Orbitofrontal and Medial Frontal Gray Matter Volumes Are Associated With Hyperactivity Traits in 2548 Nonclinical Adults

ABSTRACT

Attention‐deficit/hyperactivity disorder (ADHD) traits vary continuously across the general population, yet whether the frontal structural correlates established in clinical ADHD extend to trait variation in nonclinical adults is unknown. Based on meta‐analytic evidence, we hypothesized that orbitofrontal gray matter volumes would be negatively associated with hyperactivity traits in 2548 nonclinical Japanese adults (1488 males, 1060 females; aged 23–76 years) who underwent structural MRI and completed the Adult ADHD Self‐Report Scale. Regional gray matter volumes of 61 bilateral‐averaged regions were tested against inattention, hyperactive motor, and hyperactive verbal trait subscales using partial correlations controlling for age and sex, with false discovery rate (FDR) correction. Five associations survived correction, all linking smaller orbital or medial frontal volumes to higher hyperactivity: medial frontal cortex with hyperactive motor and hyperactive verbal traits, and anterior orbital gyrus, lateral orbital gyrus, and orbital inferior frontal gyrus with hyperactive verbal traits (r = −0.070 to −0.078). The effect sizes were within the range reported for clinical ADHD by the ENIGMA consortium, and secondary analyses found no evidence that bilateral averaging concealed an additional hemisphere‐specific association. The findings were robust across sensitivity and permutation analyses. Inattention showed no FDR‐significant associations, with Bayesian and equivalence analyses supporting near‐null effects at the benchmarked magnitude for most regions. Brain volumes did not improve individual‐level prediction of trait scores. These findings identify small orbitofrontal and medial frontal volume associations with hyperactivity along the adult ADHD trait continuum and provide a foundational nonclinical baseline for clinical and transdiagnostic studies of ADHD traits.

In 2548 nonclinical adults, smaller orbitofrontal and medial frontal cortex gray matter volumes were associated with higher hyperactivity traits (r = −0.07 to −0.08). These small, robust effects, consistent with ENIGMA ADHD findings, suggest frontal structure–trait mapping extends to the nonclinical ADHD continuum.

1. Introduction

Attention‐deficit/hyperactivity disorder (ADHD) symptoms are increasingly recognized as dimensional traits that vary continuously across the general population, rather than defining a discrete diagnostic category (Franke et al. 2018; Lubke et al. 2009; Marcus and Barry 2011). This dimensional perspective is supported by genetic evidence showing shared etiology between clinical ADHD and subclinical traits in community samples (Riglin et al. 2016), and by epidemiological data indicating that even subthreshold symptoms are associated with meaningful functional impairment (Asherson et al. 2016). Despite this, the vast majority of ADHD neuroimaging research has relied on case–control designs, comparing diagnosed patients to healthy controls. Whether the structural correlates identified in diagnosed samples extend into the nonclinical adult range therefore remains an open question that case–control designs cannot address.

Structural neuroimaging studies of ADHD have consistently reported volumetric reductions in subcortical regions (accumbens, amygdala, caudate, hippocampus, and putamen) as well as in frontal cortical areas (Frodl and Skokauskas 2012; Hoogman et al. 2017; Nakao et al. 2011). The orbitofrontal cortex (OFC) has emerged as a region of particular relevance, given its role in impulse regulation, response inhibition, and emotional decision‐making (Hesslinger et al. 2002). Recent meta‐analyses have also reported reduced OFC gray matter in ADHD (Chen et al. 2025; Yu et al. 2022). However, the effect sizes reported by the ENIGMA (Enhancing NeuroImaging Genetics through Meta‐Analysis) consortium mega‐analysis are small (Cohen's d = 0.10–0.19; Hoogman et al. 2017), and findings from clinical samples may not generalize to the broader trait continuum. At these magnitudes, brain–behavior associations are informative about population‐level structure–trait mapping but are not expected to support individual‐level prediction.

Sex is an important consideration in this context. Males and females differ in both ADHD symptom profiles, with males presenting more hyperactive–impulsive symptoms and females more inattentive symptoms (Ramtekkar et al. 2010; Willcutt 2012), and in brain morphometry, with large sex differences in total gray matter volume (d > 1.0; Ritchie et al. 2018, Ruigrok et al. 2014). Onnink et al. (2014) reported a sex‐by‐diagnosis interaction for caudate volume in adult ADHD, suggesting that sex should be accounted for in brain–ADHD analyses.

Using a subset of the present sample, we previously reported that ADHD traits mediated associations between regional gray matter volumes and traffic crash risk via path analysis (Putra et al. 2023). In related large‐scale regional gray matter volume (rGMV) studies in nonclinical adults, we have also shown that regional gray matter volumes covary with fatigue perception and exhibit sex differences (Putra, Park, and Yamashita 2022; Putra, Park, Yamashita, et al. 2022). Neither line of work, however, tested whether the frontal structural correlates established in clinical ADHD extend to dimensional trait variation in adults without a diagnosis. Testing this requires a sample large enough to detect associations of the size that clinical meta‐analyses predict (|r|≈0.05–0.10; Hoogman et al. 2017) and to separate small true associations from null effects.

