Published on 18 September 2026
Does the same disorder shape the same cortex at fifteen and at forty?
The essentials
A large imaging consortium has published maps of cortical thickness differences for four psychiatric disorders, each in two age groups. This study set those maps against a connectivity atlas, neurotransmitter maps, gene expression data and an atlas of six mitochondrial features. In adults, the distribution of thickness differences follows the brain’s connectivity: the correspondence reaches 0.65 for attention deficit hyperactivity disorder, 0.63 for bipolar disorder, 0.49 for depression and 0.16 for obsessive-compulsive disorder. That last value is significant but very weak, which the authors point out themselves. In adolescents, none of the four correspondences reaches significance. The finding is interesting, and it is fragile. The reference atlases come from adult brains and are applied to adolescent imaging, the adolescent cohorts are smaller, and the whole analysis remains correlational. It has no clinical consequence, and no level of evidence was established for this work. The document we verified is the accepted, unedited manuscript, posted online before the journal’s final editing.
Context
We have long known that psychiatric disorders come with differences in cortical thickness at the group level, and we also know that these differences are of no use for diagnosing an individual. The question asked here is a different one: are these differences scattered at random across the cortex, or do they follow some kind of organization?
The hypothesis under test is network spread. If regions that are strongly connected to each other show similar alterations, this suggests that something travels along the connections rather than striking isolated regions. The study adds a developmental dimension by asking the same question separately in adolescents and in adults.
Mechanism
The method compares two maps of the cortex. The first is a map of the difference between patients and controls, region by region. The second describes a property of the brain: its connectivity, the density of a receptor, the expression of a set of genes. When the two maps look alike, this is called spatial correspondence.
The trap in this approach is well known: any two brain maps often look alike, simply because the cortex is spatially organized. The authors guard against it with permutation tests that preserve spatial structure, and with a null model that rewires the network while keeping the distribution of degrees and connection lengths, which destroys the observed topology while preserving spatial embedding. These are the right precautions. They do not change the fact that the method cannot establish causality, and that it says nothing about what happens in an individual.
The study at a glance
| Population | |
| No individual-level data. The material consists of statistics already published by an international imaging consortium, in the form of effect size maps by cortical region, for four disorders (attention deficit hyperactivity disorder, bipolar disorder, major depressive disorder and obsessive-compulsive disorder), each in its adolescent and its adult form. Patients versus controls, as tabulated in the source: 432 vs 347 and 733 vs 539 for attention deficit hyperactivity disorder, 411 vs 1,035 and 1,837 vs 2,582 for bipolar disorder, 213 vs 294 and 1,902 vs 7,658 for depression, 407 vs 324 and 1,498 vs 1,436 for obsessive-compulsive disorder. The age limits of the so-called adolescent group vary from one disorder to another, from under 18 years for obsessive-compulsive disorder to 8-24.9 years for bipolar disorder and 15-21 years for attention deficit hyperactivity disorder. | |
| Analysis | |
| These maps were matched against a structural connectivity atlas built from two hundred and seven healthy adults aged 22 to 36 years, nineteen neurotransmitter receptor and transporter maps, a post mortem gene expression atlas and an atlas of six mitochondrial features. The cortical parcellation is Desikan-Killiany, that is, sixty-eight regions. | |
| Comparator | |
| Null models obtained by spherical permutation preserving spatial autocorrelation, and by network rewiring preserving the distribution of degrees and connection lengths. One thousand draws for each null model. Robustness examined by removing one region at a time. | |
| Outcomes | |
| Strength of the correspondence between the alteration map and connectivity, identification of the regions best placed to account for the distribution, vulnerability of the most connected regions, associations with molecular maps. | |
| Design | |
| Secondary multimodal analysis of aggregate statistics, correlational and exploratory. No clinical outcome, no longitudinal data. The Oxford level of evidence does not apply to this type of work and was not established. |
Quality control
| Item checked | Verdict |
|---|---|
| Null models | Appropriate |
| FindingSpherical permutation preserving spatial autocorrelation, and rewiring preserving the distribution of degrees and connection lengths in order to test the observed topology. These are the two precautions expected in this type of analysis, both are present, with one thousand draws each. | |
| Developmental mismatch | Central bias |
| FindingThe connectivity, neurotransmitter, gene expression and mitochondrial atlases all come from adult brains. Applying them to adolescent imaging introduces the very mismatch that the conclusion sets out to interpret. The authors acknowledge this, which does not correct it. | |
| Size of the adolescent cohorts | Smaller |
| FindingThe adolescent maps rest on smaller samples and are therefore noisier. A noisy map correlates less well with any other map. The lack of correspondence in adolescents may therefore reflect a lack of power rather than a biological difference. | |
