Demographic Variables and Clinical Measures
There were some differences in the clinical characteristics of participants from the various institutions (). For the whole sample, the age of ADHD children did not differ from that of the TDC, but the proportion of boys was higher in the ADHD sample (204 boys vs. 51 girls) than in the TDC sample (168 boys vs. 148 girls) (p < .0001). Twelve children were removed from the TDC group, 9 who were receiving psychotropic medications (4 boys and 5 girls) and 3 who had no information on medication status (2 boys and 1 girl); thus, the final TDC sample included 304 participants. In the ADHD group, 67 children were receiving psychotropic medication; the medication status was missing for 73 ADHD children; and the remaining 115 ADHD children were medication naïve.
ADHD index scores were not reported for the OHSU data set and parent-rated scores were missing for 8 TDC girls, 12 TDC boys, 2 ADHD girls, and 7 ADHD boys from the remaining institutions (KKI, NYU, and PU). The available scores of inattention, hyperactivity/impulsivity, and ADHD index were highly intercorrelated (r > .84) and showed significant correlation with IQ (r < −.2; p < .0001). Scores of inattention, hyperactivity/impulsivity, and ADHD index were significantly lower for PU than for KKI, NYU, and OHSU (p < 10−32; t test), which likely reflects differences between the ADHD Rating Scale IV (PU) and the Conners’ Parent Rating Scale–Revised, Long version and the parent and teacher Conners’ Rating Scale, Third Edition (KKI, NYU, and OHSU). Therefore, separate group comparisons of clinical measures (and the regression analysis with the clinical measures, see below) were carried out for the PU data set and for the rest of the data sets (KKI, NYU, and OHSU), which were combined.
For the PU data set, scores of inattention, hyperactivity/impulsivity, and ADHD index were higher for ADHD children than for TDC (separated analyses for boys and girls; p < 10−5; t test); IQ was significantly lower for ADHD boys than for TDC boys (p < 10−6), but this group difference was not observed in girls, likely due to the reduced number of ADHD girls in the PU data set. Gender differences in scores of inattention, hyperactivity/impulsivity, ADHD index, and IQ were not observed for ADHD children or TDC in the PU data set (p > .08). For the KKI-NYU-OHSU combined sample, scores of inattention, hyperactivity/impulsivity, and ADHD index were also higher for ADHD children than for TDC (separated analyses for boys and girls; p < 10−30; t test); the lower IQ for ADHD was more pronounced in ADHD boys (p < 10−7) than in girls (p < 10−2), but this gender × group interaction effect did not reach statistical significance. Intelligence quotient was higher for TDC boys than for TDC girls (p < .0003). Scores of inattention and ADHD index were higher for ADHD girls than for ADHD boys (p < 10−7), but these gender differences were not observed in TDC (p > .22) and the gender × group interaction effect was statistically significant (p < .0001; one-way ANOVA).
For TDC and ADHD children, the short-range FCD patterns were highly bilateral and maximal in posterior cingulate/ventral precuneus, occipital, inferior, superior and lateral parietal, ventral, and dorsolateral prefrontal cortices (). The long-range FCD patterns were also highly bilateral for TDC and ADHD children and were maximal in ventral posterior occipital, ventral prefrontal, parietal, and dorsolateral prefrontal cortices. Group comparisons showed that short-range FCD was lower in SPC (Brodmann area [BA] 7) and precuneus and higher in inferior orbitofrontal cortex (OFC)/insula, ventral striatum, and superior frontal cortex (BA 10) for ADHD children than for TDC (pFWE < .05; , ). ADHD children also had lower long-range FCD in SPC and cerebellum than TDC (pFWE < .05; , ).
Figure 1 Distribution of short-range (top panel) and long-range (bottom panel) functional connectivity density (FCD) in the human brain for 247 attention-deficit/hyperactivity disorder children and 304 typically developing children and the statistical differences (more ...)
Statistical Significance (t Scores; One-Way ANOVA) of FCD Differences Between ADHD Patients (n = 247) and Matched Healthy Control Subjects (n = 304), as well as the Corresponding Effects of Gender and Age on FCD
Region of Interest Analyses
Average FCD measures in regions of interest (ROIs) that differed between the groups showed that the strength of short- and long-range FCD varied between regions and differed between TDC and ADHD children (, ). The average strength of the short-range FCD in ROIs located in left and right OFC/insula, ventral striatum, and caudate () was 15 ± 2% higher for ADHD children than for TDC. Conversely, the short- and long-range FCD were lower in ADHD children than TDC in parietal cortex (16 ± 2% and 33 ± 4%, respectively) and cerebellum (15 ± 3%and 40 ± 10%, respectively). Short-range FCD in OFC/insula and ventral striatum was positively correlated with long-range FCD in SPC, independently for ADHD children and TDC (r > .16; p < .01).
Figure 2 Bar plot showing the average values for short-range (left) and long-range (right) functional connectivity density (FCD) across subjects and regions of interest (ROIs) for attention-deficit/hyperactivity disorder (ADHD) children (n = 247) and typically (more ...)
