Introduction
Attention-deficit/hyperactivity disorder (ADHD) is a common psychiatric disorder affecting 2.5% of adults and 5% of children worldwide.1 ADHD is typically associated with non-emotional symptoms of impulsivity, inattention, and hyperactivity.2 These symptoms have been proposed to partially arise from a deficit in ‘‘cold cognition’’, specifically low capacity in executive functioning relating to deficits in inhibition, vigilance, working memory, and planning.3–5 However, this theoretic framework does not fully capture the diagnostic complexity and heterogeneity of ADHD. Individuals with ADHD show a large symptom overlap and high comorbidity with other disorders characterized by rapid emotion shifts and emotion regulation difficulties, such as emotionally unstable (borderline) personality disorder and conduct disorder.5–7 Emotional instability has been associated with alterations in ‘‘hot cognition’’, including low capacity in regulation of conflicting and distracting information processing related to emotion and reward systems.4,5,8 Importantly, emotion dysregulation has been proposed as a core feature of ADHD, as many patients exhibit significant emotional instability even without any comorbid disorders.9,10 An effort has been made to integrate core ADHD symptoms and emotional instability symptoms within a comprehensive model that encompasses alterations in executive functioning and emotion regulation.4 It describes how similar dysfunctions in different higher order regulatory networks may lead to clinical symptoms in the respective domain. The model is based on findings that non-emotional and emotional prefrontal regulation networks overlap but are also partially distinct.4,5 This organization of top-down regulation systems is preserved in downstream basal ganglia and brainstem neuromodulatory projections.11 Finally, the model suggests that these cognitive core capacities are continuous and therefore, symptoms range from subclinical to clinical levels,4 supported by the occurrence of dimensional ADHD and emotional instability symptoms across the general population and in patients.12–15 Moreover, current findings also suggest a strong relation between these two dimensions.
ADHD has been associated with cortical and subcortical neural alterations. These include altered dorsolateral/dorsomedial prefrontal cortex and caudal anterior cingulate cortex structure—with specifically marked alterations in childhood.16–18 While the morphology in cortical regions seems to be affected in childhood and mature later than in healthy individuals,19,20 these regions show small or no changes in adulthood in individuals with ADHD.19 Altered function in these cortical regions has also been reported in ADHD patients, with more pronounced differences in children and adolescents as compared with adults.21–23 Subcortical volume differences in ADHD appear to be strongly age-dependent. In a large ENIGMA-ADHD study, widespread reductions across multiple subcortical structures were observed in children with ADHD, whereas in adolescents only the hippocampus differed, and no significant subcortical volume differences were detected in adults.24 One of the most robust and consistent functional magnetic resonance imaging (MRI) findings in ADHD is hypoactivation of the ventral striatum/nucleus accumbens during reward anticipation.25,26
Most ADHD studies have employed a case-control design based on categorical diagnoses, without adopting a dimensional approach to symptomatology. However, some exceptions have examined brain structure12,19,27,28 and function25,29–31 using dimensional measures of ADHD symptoms. Notably, the majority of these studies did not control for emotional symptoms, potentially confounding findings with emotion regulation deficits.12 Emotion regulation involves both aspects of emotional and non-emotional cognitive control.
The underlying neural networks that are specific to each of the two domains are therefore difficult to disentangle. Nevertheless, certain regions show some specificity towards emotion regulation processes rather than attentional processes. In general, prefrontal regions such as lateral orbitofrontal cortex, ventrolateral prefrontal cortex, and rostral anterior cingulate cortex exert top-down control over amygdala, ventral striatum/nucleus accumbens, and insula.4,11,32,33 Amygdala and ventral striatum/nucleus accumbens are essential in reward and emotion processing.34–36 These cortical and subcortical regions have not only been repeatedly reported to show altered structure and function in emotional instability disorders,5,22,37–41 but are also impacted in ADHD.19,24,42 Studies employing a dimensional approach to emotional instability symptoms in the general population, while controlling for co-occurring inattention symptoms, have further highlighted the involvement of the lateral orbitofrontal cortex and the rostral anterior cingulate cortex in emotional instability.12,14 Except for a small study investigating reward processing using functional MRI,31 no studies have investigated the specific relationship between emotional instability symptoms and related subcortical regions while adjusting for effects of inattention.
