When is CBT useful for depression?

Researchers have used brain scans to predict which patients with depression will respond to Cognitive Behaviour Therapy (CBT), a first line treatment that remains ineffective for more than half of patients who try it.

A team led by Filippo Queirazza and Marios Philiastides, from the University of Glasgow, UK, gave 37 people with low grade depression MRI brain scans before and after a course of computer-based CBT.

They were prompted by the rather dismal fact that CBT, which gets patients to weigh up the evidence for negative beliefs and relinquish those that don’t bear scrutiny, is only effective in up to 45% of people.

Given the labour-intensive nature of CBT – face-to-face sessions can go for 20 weeks – and the fact that some people even get worse with it, working out who will respond is crucial. 

To answer that question the researchers focused on something known to go wrong in the brains of depression sufferers; studies have found they have blunted responses to both the pleasure of reward and the pain of punishment. 

That’s a problem, because the carrot and stick of those twin motivations are a big part of how people learn from mistakes. It’s called reinforcement learning and hinges on our ability to optimise things that are good for us.

Queirazza and colleagues thought depressed people who were worse at reinforcement learning might also be bad at CBT.


It’s known that typical depressed thoughts like “I’m a loser at everything I do” are the product of negative biases. To shift such beliefs with CBT, people need to be able to detect and react to more rewarding information like, “actually, you succeeded at X”.

The researchers proposed that being more responsive to reward is key to updating negative beliefs in the face of countervailing evidence. To that end they did brain scans on participants while they did a computer task that meant calculating both probability and reward.

Tallying the scan results with how people did on CBT, a clear pattern emerged.

Fifty-one per cent of participants responded to CBT and, in that group, two brain regions showed greater activity on the pre-treatment scans.

The first was the amygdala, an area involved in identifying threat and laying down negatively biased memories in depression. The second was the striatum, a part of the midbrain replete with neurons rich in dopamine, the chemical messenger that signals “reward”. It is also an area strongly linked to reinforcement learning.

“We show that… activity in these regions is significantly and linearly correlated with the extent of post-treatment symptomatic change,” the authors write.

That finding is of particular importance because, they note, right now there is “no clinically viable neuroimaging predictor of CBT response.”

“Critical to the implementation of precision medicine is the development of predictive biomarkers of clinical outcome to enable treatment selection and prognostic stratification. Unfortunately, psychiatry is lagging behind other medical disciplines,” they add.

The researchers acknowledge the need to replicate their findings in bigger studies, but their work could just bump psychiatry forward a few places in the race to better treatment for depression.

The study appears in the journal Science Advances.

Please login to favourite this article.