Subgrouping autism and ADHD based on structural MRI population modelling centiles.
Clara Pecci-Terroba, Meng-Chuan Lai, Michael V Lombardo and 19 others
PMID 40468437WHAT IT FOUND
Brain structure subgroups in autism and ADHD depend heavily on the analysis method used, with little agreement between algorithms.
These neuroanatomical clusters did not consistently map onto clinical symptoms or IQ, meaning they currently offer no reliable tool for diagnosis or treatment planning.
Key findings
01The number and nature of identified brain subgroups changed drastically depending on whether global or regional MRI features were used, and which clustering algorithm was applied.
02Subgroups defined by opposite patterns of brain volume and surface area relative to controls showed no consistent or clinically significant differences in autism or ADHD symptoms.
03Agreement between the two different machine learning techniques used to define subgroups was minimal for regional brain data, indicating poor stability of these biological clusters.
STILL TO COME
How it was doneWhat they found
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
Clinical data (such as ADOS or ADHD rating scales) was missing for many participants, reducing the ability to link brain subgroups to specific symptoms. The study relied on multi-site data, which introduces variability in scanning and demographics that, despite harmonisation, may affect clustering stability. The number of female participants was low, preventing a robust examination of sex differences. There is no 'ground truth' for autism or ADHD subtypes, making it impossible to determine which clustering method is actually correct. The algorithms produced highly inconsistent results when applied to regional versus global data, suggesting the identified subgroups may be artefacts of the mathematical method rather than distinct biological entities.
Declared interests
The authors declare no competing interests. The work was supported by the NIH and non-US government research funds.
The easy way to misread this
Do not assume that these MRI-based subgroups represent distinct clinical conditions or diagnostic categories. The study explicitly found that the subgroups were unstable across methods and did not align with consistent behavioural or symptom profiles, so they should not be used to guide patient assessment or intervention.