In this study, we aimed to disentangle the structural neural correlates of non-emotional ADHD symptoms, represented by inattention, and emotional instability by using a dimensional approach in adults with ADHD and healthy controls. We hypothesized that inattention would be negatively associated with cortical measures in the dorsolateral/dorsomedial prefrontal cortex and the caudal anterior cingulate cortex, when controlling for emotional instability. Conversely, we hypothesized that emotional instability would be associated with reduced cortical measures of the lateral orbitofrontal cortex and the rostral anterior cingulate cortex, as well as reduced volumes of the nucleus accumbens and the amygdala, when controlling for inattention.
We further conducted exploratory analyses examining whether orbitofrontal and anterior cingulate structure was associated with performance in an emotional Stroop task as an objective index of emotion-related cognitive control.
Materials and methods
Sample
Adult patients diagnosed with ADHD (n = 71) and healthy control individuals (HC; n = 59) were recruited within the St. Göran Project from the Northern Stockholm Mental Health Service. Recruitment and assessment of study participants have been described previously.43 Briefly, ADHD patients were assessed by a clinical psychiatrist through a structured interview.
Diagnoses were established using a structured psychiatric interview, including the Affective Disorder Evaluation that had been complemented with DSM-IV diagnostic criteria for ADHD, and the Mini International Neuropsychiatry Interview.44 In addition to the anamnestic interview, the Wender Utah Rating Scale45 was used to assess childhood ADHD symptoms, and the Adult ADHD Self-Report Scale46 was used to assess adult ADHD symptoms. The final main diagnosis and psychiatric comorbid conditions were established by board-certified psychiatrists considering all available information. Healthy participants were interviewed using the same instruments to exclude any psychiatric disorder.
All participants gave oral and written informed consent. The study was approved by the regional ethics committee in Stockholm.
Non-emotional and emotional symptom assessment
The Brown Attention-Deficit Disorder Scales (B-ADD)47 self-report questionnaire was used to assess non-emotional and emotional symptoms. The questionnaire comprises 40 items rated on a scale from 0 (‘‘never’’) to 3 (‘‘almost daily’’), organized into five subscales targeting different aspects of ADHD. For the purposes of this study, two subscales were selected to represent the target symptom domains: Attention, referred to here as the Inattention score (assessing the ability to focus, sustain, and shift attention) as a proxy for non-emotional cognition, and Affect, here referred to as the Emotional instability score (assessing emotion regulation and frustration tolerance).14,31 The maximum scores are 27 for Inattention and 21 for Emotional instability, with higher scores indicating greater symptom severity.
The NecStroop-task
A subsample of our study participants (ADHD n = 25, HC n = 30) were assessed with a Neutral and Emotional Computerized Stroop task (NecStroop-task) on the same day of MRI scanning. The NecStroop-task included both neutral (buildings with numerical labels) and emotional (faces with emotional words) conditions. In each condition, trials were either congruent (stimuli matched) or incongruent (stimuli conflicted), requiring participants to focus on the relevant stimulus while ignoring the distractor.
The detailed description of the task and the sample can be found in the Supplement. Task performance was indexed using two complementary measures: (a) number of correct responses (accuracy) and (b) mean reaction time (processing speed) in the congruent and incongruent conditions, and for neutral and emotional stimuli. We tested whether the performance in the NecStroop-task was associated with cortical thickness in the following defined regions of interest (ROIs): lateral orbitofrontal cortex (OFC), rostral and caudal anterior cingulate cortex (ACC), and dorsolateral/dorsomedial prefrontal cortex (dl/dmPFC)—regions that have previously been shown to be involved in non-emotional and emotional regulation and to be dysregulated in patients with ADHD and emotional instability.4
MRI data acquisition and processing
Structural T1-weighted images were acquired on a 3T General Electric scanner (3D‐BRAVO, TR = 6.4 ms, TE = 2.8 ms, FOV = 24 cm, flip angle 12°, acquisition matrix 240 × 240 × 180) at the MR Research Centre at Karolinska University Hospital in Stockholm, Sweden between 2013 and 2017. Image processing was performed in Freesurfer 5.1 and included automated segmentation and parcellation of the ROIs, resulting in average cortical thickness and surface area of pre-specified cortical ROIs and volume of subcortical ROIs.48–50 Based on our hypotheses, we selected the following cortical ROIs for further analysis: dl/dmPFC (combining superior frontal, caudal middle frontal, and rostral middle frontal cortex), caudal anterior cingulate cortex (cACC), lateral orbitofrontal cortex (lOFC), and rostral anterior cingulate cortex (rACC) as defined previously12,51 (Supplement Fig. S1a).
