henhao
Sign in
Developmental BiologyNature Mental Health
Peer-reviewed journal article

Brain morphological pattern is associated with the presence, severity, and transition of transdiagnostic psychiatric disorders in preadolescents

Nanyu KuangiD, Christopher Joseph HammondiD, Betty Jo SalmeroniD, Xiang XiaoiD, Danni WangiD, Laura MurrayiD, Hong GuiD, Tianye ZhaiiD, Hui ZhengiD, Justine HilliD, Maria K. ScavnickyiD, Hanbing LuiD, Amy C. JanesiD, Thomas J. RossiD, Yihong YangiD

Abstract

Cognitive function, psychological processes, mental states, and behaviors are key dimensions of human subjective experience that separately relate to mental disorders across diagnostic categories. However, whether these dimensions are linked to common or distinct brain morphological patterns that convey risk or resilience for psychiatric disorders remains unclear. The current study is a longitudinal investigation on 11,875 youths from the Adolescent Brain Cognitive Development (ABCD) Study aged 9-10 years at baseline. A machine learning approach based on canonical correlation analysis was used to identify latent dimensional associations of cortical morphology (4 metrics: surface area, cortical and subcortical volume, cortical thickness, and sulcal/gyral depth) with multidomain behavioral assessments including cognitive scores and psychological measures indexing motivation, impulse control, mental states, and behaviors across a normative continuum from healthy to pathological. Across morphological measures, we identified a robust latent brain structural variate that correlated positively with cognitive performance and negatively with psychological measures indexing greater psychology. Notably, higher scores on this brain variate reflected larger cortical surface area and cortical volume--especially in the temporal gyri--together with a posterior-anterior gradient in cortical thickness, showing relatively greater thickness in occipital, parietal, and temporal cortices and lower thickness in cingulate and frontal regions. This brain variate and the related cognitive-psychological-behavioral variate remained stable at the 2-year follow-up, demonstrating temporal consistency. Importantly, the brain variate showed a dose-dependent relationship with the cumulative number of psychiatric diagnoses assessed concurrently and at 2-year follow-up, with lower brain variate scores being associated with higher numbers of comorbid diagnoses. In addition, the brain scores were associated with longitudinal transitions between healthy and diagnosed states over the 2-year study period, in which lower scores at baseline were associated with persistent psychiatric diagnoses whereas higher scores at baseline were associated with persistent healthy states, suggesting that the brain scores capture a vulnerability- resilience continuum for psychopathology. By revealing shared brain structural substrates across conventional diagnostic boundaries, these findings advance the neurodevelopmental understanding of psychiatric disorders and highlight the potential utility of morphology-informed approaches for early screening and intervention in youth.

PDFDOIData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
2
FWCI
16.32
Field percentile
top 2%

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (18 Feb 2026) · CC BY 4.0

In 8,672 ABCD preadolescents, one canonical-correlation-derived brain morphology variate links higher cognition with lower psychopathology and tracks cumulative psychiatric diagnoses and 2-year diagnostic transitions.

Diagnosis-specific neuroimaging biomarkers have proven elusive amid symptom overlap and comorbidity, and whether cognition and psychopathology share brain structural substrates in youth was unclear. Extending earlier ABCD CCA studies of cortical morphology and resting-state connectivity, this work integrates four morphology metrics and tests the variate's 2-year stability and its links to diagnostic burden and course.

Key results

  1. ShownA single cross-metric brain morphology mode generalized to held-out data, coupling better cognition with fewer psychological problems; externalizing measures loaded much more strongly than internalizing ones.

    Canonical correlation ρ = 0.32 (training) and ρ = 0.30 (holdout test), both permutation pFWER < 10^−5; Destrieux atlas reproduced the loadings. Crystallized Composite loading r = 0.82; CBCL Rule Breaking r = −0.37 and Conduct Problem r = −0.39 vs CBCL Anxious-Depression r = −0.0035 (bootstrap 5000×, FDR p < .05). · Latent Dimensions Linking the Brain Structure…

  2. ShownThe variate is not a uniform size factor: surface area and volume dominate with strongest positive temporal loadings, thickness follows a posterior-positive, anterior-negative gradient, and sulcal/gyral depth contributes weakly.

    Two-way ANOVA on loadings: Metric F = 54.36, Region F = 16.09, Metric × Region F = 5.01 (all p < 10^−5); Tukey tests showed larger surface area and volume loadings than thickness and depth (FDR p < 10^−3). Strongest subcortical loadings: left hippocampus and left ventral diencephalon. · Latent Dimensions Linking the Brain Structure…

  3. ShownBrain variate scores were highly rank-stable over 2 years, rose slightly with age in parallel across diagnostic groups, and were largely recoverable from either cognitive or psychological measures alone.

    Baseline vs projected 2-year scores r = 0.91. Mixed model: age effect F = 25.88, p = 3.7 × 10^−7; age-by-diagnosis F = 1.59, p = 0.10. Variate agreement: combined vs cognition-only r = 0.95; combined vs psychological-only r = 0.81. · Rank-order stability of the Brain Structural…; Consistency of the Cross-metric Mode Across…

  4. ShownLower variate scores accompanied more concurrent psychiatric diagnoses in a dose-dependent pattern and greater transdiagnostic burden, with comorbid externalizing plus internalizing cases lowest; several specific KSADS-5 diagnoses also showed reduced scores.

