EEG is inching closer to predicting incidence of major depressive disorder and treatment efficacy. Here’s a view of some recent evidence.

You aren’t human if you haven’t had the blues. For most of us, periods of unhappiness tend to ease over time. Your cat dies or your marriage falls apart, and life is harder for a while. Depending on the severity of the event, the intensity and duration of associated unhappiness can vary considerably. Over time, however, we return to our personal base level of happiness. This is a normal part of life and a piece of what it means to be human.

Imagine if instead of a response to a specific event that can be healed with time, sadness were simply a way of life. Your cat dies, and two weeks later you are still unable to get out of bed. Your marriage falls apart, and two years later your lust for life is still missing. Time is not healing your wounds. Or perhaps your cat is purring pleasantly in your lap and your marriage is the envy of the block. Despite this, a wave of sadness has settled over you. It has settled deep and it doesn’t seem to be going anywhere. This is what major depressive disorder looks like.

Defining Major Depressive Disorder

Major depressive disorder (MDD) is characterized by two weeks of depressed mood associated with decreased interest in activities, fatigue, and feelings of worthlessness or guilt.  One of the most prevalent mental disorders in the United States, over 10 percent of Americans aged 18- 25 suffer from MDD in any given 12 month period. Other age groups are also affected, but at a lower rate.

Yet, unlike the flu or a cold, mental illness is difficult to diagnose and tricky to treat. Our understandings of depression and related psychological illnesses is woefully limited. Though MDD can be triggered by an event or environmental shift, 40 percent of the mental disorder is thought to be genetically based. MDD is also related to suicide – but not always. More than half the people who die from suicide each year suffer from depression. But many do not.  In no small part the confusion arises because we define MDD subjectively based on a questionnaire (see related post The Difficulty of Diagnosing Depression) and we don’t know what it means at the level of the brain.

Towards a Brain Based Diagnostic

A first step towards treating depression is to be able to diagnose its myriad of subjective symptoms with a more physiological readout. Electroencephalogram (EEG) is one potential mechanism, offering an insight into the dynamics of the brains of depressed patients as compared to healthy controls.

One difference that has been identified has to do with asymmetry between the two hemispheres of the brain. Specifically, decreased power in the alpha and theta bands in the frontal region of the left hemisphere relative to the right as well as decreased synchronization. Studies are also showing that depression-indicating EEG readings shift back, more resembling healthy controls, when a patient has been successfully treated. However, distinctions along any one dimension are not perfect and results are often inconsistent.

Deep Brain Stimulation in MDD

Despite a clear understanding of the mechanisms or diagnosis of MDD, therapies abound for patients desperate for change.  Deep Brain Stimulation (DBS), for example, first developed as a treatment for Parkinson’s Disease, has seen relative success as a treatment for depression that is resistant to  less invasive methods, such as pharmaceuticals. DBS involves an advanced surgical procedure to insert a pacemaker-like device to stimulate the brain. One inserted, it can be used to apply different kinds of Yet despite being a major procedure of last resort, this treatment doesn’t always work. Parsing out the brain dynamics at work – what makes one person respond to a treatment when another person with similar symptoms is not – is a monumental task.

Linking Dynamical Features of the EEG with DBS Success

A recent Canadian study by Quraan et al., recorded the EEG signals of twelve MDD patients that had been treated with DBS surgery. Six had responded positively as measured by a significantly improved score on the Hamilton Rating Scale for Depression. The other six had not. Recordings were made both with the implanted electrodes on and with them off. The study also included recorded EEG signals of fifteen healthy volunteers to use as controls.

Researchers found that DBS responders and non-responders had statistically significant differences in hemispheric power asymmetry under both conditions (i.e. with electrodes on or off). Moreover, the healthy controls had asymmetrical power results that mirrored the responders and were opposite from the non-responders. Synchronization between alpha and theta bands was also significantly different between the two groups. The take home message here is that responders’ patterns more closely resembled the signals from healthy volunteers, and non-responders continued to exhibit tell-tale signals of MDD in their asymmetrical power patterns.

Relating Individual EEG Patterns to Treatment Outcomes

What is the next step? It is easier to find statistically significant correlations among features than well separated predictive ones. However, this is the holy grail – to find those aspects that that can predict potential treatment success.

Though it may be some time before we can predict treatment results with high accuracy, EEG can still do better than pure trial and error and is already being used to try to predict the potential efficacy of some pharmaceutical treatments.

See related post The Frustration of Treating Depression

A recent study, for example, of 25 MDD patients undergoing inpatient treatment with venlafaxine found that EEG results could predict whether the drug was going to work in the first week. In this particular case, eleven of the twelve responders showed decreased values in prefrontal theta, used as a marker of evidence that the drug was working. Though five of the thirteen non-responders showed some decrease, none of these were statistically significant.

The direction is promising and examining a larger repertoire of dynamical features has the potential to deliver even better predictive value.