| Origin of the reference maps | Single samples |
| FindingThe mitochondrial map comes from a single coronal slice of the right hemisphere of a fifty-four-year-old neurotypical man. The gene expression atlas rests on six adult donors aged 24 to 57 years, only one of them a woman, and four of them provided only the left hemisphere. These are the sources the field has, but they remain narrow ground for a biological interpretation. | |
| Multiple comparisons | High |
| FindingFour disorders, two age groups, six families of analyses: connectome constraint, epicenters, vulnerability of the most connected regions, nineteen neurotransmitter maps, transcriptome, six mitochondrial features. Corrections cover only two of them, Bonferroni for the neurotransmitter maps and false discovery rate for the functional enrichment of genes. No correction is reported for connectome constraint, epicenters, hub vulnerability or mitochondrial features. The secondary molecular findings need replication. | |
| Funding and code availability | Partial |
| FindingPublic funding declared, no competing interests declared, source data openly available. The code is stated to be available from the corresponding author on reasonable request, which falls short of a public repository. The source declares four Chinese public grants, from the National Natural Science Foundation of China (82471952 and 82371928), the Natural Science Foundation of Anhui Province (2308085MH277) and the research fund of Anhui Medical University (2022xkj143), and states that the authors declare no competing interests. | |
Results
| Result | What the data show |
|---|---|
| Correspondence with connectivity, adults | 0.648 for ADHD, 0.625 for bipolar disorder, 0.491 for depression, 0.164 for obsessive-compulsive disorder |
| InterpretationThe significance values reported against the spatial null model are 0.002, below 0.001, 0.011 and 0.012 respectively. Against the rewiring null model, they are 0.030, below 0.001, 0.014 and below 0.001. All four adult disorders are therefore positive, which is the strongest argument of the study: an artifact would hardly explain why the pattern repeats. The fourth value, however, is of an entirely different magnitude from the other three. | |
| Correspondence in adult obsessive-compulsive disorder | 0.164, significant but very weak |
| InterpretationThe fourth disorder reaches significance against both null models, but with a correlation four times weaker than that of ADHD. The authors themselves write that the constraint effect should be interpreted with caution in this condition. A significant result and a strong result are not the same thing. | |
| Correspondence in adolescents | Not found |
| InterpretationThis is the headline finding, and it is the most fragile. A correspondence that was not found is not a correspondence shown to be absent. Two non-biological explanations remain open: the smaller size of the adolescent cohorts and the use of adult atlases. No analysis rules them out, neither an equivalence test nor a power estimate. The supplementary material makes this point concrete: in the robustness analysis for adolescent obsessive-compulsive disorder, the correlation ranges from 0.587 to 0.664 depending on the region removed, a magnitude comparable to that of adult ADHD, yet only one of the sixty-eight iterations falls below 0.05 against the spatial null model. Here, significance does not track magnitude. | |
| Vulnerability of the most connected regions | Negative correlations of 0.336 in adolescent bipolar disorder, 0.354 in adult bipolar disorder, 0.390 in adult obsessive-compulsive disorder |
| InterpretationThis finding, by contrast, also appears in adolescents, which tempers the reading of an adolescent brain that would follow no network organization at all. The two analyses do not measure the same thing. Note an internal tension in the source: the results section states that no significant association is observed for adolescent obsessive-compulsive disorder, while the discussion invokes a trend toward significance in that same group to extend the conclusion to both ages. | |
| Molecular associations | Almost all concentrated in adults |
| InterpretationOf the nineteen maps tested, the associations that remain significant after Bonferroni correction break down as follows: three for adult ADHD, four for adult bipolar disorder, five for adult depression, one for adult obsessive-compulsive disorder, a single one in adolescents, for bipolar disorder, and none for the three other adolescent groups. Speaking of distinct neurochemical signatures by age therefore mostly amounts to noting that little is detected in adolescents. These associations are exploratory, rely on average maps drawn from other samples, and should be read as leads, none of them established. The transcriptomic and mitochondrial analyses, for their part, do find associations in several adolescent groups. | |
| Robustness | Results stable when one region at a time is removed |
| InterpretationThe supplementary material reports the sixty-eight iterations for each of the eight groups. The four adult connectome constraint results remain significant against the spatial null model in all sixty-eight iterations for three of them and in sixty-seven out of sixty-eight for obsessive-compulsive disorder, and the three hub vulnerability results remain significant in all sixty-eight iterations. This check rules out the possibility that a single region carries the whole signal. It says nothing about the developmental mismatch or about the difference in power between age groups. | |
Critical appraisal
| Domain | Judgment |
|---|---|