    Count groups at baseline (0/1/2/≥3 diagnoses, N = 6262/1211/414/252): F = 16.71, p < .00001; 2-year F = 4.34, p < .01. Spectrum groups: F = 10.80 baseline, F = 2.79 2-year; externalizing vs internalizing not different. Mania, psychosis, ADHD and others lower at baseline; conduct, eating disorders, specific phobia at 2-year (FDR). · Association between Brain Structural Variate…; Association between Brain Structural Variate…; Association between Brain Structural Variate…

  5. ShownBaseline variate scores and their 2-year change differed by diagnostic course: persistently healthy youth scored highest and rose most, whereas persistently diagnosed youth scored lowest and declined on average.

    Groups: Healthy Persistent N = 5082, Disorder Remitted N = 893, Disorder New Onset N = 702, Disorder Persistent N = 893. One-way ANOVA on baseline score F = 7.26, p < .0001; on score change F = 5.01, p < .01. Baseline ordering: Healthy Persistent > New Onset > Remitted > Persistent. · Association between Brain Structural Variate…

Methods, data and reproducibility

Four morphology metrics (surface area, cortical/subcortical volume, thickness, sulcal/gyral depth) were concatenated and related by canonical correlation analysis (CCA) to 13 cognitive and 31 psychological measures using permutation-tested train/test splits; baseline weights were projected onto 2-year data and scores compared across KSADS-5 diagnostic groups.

  • ABCD Study, NIMH Data Archive release 5.1: US youths aged 9-10 at baseline recruited at 21 sites, reassessed at 2-year follow-up · human · n = 11,875 enrolled; 8,672 baseline and 6,218 2-year in CCA analyses
  • Structural MRI morphology: area, volume, thickness and sulcal/gyral depth for 68 Desikan-Killiany cortical regions plus 16 Fischl subcortical volumes · human · n = 8,672 (baseline); 6,218 (2-year)
  • Destrieux atlas parcellation (148 cortical regions) used to replicate the CCA mode · human
  • Cognitive data: 13 scores from 10 neurocognitive tests (including NIH Toolbox) and 3 neuroimaging tasks · human
  • Psychological data: 31 youth- and parent-reported scores from BIS/BAS, UPPS-P, CBCL, PGBI-10 and PQ-B · human
  • KSADS-5 current diagnoses grouped by cumulative count (0, 1, 2, ≥3) at baseline and 2-year follow-up · human · n = baseline 6262/1211/414/252; 2-year 3254/655/186/104
  • Transdiagnostic spectrum groups (externalizing, internalizing, comorbid) compared with healthy controls · human · n = Ext 707/368; Int 505/271; comorbid 234/90 (baseline/2-year)
  • Diagnostic transition groups defined from baseline to 2-year follow-up status · human · n = healthy 5082; remitted 893; new onset 702; persistent 893
Methods & tools
Canonical correlation analysis (CCA) with permutation testing, Desikan-Killiany atlas (68 cortical regions), Destrieux atlas (148 cortical regions), Fischl subcortical atlas, NIH Toolbox Cognition Battery, KSADS-5 computerized semi-structured diagnostic interview, Child Behavioral Checklist (CBCL), BIS/BAS, UPPS-P, PGBI-10 and PQ-B questionnaires, Linear mixed-effects model, Two-way ANOVA with Tukey post hoc tests, One-way ANOVA with Games-Howell post hoc comparisons, Welch two-sample t-tests with FDR correction
Data & accessions
https://abcdstudy.org/https://data-archive.nimh.nih.gov/abcdhttps://wiki.abcdstudy.org/release-notes/start-page.html
Reproducibility gaps
No analysis code or repository is reported; ABCD data require NIMH Data Archive registration and access approval; CCA and quality-control details are deferred to Supplementary Information.
Limitations (5)
  • Being a secondary ABCD analysis, the work inherits design and measurement decisions made by the original study team. · stated by the authors
  • Psychopathology measures were largely parent-reported at this wave and need cross-validation against multi-informant assessments from later ABCD waves. · stated by the authors
  • Most participants scored in normal ranges and had no diagnosis at either time point; generalizability, pathophysiological specificity and translational value require testing in clinical cohorts with more severe, heterogeneous psychopathology. · stated by the authors
  • Follow-up analyses relate 2-year scores to 2-year diagnoses (concurrent), and no classification accuracy, AUC or out-of-sample diagnostic prediction is reported, so the suggested screening utility is not directly tested. · noted in this summary
  • Main text gives no effect sizes for group differences, does not describe adjustment for site, sex or intracranial volume, and reports 11,875 participants in some sections but 11,985 in another; the identical N = 893 for Remitted and Persistent groups may merit checking. · noted in this summary

Who should read:Read this if you study transdiagnostic neurodevelopmental markers in youth and want a multi-metric structural MRI CCA analysis of ABCD linking brain morphology to cognition, psychopathology, and diagnostic course.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISClauss '26Ramduny '26Xiao '21Deng '24Royer '24Li '26
78% semantic matchNeuroscience
Jacqueline Alexandra Clauss, Dongmei Zhi, Sofía Álvarez Perdomo+25 more·bioRxiv·Oct 11, 2026· 1 citations