Highlights
- Identification of three distinct neurobiological ADHD biotypes: Severe-combined (emotional dysregulation), Hyperactive/Impulsive, and Inattentive, based on brain morphometry.
- The study utilized Normative Modeling of Morphometric Similarity Networks (MSNs) to map individual deviations from a healthy developmental trajectory.
- Significant case-control differences were localized to atypical hub organization in the orbitofrontal cortex, a region critical for executive control and reward processing.
- Neural profiles of these biotypes correlate with specific neurochemical systems and longitudinal cognitive outcomes, demonstrating high clinical and biological validity.
Background
Attention-deficit/hyperactivity disorder (ADHD) is one of the most prevalent neurodevelopmental disorders, affecting approximately 5% to 7% of children worldwide. Despite its high prevalence, the clinical management of ADHD is challenged by its extreme heterogeneity. Currently, diagnosis relies heavily on behavioral observations codified in the DSM-5 or ICD-11, which categorize patients into three clinical presentations: predominantly inattentive, predominantly hyperactive-impulsive, and combined. However, these symptomatic categories often fail to reflect the underlying neurobiological diversity of the disorder. Many patients exhibit overlapping symptoms, and clinical presentations often shift over development, suggesting that current frameworks do not capture the biological ‘ground truth’ of ADHD.
To move toward precision psychiatry, there is an urgent need for biomarkers that can stratify patients into neurobiologically homogeneous subgroups (biotypes). Previous attempts using traditional structural or functional MRI have often produced inconsistent results due to small sample sizes and the ‘average’ effect, where subtle individual differences are washed out in group comparisons. The study by Pan et al. (2026) addresses these limitations by integrating normative modeling—an approach analogous to pediatric growth charts—with morphometric similarity networks (MSNs) to map the landscape of ADHD heterogeneity.
Key Content
Methodological Innovation: Morphometric Similarity Networks and Normative Modeling
The core of this research lies in the construction of Morphometric Similarity Networks (MSNs). Unlike traditional structural imaging that looks at volume or thickness in isolation, MSNs quantify the similarity between different brain regions across multiple morphometric features (e.g., cortical thickness, surface area, volume, Gaussian curvature). This provides a proxy for cortical organization and ‘wiring’ that is more sensitive to individual variation than single-feature metrics.
By applying Normative Modeling to these MSNs, researchers established a reference range for ‘typical’ brain development using a large control dataset. Each child with ADHD was then mapped against this norm, allowing for the calculation of individual ‘deviation scores’ (Z-scores) at each brain region. This shift from ‘group-average’ to ‘individual-deviation’ is crucial for capturing the true heterogeneity of the disorder.
Orbitofrontal Cortex: The Common Hub of Deviation
Before subtyping, the study investigated whether a common neural signature exists for ADHD. The results identified a central ‘hub’ of atypical organization in the orbitofrontal cortex (OFC). Across all three topological metrics—degree centrality, nodal efficiency, and participation coefficient—children with ADHD showed significant deviations in the OFC. This finding aligns with the executive dysfunction and reward-circuitry models of ADHD, as the OFC is vital for decision-making and impulse control. However, the researchers noted that while the OFC was a common site of deviation, the direction and extent of these deviations varied significantly across the cohort, necessitating further stratification.
Characterizing the Three Biotypes
Using semisupervised clustering of the topological deviation maps, the study delineated three distinct biotypes, each with unique neural, clinical, and longitudinal profiles:
- Biotype 1: Severe-Combined with Emotional Dysregulation (n=142). This group exhibited the most widespread deviations, particularly in the medial prefrontal cortex-pallidum circuit. Clinically, these children presented with the most severe symptoms across both inattentive and hyperactive domains, frequently accompanied by high levels of emotional dysregulation. Longitudinal data suggested a more persistent symptom trajectory for this group.
- Biotype 2: Predominantly Hyperactive/Impulsive (n=177). This biotype was characterized by localized alterations in the anterior cingulate cortex (ACC)-pallidum circuit. These regions are integral to motor control and conflict monitoring. Clinically, these children scored highest on hyperactivity-impulsivity scales but had relatively preserved attention compared to Biotype 1.
- Biotype 3: Predominantly Inattentive (n=127). The neural hallmark of this group was alterations in the superior frontal gyrus, a key node in the dorsal attention network. These children showed significant deficits in sustained attention and executive function, with lower scores on hyperactive-impulsive metrics.
Neurochemical and Functional Context
To provide a biological bridge between brain structure and clinical symptoms, the researchers contextualized these biotypes using large-scale neurochemical databases. Biotype 1 (Severe-combined) showed a strong spatial correlation with the distribution of dopamine and serotonin transporters, suggesting that their emotional and behavioral symptoms might be driven by dysregulation in these monoamine systems. In contrast, Biotype 3 (Inattentive) was more closely linked to norepinephrine system distributions, highlighting potential targets for pharmacological interventions (e.g., atomoxetine) specific to that subgroup.
Generalizability and Validation
A significant strength of this study is the use of an independent transdiagnostic validation cohort. The biotypes identified in the discovery group were successfully replicated in a separate dataset, demonstrating that these MSN-based markers are robust across different scanners, sites, and clinical populations. This generalizability is a prerequisite for any biomarker intended for clinical use.
Expert Commentary
The work by Pan et al. represents a significant leap forward in the application of computational psychiatry to neurodevelopmental disorders. By moving away from symptom-based classification and toward ‘biotyping,’ we are finally beginning to see the ‘how’ and ‘why’ behind the clinical diversity of ADHD.
One of the most compelling aspects of this study is the integration of normative modeling. In traditional medicine, we do not diagnose a child with ‘short stature’ by comparing them to an average of all children; we use growth charts that account for age and sex. This study brings that same rigor to neuroimaging. The identification of the orbitofrontal cortex as a common site of atypicality, while finding divergent pathways for biotypes, explains why ADHD can look so similar yet respond so differently to treatment.
However, limitations remain. While the biotypes are statistically robust, the ‘clinical utility’—that is, whether knowing a child’s biotype changes their treatment outcome—has yet to be tested in a prospective clinical trial. Furthermore, the reliance on high-quality structural MRI and complex topological modeling may currently limit the accessibility of this approach in primary care settings. Future research should focus on whether these biotypes can be mapped using more accessible markers, such as EEG or abbreviated MRI protocols.
Conclusion
This study provides evidence that ADHD heterogeneity is not merely ‘noise’ but is organized into distinct neurobiological biotypes. The use of morphometric similarity networks and normative modeling allowed for the identification of three clusters with unique neural architectures and clinical paths. These findings lay the groundwork for a more personalized approach to ADHD, where neuroimaging might eventually guide the selection of pharmacotherapy or behavioral interventions based on an individual’s specific biological profile. As we move closer to the era of precision medicine, these insights will be instrumental in improving the long-term outcomes for children with ADHD.
References
- Pan N, Long Y, Qin K, et al. Mapping ADHD Heterogeneity and Biotypes by Topological Deviations in Morphometric Similarity Networks. JAMA Psychiatry. 2026 Feb 25:e260001. doi: 10.1001/jamapsychiatry.2026.0001. PMID: 41739459.
- Marquand AF, Rezek I, Buitelaar J, Beckmann CF. Understanding Heterogeneity in Clinical Cohorts Using Normative Models: Beyond Case-Control Studies. Biol Psychiatry. 2016;80(7):552-561.
- Seidlitz J, Váša F, Shinn M, et al. Morphometric Similarity Networks Detect Microscale Cortical Organization and Socio-Cognitive Effects. Neuron. 2018;97(1):231-247.e7.