| Nature of the design | Correlational |
| FindingSecondary analysis of aggregate statistics, with no individual-level data and no clinical outcome. No causal inference is possible, and the interpretation in terms of spread remains a reading, not a result. | |
| Mismatch between atlas and population | Not corrected |
| FindingThis is the point that governs how the whole study should be read. Applying adult references to adolescents, then concluding that adolescents behave differently from adults, runs a risk of circularity that this work does not rule out. | |
| Interpretation of a non-significant result | Point of caution |
| FindingThe central claim rests on an absence of correspondence in adolescents. Yet no equivalence analysis and no power estimate support this reading. Failing to detect a signal and showing that there is no effect are not the same thing. | |
| Statistical precision | Adequate |
| FindingLarge source cohorts, suitable null models drawn a thousand times, a detailed region-by-region robustness check. The statistical handling is the strength of the study. One reservation: correction for multiple comparisons covers only the neurotransmitter maps and the functional enrichment of genes. | |
| Fit between claim and evidence | Strained |
| FindingThe wording stays cautious, but the structure of the article makes the conclusion hinge on a negative result whose alternative explanations have not been ruled out. Careful wording does not make up for the weight given to that result. | |
| Clinical relevance | None |
| FindingNo biomarker, no test, no change in management. This work belongs to translational research and stops there. | |
Level of evidence
Confidence is reasonable that, in adults, the distribution of cortical thickness differences corresponds to the organization of connectivity. All four disorders point the same way, with appropriate null models. The reservation concerns magnitude: for obsessive-compulsive disorder, the correlation is only 0.164, and the authors themselves call for caution on this point.
Confidence is low for the developmental reading, even though that is what the article puts forward. A correspondence not found in adolescents is explained just as well by smaller cohorts and ill-suited atlases as by a different biology. Until these hypotheses are told apart, the conclusion remains a hypothesis.
Confidence is nil for any application. This work provides no marker, no test and no therapeutic direction, and it does not claim to. No Oxford level of evidence was established for this type of analysis, so the field was left blank rather than guessed.
The colleague test
What an experienced colleague would say after hearing about this study for two minutes, between two appointments.
“Nothing changes on Monday. It’s a nice idea: in adults the abnormalities follow the wiring, in teenagers they don’t. But they’re holding teenagers up against adult atlases, with fewer people in the teenage groups, and then concluding that teenagers are different. Fine, except that’s also exactly what you’d get if the measurement were just worse.”
What it means in practice: no consequence for a consultation. What you can take from it fits in one sentence worth passing on to families: a teenager’s brain is not a scaled-down adult brain, and a disorder does not leave the same imprint on it.
What you can do with this
- Have an accurate way of putting it when a family asks what can be seen in a teenager’s brain. The differences observed are differences in group averages, they cannot be read on any individual scan, and they are not used for diagnosis.
- Be wary of extrapolating adult data directly to adolescents, including when reading neuroimaging papers. This may be the only transferable lesson of this work, and it is a methodological one.
- Know how to answer a patient who has read that a brain scan could diagnose ADHD or bipolar disorder. No imaging can do that today, and this work does not point in that direction.
- Remember the difference between significant and strong. In adults, all four disorders reach the threshold, but the correspondence for obsessive-compulsive disorder is four times weaker than for ADHD, and the authors say so.
- Do not read this finding as proof that the adolescent brain escapes all network organization. Another analysis in the same article finds vulnerability of the most connected regions in adolescents with bipolar disorder.
Frequently asked questions
Does this change anything in my practice?
No. The study looks at no clinical outcome, uses no individual-level data and proposes no marker. It belongs to translational research.
So is the adolescent brain organized differently?
That is the hypothesis the authors put forward, and it is plausible given what we know about cortical maturation. But the study does not establish it, because two technical explanations remain open: smaller adolescent cohorts and reference atlases built on adult brains.
What does a correlation of 0.65 between two brain maps mean?
That the regions where thickness differs most between patients and controls tend to be those occupying a particular position in the network of connections. It is a resemblance between two average maps, not a measurement taken in a person, and it says nothing about which way things happen.
Why is the level of evidence field empty?
Because the Oxford levels of evidence scale is built for questions of diagnosis, prognosis or treatment. It does not apply to a correlational analysis of aggregate statistics. Filling in a value here would have been an invention.
Does this point to a therapeutic lead?
No. The associations with neurotransmitter maps are exploratory and rest on average data drawn from other samples. None has been tested, and none identifies a usable target.
