Siemens). We obtained T1-weighted images using 3-dimensional magnetization-prepared rapid acquisition gradient echo for anatomical registration (eAppendix in the Supplement).
To enhance signal to noise, analyses were conducted only in children who completed 2 DTI scans. Analyses were conducted with Functional Magnetic Resonance Imaging of the Brain (FMRIB) Software Library version 5 (http://www.fmrib.ox.ac.uk). Quality assurance involved eddy current and motion corrections, as well as removal of nonbrain tissue. Our head motion index was the mean absolute intervolume displacement with respect to the first image of each run; root-mean-square (RMS) deviation was calculated using FSL (the FMRIB Software Library) rmsdiff.39 Given that motion introduces artifacts in DTI metrics,40 we only included data with a mean absolute RMS less than 5 mm and image quality passing visual inspection. As a result, 69 of 83 children with ASD, 66 of 82 children with ADHD, and 68 of 79 TDC passed quality assurance; they did not differ from those excluded in demographic or primary clinical measures (eTable 2 and eTable 3 in the Supplement). As detailed in the eAppendix in the Supplement, to match diagnostic groups by age and sex, 11 children with ADHD and 18 TDC were further excluded, yielding a final sample of 174 children.
First, nonlinear registration to a common space was conducted by aligning each participant’s FA image to the Montreal Neurologic Institute 152 space template. Then, a mean FA image and a mean FA skeleton of the aligned images were created. Voxelwise analyses were subsequently conducted for skeleton areas with an FA of at least 0.2. Analyses with more stringent FA thresholds (ie, ≥0.25 and ≥0.3) yielded similar results (eFigure 2 in the Supplement). Voxelwise analyses for secondary DTI metrics were conducted using values projected onto the mean FA skeleton.
To examine the main influence of diagnosis on DTI metrics, we performed an F test using FSL Randomize.41 Age, sex, and motion were included as nuisance covariates. Statistical significance was set at threshold-free cluster enhancement (TFCE) P < .05 to control for familywise error (FWE) rate (α = .05), thus accounting for multiple comparisons.42 Post hoc pairwise group comparisons were conducted for clusters showing significant main group associations for each metric separately.
We assessed the association between DTI metrics and ASD traits or ADHD traits (indexed by SRS-P or CPRS-R:LV DSM-IV total T scores, respectively) using FSL Randomize across all participants, regardless of diagnosis (ie, ASD, ADHD, and TDC). Nuisance covariates were age, sex, and motion. In addition, to identify their unique contributions to brain-behavior relationships and given their significant relationship (r = 0.59, df = 169, P < .001), we included both SRS-P and CPRS-R:LV DSM-IV total T scores in the same model. There was no multicollinearity across covariates (variation inflation factor <4).43 Analogous to the categorical approach, statistical significance was set at TFCE FWE corrected 1-sided P < .05. For interpretation, only TBSS results within the International Consortium for Brain Mapping DTI-81 atlas44 are reported and labeled accordingly (raw images are shown in eFigures 3, 4, and 5 in the Supplement).
Values in clusters located in the genu, body, and splenium of the CC differed across diagnostic groups (overall 4179 voxels; Figure 1 and eTable 4 in the Supplement). Post hoc analyses revealed significantly lower FA among children with ASD compared with both children with ADHD and TDC in these clusters. Children with ADHD and TDC did not differ significantly in regard to the mean FA for these clusters (eFigure 6 in the Supplement).
A significant influence of ASD diagnosis was also evident in several portions of the CC for all DTI metrics examined except MA (Figure 1 and eTable 4 in the Supplement). Similar to the FA results, children with ADHD did not differ significantly from TDC in any of the diffusivity metrics or MA. In these clusters, children with ASD had significantly higher mean MD, RD, and AD compared with those having ADHD and TDC (eFigure 6 in the Supplement). The number of voxels identified by these categorical TBSS analyses of MD, RD, and AD was 1280, 2721, 523, respectively. Across these DTI metrics, ASD-related abnormalities converged on the midbody and anterior portions of the CC (Figure 1).
