The participants were 15 adult men with ADHD (5 combined, 10 predominantly inattentive subtype) and 18 male healthy controls, all of whom took part in a previously reported study of DA function in treatment-naïve ADHD (Cherkasova et al., 2014). In brief, the diagnosis of ADHD (DSM-IV-TR) was ascertained by a research psychiatrist (CB, LH, RJ), and ADHD symptoms were measured in both groups using the Conners' Adult ADHD Rating Scale (CAARS) (Conners et al., 1999) (Table 1). Participants were free from any current or past Axis I disorder other than ADHD (Structured Clinical Interview for DSM-IV Axis I disorders First et al., 1996), except two ADHD participants who reported a single mild depressive episode occurring ≥2 years in the past. Other exclusion criteria were: a first-degree relative with a history of substance dependence; current use of psychotropic medications; Beck Depression Inventory (BDI) (Beck and Steer, 1993) score > 13, estimated IQ < 80; a neurological history; reported history of serious head injury; history of any physical disorder (e.g., cardiovascular) contradictory to participation; and a positive toxicology screen as per the Triage Drugs of Abuse urine test (Biosite Inc., San Diego). Controls were additionally excluded for a reported ADHD diagnosis in a first-degree relative. All ADHD participants were stimulant treatment-naïve except one, who 2 years prior to his participation underwent a 6-month methylphenidate trial. Excluding his data did not change the results. Lifetime stimulant exposure did not exceed two uses for any other participants.

The study was carried out in accordance with the Declaration of Helsinki and was approved by the Research Ethics Board of the Montreal Neurological Institute. All participants gave written informed consent.

Participants first underwent an assessment to determine their eligibility to take part in the study based on the criteria described above. Eligible participants underwent two double-blind, randomized, and counterbalanced [¹¹C]raclopride PET scans, one following a lactose placebo and the other following 0.3 mg/kg p.o. of d-amphetamine; capsule administration occurred 60 min before tracer injection. PET scans occurred at least 3 days apart. The mean interval between the two PET scans was longer the ADHD participants (23.1 days, SD = 25.1) than controls (8.4 days, SD = 5.5) [U₍₃₂₎ = 70, p =0.03]. The longer interval for the ADHD group resulted from several outliers, and the median interval did not differ between groups (ADHD: 15 days; controls: 7 days, p = 0.30). Before each scan, participants were asked to abstain from food, caffeine and smoking for 4 h and from alcohol for 24 h. A structural MRI was obtained on a separate day. The average interval between the PET and MR scans was 22.0 (SD = 21.9) for the ADHD participants and 21.8 (SD = 35.1) for the controls (p = 0.52). Toxicology screening occurred on the initial screening interview and prior to both PET scans.

Participants were scanned on a Siemens ECAT HR + PET scanner (CTI/Siemens, Knoxville, Tennessee) with lead septa removed (63-slice coverage), with a maximum resolution 4.2-mm, full width at half maximum (FWHM) in the center of the field of view. Attenuation correction was performed using a 12-min ⁶⁸Ga transmission scan immediately prior to tracer injection. The emission scan started simultaneously with the injection of [¹¹C]raclopride, as an i.v. bolus, and data were acquired for 60 min in 26 time frames of progressively longer duration. Vital signs were monitored and blood samples for plasma amphetamine collected just prior to capsule administration, at the time of tracer administration, mid-scan, and at the end of scan.

High-resolution (1 mm) T1-weighted magnetic resonance images (MRI) were obtained on a 1.5-Tesla Siemens scanner, using gradient echo pulse sequence (TR = 22 ms, TE = 9.2 ms, flip angle = 30°, FOV = 256 mm, and matrix 256 × 256).

Native T1 weighted MRIs were processed through the CIVET automated pipeline (version 1.1.11, in-house software developed in the lab of Alan C. Evans, Montreal Neurological Institute) (Benedetti et al., 2006). CIVET outputs cortical surfaces and CT measurements at 40,962 vertices per hemisphere (Collins et al., 1995; MacDonald et al., 2000; Kim et al., 2005; Benedetti et al., 2006; Lyttelton et al., 2007).

