Depression · Replication
Depression "biotypes" did not survive permutation testing or cross-validation
Save this development or follow its topic to track what changes.
Short answer
Proposed connectivity-based depression subtypes were indistinguishable from what a single continuous distribution produces once feature selection was properly accounted for.
What happened
Dinga and colleagues applied the original pipeline to 187 patients with depression or anxiety: select the 150 connectivity features correlating best with 17 clinical symptoms, run canonical correlation analysis, then cluster. The initial correlations looked strong. Permutation tests that account for the feature-selection step found them non-significant, and they effectively disappeared under cross-validation. The three-cluster solution scored 109 on the Calinski-Harabasz index, but simulations showed a single continuous distribution produces clustering that looks just as distinct.
Why it matters
The failure is not in the data but in the procedure: selecting features by their correlation with the outcome and then testing that correlation inflates significance. Because clustering algorithms return clusters whether or not any exist, a cluster index alone cannot establish that categories are real.
Evidence
- Study type
- Methodological replication of a published biotyping pipeline with added permutation tests and cross-validation
- Sample
- 187 patients with depression or anxiety
- Journal
- NeuroImage: Clinical · peer reviewed
- Replication
- Both the canonical correlations and the cluster structure failed to replicate once the added checks were applied
- Limitations
- The replication sample was clinically broader than the original, so sample differences contribute. A null here cannot rule out that other imaging measures or clinical instruments would reveal genuine subgroups.
What this connects to
Sources
The 2 studies this explanation is built from, by the role each plays. Every source links to PaperFren’s explanation of it and to the original paper.
Primary study
- Can brain scans divide depression into distinct biological types?
A rigorous re-analysis shows that previously proposed biological subtypes of depression may be statistical artifacts of overfitting rather than true, distinct categories.
What it does not showLimitations
This study does not prove that biological subtypes of depression do not exist; rather, it shows that current methods are too weak to prove they do. Because the replication sample included a broader clinical group than the original study, some differences could stem from sample variation. Additionally, the study cannot rule out whether alternative imaging methods or different clinical measures might reveal genuine biological groupings.
PaperFren explanationStudy with cards and a quizOriginal paper (DOI)cc by
Supporting evidence
- How do antidepressants affect the resting brain in healthy people?
Short-term use of antidepressant medications decreases communication between deeper emotional brain structures and the prefrontal cortex during rest.
What it does not showLimitations
This study only evaluated healthy volunteers, meaning we cannot conclude whether these exact connectivity reductions would occur or improve symptoms in clinically depressed patients. Additionally, the drug administration period lasted for only 7 days, which is much shorter than the typical multi-week timeline required to see clinical therapeutic effects in patients. Finally, because the resting-state scan was performed shortly after participants completed cognitive tasks, residual brain activity from those tasks might have influenced the resting patterns.
PaperFren explanationStudy with cards and a quizOriginal paper (DOI)cc by
Before
Resting-state connectivity was reported to divide depression into distinct neurophysiological biotypes with different treatment responses, a result that drew substantial attention.
Now
The brain–symptom associations and the cluster structure both fail under appropriate validation. This shows the method cannot support the claim — not that biological subtypes of depression do not exist.