Dimensional analyses revealed a significant relationship between FA and SRS-P total T scores but not for CPRS-R:LV DSM-IV total T scores. Specifically, across participants, FA was significantly and negatively related to SRS-P scores in clusters overall encompassing 12121 voxels extending from anterior to posterior regions of the CC and other tracts not identified by the categorical approach (Figure 2 and eTable 5 in the Supplement). These tracts extending within and outside the CC encompassed the anterior limb of the internal capsule, the inferior longitudinal fasciculus, and the corona radiata.
Among our 4 complementary DTI metrics, MD and RD were significantly positively related to SRS-P total T scores in clusters of 18058 and 15441 voxels, respectively. These relationships converged in the CC, corona radiata, and inferior longitudinal fasciculus areas that were also identified in FA analyses (Figure 2 and eTable 5 in the Supplement).
To ensure that the above associations in the Dimensional Approach subsection did not simply depend on ASD diagnosis, we assessed the relationship between SRS-P total T scores and FA measures at the clusters identified in TBSS analyses across all children after regressing out the nuisance covariates and ASD diagnostic status (ie, ASD is 1, and non-ASD is 0). The pattern of results was similar to that observed in primary analyses (β6,164 = −0.27, P = .01). Similar findings were obtained for MD (β6,164 = 0.32) and RD (β6,164 = 0.31) (P = .005 for both). To verify this pattern’s robustness to TFCE FWE correction, we repeated a TBSS dimensional DTI-SRS analysis adding ASD membership as a nuisance covariate. The pattern of DTI-SRS relationships was similar to the patterns identified in primary analyses, particularly at corrected P < .10 (eFigure 7 in the Supplement).
Furthermore, given concerns that SRS-P total T scores may be confounded by non-ASD behavioral symptoms,45 secondary analyses examined the reproducibility of these relationships using an alternative, independent measure of ASD traits. We focused on the 5 subscales of the CCC-2 capturing pragmatic aspects of language and behaviors commonly impaired in ASD. After regressing the influences of age, sex, motion, and CPRS-R:LV DSM-IV total T scores, we tested the relationships between these CCC-2 scaled scores and the mean FA, MD, and RD values in the clusters identified in primary voxelwise dimensional analyses. Results were in line with the pattern observed with SRS-P total T scores. Use of context, stereotyped language, and nonverbal communication subscale scores were significantly and positively related to FA in these clusters(β5162 = 0.22, 0.22, 0.18, and P = .01, 0.01, 0.05, respectively); use of context and stereotyped language were negatively associated with RD (β5162 = −0.25 and −0.20, P = .004 and 0.02, respectively) and MD (β5162 = −0.27 and −0.18, P = .002 and 0.04, respectively) (eFigure 8 in the Supplement).
Because no significant DTI relationships were identified in regard to CPRS-R:LV DSM-IV total T scores for any DTI metrics examined, we secondarily explored whether subscores of the ADHD subdomains (ie, inattention and hyperactivity/impulsivity) may provide additional information. We first performed dimensional TBSS analyses of CPRS-R:LV DSM-IV inattention and hyperactivity/impulsivity T scores separately (controlling for each other in the same model while also including age, sex, and motion). Results indicated a significant relationship of CPRS-R:LV DSM-IV inattention T scores with AD in a cluster of 716 voxels localized the anterior and middle CC (eTable 6 and eFigure 9A in the Supplement). Given that SRS-P total scores were positively associated with both CPRS-R:LV DSM-IV subdomains, to determine the specificity of the AD-inattention relationship, we repeated analyses, including SRS-P in the model. Results indicated significant effects of CPRS-R:LV DSM-IV inattention T scores for AD in clusters of overall 6142 voxels centered in CC, as well as for MD in a small cluster with 37 voxels in the posterior CC (eTable 7 and eFigure 9B in the Supplement). No significant relationships were detected with CPRS-R:LV DSM-IV hyperactive/impulsivity T scores.
To further address concerns about head motion, we repeated group TBSS analyses after automatic artifact correction using DTIPrep (eAppendix in the Supplement).46 The pattern of results was similar to the patterns reported above in the Categorical Approach and Dimensional Approach subsections and in Exploring the ADHD Subdomains in the Follow-up Analyses subsection (eFigures 5, 10, and 11 in the Supplement).