Annotated bibliography
Source study. Cui S, Tao C, Xu S, Han X, Li K, Yu Y, Zhu J. Biological factors contributing to divergent cortical abnormalities in adolescent and adult psychiatric disorders. Transl Psychiatry. 2026. doi: 10.1038/s41398-026-04247-4. Received February 22, 2026, revised June 9, 2026, accepted June 30, 2026. Editorial status: the journal states that this is an unedited version of the manuscript, released to provide early access to the findings, and that the text will undergo further editing before final publication, so errors may remain. The document carries the label Article in Press. Volume, issue and pagination or article number have not yet been assigned on this version.
Funding and competing interests. The source declares support from the National Natural Science Foundation of China (grants 82471952 and 82371928), the Natural Science Foundation of Anhui Province (2308085MH277) and the scientific research fund of Anhui Medical University (2022xkj143). The competing interests section states that the authors declare no competing interests. The acknowledgements section is empty. The availability statement specifies that all data analyzed come from open-access sources and that the analysis code is available from the corresponding author on reasonable request. The PubMed identifier 42425965 was established on August 31, 2026 by searching bibliographic databases, as the document consulted carried none. The supplementary material was supplied and consulted, in five files. The following were checked there: the method used to build the normative connectome, Table S1 with the sample sizes and age limits of the eight cohorts, Table S2 with the nineteen receptor and transporter maps and their tracers and samples, Table S3 with the six donors of the gene expression atlas, the file of correlations between neurotransmitters and cortical thickness for the eight groups, the gene functional enrichment file for the five groups concerned, and the two robustness files for the removal of one region, that is, sixty-eight iterations for each of the eight groups and for each of the two analyses. The following could not be used in the format supplied: Figures 1 to 5, which are images rather than text, and therefore the correlation values for the eight adolescent and adult groups that appear only in the figures, as well as the detailed coordinates of the epicenter regions.
Background references. All the references below were read in the bibliography of the verified source. The cortical thickness maps come from four consortium studies: Hoogman M, et al. Brain imaging of the cortex in ADHD: a coordinated analysis of large-scale clinical and population-based samples. Am J Psychiatry. 2019;176(7):531-542, which supplies the attention deficit hyperactivity disorder data, with the limitation of being a multisite meta-analysis with small effect sizes; Hibar DP, et al. Cortical abnormalities in bipolar disorder: an MRI analysis of 6503 individuals from the ENIGMA Bipolar Disorder Working Group. Mol Psychiatry. 2018;23(4):932-942, for bipolar disorder, about which the source itself notes that adolescent abnormalities were limited to the right supramarginal gyrus; Schmaal L, et al. Cortical abnormalities in adults and adolescents with major depression based on brain scans from 20 cohorts worldwide in the ENIGMA Major Depressive Disorder Working Group. Mol Psychiatry. 2017;22(6):900-909, published online in 2016, for depression, with a very small adolescent cohort; Boedhoe PSW, et al. Cortical abnormalities associated with pediatric and adult obsessive-compulsive disorder: findings from the ENIGMA Obsessive-Compulsive Disorder Working Group. Am J Psychiatry. 2018;175(5):453-462, for obsessive-compulsive disorder. The contextualization maps come from Hansen JY, et al. Mapping neurotransmitter systems to the structural and functional organization of the human neocortex. Nat Neurosci. 2022;25(11):1569-1581, whose limitation is that it pools heterogeneous adult positron emission tomography samples; from Hawrylycz MJ, et al. An anatomically comprehensive atlas of the adult human brain transcriptome. Nature. 2012;489(7416):391-399, limited to six adult donors; and from Mosharov EV, et al. A human brain map of mitochondrial respiratory capacity and diversity. Nature. 2025;641(8063):749-758, limited to one coronal slice from a single brain. The statistical framework rests on Alexander-Bloch AF, et al. On testing for spatial correspondence between maps of human brain structure and function. NeuroImage. 2018;178:540-551, which introduced spherical permutation, and on Betzel RF, Bassett DS. Specificity and robustness of long-distance connections in weighted, interareal connectomes. Proc Natl Acad Sci U S A. 2018;115(21), for distance-controlled rewiring. The connectivity data come from Van Essen DC, et al. The Human Connectome Project: a data acquisition perspective. NeuroImage. 2012;62(4):2222-2231, a cohort of young adults aged 22 to 36 years, which is precisely the source of the developmental mismatch discussed above. Finally, the hub vulnerability hypothesis is set out in Fornito A, Zalesky A, Breakspear M. The connectomics of brain disorders. Nat Rev Neurosci. 2015;16(3):159-172, which is a review and not a demonstration. The other references in the source have not been included here, as their content could not be verified beyond their titles.