“In the clinic, we think about depression with a single label,” says Leonardo Tozzi, a neuroscientist and lead author of the study. He was at Stanford before moving to a neurodiagnostic company in the US last year.

The brain is the seat of our minds, and in people with depression, it manifests itself in different ways. These manifestations appear as faulty brain patterns that psychiatrists often fail to take into account, according to the study’s researchers.

Dr. Tozzi et al. showed that these patterns can be classified into six unique subtypes of depression. The team also found that at least three of these subtypes could predict the antidepressant drugs and/or types of therapy that might be effective in treating these people.

Biomarkers for the brain

Dr. Tozzi joined Stanford Medicine’s Center for Precision Mental Health and Wellness as a postdoctoral fellow in 2018. He compared the purpose of the study to cardiologists who use EKG data to assess a patient’s heart condition. “We’re trying to turn psychiatry into something similar.”

Like the heart, the brain exhibits electrical activity. A functional magnetic resonance imaging (fMRI) machine can detect this activity and how it changes over time through electrical signals.

In people with mental illness, the underlying brain circuits that connect different regions are not activated normally. A region may show more intense electrical activity than in a healthy person.

Different people exhibit different patterns, both normal and abnormal. When some of these appeared in people with a particular mental illness, researchers called this a subtype.

Many studies have divided depression into subtypes based on brain activity, but the new study used a “theory-driven” approach to create subtypes that are also clinically relevant.

Data analysis

In 2021, Dr. Tozzi and co. began working on the study, combing through 801 patient records from clinical trials conducted over the past two decades. All of these people had been diagnosed with depression and/or an anxiety disorder but had not undergone treatment at the time. “Many had comorbid anxiety,” Dr. Tozzi said.

The recordings included patients’ responses to symptom questionnaires and behavioral outcomes, as well as fMRI data – lots of it. Scanning each brain volume requires tracking changes in 100,000 data entries over 20 minutes, the typical scan time in the studies.

“We had to do some kind of compression and reduce the data to a meaningful, smaller set of variables,” said Dr. Tozzi.

They limited their analysis to six brain circuits that had been linked to depression and anxiety disorders in previous studies. But these circuits did not light up consistently or simultaneously in the fMRI data.

Instead, a person’s attention, salience, and “default mode” circuits were activated when the person was sitting idle and lost in thought. Elucidating the other three circuits – cognitive, positive, and negative affect – required a series of tasks.

During clinical testing, participants were scanned with an fMRI machine when they were not tasked with a task and then again when they responded to smiling or sad faces or followed commands that flashed on a computer screen.

Dr. Tozzi recalled a “eureka” moment when the team realized how to best use a machine learning algorithm to detect faulty brain activity in people. “After that, everything else just kind of fell into place,” he said, as the algorithm revealed the six subtypes.

Overcoming depression

The team was then able to prove that the subtypes were real data and not artifacts by validating them with the symptoms reported by those affected and their behavior during behavioral tests.

The team also analyzed treatment response data from 250 participants in the clinical trials. These people were randomly assigned to receive some common antidepressants – sertraline, venlafaxine XR and escitalopram – and therapy. The team members found that people with three of the six depressive subtypes could receive treatment and expect their symptoms to improve.

One subtype the study discovered is characterized by a hyperactive cognitive circuit. The researchers found that people with this subtype reported high levels of anxiety, a lack of interest in interacting with the outside world, and general feelings of threat. Their analysis suggested that they might respond better to Venlafaxine XR.

People with the other two subtypes may respond well to therapy, according to the analysis, although one of them, who had a hyperactive attention circuit, “responded less well to behavioral therapy,” the article says.

For people in the other three subtypes, researchers were unable to match treatment options that might help them. However, for one of these subtypes, researchers were unable to find a large amount of associated faulty brain activity, and there was not enough data to reliably measure treatment responses for the other two.

Dr. Tozzi said the next step is to find more treatments that could address the symptoms of depression across all subtypes. In fact, he said, Stanford has been running a clinical trial for a year now — after the team’s study was completed — to test the ability of the subtypes to predict treatment effects.

“People will be scanned, their (subtype) will be determined and then they will be given a drug that targets that specific (subtype),” Dr. Tozzi said of the vision of the new study.

Karthik Vinod is a freelance science journalist and co-founder of Ed Publica. He holds master’s degrees in astrophysics and science, technology and society.