Amygdala and nucleus accumbens (NAcc) were selected as subcortical ROIs (Supplement Fig. S1b). Although not related to our main hypotheses, caudate, putamen, and hippocampus24 were analysed in exploratory tests for completeness.
Right and left ROIs were combined since there was no hypothesis regarding lateralization. We further performed a surface-based vertex-wise analysis to complement our main hypotheses (see Supplement for details).
Statistical analysis
Statistical analyses were conducted using R version 4.0.2 and Python version 3.11.1. Demographic variables were compared between the patient and healthy individuals using Levene's test to test for unequal variance of variables in the two groups, and Student's t-test or Welch's t-test were used where appropriate.
To test our hypotheses, we ran multiple regression analyses across the full sample using a linear model with cortical thickness of the defined ROIs as dependent variable and Emotional instability and Inattention B-ADD scores as independent variables. When investigating ROIs hypothesized to be related to emotional processes, Emotional instability B-ADD score was the variable of interest, whereas Inattention B-ADD score was treated as nuisance variable of no interest to remove any variance explained by this domain; and vice versa for investigating non-emotional ROIs. Age and sex were additional regressors in all analyses.
In additional tests, we controlled for intracranial volume (ICV) to confirm that ICV was not a confounding factor.
Multiple comparison correction was performed within each imaging phenotype using Bonferroni's Dubey Armitage-Parmar/Sidak's adjustment of α-level considering the number of tests (n = 4 cortical ROIs, n = 2 subcortical ROIs) and the inter-correlation between the dependent variables.52 We treated analyses of cortical and subcortical phenotypes as separate scientific questions as commonly done in previous studies by us and others.19,24 Although we had directed hypotheses, we report results from two-sided tests.
Unless otherwise stated, all primary ROI and vertex-wise analyses were conducted across the full sample of individuals with ADHD and healthy controls, using dimensional B-ADD symptoms scores as predictors. In follow-up analyses, we tested for a diagnosis-by-B-ADD interaction effect and main effect of diagnosis to examine whether the observed dimensional associations differed by diagnostic group, and to rule out that our findings were driven by case-control differences unrelated to the B-ADD symptom measures of interest.
We also tested whether other clinical variables—specifically co-morbid diagnoses and medication (see Supplement Table S1)—confounded the observed correlations, by (a) adding those (one at a time) as independent variables in the multiple regression model, or (b) excluding cases with a specific condition where the number of participants was small (n < 10).
Complementing the ROI analyses and to detect potential effects not captured by predefined regions, we performed complementary vertex-wise analyses of cortical thickness and surface area investigating associations between brain structure and B-ADD scores. We computed two separate models: first, associations with Inattention, correcting for age, sex, and Emotional instability, and second, associations with Emotional instability while correcting for age, sex, and Inattention. Each analysis was conducted separately across the left and right hemispheres, resulting in a total of eight vertex-wise models.
Given the exploratory nature of this analysis, we applied an uncorrected cluster-wise significance threshold of −log10(p) > 1.3, corresponding to p < 0.05. Statistical maps were generated using Freesurfer's QDEC tool. Only clusters exceeding the threshold were visualized.
In a subsample of our study participants (see Supplement Table S5), we explored associations of the performance in the NecStroop-task with brain ROIs. The measures (mean reaction time and accuracy) were entered in a regression model testing the interaction term group-by-stroop performance to identify group-specific effects, which were followed up by partial correlations (adjusting for age and sex) in individuals with ADHD and in HC alone. These analyses were not part of the primary hypotheses and intended to generate hypotheses future investigations.
Results
Demographic and clinical data
We included 71 patients with ADHD (34 female, 37 male, age 34.1 ± 10.00) and 59 healthy individuals (29 female, 30 male, age 43.5 ± 12.7). Individuals with ADHD were younger than controls (t(128) = 4.71, p < 0.001), yet all analyses were adjusted for age. A detailed description of the sample, medication, and comorbidities is presented in Supplementary Table S1.