The short-range FCD in ADHD boys from the PU data set showed negative correlations with hyperactivity/impulsivity and the ADHD index in cerebellum, SPC, and precuneus (r < −.24; p < .05; ) and for the TDC boys positive correlations were observed in ventral striatum with scores of inattention and the ADHD index, in left OFC/insula with hyperactivity/impulsivity and the ADHD index, and in superior frontal gyrus (BA 9) with hyperactivity/impulsivity scores (r > .25; p < .05; ). The long-range FCD in ADHD boys from the PU data set showed negative correlations in SPC with inattention, hyperactivity/impulsivity, and the ADHD index scores (r < −.32; p = .01; ).
Figure 3 Scatter plots showing correlations between rating scores and average values of short- and long-range functional connectivity density (FCD) in parietal (top panels) and reward-motivational (bottom panels) regions of interest (ROIs). Sample: 67 attention-deficit/hyperactivity (more ...)
For the KKI-NYU-OHSU combined sample, rating scores did not show significant correlations with FCD in the ROIs (separate analyses for TDC and ADHD boys and girls; |r| < .2). Correlations with scores of IQ were not significant for any ROIs.
We mapped the networks functionally connected to the left SPC (cubic ROI seed; ROI volume = 125 voxels; center coordinates: x, y, z = [−24, −66, 63] mm), an attention region that showed lower short- and long-range FCD for ADHD than for TDC, and that of the ventral striatum (ROI coordinates: x, y, z = [9, 15, −3] mm), a reward-motivation region that showed higher short-range FCD for ADHD than for TDC.
For both ADHD and TDC, the positive SPC-RSFC mapped into a bilateral network that included other SPC regions (BAs 2, 5, 7, and 40); superior, middle, and inferior occipital gyri (BA 19); inferior temporal cortex (BA 37); and the superior frontal gyrus (frontal eye field, BA 6) (pFWE < .001; ); the negative RSFC mapped into anterior and posterior insula; anterior cingulum (BAs 24 and 32); superior frontal gyrus (BA 9) and OFC (BAs 10 and 11); inferior frontal (BAs 45 and 47), supramarginal gyrus (BA 40), angular gyrus (BA 39), precuneus (BA 39), superior (BA 22), middle (BA 21), and inferior (BA 20) temporal gyri; temporal pole (BA 21); parahippocampal gyrus (BAs 30 and 36); midbrain; dorsal striatum (caudate, putamen, and pallidum); thalamus; and vermis. The positive strength of the RSFC was lower for ADHD children than for TDC in precuneus and SPC, and the negative RSFC strength was lower in medial OFC and temporal cortices (pFWE < .05; ). Thus, the lower SPC-RSFC for ADHD mapped into the attention network regions that showed lower FCD and the reward-cin TDC.
Figure 4 Statistical significance (color-coded t score) of resting-state functional connectivity patterns for the superior parietal cortex (SPC) seed (cubic region of interest [ROI] centered at x, y, z = (−24, −66, 63) mm; ROI volume = 125 imaging (more ...)
For both ADHD and TDC, the positive RSFC of the ventral striatum mapped into a bilateral network that included middle OFC (BA 11), anterior cingulum (BA 32), superior frontal gyrus (BA 10), OFC/insula, middle temporal gyrus (BA 22), temporal pole (BA 38), anterior thalamus, hypothalamus, caudate, hippocampus, precuneus, temporal pole, and parahippocampal gyrus (pFWE < .001). The negative RSFC of the ventral striatum mapped into the primary visual cortex (BA 17). The strength of the RSFC with ventral striatum was lower for ADHD children than for TDC in precuneus, temporal pole, and parahippocampal gyrus (pFWE < .05).
ADHD children that were not using psychotropic medications had higher short-range FCD than TDC in OFC/insula and ventral striatum (p < .05; ROI analyses), but this effect was not observed in medicated ADHD children (), suggesting that ADHD medications could normalize the short-range FCD in reward-motivation regions. However, the short-range FCD differences between medicated and unmedicated ADHD children were not statistically significant (p > .1). On the other hand, the medicated ADHD group exhibited lower long-range FCD than TDC in cerebellum and SPC (p < .05), but this effect was not observed in ADHD children that were not using psychotropic medications ().
Figure 5 Effect of medications on short-range functional connectivity density (FCD) in reward-motivational regions of interest (ROIs) (right orbitofrontal cortex [OFC]/insula and ventral striatum [VS]) and of long-range FCD in cerebellum and superior parietal (more ...)
Both genders demonstrated higher short-range FCD in ventral striatum for the ADHD group compared with TDC (p
< .05; ROI analyses), but the higher short-range FCD in OFC/insula for ADHD was pronounced for boys but not for girls (p
< .001; Figure S3 in Supplement 1
). The long-range FCD in SPC and cerebellum was lower for ADHD than for TDC for boys (p
< .001) but not for girls and showed a gender × diagnosis interaction effect (p
< .05; one-way ANOVA with gender as a covariate).