A machine learning algorithm can predict whether people with major depression will respond to a common antidepressant after only 1 week.
Credit: Yuuji/Getty Images

Research led by Amsterdam UMC shows artificial intelligence (AI) can help predict whether patients will respond to antidepressants after only one week.

As reported in the American Journal of Psychiatry, the researchers used a combination of machine learning, brain scans and clinical information to predict a patient’s response to the antidepressant sertraline, a selective serotonin reuptake inhibitor.

Although there are now many antidepressant medications on the market, their efficacy among people with major depressive disorder varies considerably. Normally, it can take up to eight weeks to assess whether a person with depression is responding to an antidepressant drug. If they do not have an improvement in symptoms at this point then a new medication is started, which means it can take months for someone to get relief from their symptoms.

However, the machine learning tool developed by Liesbeth Reneman, professor of neuroradiology at Amsterdam University Medical Center, and colleagues managed to give accurate responses about efficacy within a week.

The researchers used data from 229 patients with major depression who had clinical information collected and brain scans completed both before and after a week of treatment with sertraline, one of the most commonly prescribed antidepressants in the U.S. and Europe. Information on how effective the drug was in the longer term (more than eight weeks) was also available for these patients.

Reneman and team developed a multimodal machine learning algorithm to predict whether sertraline would be effective long-term using the data collected one week after the start of treatment. Using the original dataset and external validation data, the AI was able to predict responses to sertraline with a good degree of accuracy (balanced accuracy test: 62–68% and area under the receiver operator curve test: 0.66–0.73).

“The algorithm suggested that blood flow in the anterior cingulate cortex, the area of brain involved in emotion regulation, would be predictive of the efficacy of the drug. And at the second measurement, a week after the start, the severity of their symptoms turned out to be additionally predictive,” said co-author Eric Ruhé, a psychiatrist at Radboud University Medical Center, in a press statement.

Overall, the study results showed that only around one third of patients respond to sertraline, whereas the remaining two-thirds do not. Although sertraline can be very effective in some people, it is important that patients who do not or will not respond to the drug stop taking it as soon as possible so they can find a more effective medication more quickly.

“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. Because the drug also has side effects,” noted Reneman.

Also of Interest