Participant Demographics
A total of 218 TD participants (median age = 9.80 years; 50.5% male) and 232 participants diagnosed with ADHD (median age = 9.82 years; 77.6% male) were recruited. Of the 232 children and adolescents diagnosed with ADHD, 108 participants had the ADHD-C subtype (median age = 8.80 years; interquartile range [IQR] = 7.26–11.24 years; 79.6% male) and 124 participants had the ADHD-I subtype (median age = 10.60 years; IQR = 8.58–13.05 years; 75.8% male). Participant characteristics can be found in Table 1.
Problem Behaviours, Peer Victimization and Diagnostic Group
In our first aim, we examined the relationship between problem behaviours from the CBCL, including social problems, rule breaking behaviour and aggressive behaviour, and diagnostic group (TD, ADHD-I, and ADHD-C). We also examined how the experience of peer victimization differs by diagnostic group (TD, ADHD-I, and ADHD-C). Age, sex, SES, and study site were adjusted for in all the models. In the problem behaviours models 433 participants were included in analysis, 204 TD children, 123 ADHD-I children, and 106 ADHD-C children.
Seventeen participants were excluded from the problem behaviour analyses due to missing data. In the peer victimization model a total of 431 total participants were included in the analysis, 203 TD children, 123 ADHD-I children, and 105 ADHD-C children.
Compared to the TD group, children and adolescents from both the ADHD-C (B = 5.42, 95%CI = 4.26–6.58, p < 0.001) and ADHD-I (B = 2.41, 95%CI = 1.32–3.51, p < 0.001) groups had significantly higher scores on the aggressive behaviour subscale. ADHD-C participants also had significantly higher scores than the TD participants on the rule breaking behaviour (B = 1.95, 95%CI = 1.39–2.50, p < 0.001) and social problems (B = 2.16, 95%CI = 1.53–2.78, p < 0.001) subscales. No significant differences were found between the TD and ADHD-C participants on the withdrawn (B = 0.51, 95%CI = 0.03–0.99, p = 0.038) subscale. Nor were significant differences found between the TD and ADHD-I participants on the rule breaking behaviour (B = 0.62, 95%CI = 0.10–1.15,p = 0.021), social problems (B = 0.75, 95%CI = 0.16–1.33, p = 0.013), or withdrawn (B = 0.49, 95%CI = .04-0.94, p = 0.033) subscales. In comparing ADHD-C to ADHD-I children, we found that the ADHD-C children had significantly higher scores on aggressive behaviour, rule breaking behaviour, and social problems (all, p < 0.001).
When examining the peer victimization subscale, children with ADHD-C had significantly higher levels of peer victimization compared TD children (B = 0.73, 95%CI = 0.47–0.98, p < 0.001), but there was no significant difference found for peer victimization between the ADHD-I and TD groups (B = 0.09, 95%CI=-0.16-0.33, p = 0.483). In addition, the ADHD-C children had significantly higher peer victimization scores than the ADHD-I children (p < 0.001). Refer to Fig. 1 for a summary of the results.
Problem Behaviours, Peer Victimization and Hippocampal Subfield Volumes
In our second aim, we examined whether hippocampal subfield volumes could predict problem behaviours and levels of peer victimization. The problem behaviours examined in this aim included aggressive behaviour, rule breaking behaviour, and social problems subscales, adjusting for age, sex, SES, MRI site, diagnostic group, and TCV. The analysis was run with a total of 282 participants, including 136 TD children, 83 ADHD-I children, and 63 ADHD-C children.
Left CA3 volume was positively associated with peer victimization (B = 0.019, 95%CI = 0.005–0.034, p = 0.010). Left CA3 volume showed a positive but not-significant association with rule breaking behaviour (B = 0.034, 95%CI = 0.003–0.066, p = 0.032), aggressive behaviour (B = 0.054, 95%CI=-0.010–0.119, p = 0.097), and social problems (B = 0.029, 95%CI=-0.006-0.064, p = 0.107).
We subsequently examined the interaction between left CA3 volume and diagnostic group (left CA3 volume x diagnostic group) in the peer victimization model. We found that left CA3 volume was significantly and positively associated with peer victimization in in TD children (B = 0.018, 95%CI = 0.003–0.33, p = 0.017), ADHD-C children (B = 0.022, 95%CI = 0.007–0.036, p = 0.004), and ADHD-I children (B = 0.017, 95%CI = 0.003–0.032, p = 0.022).
Hippocampal Subfield Volumes and Working Memory Ability
This analysis included a total of 243 participants, including 109 TD children, 76 ADHD-I children, and 58 ADHD-C children. When adjusting for age, sex, SES, MRI site, TCV, and hippocampal subfield volumes the ADHD-I group had significantly poorer WM scores than the TD group (B=-5.49, 95%CI = -9.75- -1.23, p = 0.011), but the ADHD-C group did not significantly differ from the TD group (B=-4.31, 95%CI= -8.95- 0.33, p = 0.069).
We examined the relationship between hippocampal subfield volumes and WMI (WISC-V). Left CA3 volume was found to be significantly and positively associated with WMI (B = 0.233, 95%CI = 0.042–0.424, p = 0.017). Statistical interactions were tested in the WMI model between left CA3 volume and diagnostic group (left CA3 x diagnostic group). Left CA3 volume was significantly and positively associated with WMI in TD children (B = 0.249, 95%CI = 0.59 − 0.440, p = 0.010), ADHD-C children (B = 0.230, 95%CI = 0.038–0.422, p = 0.019), and ADHD-I children (B = 0.220, 95%CI = 0.30-0.411, p = 0.023).