Individuals with ADHD had significantly higher B-ADD Inattention scores (ADHD: 16.41 ± 5.08, HC: 5.69 ± 4.13, t(124) = 12.85, p < 0.001) and B-ADD Emotional instability scores (ADHD: 9.16 ± 4.18, HC: 2.98 ± 2.47, t(108) = 10.28, p < 0.001) than controls. Figures 1a and 1b illustrate the distribution of Inattention and Emotional instability across the whole sample and within each group. B-ADD subscales Inattention and Emotional instability showed a strong correlation (r = 0.73, p < 0.001) as shown in Fig.
1c. Individuals with ADHD had significantly higher MADRS scores than controls (ADHD: 4.63 ± 2.46, HC: 1.57 ± 1.57, t(127) = 7.04, p < 0.001), although still on a non-clinical level.
Correlations between cortical brain measures and B-ADD Inattention and Emotional instability
There were no significant correlations between Inattention scores and cortical thickness in dl/dmPFC or cACC ROIs after adjusting for Emotional instability, nor were there any significant correlations between Emotional instability and cortical thickness in rACC or lOFC ROIs when adjusting for Inattention (see Supplement Table S2). We observed a positive correlation between bilateral lOFC thickness and Inattention when controlling for Emotional instability (standardized regression coefficient β = 0.24, p = 0.042), which did not withstand correction for multiple comparisons (inter-correlation coefficient of 0.532 led to an adjusted α-level of 0.026 for statistical significance; see Supplement Table S2). Controlling for ICV did not change these results (β = 0.24, p = 0.044).
There was no group-by-Inattention interaction, or main effect of group (β = 0.04, p = 0.784), i.e., ADHD diagnosis status did not explain a significant part of the variance. Follow-up analyses controlling for co-morbid diagnoses and medication as potential confounders did not change the results. No significant correlations were found for cortical surface area and B-ADD scores (see Supplement Table S3).
Correlations between subcortical brain measures and B-ADD Inattention and Emotional instability
In line with our hypothesis, we found a significant negative correlation between NAcc volume and Emotional instability when adjusting for Inattention, age, and sex (β = −0.29, p = 0.016, Fig. 2, Supplement Table S4). This correlation remained significant after correction for multiple testing (inter-correlation coefficient 0.37 rendered p < 0.032 as statistically significant).
The negative correlation also remained significant when controlling for ICV (β = −0.25, p = 0.029). There was no interaction effect of group-by-Emotional instability (p = 0.27), and group status did not predict NAcc volume (β = 0.21, p = 0.12). We conducted follow-up analyses to test potential confounders.
When excluding seven ADHD patients with a comorbid diagnosis of social anxiety disorder, the negative correlation between Emotional instability and NAcc volume was no longer statistically significant (β = −0.18, p = 0.14, Supplement Fig. S2).
There were no correlations between bilateral NAcc and Inattention when adjusting for Emotional instability (Supplement Table S4). There were no correlations between either Inattention or Emotional instability and amygdala volume (Supplement Table S4). However, Emotional instability correlated negatively with bilateral caudate volume, controlling for Inattention, age, and sex (β = −0.30, p = 0.018).
The negative correlation remained significant when controlling also for ICV (β = −0.23, p = 0.038). There was no group-by-Emotional instability interaction effect. There were no correlations of bilateral caudate and Inattention, controlling for Emotional instability, age, and sex.
Volume of bilateral hippocampus and putamen did not correlate either with Inattention or Emotional instability controlling for the other B-ADD dimension, age, and sex.
After correction for multiple comparisons, the association between Emotional instability and NAcc volume remain significant, whereas all other ROI associations from our main hypotheses were non-significant or did not survive correction.
Vertex-wise analysis of cortical thickness and surface area
The results from the exploratory vertex-wise analyses yielded patterns broadly consistent with the ROI-based findings, but also revealed additional regionally specific associations. While no significant associations of the dl/dmPFC and symptom scores was found in the ROI analysis, the vertex-wise analysis revealed a cluster in surface area comprising the left dlPFC and superior frontal cortex, the left middle temporal, and the left superior/inferior parietal cortex indicating negative correlations between Emotional instability and surface area after adjusting for Inattention (see Supplement Fig. S3).
After correcting for Emotional instability, Inattention associated positively with cortical thickness of a cluster comprising the bilateral, lateral orbitofrontal cortex, pars orbitalis, insula, and cuneus (see Supplement Fig. S4). Furthermore, a positive correlation was found between Inattention and surface area in a cluster comprising the precentral and superior frontal cortex (see Supplement Fig.
S5). Note that vertex-wise analyses were conducted using an uncorrected cluster-wise threshold of p < 0.05 as they were considered additional and exploratory.
Association of NecStroop-task performances and orbitofrontal cortex thickness
Individuals with ADHD displayed longer reaction times (conditions incongruent neutral and emotional) and fewer number of correct hits (conditions congruent and incongruent neutral) than HC (see Supplement Table S5 and S6).
We found interaction effects of reaction time and group for bilateral lOFC and rostral ACC cortical thickness (see Supplement Table S7.1). Group-specific analyses indicated differential directions of these associations: In ADHD patients, longer reaction time associated with more bilateral lOFC (congruent neutral: r = 0.49, p = 0.012; incongruent neutral: r = 0.44, p = 0.027) and rostral ACC cortical thickness (congruent neutral: r = 0.49, p = 0.013). In HC, bilateral lOFC cortical thickness associated negatively with the reaction time in the congruent emotional (r = −0.57, p < 0.001) and the incongruent neutral condition (r = −0.40, p = 0.028; see Supplement Table 7.2 and Fig.
S6).
Discussion
This study examined how non-emotional inattention symptoms and emotional instability symptoms relate to cortical and subcortical brain structures while controlling for the respective other symptom domain in a sample of adults with ADHD and healthy controls. This approach allows for discrimination of brain correlates uniquely associated with inattention or emotional instability.
In line with our hypotheses, NAcc volume correlated negatively with Emotional instability score when controlling for Inattention score. In contrast to our hypothesis, we did not observe any associations between dorsolateral/dorsomedial prefrontal cortex (dlPFC) or cACC structure and Inattention when controlling for Emotional instability, nor between lOFC, rACC, or amygdala structure and Emotional instability when controlling for Inattention. In additional exploratory analyses, we observed a negative correlation between caudate volume and Emotional instability while controlling for Inattention, whereas amygdala, putamen, and hippocampus showed no associations with either Inattention or Emotional instability scores.
Complementary whole-brain analyses further suggested that higher emotional instability was associated with smaller prefrontal surface area across regions implicated in emotional regulation, social cognition, and multisensory integration. Inattention was positively associated with cortical thickness and surface area in a cluster encompassing the bilateral lateral orbitofrontal cortex, pars orbitalis, insula, and cuneus, and with surface area in a cluster comprising the precentral and superior frontal cortex.
Although ADHD is classically associated with non-emotional symptoms and executive function deficits, it has been shown that ADHD relates to smaller NAcc volume in children24 and to lower VS/NAcc functional activation during reward anticipation in children and in adults.5,25–27 In the present study, smaller NAcc volume was associated with higher levels of emotional instability after accounting for inattention, suggesting that NAcc morphology may be more closely related to emotional than to non-emotional symptom variance in adulthood. This interpretation is consistent with clinical observations of emotion regulation difficulties in adults with ADHD.9,10 However, exclusion of ADHD participants with comorbid social anxiety disorder attenuated this association, highlighting the difficulty of disentangling emotional instability from anxiety-related symptomatology using categorical diagnostic boundaries. Consequently, the present data do not allow determination of whether the observed association reflects effects of emotional instability per se, comorbid anxiety symptoms, or variance shared between these dimensions, and the finding should therefore be interpreted cautiously.
This ambiguity is further complicated by the substantial phenomenological overlap between ADHD and anxiety disorders, including shared features such as emotional dysregulation and rejection sensitivity. Excluding individuals based on a categorical diagnosis of social anxiety disorder may thus preferentially remove participants with higher overall emotional symptom burden rather than isolating a distinct comorbid condition. In this context, large-scale volumetric studies of social anxiety disorder have reported smaller putamen volumes but no differences in NAcc volume relative to healthy controls, suggesting that anxiety-related subcortical alterations may be regionally specific.53 Nevertheless, given the overlap between ADHD and anxiety symptoms, the present findings cannot be attributed to a single symptom domain.
In line with this, patients with emotional instability disorders, characterized primarily by emotional rather than non-emotional problems, often present with altered ventral striatal function54,55 and connectivity.56 A recent study in ADHD patients demonstrated that functional connectivity of the NAcc—particularly with the salience and default mode network—predicted ADHD symptom trajectories, suggesting a key role of NAcc connectivity in emotional processing and long-term clinical outcomes.57 Moreover, results of the present study also support our previous fMRI study reporting a negative correlation of Emotional instability and VS activation during reward anticipation in healthy females.31 The present results further highlight the need to break down individual psychiatric diagnoses into continuous transdiagnostic symptom domains, as this approach might capture the dimensional nature of psychiatric symptoms more accurately and may reveal shared mechanisms that are obscured by categorical diagnoses.
Furthermore, exploratory tests suggested a negative correlation between caudate volume and Emotional instability, while controlling for the effect of Inattention. Previous studies examining caudate structure in relation to ADHD have reported mixed results.24,58,59 The largest study to date reported smaller caudate volumes in children with ADHD, whereas no significant case-control differences were observed in adolescents or adults.24 Given that the present study focuses exclusively on adults, these null findings in adult ADHD represent the most relevant point of comparison. Rather than reflecting persistent case-control differences in adulthood, our results suggest that variability in subcortical structure may be more closely related to dimensional aspects of emotional instability than to inattention symptoms.
This interpretation aligns with meta-analytic evidence implicating the caudate in emotionally related disruptive behaviour disorders60. Accordingly, the caudate should be considered in future studies investigating subcortical correlates of both non-emotional and emotional symptoms in ADHD. Supporting this view, in a previous study focusing exclusively on healthy adults, we observed associations between emotional traits and subcortical structure in a similar direction, further suggesting that emotional instability indexes neurobiological variability that extends across diagnostic boundaries.14
It has previously been shown that ADHD patients display larger lOFC compared to controls,18 and that Inattention scores are associated with increased lOFC volume in healthy participants.14 It has been suggested that a larger lOFC in ADHD patients could potentially compensate for a deficient function of the parallel non-emotional executive function system.4 In line with that, we found a positive correlation between lOFC thickness and the Inattention score when controlling for Emotional instability. However, this result should be treated with caution since the correlation was not significant after correction for multiple testing and as this was an exploratory analysis.
We did not observe any correlations between either dlPFC or cACC structure and Inattention score, or between lOFC or rACC and Emotional instability score in our sample when controlling for the respective other domain. However, our additional whole brain analysis highlighted distinct cortical patterns: higher levels of emotional instability were linked to smaller prefrontal surface area across widespread regions involved in emotional regulation, social cognition, and integration of sensory information. Inattention was positively associated with cortical thickness in a cluster, including the bilateral lOFC, pars orbitalis, insula, and cuneus, and related to surface area in a cluster comprising the precentral and superior frontal cortex.
These findings suggest that structural enlargement in regions supporting cognitive control and attentional processes may relate to attentional difficulties. Some medium large studies and meta-analyses reported altered brain morphology in dlPFC and cACC as well as in lOFC and rACC in both ADHD5,16,42,61,62 and emotional instability disorders.22,38,39 Still, other large multi-site studies and meta-analyses report no cortical differences between adult ADHD patients and controls in these regions.19,63 Instead, ADHD appears to be characterised by a delay in prefrontal cortex development,20 rather than a persistent difference between adults with and without ADHD.
These divergent findings may also suggest that differences in brain structure are not simply related to a categorical diagnosis per se. However, a dimensional approach as applied here could also not confirm a relation between Inattention or Emotional instability score in these brain regions. One explanation for this could be the fact that we used B-ADD subscales as a proxy for non-emotional and emotional symptoms, only indirectly reflecting the capacity in non-emotional and emotional top-down regulation.
We acknowledge that “cold” cognition encompasses a broader set of functions—including sustained attention, working memory, and inhibition—which are not fully captured by inattention scores alone. Besides, the pre-specified ROIs were relatively large, and future studies may use higher regional resolutions to investigate more focal effects. In addition, cortical differences seem to be less prominent than subcortical alterations in relation to ADHD in adults,62,63 and may also be affected by long-term use of central stimulant medication.16
We did not find any association between amygdala volume and Emotional instability while controlling for Inattention. The amygdala is central to emotion processing and regulation,36 and altered amygdala structure and function have been reported in disorders characterized by emotional dysregulation.38,39 In ADHD, large-scale data showed smaller amygdala volume overall, but age-stratified analyses indicated that these differences were most prominent in childhood and not significant in adulthood.24 Thus, any structural amygdala differences in ADHD should be interpreted as age-dependent rather than uniform across the lifespan. Structural and functional amygdala alterations have also been reported in major depressive disorder (MDD), post-traumatic stress disorder, and in relation to childhood trauma.64Although our study could not relate amygdala structure to Emotional instability score, future studies should focus on disentangling the effects of emotional instability and long-term mood alterations.
Furthermore, we report that the performance in an emotional/non-emotional Stroop task related to cortical thickness in the lOFC with both shared and disorder-specific effects. These results suggest that structural features of the lOFC may differentially impact attentional and emotional performance across groups, potentially reflecting distinct compensatory or maladaptive mechanisms in ADHD versus healthy individuals. However, this conclusion cannot be derived from this study alone and needs further investigation.
Strengths and limitations
A key strength of this study is its dimensional approach, focusing on symptom-related neurobiological variation across individuals rather than exclusively on categorical diagnostic contrasts. Nevertheless, several limitations should be acknowledged. First, a large proportion of participants with ADHD were using central stimulant medication (n = 59) and/or antidepressants (n = 23).
Although associations between brain measures and B-ADD symptom scores were comparable between patients and healthy controls and remained similar when controlling for current medication status, we acknowledge that information on lifetime medication exposure, treatment duration, and cumulative dosage was not available. As prolonged psychostimulant or antidepressant treatment may influence brain morphology, particularly in frontostriatal and limbic regions, residual medication effects related to past exposure cannot be excluded. A second, important limitation is the age difference between patients and controls.
Although age was included as a covariate in all analyses, brain structural measures do not change linearly across adulthood, and individuals with ADHD and healthy controls might follow different developmental trajectories. Further, our cross-sectional design precludes any conclusions about causality or longitudinal changes in brain morphology. Third, it should be noted that morphometric estimates derived from structural MRI can vary across FreeSurfer versions due to ongoing algorithmic refinements.
Although such version-dependent differences are unlikely to affect the relative group comparisons and statistical inferences central to the present study, future work could assess the robustness of the findings using more recent FreeSurfer releases. Fourth, while structural MRI provides valuable information on brain anatomy, it does not address functional dynamics. While we observed correlations between cognitive outcomes and brain structure, future studies using functional MRI are needed to investigate the relationship between symptom levels and neural activity.
Further, we operationalized emotional instability using the affect subscale of the B-ADD, which captures clinically relevant aspects of emotional instability in ADHD but does not encompass the full breadth of emotion dysregulation as assessed by more comprehensive instruments such as the DERS. However, it should be noted that DERS captures how individuals cope with emotions, rather than how emotions vary in intensity and frequency.
In ADHD, only a subset of individuals shows pronounced objective executive function deficits, whereas others may present primarily with motivational or emotional dysregulation, highlighting the heterogeneity of cognitive and affective profiles within the disorder.65 In this context, the high rate of comorbidity with MDD in the present ADHD sample warrants careful consideration. Emotional dysregulation is a prominent feature of mood disorders, and such comorbidities may contribute to the severity of emotional instability symptoms observed in our ADHD cohort. While our dimensional analytic approach adjusts for ADHD diagnosis, it cannot fully disentangle the unique contributions of comorbid psychiatric conditions.
Nevertheless, the present cohort represents a clinically and ecologically valid sample. Future studies should aim to clarify the distinct and shared neurobiological mechanisms underlying emotional dysregulation across ADHD and commonly comorbid conditions, including mood and anxiety disorders.
Conclusion
Our study suggests a negative association between NAcc volume and emotional instability, which is unrelated to inattention symptoms, in a sample spanning clinical ADHD patients and healthy individuals. This indicates that VS/NAcc alterations commonly attributed to an ADHD diagnosis might relate to the level of emotional rather than non-emotional symptoms regardless of diagnosis. Future studies should include individuals with emotional instability, ADHD, and healthy controls, while assessing continuous symptom levels across transdiagnostic domains.
This approach could help disentangle emotional from non-emotional symptoms and reveal distinct neural mechanisms underlying specific dimensions.