Statistical analyses were implemented using SurfStat, a statistical toolbox created in MATLAB by Dr. Keith Worsley (http://www.math.mcgill.ca/keith/surfstat/). Absolute native-space CT values (blurred to 20 mm) for each subject were entered into general linear models (GLM) predicting CT at each vertex from BP_ND and ΔBP_ND in the three striatal ROIs outlined above (LST, AST, SMST), Group membership (ADHD vs. Control), and the interaction of these terms, with total brain volume as a covariate:

Total brain volume, age and BDI scores were all considered as potential covariates. Their inclusion was based on whether adding each covariate to the model significantly improved its fit. This was evaluated using the SurfStatF. Only total brain volume significantly improved the model's fit, so this was the only covariate used.

The effect of group membership on the relationship of CT with BP_ND and ΔBP_ND was examined by testing the significance of the BP_ND^*Group and ΔBP_ND^*Group interaction terms: their significance indicated that regression slopes describing the relationship between CT and BP_ND/ΔBP_ND differed significantly between the groups. Vertex and cluster significance was determined using Random Field Theory (RFT) for non-isotropic images (Worsley et al., 1999) with significance threshold of p = 0.05, corrected. For each cluster where the interaction was significant, mean CT was computed for each participant, and its relationship with BP_ND/ΔBP_ND characterized in each group using bivariate correlations (using partial correlations with total brain volume as a covariate did not change the pattern of results). The significance thresholds for these latter correlations were uncorrected, as RFT correction was not possible in these analyses.

To test the robustness of these correlations, confirmatory post-hoc analyses were performed predicting vertex-wise CT and from BP_ND/ΔBP_ND in each group using the following GLM:

This was done to identify associations that, despite contributing to an interaction, are weak and do not survive the RFT threshold on their own. This could occur (a) if an interaction is primarily driven by a strong association in one group and a weak opposite association in the other group (significant uncorrected, but no longer significant using the RFT threshold); (b) if an interaction is driven by two such weak associations, in which case the meaningfulness of the interaction is questionable. If a cluster showing a significant Group^*BP_ND/ΔBP_ND interaction remained significant in the confirmatory group-wise analysis, it was considered a valid finding.

Because possible group differences in CT between ADHD and Control groups, which have been reported previously (Bush et al., 2008; Proal et al., 2011; Duerden et al., 2012), could bear on the relationship between CT and BP_ND/ΔBP_ND, we also compared CT between ADHD and Control groups using both vertex-wise and a region of interest (ROI) analyses. The latter was performed because the significance threshold of the vertex-wise analysis (based on RFT) may be too conservative to detect significant differences in our relatively small sample. The vertex-wise analysis used the following GLM:

We did not covary out total brain volume in this analysis. We considered age as a covariate, but it did not improve the fit of the model. For the ROI analysis, mean CT-values for each participant were extracted from the following ROIs defined on the average cortical surface using the automated anatomical labeling set (Tzourio-Mazoyer et al., 2002): frontal (labels 1–28), insula (labels 29 and 30), limbic (labels 31–40), occipital (labels 43–56), parietal (labels 57–70), and temporal (labels 79–90). Between-group comparisons of these values were then made.

The ADHD group did not differ significantly from Controls on any of the demographic variables. Estimated IQ was marginally higher in the control group (p = 0.06) (Table 1). Although no participant was clinically depressed, the ADHD BDI scores at intake were higher than those of Controls (p_s < 0.0005). Because of this group difference and a significant correlation between BDI and ΔBP_ND in ADHD participants only (r = −0.68 p = 0.007), BDI was considered as a covariate, but it was not used in the final models since it did not improve the fit.