Left CA4 volume was found to be significantly and negatively associated with WMI (B=-0.626, 95%CI=-1.14- -0.107, p = 0.018). Statistical interactions were also tested in the WMI model between left CA4 and diagnostic group. Left CA4 volume was significantly and negatively associated with WM in TD children (B=-0.608, 95%CI=-1.128- -0.089, p = 0.022), ADHD-C children (B=-0.625, 95%CI=-1.141 - -0.108, p = 0.018), and ADHD-I children (B=-0.632, 95%CI=-1.151- -0.113, p = 0.017).
Post-Hoc Cluster Analysis
To examine the associations amongst diagnosis, hippocampal volumes, WM ability as well as ADHD symptomatology, a K-means cluster analysis was performed with 245 participants: 111 TD, 58 ADHD-C, and 76 ADHD-I. The analysis included variables of brain morphology, behaviour, and cognition (Z-scored). The model included, left CA3 volume (p = 0.018), peer victimization score (p < 0.001), WMI (p < 0.001), and the hyperactivity-impulsivity and inattention subscales from the SWAN (both p < 0.001). Refer to Table 2 for a summary of the clusters.
Table 2
Three-Cluster Model Participant Demographics.
|
Cluster 1
n = 41
|
Cluster 2
n = 54
|
Cluster 3
n = 150
|
P value
|
Left CA3 (mm3),
Median [IQR]
WMI
Median [IQR]
Peer Victimization
Median [IQR]
SWAN
Hyperactivity-Impulsivity
Median [IQR]
Inattention
Median [IQR]
|
180.00
[155.82–196.24]
107.00
[98.50–122.00]
0.00
[0.00–0.00]
-1.22
[-2.78- -0.56]
-1.56
[-2.22 - -0.78]
|
183.01
[169.90–207.00]
100.00
[88.00–110.00]
2.00
[2.00-3.25]
0.72
[0.22–1.44]
1.00
[0.22–1.67]
|
191.08
[172.81- 211.52]
97.00
[88.00-104.00]
0.00
[0.00–0.00]
0.22
[0.00-0.78]
0.78
[0.11–1.44]
|
0.018
< 0.001
< 0.001
< 0.001
< 0.001
|
Cluster 1 included a total of 41 participants and is characterized as children with small left CA3 volumes, high WMI scores, low peer victimization scores, and low hyperactivity-impulsivity and inattention symptomatology scores. This group is made up of 40 TD participants, and 1 participant from the ADHD-I group. All but 3 participants in this cluster scored 0.5 SD below the mean on the peer victimization scale. In this cluster 56% of the participants had left CA3 volumes that were below the average and 75% of the participants had higher average WMI. In regard to symptoms of hyperactivity-impulsivity and inattention, 39% of the participants in cluster 1 scored 1 SD below the mean and 34% scored below 2 SD below the mean.
Cluster 2 included a total of 54 participants. Participants in this group have average left CA3 volumes, average WMI scores, high peer victimization scores, and high hyperactivity-impulsivity and inattention symptomatology. This group is made up of 19 TD participants, 24 ADHD-C participants and 11 ADHD-I participants. All participants in this cluster were above average in peer victimization scores, with approximately 41% of the cluster at least 2 SD above the mean, and 56% above 1 SD. In this cluster 54% of participants had left CA3 volumes that were below average and 20% of participants had WMI at least 1 SD below the average. In regard to symptomatology, 20% of the participants had hyperactivity-impulsivity scores at least 1 SD above the mean, and 28% had inattention scores at least 1 SD above the mean.
Cluster 3 included a total of 150 participants and can be described as participants with large left CA3 volumes, low WMI scores, low peer victimization scores, and average hyperactivity-impulsivity and inattention symptomatology. This cluster is made up of 52 TD participants, 34 ADHD-C participants, and 64 participants from the ADHD-I group. In cluster 3, approximately 79% of the participants had peer victimization scores that were 0.5 SD below the mean, A total of 60% of the participants had left CA3 volumes that were above the mean, with 21% being least 1 SD above the average. Approximately 30% of this cluster had WMI above the mean, and most participants in this cluster were above average in hyperactivity-impulsivity and inattention symptomatology, 57% and 60% respectively.
When examining cluster membership by diagnostic groups, we found 36% of the total TD sample (n = 111) was grouped into cluster 1, 17% were grouped into cluster 2, and 47% were grouped into cluster 3. Of the 58 total ADHD-C participants, 41% were grouped into cluster 2, and 59% were grouped into cluster 3. Of the 76 total ADHD-I participants, 1% belonged to cluster 1, 14% were grouped into cluster 2, and 84% were grouped into cluster 3. Refer to Fig. 2 for a summary of the cluster analysis.
A one-way ANOVA was conducted to examine the differences between the clusters on all the variables. Significant between-group differences were found for all the variables: left CA3 volume (F(2, 242) = 4.07, p = 0.018), WMI (F(2,242) = 15.44, p < 0.001), peer victimization (F(2,242) = 364.86, p < 0.001), hyperactivity-impulsivity (F(2,242) = 103.63, p < 0.001), and inattention (F(2,242) = 112.24, p < 0.001). Post hoc multiple comparison using Bonferroni correction revealed the differences between the clusters. Significant differences were found between Cluster 1 and Cluster 2 on WMI, peer victimization, hyperactivity-impulsivity, and inattention (all p < 0.001). Significant differences were found between Cluster 1 and 3 on Left CA3 volume (p = 0.015), WMI, hyperactivity-impulsivity, and inattention (all p < 0.001). Significant differences were found between Cluster 2 and 3 on peer victimization (p < 0.001) and hyperactivity-impulsivity (p = 0.001).