The test for fixed effects in a mixed model revealed a highly significant group × region interaction (F112=7.96; P<.006), demonstrating a regional specificity in group differences of amygdalar and hippocampal volumes.
Post Hoc Analyses
Post hoc assessment of the origin of this regionally specific difference between groups in volume, using a test of differences in least-square means, indicated that the hippocampus was larger bilaterally in the children with ADHD than in the controls (3384.2 mm3 vs 3164.1 mm3; t112=3.35; P<.002). Amygdalar volumes did not differ significantly across diagnostic groups (ADHD, 2062.6 mm3 vs 2106.0 mm3; t112=−0.64; P=.53). Other significant covariates in the model were WBV (F111=39.8; P<.0001), indicating the presence of significant scaling effects, and hemisphere (F113=6.1; P<.02), reflecting significantly larger volumes in the right hemisphere. A group × region × hemisphere interaction was not significant (at the point of elimination, F111=0.3; P=.62), indicating the absence of significant lateralizing effects across groups. The group × region × age interaction also was not significant (F322=0.2; P=.70), indicating the stability of findings across the age range of children studied. The variables sex (F109=0.5; P=.48) and age (F109=0.7; P=.42) were conservatively retained in the final model because of the biological plausibility that these variables could influence the overall findings.
The group × region interaction remained significant (F82=4.37; P<.04), with a larger hippocampus in boys with ADHD compared with controls (3398.8 mm3 vs 3222.6 mm3; t82=2.23; P<.03).
Statistical maps revealed that global differences in hippocampal volume between groups arose mainly from enlargement of the anterior hippocampus in children with ADHD (), particularly over the anatomical subfields cornu ammonis (CA) and dentate gyrus (DG) (). In posterior portions of the hippocampus, in contrast, smaller indented regions bilaterally suggested the presence of reduced volumes locally in underlying tissue in the ADHD group.
Figure 2 Group differences in surface measures of the hippocampus and amygdala. A, Views of the right (upper row) and left (lower row) hippocampus, anterior (A) and posterior (P) orientation. Arrows point to the main protrusion, most visible laterally in the right (more ...)
Figure 3 Subregions of the hippocampus and amygdala. A, Subregions of the hippocampus showing the head of the hippocampus (HH), the digitationes hippocampi (DH), the hippocampal body (HB), the hippocampal tail (HT), the terminal segment of the HT (TS), the dentate (more ...)
Several portions of the surface of the amygdala suggested the presence of locally reduced volumes in the ADHD group that were not evident in the more conventional measures of overall volume of this structure, with several clusters of voxels reaching P values <.0001 (). Differences in size were located primarily over the basal nucleus of the right amygdala and lateral nucleus of the left.
Gaussian random field–based corrections for multiple comparisons produced clusters of significant voxels that were similar in location to, but smaller in size than, clusters identified in uncorrected comparisons at a threshold of P<.0001 (, , and ).
Figure 4 Symptom correlations with hippocampus surface morphology in children with attention-deficit/hyperactivity disorder. Correlation of surface measures, in signed Euclidean distances, with current hyperactivity scores (A), inattention scores (B), and total (more ...)
Figure 5 Symptom correlations with amygdala surface morphology in children with attention-deficit/hyperactivity disorder. A, Correlation of surface measures, in signed Euclidean distances, with current hyperactivity scores. B, Correlation of surface measures, (more ...)
Correlations With Symptom Severity
In children with ADHD only, while controlling for WBV, sex, and age, a statistical trend was detected for an inverse correlation of hippocampal volume with ratings of the severity of ADHD symptoms in the right (r=−0.29; P=.06) and left (r=−0.27; P<.07) hemispheres (). Surface analyses also suggested that symptom severity correlated inversely with the local features of hippocampus morphology, particularly in portions that were enlarged relative to controls ().
Figure 6 Scatterplots demonstrating correlations with symptom severity. A, Right hippocampal volumes (adjusted for age, sex, and whole brain volume) correlate inversely with Conners Parent Rating Scale43 scores (n=47) (r=0.3; P<.06). B, Right amygdalar (more ...)
In children with ADHD only, volumes of the left (r=0.3; P<.07) and right (r=0.3; P<.06) amygdala showed strong trends toward positive correlations with hyperactivity scores (). Supporting the validity of these trends detected for overall volumes, analyses of symptom severity with surface features exhibited large clusters of positive correlations for hyperactivity scores bilaterally (). Inattention scores, in contrast, correlated inversely with surface morphology mainly in the left amygdala (). Symptoms of anxiety and depression did not correlate significantly with amygdalar or hippocampal volumes.
Group Comparisons of Prefrontal Volumes
The ADHD group had significantly smaller volumes of the left OFC gray matter (ADHD, 10.535 cm3 vs controls, 11.979 cm3; t=2.24; P<.03) and a trend toward lower mean volumes of right OFC gray matter (ADHD, 10.973 cm3 vs controls, 12.033 cm3; t=1.68; P<.09) (). Groups did not differ in volumes of DPFC gray matter.
Comparison of Brain Morphometric Measures*
Correlations With Volumes of PFC Gray Matter
Interregional correlation analyses revealed positive correlations in the control group (n=56) for the right and left amygdala with OFC gray matter (right, r=0.66; P<.001; left, r=0.48; P<.001) (). None of these correlations were significant in the ADHD group (n=47). The test statistic D for comparing 2 Pearson correlation coefficients confirmed significant group differences for these correlations for the left (P<.02) and right (P<.001) amygdala.
In separate assessments of the statistical model used for hypothesis testing, none of the possible confounds reached statistical significance: lifetime diagnosis of depression (F111=0.3; P=.61), oppositional defiant disorder (F111=0.2; P=.67), specific developmental disorder (F111=1.96; P=.17), full-scale IQ (F99=0.3; P=.62), handedness (F109=1.5; P=.22), socioeconomic status (F101=1.0; P=.31), medication status (F110=1.7; P=.20), and stimulant medication (F96=2.2; P=.14). In addition, verbal (right hippocampus, r=−0.10; P=.32; left hippocampus, r=−0.10; P=.35; right amygdala, r=0.17; P=.11; left amygdala, r=0.15; P=.16), performance (right hippocampus, r=−0.10; P=.35; left hippocampus, r=−0.19; P=.07; right amygdala, r=0.15; P=.15; left amygdala, r=0.07; P=.47), and full-scale IQs (right hippocampus, r=−0.10; P=.33; left hippocampus, r=−0.14; P=.16; right amygdala, r=0.17; P=.09; left amygdala, r=0.12; P=.22) did not correlate significantly with regional volumes of either the hippocampus or amygdala while controlling for WBV, sex, and age, further suggesting that IQ measures did not unduly influence findings of the primary analyses.