In the healthy controls, there was a significant d-amphetamine-induced decrease in BP_ND in LST [t₍₁₇₎ = 5.32, p < 0.0005], but not in AST or SMST (p_s > 0.1). In ADHD subjects, there were significant BP_ND decreases in all three ROIs [AST: t₍₁₄₎ = 2.15, p = 0.05; SMST: t₍₁₄₎ = 3.24, p = 0.006; LST: t₍₁₄₎ = 2.36, p = 0.03), and these effects were significantly greater than those seen in the controls within both the AST and SMST [Group × ROI interaction: F_{(1.36, 39.70)} = 4.07; p = 0.04; AST: F_{(1, 30)} = 4.24, p = 0.05; SMST: F_{(1, 30)} = 4.73, p = 0.04]. See Cherkasova et al. (2014) for more detail.

BP_ND-values on the placebo and d-amphetamine scans and ΔBP_NDs are given in Table 2. As reported previously, the groups did not differ significantly on baseline BP_ND (p_s ≥ 0.07).

We examined whether cortical thickness differed between groups using vertex wise and ROI analyses. Vertex-wise analysis did not reveal any significant clusters. ROI analyses found greater CT in controls than ADHD participants in the right frontal lobe [t₍₃₁₎ = 2.04, p = 0.05], left insula [t₍₃₁₎ = 2.19, p = 0.04], and left and right temporal lobes [left: t₍₃₁₎ = 2.26, p = 0.03; right: t₍₃₁₎ = 2.52, p = 0.02] but these trends did not meet family-wise correction for multiple comparison (p < 0.004).

ΔBP_ND values in the LST interacted with Group to predict cortical thickness in a cluster in the right middle frontal gyrus (p < 0.00005, RFT-corrected) (Figure 1, Table 3). Group-wise correlations to characterize the relationship between LST ΔBP_ND and mean CT in this cluster showed that in the healthy controls, smaller LST ΔBP_ND values were associated with a thicker middle frontal gyrus [r₍₁₈₎ = −0.55; p = 0.02, uncorrected), whereas the converse was seen in the ADHD participants [r₍₁₅₎ = 0.66; p = 0.008, uncorrected]. Separate vertex-wise analyses in each group using RFT thresholding showed that the interaction was most prominently driven by a negative association between LST ΔBP_ND and CT in the control subjects [r₍₁₈₎ = −0.57, p = 0.001, corrected; Table 4). There was also a significant interaction of SMST ΔBP_ND values with Group on CT in the left supplementary motor area (SMA) (Figure 1, Table 3). As in the LST, this interaction reflected divergent associations in the controls and ADHD participants: a trend for a negative association between ΔBP_ND and mean CT was seen in the healthy volunteers [r₍₁₈₎ = −0.36; p = 0.14, uncorrected] while in ADHD participants the association was positive [r₍₁₅₎ = 0.64; p = 0.01, uncorrected; Table 3]. The SMA cluster did not emerge as significant in group-wise confirmatory analysis. The pattern of associations of ΔBP_ND with mean CT was not altered by including total brain volume as a covariate in these correlation analyses (data not shown). Together, the analyses identified evidence of converse associations between CT and striatal DA release in healthy controls (negative correlations) and volunteers with ADHD (positive correlations).

