Brain Waves May Help Predict Who Will Benefit From an Antidepressant

AI analysis of EEG might reduce trial-and-error in treating depression.

Posted Mar 12, 2020

Treating depression with medications such as SSRIs (e.g. Prozac and Zoloft) is a hit or miss proposition, with an estimated 20% of patients getting more benefit than placebo. And up until now, it has been difficult to predict which patients are likely to be among the fortunate ones.

But Stanford University researchers have described a test that can help predict who will and will not benefit from a common antidepressant, Zoloft.

Post-doctoral researcher Wei Wu and colleagues, writing in the February 10 issue of Nature Biotechnology, reported that electroencephalograph (EEG) data were predictive of which patients improved with the widely prescribed antidepressant.

Using a machine learning (ML) system called Latent Space Analysis, the Stanford team compared pre-treatment resting-state EEGs (EEGrs) of depressed patients who were then given either placebo or Zoloft. Following treatment, the researchers analyzed which patients in both the placebo and Zoloft-treated groups improved, using the Hamilton Depression Rating Scale.

The Latent Space ML system uncovered a statistically significant .6 correlation between patient’s EEG patterns and their degree of reduction of depressive symptoms.

Although EEGs were not a perfect predictor of treatment outcomes, ML brain wave analysis performed significantly better than chance, suggesting that, in the future, physicians may be able to make more informed choices about which medications to prescribe for their depressed patients, instead of having to wait eight weeks or longer to learn whether or not a particular drug will reduce depressive symptoms.

The Stanford work on using EEG to predict treatment outcomes is just one of several new diagnostic methods being developed to determine which patients will respond to which drugs. For example, research in the emerging field of pharmacogenetics suggests that genetic screening of depressed patients may ultimately help predict not only which patients will best respond to different medications, but also which patients are most likely to experience adverse side effects.

Like EEG analysis, pharmacogenetics is far from perfect. But such emerging diagnostic approaches could eventually significantly reduce the lengthy trial-and-error period that is common now in the treatment of depression and more quickly put patients on the path to recovery.