We assessed white matter organization within and beyond the diagnostic boundaries of ASD and ADHD using TBSS analyses in a moderately large, well-characterized, and low-motion sample of 174 school-aged children with ASD, those having ADHD, or TDC. Taken together, our results indicate that white matter organization was affected by both ASD diagnosis and ASD traits across diagnoses. While categorical analyses revealed white matter abnormalities only in children with ASD, dimensional analyses provided a more complete picture of the influence of ASD or ADHD symptoms on white matter. By indexing individuals as a function of ASD severity, our approach revealed brain-behavior relationships, regardless of diagnostic status. Notably, these relationships were observed across distinct measures of ASD traits (SRS-P and CCC-2) while controlling for ADHD severity, thereby underscoring the specific role of autistic traits. Although ADHD total T scores were not significantly related to any DTI metrics, ADHD subdomain analysis revealed a significant dimensional association, primarily between AD and inattention, mostly localized in the CC.
Consistent with models emphasizing the role of abnormal interhemispheric interactions in neurodevelopmental disorders,47-51 results from both our categorical and dimensional approaches converged on the CC. Lower FA, along with greater MD and RD, in multiple CC regions both characterized ASD diagnosis, as well as ASD traits across children. While atypical DTI findings in CC regions have been consistently reported in prior studies15,19,20 comparing ASD or ADHD vs TDC separately, our assessment of both ASD and ADHD dimensions across groups provided a novel perspective on the influence of ASD traits on white matter organization in both disorders. Indeed, emerging clinical evidence highlights the significance of autistic traits in a substantial group of children with ADHD and vice versa.5,6 Still, their underlying neural mechanisms have been virtually ignored to date. Findings of brain-behavior relationships that are specific to the ASD domain and yet shared across disorders suggest that these are potentially shared biomarkers. Future longitudinal studies may elucidate whether common developmental pathways exist.
The wide spatial extent of significant findings emerging from dimensional analyses of ASD symptoms likely reflects the multifaceted nature of ASD-related impairments. The posterior CC has been shown to be involved in sensory and visuospatial processing.52,53 In contrast, consistent with its role of connecting the bilateral frontal cortex,54 the anterior CC has been associated with social functioning impairment in ASD.17 The midbody of the CC connects the bilateral premotor, primary motor, and primary sensory cortex,55 suggesting that atypicalities in this CC region are related to sensory and motor processing abnormalities. Social, sensorimotor, and language impairments have also been previously reported in children with ADHD, but their nature remains unclear.56-58 Different dimensional impairments common to children with ASD or ADHD may account for our findings. Future studies that include fine-grained measures of ASD subdomains and are obtained from multiple sources and in large samples are needed to differentiate the specific functional roles of these tracts in children with ASD and ADHD.
Although dimensional analyses with combined ADHD (ie, inattention and hyperactivity/impulsivity combined) revealed no significant relationships, those exploring inattention did. These findings converged on the CC and affected AD, which did not relate to ASD symptoms. A tantalizing hypothesis is that abnormalities in different aspects of white matter structure might correspond to distinct psychopathological profiles. This hypothesis can only be tested with larger samples and more advanced diffusion methods able to capture finer aspects of white matter organization.59,60 Overall, the results herein suggest that brain-behavior relationships vary depending on the ADHD subdomains; therefore, inclusion of ADHD heterogeneous samples without accounting for their core subdomains may have contributed to prior conflicting findings.
