Does Depression Age the Brain? A 7,000-Person Study Finds a Real Signal That’s Too Small to Diagnose With
A large multi-cohort study finds that depressed brains look about a year older on average, but the bigger story is how much individuals vary.
Over the past few years, “brain age” has become a popular metric in neuroimaging. The idea is simple. You train a model on structural brain data from a large group of healthy people so that it learns to predict a person’s chronological age from their brain alone. Then you apply that model to new people.
Subtract someone’s actual age from their predicted brain age and you get the brain-predicted age difference, or brain-PAD. A positive value means the person’s brain looks older than they are.
A scattering of earlier studies hinted that people with depression might have older-looking brains. But most of those studies had only one or two hundred participants, and their findings didn’t agree. This new work set out to settle the question with a sample large enough to give a clear answer.
Going big: 19 cohorts, nearly 7,000 brains
The researchers pooled T1-weighted structural MRI scans from 19 cohorts around the world, covering nearly 7,000 people. Of these, 2,675 had major depressive disorder.
The features came from FreeSurfer: 77 structural measures spanning cortical thickness, cortical surface area, subcortical volumes, ventricular volumes, and total intracranial volume.
Get ComfyAI’s stories in your inbox
Join Medium for free to get updates from this writer.
The model was trained only on healthy participants, with separate models for men and women. Using ridge regression, it learned to predict age from these features and was then applied to held-out healthy controls and to the patients. The model’s mean absolute error was roughly 6 to 7 years.
The headline result: older, but only slightly
After adjusting for age, sex, and scanning site, the depression group’s brain-PAD was on average 1.08 years higher than that of controls (p < 0.0001). So the direction is clear: depressed brains do look older.
But the effect is small. The Cohen’s d is just 0.14.
No clinical feature seemed to drive it
What stands out even more is that the researchers couldn’t pin the difference on any particular clinical characteristic. They looked at recurrent versus first-episode depression, medication use, remission status, age of onset, and symptom severity. Every subgroup showed the same slight shift toward an older-looking brain, and none differed significantly from the others.
Watch the variance
The within-group spread also deserves attention. The brain-PAD distributions of patients and controls overlap almost completely. That one-year average gap between groups is swamped by individual variation of more than a decade.
So this paper does not support a headline like “Depression ages your brain by a year.” What it shows is that at the population level the signal is real and consistent in direction. The effect is also genuinely weak, and individual differences are so large that brain age can’t currently serve as a diagnostic marker for depression.
Where brain age might actually be useful
Its more likely value lies elsewhere: helping identify which patients are at higher risk of later problems such as brain atrophy or cognitive decline. Testing that will require longitudinal follow-up.
| Find papers faster on arXivSub with AI summary (CVPR、ICCV、ECCV、ICML、ICLR、NeurIPS、AAAI、MICCAI、CoRL、RSS、ACL、EMNLP、ICRA、IROS、IJCAI)