The present study examined this question in 2548 nonclinical adults who underwent structural MRI and completed the Adult ADHD Self‐Report Scale as part of a brain health check‐up program in Japan. ADHD traits were decomposed into three subscales (inattention, hyperactive motor, and hyperactive verbal), following the factor structure established by Stanton et al. (2018) and consistent with the three‐community structure reported by Panagiotidi et al. (2024). Based on meta‐analytic evidence of OFC involvement in ADHD (Chen et al. 2025; Nakao et al. 2011; Yu et al. 2022) and the role of the OFC in impulse regulation and the suppression of inappropriate responses (Hesslinger et al. 2002), we hypothesized that orbitofrontal gray matter volumes would be negatively associated with hyperactivity traits in this nonclinical population. Because clinical meta‐analyses indicate that such effects are small, findings were interpreted against an ENIGMA‐informed benchmark of (|r| = 0.05). Whether inattention shows volumetric correlates in nonclinical adults, whether associations extend beyond the OFC, and whether any associations are subscale‐specific were treated as open empirical questions. Hemispheric lateralization and sex were examined in secondary analyses, because many frontal functions are lateralized (Aron et al. 2004) and because males and females differ in symptom profiles and brain morphometry (Ruigrok et al. 2014; Willcutt 2012).

2. Methods

Figure S1 provides an overview of the analytical pipeline, from data acquisition through the primary and secondary analyses.

2.1. Participants

The dataset available for the present analyses comprised 2695 participants before MRI quality‐control exclusion, and the final analytic sample included 2548 nonclinical adults (1488 males, 1060 females; aged 23–76 years, M = 52.8, SD = 8.6) recruited through brain health check‐up programs (“Brain Dock”) in Kochi, Japan. All retained participants were right‐handed, aged 23 years or older, had a Mini‐Mental State Examination (MMSE) score ≥ 28, and completed the Adult ADHD Self‐Report Scale questionnaire. Participants without consent for research use of their data were excluded. MRI scans were acquired at Kochi Kenshin Clinic by trained clinical staff under the supervision of a medical specialist (K.P.), who was responsible for MRI operations and visual quality control. Participants with imaging artifacts, including artifacts related to dental treatment, and those with cerebrovascular or neurological conditions likely to affect regional gray matter volume measurements, such as cerebral infarction, high‐grade leukoaraiosis, brain tumors, arachnoid cysts, hydrocephalus, or psychotic disorders including schizophrenia, were excluded. Exact counts for each exclusion reason were not retained in the exported analysis dataset. This study was conducted in compliance with the Ethics Guidelines for Medical and Health Research Involving Human Subjects and the Declaration of Helsinki. All participants provided written informed consent before participation. Participants were informed that their data would be kept confidential and used solely for the scientific objectives of this study. Ethical approval was obtained from the institutional review board at Kochi University of Technology (application no. 213).

2.2. ADHD Trait Assessment

ADHD traits were measured using items from the Adult ADHD Self‐Report Scale v1.1 (ASRS; Kessler et al. 2005), an 18‐item World Health Organization instrument whose items correspond to the 18 DSM symptom criteria for ADHD. The full scale comprises Part A, a 6‐item subset selected for maximal predictive value when screening for probable clinical ADHD, and Part B, the remaining 12 symptom‐frequency items (items 7–18 of the full scale). The Brain Dock questionnaire protocol, established before the present analysis, used the 12 Part B items to measure dimensional symptom variation in a nonclinical population and did not include the Part A screening subset. Each administered item was scored on a 5‐point frequency scale (1 = never to 5 = very often). The Japanese version of the full 18‐item instrument (ASRS‐J) has been validated by Takeda et al. (2017), and the items administered here are drawn from that validated translation. Within the original scale, Part B comprises five inattention items and seven hyperactivity/impulsivity items (Kessler et al. 2005). Items were grouped into three subscales following the factor structure established by Stanton et al. (2018), whose analysis of all 18 ASRS items identified distinct inattention, motor hyperactivity/impulsivity, and verbal hyperactivity/impulsivity dimensions; restricted to Part B, their item assignments correspond exactly to the three subscales below. Exploratory graph analysis has also supported a three‐dimensional structure for the ASRS (Panagiotidi et al. 2024), with a hyperactivity grouping that comprises the same items as the motor dimension used here. While alternative factor structures remain in use, we adopted this three‐factor solution to enable finer‐grained trait mapping rather than to adjudicate competing measurement models:

• Inattention (Part B items 1–5, corresponding to full‐scale items 7–11; range 5–25): difficulty focusing, organizing, and remembering.

• Hyperactive motor (Part B items 6–8, full‐scale items 12–14; range 3–15): fidgeting, restlessness, inability to relax.

• Hyperactive verbal (Part B items 9–12, full‐scale items 15–18; range 4–20): talking excessively, interrupting, difficulty waiting.

Internal consistency (Cronbach's α) was 0.696, 0.608, and 0.632 for Inattention, Hyperactive motor, and Hyperactive verbal, respectively.

2.3. MRI Acquisition and Brain Volumetry

T1‐weighted MRI scans were acquired using a 1.5‐T ECHELON Vega system (Hitachi, Tokyo, Japan) with a three‐dimensional gradient echo with inversion recovery sequence (repetition time = 9.2 ms, echo time = 4.0 ms, inversion time = 1000 ms, flip angle = 8°, field of view = 240 mm, in‐plane resolution = 0.9375 × 0.9375 mm, slice thickness = 1.2 mm, single excitation). Imaging data were screened for brain pathologies, motion artifacts, and anomalies potentially affecting volumetric measurements.

Voxel‐based morphometry (VBM) analysis was performed using the VBM8 toolbox (http://dbm.neuro.uni‐jena.de/vbm/) for SPM12 (RRID:SCR_007037). Segmentation of gray matter, white matter, and cerebrospinal fluid was executed using the maximum a posteriori approach. Gray‐ and white‐matter images were further processed to establish morphological correspondence between the template and participant brains using a high‐dimensional nonlinear warping algorithm. The estimated nonlinear warp was applied inversely to an atlas defined in template space to facilitate anatomical parcellation. An adapted neuromorphometrics atlas within SPM12 was used for parcellation. Regional volumes were estimated by summing tissue densities within anatomical regions, then normalized by intracranial volume (ICV) to obtain proportional regional volumes for each participant. After excluding nongray‐matter structures (white matter, ventricles, cerebrospinal fluid, vessels, brain stem) and two atlas labels lacking anatomical designation (labels 244 and 245), 119 gray matter regions remained for analysis.

2.4. Data Preparation

2.4.1. Bilateral Averaging

To reduce multicollinearity and the number of statistical tests, left and right homologous regions were averaged, yielding 58 bilateral and three midline regions (61 total). This approach is consistent with practices employed by the ENIGMA consortium (Hoogman et al. 2017); the high shared variance among regional volumes that motivates this reduction is illustrated in Figure S2.

2.5. Statistical Analysis

2.5.1. Primary Analysis

Partial Pearson correlations were computed between each bilateral‐averaged region and each ADHD subscale, controlling for age and sex (61 × 3 = 183 tests). Multiple comparisons were controlled using the Benjamini–Hochberg false discovery rate (FDR) at α = 0.05. 95% confidence intervals were obtained via Fisher's z‐transformation.

2.5.2. Sensitivity Analyses

To assess robustness, FDR‐significant findings were re‐tested using: (1) Spearman rank correlations controlling for age and sex; (2) partial correlations additionally controlling for total brain volume; and (3) partial correlations additionally controlling for the Zung Self‐Rating Anxiety Scale (SAS).

2.5.3. Robustness Checks

Permutation testing (5000 iterations) assessed the significance of each FDR‐surviving association against a null distribution generated by shuffling ADHD scores. An internal split‐sample stability analysis was conducted using a stratified 60/40 random split (discovery n = 1528; replication n = 1020), with support in the held‐out split defined as a same‐direction association at one‐tailed p < 0.05.

2.5.4. Cross‐Validated Prediction

Statistically significant correlations do not necessarily translate to individual‐level prediction, particularly at small effect sizes (Marek et al. 2022). To evaluate the practical predictive utility of brain volumes for ADHD traits, we performed Ridge regression with 10‐fold cross‐validation (repeated 10 times) for each subscale. A baseline model (age and sex) was compared with a full model (age, sex, and 61 brain regions); statistical significance was assessed using 1000‐iteration permutation tests. The Ridge regularization parameter was determined via internal leave‐one‐out cross‐validation.

2.5.5. Exploratory Cross‐Sectional Mediation

Because the OFC is known to undergo age‐related volumetric decline, we conducted an exploratory cross‐sectional mediation analysis (Preacher and Hayes 2008) to characterize shared age‐related variance between regional brain volume and ADHD traits. For each FDR‐significant region–subscale pair, the pathway Age → brain volume → ADHD trait was modeled, controlling for sex. Bootstrap confidence intervals (5000 iterations, percentile method) were computed for the indirect effect (ab), with FDR correction applied across five tests.

2.5.6. Bayesian and Equivalence Testing

To quantify evidence for the null hypothesis, Bayesian information criterion (BIC)‐approximated Bayes factors (Wagenmakers 2007) were computed for each brain–ADHD association. Additionally, Two One‐Sided Tests (TOST) equivalence testing (Gerchen et al. 2021; Lakens 2017) was applied to inattention–brain associations, with the smallest effect size of interest set to |r| = 0.05, corresponding to the ENIGMA ADHD effect size range.

2.5.7. Lateralized and Asymmetry Analyses

Because many cortical functions, including several linked to the frontal lobe, are lateralized, bilateral averaging could in principle mask hemisphere‐specific associations. Two secondary analyses addressed this possibility. First, partial correlations controlling for age and sex were recomputed for all 119 lateralized regions and each ADHD subscale (119 × 3 = 357 tests, FDR‐corrected), in the pooled sample and within each sex. Second, a hemispheric asymmetry index, $\mathrm{AI}=\left(L-R\right)/\left(\left(L+R\right)/2\right)$, was computed for each of the 58 bilateral pairs following the ENIGMA consortium standard (Kong et al. 2018) and tested against each subscale (58 × 3 = 174 tests, FDR‐corrected).

2.5.8. Supplementary Analyses

Additional analyses reported in the Supporting Information include: a literature‐defined a priori analysis of 10 regions from prior meta‐analyses (Nakao et al. 2011; Yu et al. 2022) (Table S1); sex‐stratified partial correlations and sex × brain interaction models (Tables S2 and S3); a priori precuneus analysis based on default mode network–ADHD literature (Castellanos et al. 2008; Cavanna and Trimble 2006; Putra et al. 2023); attenuation correction for measurement unreliability (Spearman 1904) (Table S4); and multiple regression with all significant regions entered simultaneously.

All statistical analyses were performed in Python 3.13 (RRID:SCR_008394) using SciPy (RRID:SCR_008058), statsmodels (RRID:SCR_016074), and scikit‐learn (RRID:SCR_002577). Analysis scripts and derived summary outputs used to generate the reported tables and figures are available to editors and reviewers from the corresponding author on request during review. Broader sharing will be considered upon publication, subject to institutional and data‐governance constraints.

2.5.9. Use of AI‐Assisted Tools

During the preparation of this manuscript, the authors used Claude (Anthropic) to assist with statistical analysis code development and LaTeX formatting, ChatGPT (OpenAI) for English language editing, and Perplexity for reference verification. All AI‐generated outputs were critically reviewed, verified, and edited by the authors, who take full responsibility for the content of this work. These tools were not used for study design, data collection, data interpretation, or the formulation of scientific conclusions.

3. Results

3.1. Participant Characteristics

Table 1 summarizes the demographic and ADHD trait characteristics. Males scored higher on hyperactive motor traits (d = 0.12, p = 0.002), while females scored higher on Inattention (d = −0.21, p = 0.001). Hyperactive verbal traits did not differ by sex (p = 0.533). Subscale score distributions by sex are shown in Figure S3.

TABLE 1.

Participant demographics and ADHD trait descriptive statistics.

	All	Male	Female
Variable	(N = 2548)	(n = 1488)	(n = 1060)	t	p	d
Age (years)	52.8 (8.6)	52.3 (8.8)	53.4 (8.4)	−3.27	0.001	−0.13
ADHD subscales (ASRS Part B)
Inattention	11.60 (2.49)	11.38 (2.48)	11.91 (2.47)	−5.33	< 0.001	−0.21
Hyp. motor	4.86 (1.53)	4.94 (1.53)	4.75 (1.52)	3.10	0.002	0.12
Hyp. verbal	7.46 (2.00)	7.44 (2.03)	7.49 (1.95)	−0.62	0.533	−0.03
Internal consistency (Cronbach‘s α)
Inattention (five items)	0.696
Hyp. motor (three items)	0.608
Hyp. verbal (four items)	0.632

Note: Values are mean (SD). t = independent t‐test; d = Cohen‘s d (male − female). ADHD traits from ASRS Part B; three‐subscale structure per Stanton et al. (2018) and Panagiotidi et al. (2024).

3.2. Primary Analysis

Figure 1 shows the anatomical localization of the significant regions. Of 183 tests, five associations survived FDR correction (Table 2 and Figures 1 and 2). All five involved hyperactivity traits and orbital or medial frontal regions:

1. Bilateral medial frontal cortex (MFC) was negatively associated with hyperactive verbal ($r=-0.078$, 95% CI [−0.117, −0.039], $p_{\mathrm{FDR}}=0.007$) and hyperactive motor ($r=-0.076$, $p_{\mathrm{FDR}}=0.007$) traits.

2. Bilateral anterior orbital gyrus ($r=-0.077$, $p_{\mathrm{FDR}}=0.007$), lateral orbital gyrus ($r=-0.075$, $p_{\mathrm{FDR}}=0.007$), and orbital inferior frontal gyrus ($r=-0.070$, $p_{\mathrm{FDR}}=0.015$) were negatively associated with hyperactive verbal traits.

FIGURE 1.

Atlas‐volume localization of the FDR‐significant neuromorphometrics parcels overlaid on the SPM12 avg152T1 template. The left panel shows a mid‐sagittal slice highlighting medial frontal cortex, which was associated with both hyperactive motor and hyperactive verbal traits. The center and right panels show right sagittal and coronal orbitofrontal slices highlighting anterior orbital gyrus, lateral orbital gyrus, and the orbital part of the inferior frontal gyrus, which were associated with hyperactive verbal traits. Amber contour = Hyperactive motor; green fill = Hyperactive verbal.

TABLE 2.

Brain regions with FDR‐significant associations with ADHD trait subscales.

Brain region	ADHD subscale	r	95% CI	p	$p_{\mathrm{FDR}}$
Medial frontal cortex	Hyp. verbal	$-0.078$	[$-0.117$, $-0.039$]	$<0.001$	$0.007$
Anterior orbital gyrus	Hyp. verbal	$-0.077$	[$-0.116$, $-0.039$]	$<0.001$	$0.007$
Medial frontal cortex	Hyp. motor	$-0.076$	[$-0.115$, $-0.038$]	$<0.001$	$0.007$
Lateral orbital gyrus	Hyp. verbal	$-0.075$	[$-0.113$, $-0.036$]	$<0.001$	$0.007$
Inf. frontal gyrus (orb.)	Hyp. verbal	$-0.070$	[$-0.108$, $-0.031$]	$<0.001$	$0.015$

Note: N = 2548. Partial Pearson r controlling for age and sex. ICV‐normalized bilateral‐averaged volumes. FDR = Benjamini–Hochberg across 183 tests.

FIGURE 2.

Forest plot of FDR‐significant brain–ADHD associations. Bars show partial Pearson r (controlling for age and sex) with 95% confidence intervals. Amber = Hyperactive motor; green = Hyperactive verbal.

These associations correspond to R ² = 0.005–0.006, or less than 1% of trait variance. No region was significantly associated with inattention after FDR correction. The full pattern of associations is shown in Figure 3, and participant‐level scatter plots for the five FDR‐significant associations are shown in Figure S4. A targeted, literature‐defined analysis restricted to 10 orbitofrontal, anterior cingulate, and striatal regions identified by prior meta‐analyses (Nakao et al. 2011; Yu et al. 2022) recovered the same five associations and additionally identified anterior cingulate gyrus ($r=-0.055$) and gyrus rectus ($r=-0.054$) associations with hyperactive verbal traits (Table S1).

FIGURE 3.

Heatmap of partial correlations (r) between 61 brain regions and three attention‐deficit/hyperactivity disorder (ADHD) trait subscales. Rows are brain regions and columns are the three subscales. The regions are divided between two panels for legibility only: Panel A shows the first 31 regions in atlas label order and Panel B the remaining 30. Cell colour encodes the partial correlation (blue = negative, red = positive) on the shared scale below the panels. Asterisks denote false discovery rate (FDR) < 0.05. Partial correlations control for age and sex.

3.3. Sensitivity Analyses

All five associations were robust across methods (Table 3). Five of five survived Spearman nonparametric analysis; all five survived when additionally controlling for total brain volume (changes in $\mid r\mid \le 0.006$) and when controlling for anxiety (SAS; changes in $\mid r\mid <0.005$). A supplementary sensitivity model controlling simultaneously for all available health and lifestyle covariates, namely anxiety (SAS), smoking, drinking, hypertension, hyperlipidemia, diabetes, hyperuricemia, heart disease, and arrhythmia ($N=2545$), also retained all five associations (Table S5). A consolidated robustness dashboard summarizing effect sizes across all analytical approaches, permutation results, and Bayes factors is provided in Table S6.

TABLE 3.

Sensitivity analysis: FDR‐significant associations across analytical approaches.

Brain region	ADHD subscale	Partial correlation (r)
		Pearson (Age + Sex)	Spearman (Age + Sex)	+TBV (Age + Sex + TBV)	+SAS (Age + Sex + SAS)
Medial frontal cortex	Hyp. verbal	−0.078**	−0.071**	−0.076**	−0.077**
Anterior orbital gyrus	Hyp. verbal	−0.077**	−0.080**	−0.075**	−0.079**
Medial frontal cortex	Hyp. motor	−0.076**	−0.064*	−0.082**	−0.075**
Lateral orbital gyrus	Hyp. verbal	−0.075**	−0.075**	−0.073**	−0.078**
Inf. frontal gyrus (orb.)	Hyp. verbal	−0.070*	−0.070*	−0.067*	−0.074**

Note: Raw bilateral‐averaged DICV, controlling age and sex. +TBV = total brain volume added. +SAS = anxiety scale added. DICV = regional volume divided by intracranial volume.

^* $p_{\mathrm{FDR}}<0.05$.

^** $p_{\mathrm{FDR}}<0.01$.

Formal sex × region interaction tests for the five headline associations did not survive FDR correction, and within‐sex analyses yielded no FDR‐significant findings (Tables S2 and S3; sex‐stratified scatter plots are shown in Figure S5).

3.4. Lateralized and Asymmetry Analyses

Hemisphere‐specific analyses did not identify associations outside the regions represented among the primary bilateral findings (Tables S7 and S8). In the lateralized analysis (357 tests), five individual hemisphere–trait associations survived FDR correction; the corresponding contralateral estimates were all in the same direction with similar magnitudes, although none of them individually survived correction. Within‐sex lateralized analyses yielded no FDR‐significant associations. None of the 174 asymmetry‐index tests survived FDR correction (smallest $p_{\mathrm{FDR}}=0.438$). These results provide no evidence that bilateral averaging concealed an additional hemisphere‐specific association, but they do not establish exact equivalence between hemispheres.

3.5. Permutation Testing and Split‐Sample Stability

Permutation tests (5000 iterations) confirmed all five associations ($p_{\text{perm}}<0.001$; the null distribution for the strongest association is shown in Figure S6). In the split‐sample stability analysis, four of five associations (80%) showed same‐direction support in the held‐out sample (Table S9).

3.6. Bayesian and Equivalence Testing

Bayesian analysis using BIC‐approximated Bayes Factors provided strong evidence for the null hypothesis (${\mathrm{BF}}_{10}<1/10$) for inattention in 57 of 61 brain regions (Table S10). TOST equivalence testing further demonstrated that 40 of 61 inattention–brain correlations were statistically equivalent to zero within the ENIGMA benchmark ($\mid r\mid <0.05$; Table S11). Together with the absence of FDR‐significant associations, these results support near‐null inattention effects for most regions at the present measurement precision and effect‐size benchmark; they do not establish equivalence for every region. For the five significant associations, BF₁₀ ranged from 10.2 to 47.7, indicating strong evidence for the alternative hypothesis.

3.7. Cross‐Validated Prediction

Permutation tests of the 61‐region models yielded $p=0.004$ for inattention, $p=0.006$ for hyperactive motor, and $p=0.154$ for hyperactive verbal. However, adding the 61 brain regions to a baseline model containing age and sex did not improve out‐of‐sample prediction for any subscale ($\mathrm{\Delta cv}‐R^2<0$). Thus, the brain–volume features provided no incremental predictive value at the individual level.

3.8. Exploratory Cross‐Sectional Mediation

Exploratory bootstrap mediation analysis identified a statistical pattern consistent with age‐related OFC/medial frontal volume decline accounting for a portion of age‐related hyperactivity variance. All five indirect effects were significant after FDR correction ($p<0.001$; Table 4). Indirect effects (ab = 0.005–0.007) were positive, and total effect decomposition confirmed $c\approx c^{\prime }+\mathit{ab}$ for all five tests.

TABLE 4.

Exploratory cross‐sectional mediation: Age → OFC/MFC volume → ADHD hyperactivity.

Region	ADHD subscale	Indirect (ab)	95% CI	$p_{\mathrm{FDR}}$
Medial frontal cortex	Hyp. verbal	0.0067	[0.003, 0.010]	< 0.001
Anterior orbital gyrus	Hyp. verbal	0.0055	[0.003, 0.008]	< 0.001
Medial frontal cortex	Hyp. motor	0.0050	[0.002, 0.008]	< 0.001
Lateral orbital gyrus	Hyp. verbal	0.0062	[0.003, 0.009]	< 0.001
Orbital IFG	Hyp. verbal	0.0057	[0.003, 0.009]	< 0.001

Note: Five thousand bootstrap iterations, percentile CI. All CIs exclude zero. FDR applied across five indirect effects. Cross‐sectional path a: Age → brain; path b: brain → ADHD.

4. Discussion

In a large sample of nonclinical adults, we found that smaller orbitofrontal and medial frontal gray matter volumes were associated with higher hyperactivity traits, particularly verbal hyperactivity/impulsivity, after controlling for age and sex. Five associations survived whole‐brain FDR correction across 183 tests, all involving orbital or medial frontal regions and hyperactivity subscales. Inattention showed no whole‐brain FDR‐significant associations, with Bayesian and equivalence testing suggesting that any inattention–volume associations in this dataset were small. Cross‐validated prediction indicated limited individual‐level utility, and exploratory cross‐sectional mediation suggested shared age‐related variance involving frontal volume and hyperactivity traits.

4.1. Orbitofrontal Gray Matter and Hyperactivity

The involvement of the OFC in our findings is consistent with converging evidence from clinical ADHD research. Meta‐analyses of VBM studies have repeatedly identified reduced gray matter in the OFC of individuals with ADHD (Chen et al. 2025, Nakao et al. 2011; Yu et al. 2022), and the OFC is a core component of frontostriatal circuits that mediate inhibitory control and reward‐based decision‐making (Hesslinger et al. 2002; Rubia 2018). The present findings extend this literature by showing that similarly directed, small structure–trait associations can also be detected along the adult trait continuum in a nonclinical sample. This supports dimensional models of ADHD (Franke et al. 2018) and aligns with the Research Domain Criteria (RDoC) framework, which emphasizes dimensions that cut across diagnostic boundaries.

That the FDR‐significant associations were confined to hyperactivity, and particularly to verbal hyperactivity (excessive talking, interrupting, difficulty waiting), is notable. The three‐factor decomposition of ADHD symptoms adopted here (Panagiotidi et al. 2024; Stanton et al. 2018) allows for finer‐grained mapping of brain–behavior associations than the conventional two‐factor (inattention/hyperactivity‐impulsivity) model, although the present study was not designed to adjudicate between competing factor structures. The OFC is implicated in regulating social behavior and suppressing inappropriate verbal responses (Aron et al. 2004; Hesslinger et al. 2002), functions that align closely with the content of the verbal hyperactivity subscale. By contrast, inattention may depend more on distributed attentional networks and frontoparietal systems that are less well captured by volumetric measures at this spatial resolution.

The hemisphere‐specific estimates were concordant in direction, and the five FDR‐significant lateralized tests involved the same orbital and medial frontal regions represented among the primary bilateral findings. Both hemispheres were represented among the significant tests, although each association reached significance in only one hemisphere. Moreover, no asymmetry‐index association survived FDR correction. Several frontal functions are strongly lateralized; response inhibition is preferentially associated with the right inferior frontal cortex (Aron et al. 2004), and language production with the left hemisphere. The present analyses nevertheless provide no evidence that bilateral averaging concealed an additional hemisphere‐specific correlate of hyperactivity. This pattern is consistent with meta‐analytic evidence that orbitofrontal structural alterations in ADHD are often reported bilaterally (Nakao et al. 2011, Yu et al. 2022), while stopping short of establishing exact equivalence between hemispheres.

Although sex differences in ADHD presentation motivated covariate adjustment and supplementary stratification, formal sex × region interaction tests did not survive FDR correction, and within‐sex analyses yielded no FDR‐significant findings. Any sex moderation of these small structure–trait effects was therefore below detection in the present sample.

4.2. Effect Sizes in Context

The effect sizes we observed (|r| = 0.070–0.078) are small by conventional standards. However, it is important to place them in the context of the broader ADHD neuroimaging literature. The ENIGMA consortium mega‐analysis (Hoogman et al. 2017), the largest and most well‐powered study of brain structure in ADHD, reported effect sizes of d = 0.10–0.19 for subcortical volumes (equivalent to r = 0.050–0.095). Our observed effects fall within this range. After correcting for the attenuating effect of measurement unreliability in the ADHD subscales (Cronbach's α = 0.608–0.696), the estimated true effects are r = 0.088–0.098 (Table S4). Importantly, these effects remained directionally consistent across sensitivity analyses, survived 5000‐iteration permutation tests, and showed split‐sample support for four of five associations in a held‐out sample. At the same time, effects of this magnitude should be interpreted as population‐level signals rather than clinically actionable individual markers. Similar small but robust rGMV–trait associations have been observed in our previous large‐scale work on fatigue perception in healthy adults using the same imaging protocol (Putra, Park, and Yamashita 2022; Putra, Park, Yamashita, et al. 2022).

4.3. Interpreting the Inattention Null

A distinctive feature of this study is that inattention showed no whole‐brain FDR‐significant associations and substantial Bayesian/TOST support for effects close to zero at the ENIGMA‐informed benchmark. Rather than proving an absolute absence of structural correlates, these analyses suggest that any inattention–volume associations in this dataset are smaller than the benchmarked ENIGMA‐sized effects or are obscured by the measurement precision of the scale. This finding is still informative: it suggests that the neural basis of inattention in the general population may lie in functional or microstructural properties not captured by macroscopic volumetric measures. This is consistent with proposals that attentional deficits in ADHD are more closely linked to functional connectivity and network dynamics than to gray matter volume per se (Castellanos and Proal 2012). Importantly, the null‐leaning pattern for inattention is unlikely to be driven by measurement imprecision alone: inattention had the highest reliability of the three subscales ($\alpha =0.696$) yet showed the weakest brain associations, whereas hyperactive motor, the subscale with the lowest reliability ($\alpha =0.608$), showed FDR‐significant effects. This pattern argues against a purely measurement‐driven explanation, although it does not eliminate that possibility entirely.

4.4. Cross‐Validated Prediction and Practical Significance

The regional brain–ADHD associations did not translate to individual‐level predictive utility. Cross‐validated Ridge regression showed that adding 61 brain regions to a model containing age and sex did not improve out‐of‐sample prediction of any ADHD trait subscale ($\mathrm{\Delta cv}‐R^2<0$). This finding is consistent with Marek et al. (2022), who showed in over 50,000 participants that brain–behavior associations at typical effect sizes ($r\sim 0.05-0.10$) do not support individual‐level prediction. The partial correlations observed here should therefore be interpreted as population‐level associations rather than clinically useful individual markers.

4.5. Exploratory Age‐Related Shared Variance

Exploratory cross‐sectional mediation analysis identified a statistical pattern consistent with age‐related OFC and medial frontal volume decline accounting for a portion of age‐related hyperactivity variance. All five indirect effects were significant ($p_{\mathrm{FDR}}<0.001$), with bootstrap confidence intervals excluding zero. This pattern (aging, smaller frontal volume, and higher hyperactivity traits) is consistent with the established role of the OFC in impulse control (Aron et al. 2004; Rubia 2018) and with recent reports of brain structure mediating environmental or demographic influences on ADHD‐related outcomes (Machlin et al. 2020; Shou et al. 2025). However, cross‐sectional mediation estimates may be biased relative to longitudinal estimates (Maxwell and Cole 2007), and the present results should be interpreted as hypothesis‐generating shared‐variance patterns rather than proof of temporal mediation. Longitudinal data will be needed to test whether this age‐related pattern reflects a true mediating process.

4.6. Limitations

Several limitations should be acknowledged. First, the cross‐sectional design precludes causal inference; while our exploratory mediation analysis models the pathway age → brain → ADHD, temporal ordering cannot be established without longitudinal data, and cross‐sectional mediation estimates may diverge from longitudinal effects (Maxwell and Cole 2007). Second, the ADHD subscale reliabilities, while acceptable for brief community measures, are modest (Cronbach's α = 0.608–0.696), which attenuates observed correlations and limits how strongly null‐leaning results can be interpreted. Relatedly, only the 12 Part B items of the ASRS were administered, by protocol design; because the 6‐item Part A screener was not included, no screening score is available for estimating probable ADHD caseness, and scores are not directly comparable with studies administering the full 18‐item scale. Third, the sample was recruited through brain health check‐ups (“Brain Dock”) in Kochi, Japan, and likely reflects relatively health‐conscious middle‐aged and older adults, limiting generalizability to younger, more diverse, or clinical populations. Some participants may also meet diagnostic criteria for ADHD despite no clinical diagnosis; this is inherent to the dimensional approach, and we cannot rule out the possibility that a subset of participants with undiagnosed clinical‐level ADHD contributed disproportionately to the observed dimensional associations. Fourth, MRI was acquired at 1.5‐T and processed with the VBM8 toolbox. This historically consistent pipeline enabled uniform processing across the full acquisition period, but 1.5‐T imaging provides lower signal‐to‐noise than contemporary 3‐T studies, and VBM8 lacks some of the segmentation and quality‐control features available in newer toolboxes such as CAT12. Atlas‐based parcellation (neuromorphometrics), rather than surface‐based methods, may also limit sensitivity to focal effects. Fifth, education and socioeconomic data were not available for this sample, and additional potentially relevant lifestyle or treatment‐related covariates were only available in limited harmonized form. A supplementary model including available anxiety, smoking, drinking, and vascular/medical covariates did not materially alter the five headline associations, but residual confounding remains possible. Sixth, we did not examine other morphometric features (cortical thickness, surface area, gyrification) that may show complementary associations. Seventh, psychiatric comorbidity was not formally assessed. Psychotic disorders were exclusionary at recruitment and anxiety symptoms were measured by self‐report (SAS), but no depression or mood‐disorder measure was collected, and MMSE scores, which were used for eligibility screening at a threshold of 28 or above, were not retained in the analysis dataset, precluding covariate adjustment for residual variation in global cognition. Mood disorders are among the most common psychiatric comorbidities of adult ADHD, with prevalence in clinical samples comparable to or exceeding that of anxiety disorders (Kessler et al. 2006; Sobanski 2006), and depressive symptomatology has itself been associated with prefrontal, including orbitofrontal, structural differences (Schmaal et al. 2017). Unmeasured mood‐related variance could therefore contribute to the frontal associations with hyperactivity observed here. The present study is therefore best regarded as foundational: it characterizes trait–structure covariation in a nonclinical adult sample and provides a baseline for studies of clinical ADHD and of populations with formally assessed psychiatric comorbidity.

4.7. Future Directions

The present results provide a population‐level reference for frontal structure–trait covariation in nonclinical adults. Several next steps follow. First, longitudinal designs are needed to test whether the age‐related shared‐variance pattern reflects a true mediating process in which frontal volume decline contributes to hyperactivity change (Maxwell and Cole 2007). Second, the nonclinical baseline established here can serve as a reference for studies of clinically diagnosed ADHD and of populations with elevated ADHD traits, including samples with formally assessed mood and anxiety comorbidity, to test whether clinical status shifts or amplifies the trait–structure associations observed along the continuum. Third, studies that measure ADHD and autism‐spectrum traits together could test whether the present associations are specific to hyperactivity or extend across overlapping neurodevelopmental dimensions. This question is relevant because ADHD and autism spectrum disorder co‐aggregate within families and share etiological influences (Ghirardi et al. 2018), while orbitofrontal and related frontostriatal circuitry has also been implicated in repetitive behavior in autism (Ecker et al. 2015). Finally, surface‐based morphometry, microstructural, and functional connectivity measures may capture inattention‐related variance that volumetric measures do not.

4.8. Conclusions

In 2548 nonclinical adults, smaller bilateral orbitofrontal and medial frontal gray matter volumes were associated with higher hyperactivity traits, particularly verbal hyperactivity, with small effect sizes consistent with the ENIGMA range for clinical ADHD and no evidence that bilateral averaging concealed an additional hemisphere‐specific association. Inattention showed no whole‐brain FDR‐significant structural associations, with Bayesian and equivalence analyses supporting near‐null effects for most regions. Adding brain volumes did not improve individual‐level prediction, while exploratory cross‐sectional mediation suggested an age‐related pattern that warrants longitudinal confirmation. These findings support dimensional models of ADHD by identifying small, population‐level frontal gray matter associations with hyperactivity traits in this adult Japanese sample, and they provide a foundational nonclinical reference point for future clinical and transdiagnostic studies of ADHD traits.

Author Contributions

Handityo Aulia Putra: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, writing – original draft. Kaechang Park: conceptualization, data curation, funding acquisition, investigation, project administration, resources, supervision, writing – review and editing.

Funding

This work was supported by Japan Society for the Promotion of Science KAKENHI (25H00762, 23K17330, and 25K00301).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Analytical pipeline for the main manuscript. Participants from the Brain Dock cohort underwent 1.5‐T structural MRI and completed the 12 Part B items of the Adult ADHD Self‐Report Scale (see Section 2). Gray‐matter volumes were extracted from the SPM12 neuromorphometrics atlas using VBM8, normalized by intracranial volume, bilateral homologous regions were averaged, and regional associations with three ADHD subscales were tested using age‐ and sex‐adjusted partial correlations with false discovery rate correction. Downstream validation included sensitivity analyses, permutation testing, split‐sample stability, Bayesian and equivalence testing, cross‐validated prediction, and exploratory cross‐sectional mediation.

Figure S2: PCA scree plots for age‐residualized bilateral brain regions, separately for males (blue, n = 1488) and females (vermillion, n = 1060). Bars show variance explained by each component.

Figure S3: Distributions of attention‐deficit/hyperactivity disorder (ADHD) subscale scores for males (blue) and females (vermillion). Density histograms with independent t‐test statistics and Cohen's d effect sizes.

Figure S4: Scatter plots for the five FDR‐significant brain region–ADHD associations. Blue dots = males (n = 1488), vermillion dots = females (n = 1060). Black line = OLS regression.

Figure S5: Sex‐stratified scatter plots for the top brain region–ADHD associations in males (blue, n = 1488) and females (vermillion, n = 1060), controlling for age. None survived false discovery rate correction within sex.

Figure S6: Permutation null distribution (5000 iterations) for the top FDR‐significant association (bilateral medial frontal cortex × hyperactive verbal). Vermillion line shows the observed partial r. Permutation p < 0.001.

Table S1: A priori analysis: orbitofrontal, medial frontal, anterior cingulate, and striatal regions.

Table S2: Top 10 brain region–ADHD associations for males and females separately.

Table S3: Formal sex‐by‐region interaction tests for the five headline associations.

Table S4: Attenuation‐corrected effect sizes for FDR‐significant associations.

Table S5: Health/lifestyle covariate sensitivity for the five headline brain–ADHD associations.

Table S6: Robustness dashboard: convergent validation of the five FDR‐significant associations.

Table S7: ENIGMA asymmetry index analysis: top 10 associations with ADHD traits.

Table S8: Bilateral and hemisphere‐specific estimates for the five primary associations.

Table S9: Internal split‐sample stability analysis for FDR‐significant associations.

Table S10: Bayesian evidence for brain region–ADHD trait associations.

Table S11: TOST equivalence testing for inattention–brain associations.

Data Availability Statement

Participant‐level data supporting the findings of this study were collected through the Brain Dock program at Kochi University of Technology and are not publicly available because they contain sensitive clinical information. De‐identified data may be available from the corresponding author upon reasonable request and subject to institutional approval and data‐use agreements. Analysis scripts and derived summary outputs used to generate the reported tables and figures are available to editors and reviewers from the corresponding author on request during review. Broader sharing will be considered upon publication, subject to institutional and data‐governance constraints.