Lack of regional differences associated with ADHD diagnosis is not consistent with results of some prior TBSS studies, including meta-analyses.20,61 Several scenarios can explain this inconsistency. First, the effect size of white matter abnormalities in ADHD may be small and thus missed in multiple-group comparisons. Second, as discussed in the previous paragraph, more homogeneous ADHD presentations may be necessary to detect group differences. Third, negative findings for ADHD diagnosis may reflect our emphasis on groups not differing in head motion. Indeed, our negative findings are consistent with recent reports38,62,63 verifying lack of group differences in head motion.20 Fourth, other additional factors may contribute to discrepancies related to ADHD diagnosis. Similar to the present work, a recent study22 compared multiple disorders using a broader range of diagnoses (ie, ASD, ADHD, and obsessive-compulsive disorder). While the authors reported low FA in the posterior CC for each of the 3 clinical groups relative to controls, our analyses only revealed low FA with respect to ASD diagnosis and traits. In comparing that study and the present study, we note that our ADHD sample had a lower rate of medication use and lower prevalence of psychiatric comorbidities, both of which have been reported to affect diffusion findings in ADHD.62,64 Our results underscore the need for future comprehensive investigations of factors contributing to clinical heterogeneity in substantially large samples, by using multimodal objective phenotypic assessments and approaches that facilitate replication, such as those providing for open data sharing.65,66
Our findings should be interpreted in light of some limitations. First, although we assessed both categorical and dimensional models, we did not test their interactions (hybrid analyses) as done in prior studies67,68 that only examined either ASD or ADHD. Measuring interactions between diagnostic status and 2 psychopathological dimensions requires larger samples than have been seen to date to capture optimal distributions of both traits within each group. Ideally, these studies should use gold standard diagnostic measures of ASD (ie, the Autism Diagnostic Observation Schedule28,29 or the Autism Diagnostic Interview–Revised30,31) across all groups to confirm exclusion of ASD in ADHD and TDC, as well as providing independent metrics of ASD severity. Unfortunately, the extensive training requirements and lengthy administrations limit their use in non-ASD populations. Abbreviated assessments, such as the recently validated Autism Symptom Interview, School-Age,69 may bypass this concern. Second, although the diagnostic groups did not differ significantly in sex distribution, most participants were male, reflecting the higher prevalence of boys in ASD and ADHD.70 Therefore, results herein may not generalize to girls.
In summary, our dimensional approach was more sensitive in detecting brain-behavior relationships with ASD traits and ADHD traits than the categorical approach. The clinical observation that these traits extend across diagnostic categories4,6 likely reflects shared underlying neural mechanisms. Therefore, this study emphasizes investigations of constructs and domains transcending traditional categorical boundaries, with the ultimate goal of identifying biomarkers on the path toward precision medicine.71,72
Accepted for Publication: June 29, 2017.
Corresponding Author: Adriana Di Martino, MD, Department of Child and Adolescent Psychiatry at NYU Langone Medical Center, One Park Avenue, Seventh Floor, New York, NY 10016 (firstname.lastname@example.org).
Published Online: September 6, 2017. doi:10.1001/jamapsychiatry.2017.2573
Author Contributions: Drs Aoki and Di Martino had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Aoki, Yoncheva, Milham, Di Martino.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Aoki, Di Martino.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Aoki, Yoncheva, Nath, Milham, Di Martino.
Obtained funding: Aoki, Di Martino.
Administrative, technical, or material support: Aoki, Chen, Nath, Sharp, Velasco, Di Martino.
Study supervision: Milham, Di Martino.
Conflict of Interest Disclosures: Dr Di Martino reported receiving royalties from the publication of the Italian version of the Social Responsiveness Scale–Child Version. No other disclosures were reported.
Funding/Support: This work was supported in part by the Japan Society for the Promotion of Science, the Kanae Foundation for the Promotion of Medical Science, and the Uehara Memorial Foundation (Dr Aoki). Funding was also provided by grants K23MH087770 and R01MH105506 (Dr Di Martino) and grants R01MH081218 and R01HD065282 from the Leon Levy Foundation (Dr Di Martino).
Role of Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: F. Xavier Castellanos, MD (Child Study Center at NYU Langone Medical Center and The Nathan S. Kline Institute for Psychiatric Research) contributed helpful discussions and editorial suggestions (no compensation was received). We are grateful to the children and parents who generously contributed their time to this research. We thank the supporting staff at the New York University Center for Brain Imaging for their technical support. We also thank the research staff at the Autism Research and Clinical Program and the Phyllis Green and Randolph Cowen Institute for Pediatric Neuroscience of the Child Study Center at New York University Langone Medical Center for their help in aspects of participant recruitment, assessment, data collection, and data entry.
Additional Information: Some data included in this article were deposited as fully anonymized data in the Autism Brain Imaging Data Exchange (http://fcon_1000.projects.nitrc.org/indi/abide/) or the National Database for Autism Research (http://ndar.nih.gov/).
Create a personal account or sign in to: