Mental Health

AI breakthrough rapidly determines if an antidepressant will work

AI breakthrough rapidly determines if an antidepressant will work
An algorithm using patient data may save patients months of trial and error with new medications
An algorithm using patient data may save patients months of trial and error with new medications
View 1 Image
An algorithm using patient data may save patients months of trial and error with new medications
1/1
An algorithm using patient data may save patients months of trial and error with new medications

Months of trial and error, as well as adverse side effects, in an effort to find the right antidepressant may soon be a thing of the past for those with major depressive disorder, with a novel artificial intelligence model able to determine a drug's efficacy on an individual within just one week.

"This is important news for patients," said Liesbeth Reneman, Professor of Neuroradiology at Amsterdam University Medical Center (UMC). "Normally, it takes six to eight weeks before it is known whether an antidepressant will work."

In the latest clinical application of AI, Amsterdam University Medical Center (UMC) and Radboud UMC researchers developed an algorithm that, based on patient MRI scans and other data, could determine if a particular antidepressant would be effective in the long run. While an estimated 11% of the US population has a prescription to manage their depression, around 60% won't find a suitable medication on their first attempt. Because of the time invested and side effects experienced, this can be enough for many to not try a second, or third, time.

The researchers set out see if such an AI model could first work for predicting the efficacy of selective serotonin reuptake inhibitor (SSRI) sertraline, AKA Zoloft, the most commonly prescribed antidepressant in the US.

For this study, the team took the results from a previous US study of 229 patients with depression, using MRI brain scans and clinical data recorded before administering either sertraline or a placebo. This data was then fed into the AI, with the algorithm looking specifically at the anterior cingulate cortex, as well as symptom severity.

"The algorithm suggested that those who had a lot of blood flow in the anterior cingulate cortex, the area of brain involved in emotion regulation, would be helped by the drug," said Eric Ruhé, psychiatrist at Radboud UMC. "And at the second measurement, a week after the start, this turned out to be the severity of their symptoms."

Through this, the AI determined that sertraline would only work for a third of the participants, essentially saving two-thirds from up to two months of 'wait and see'. While it can take up to six months for an antidepressant to be most effective, serious side effects can linger for a significant time and have a huge impact on day-to-day life, often as much as the disorder itself.

"With this method, we can already prevent two-thirds of the number of 'erroneous' prescriptions of sertraline and thus offer better quality of care for the patient," said Reneman. "Because the drug also has side effects."

Treating major depressive disorder is challenging because of its complexities. It's made even more difficult because of the broad medical interventions available, including SSRIs, serotonin and norepinephrine reuptake inhibitors (SNRIs), atypical antidepressants, tricyclic antidepressants and monoamine oxidase inhibitors (MAOIs).

While this algorithm is sertraline specific, the researchers hope to not just adapt it to become more personalized, but to apply the same kind of model to a wide range of medications used to treat depression.

The study was published in the American Journal of Psychiatry.

Source: Amsterdam University Medical Center

1 comment
1 comment
paul314
In the US, at least, not a lot of people will be able to make use of this unless the cost of MRIs comes down substantially.