Group interacted with BP_ND in the SMST and AST to predict CT in overlapping regions, such that higher BP_ND was associated with thicker cortex in ADHD but with thinner cortex in controls (Figures 2, 3, Table 4). In the SMST, the interaction was significant in right dorsal anterior cingulate, right cerebellum, left cuneus, and left posterior insula (p_s ≤ 0.03, corrected). In the AST, the interaction was significant in right dorsal anterior cingulate, right cerebellum, left cerebellum and left cuneus (p_s ≤ 0.02, corrected; Table 4). As with the ΔBP_ND-values, these interactions reflected opposite associations in the two groups. In the healthy controls, BP_ND was inversely associated with CT [SMST: right cerebellum r₍₁₈₎ = −0.58, p = 0.01; right anterior cingulate r₍₁₈₎ = −0.43, p = 0.08; left cuneus r₍₁₈₎ = − 0.55, p = 0.02; left posterior insula r₍₁₈₎ = −0.44, p = 0.07; AST: left cerebellum r₍₁₈₎ = −0.34, p = 0.17; right cerebellum r₍₁₈₎ = −0.47, p = 0.051; left cuneus r₍₁₈₎ = −0.58, p = 0.01; right dorsal anterior cingulate r₍₁₈₎ = −0.48, p = 0.04, all p_s uncorrected). Conversely, CT in these clusters was positively associated with BP_ND in the ADHD subjects [SMST BP_ND: r_s(15) ≥ 0.75; AST BP_ND: r_s(15) ≥ 0.76]. Separate vertex-wise analyses in each group showed that the interactions were primarily driven by positive associations between BP_ND and CT in the ADHD group. In these confirmatory analyses, clusters in the right (SMST) and left (AST) cerebellum (for AST and SMST BP_ND, respectively), right anterior cingulate (SMST and AST), left posterior insula (SMST), and left occipital clusters (SMST and AST) and left posterior insula (SMST) remained significant in the ADHD group [r_s(15) ≥ 0.82, p ≤ 0.03, corrected; Figures 2, 3, Table 4]; no clusters remained significant in the control group. BP_ND in the ADHD group also showed highly significant associations with cortical thickness in the insula (particularly on the right) [r_s(15) ≥ 0.85, p_s ≤ 0.001, corrected]. Because of the size of these clusters and the strength of the associations, we report them in Table 4.

For BP_ND in the LST, the interaction was significant in three clusters located in the right posterior cingulate, left posterior cingulate and supramarginal gyrus (p_s ≤ 0.03, corrected). In these clusters BP_ND-values were inversely associated with CT in the healthy controls [right posterior cingulate: r₍₁₈₎ = −0.60; p = 0.009; left posterior cingulate: r₍₁₈₎ = −0.47; p = 0.05; right supramarginal gyrus: r₍₁₈₎ = −0.72; p = 0.001, uncorrected] but positively associated in ADHD [right posterior cingulate: r₍₁₅₎ = 0.68; p = 0.005; left posterior cingulate: r₍₁₅₎ = 0.74; p = 0.002; right supramarginal gyrus: r₍₁₅₎ = 0.53; p = 0.04, uncorrected; Table 4]. However, these clusters did not emerge as significant in the confirmatory vertex-wise analyses in each group using RFT thresholding. As for ΔBP_ND, the pattern of associations of mean CT with BP_ND was not altered by including total brain volume as a covariate (data not shown).

To summarize, we found divergent associations in healthy volunteers (negative correlations) and ADHD subjects (positive correlations) between baseline BP_ND in SMST and AST and CT in anterior cingulate, cerebellar, and occipital regions.

As a final analysis, we examined whether cortical thickness in clusters that were associated with binding measures co-varied with symptom severity. Due to the number of clusters showing significant associations with baseline BP_ND, we used a principal component analysis (PCA) to reduce the number of variables. We entered the clusters showing significant associations with BP_ND in ADHD subjects (including both the clusters where we found a significant interaction with Group and the clusters that emerged in the post-hoc analysis) into a PCA with an oblique rotation; Kaiser-Meyer-Olkin (KMO) value of 0.74 confirmed sampling adequacy. The PCA yielded a single factor with the eigenvalue of 8.73 and accounting for 79.3% of the variance; factor loadings ranged from 0.84 to 0.93. We then asked whether scores on this factor indexing cortical thickness associated with baseline BP_ND were inversely related to ADHD symptom severity. As the measure of ADHD symptom severity, we used the ADHD Index scores of the CAARS. The ADHD Index is a global composite scale of this instrument containing the items that best distinguish ADHD adults from non-clinical adults. In the ADHD subjects, higher ADHD symptom scores were associated with lower scores on the cortical thickness factor, indicating thinner cortex [r₍₁₅₎ = −0.54, p = 0.04]. In the healthy controls, factor scores did not correlate with ADHD Index [r₍₁₃₎ = 0.05; p = 0.86]. Cortical thickness in the middle frontal gyrus cluster (that was associated with ΔBP_ND in controls) was not significantly correlated with ADHD index. In summary, regional cortical thickness associated with higher BP_ND correlated with severity of ADHD symptoms in ADHD subjects.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.