Beyond Diagnosis: Time for a New Paradigm

Copyright (c) 2009 by Kevin Thompson

The Meaning of Diagnosis

In the last issue, we reviewed the concept of diagnosis, and discovered that the popular conception of diagnosis does not correspond very clostly to the reality. The popular understanding is that a diagnosis describes the cause of an illness (such as a particular bacterial infection), and provides guidance about the best type of treatment. While this description is sometimes correct (such as for strep throat), it often is not. Instead, a diagnosis is often a label applied to a set of symptoms, such as bronchitis. Bronchitis simply means an inflammation of the bronchi (the major passages leading to the lungs), but says nothing about the cause of this inflammation. The term doesn't even specify whether the inflammation is due to a bacterial or viral infection, which is an important distinction, given that antibiotics are useful for the former, but not the latter. A physician may have reasons to suspect viral or bacterial origins, but will not be able to identify a specific viral or bacterial strain.

Fortunately, these limitations are not critical for bronchitis, as the viral form usually goes away in time without treatment, and antibiotics usually suffice to deal with the bacterial form independent of which bacterium is responsible. Unfortunately, this limitation becomes more severe when applied to mental illness.

Diagnosis in Mental Illness

Unlike bronchitis, mental illnesses rarely come with any physical symptoms that can be observed by a physician. Symptoms are almost entirely internal to the patient's subjective world (perceptions, thoughts, feelings), or behavioral in nature (angry outbursts, spending sprees). Worse, the causes of mental illness (believed to be abnormalities in brain chemistry, receptor concentrations, and so forth) are not known at a level that is detailed enough to provide an obvious connection between symptoms and causes. Worst of all, no standard medical tests exist to detect the underlying causes anyway.

Given the scarcity of useful knowledge and diagnostic tests, it is not surprising that diagnosis is solely symptomatic. It is also dismayingly subjective, as different psychiatrists can justifiably interpret the same symptoms as indicating such widely different diagnoses as anxiety and bipolar disorders. The fault here is not the physician's, but rather the arbitrariness of the diagnostic categories, and the lack of any objective criteria for measurement.

The final "dirty secret" of psychiatry is that medications used to treat mental illness have a relatively low probability of success. The probability that someone with depression will response well to any single antidepressant, for example, is around 30%. Thus the average patient may have to try a few different antidepressants before he gets much relief.

Any concept of diagnosis as a precise description of an illness, which serves as a useful guide to treatment, has almost disappeared from this picture. In fact, the picture is so poor that any reasonable person would shake his head and wonder why the mental-health profession persists with its practice of fuzzy diagnoses and low-reliability treatments.

The answer, of course, is one that psychiatrists understand all too well: as inadequate as this approach to diagnosis and treatment may seem, it is the best that medical science can do.

Or is it?

Correlation in the Treatment of Illness

Let's pull back and remember that what the patient really wants is to feel better. He cares about a diagnosis primarily because he sees it as a necessary step to feeling better. However, if we have reached the point where diagnosis is an unsatisfactory guide to treatment, perhaps it is time to look at alternative strategies for meeting the patient's needs.

What we really want is a predictive model than can determine which treatment will work best for each patient. If a classic diagnosis assists this result, then so be it, but if not, we should not allow the expectation that diagnosis is necessary to get in the way.

The concept of a predictive model is simple: Given an initial state, the model predicts the effects of different treatments, and we select the treatment that produces the best results. To do this, we need a quantitative description of the patient's state, and a method for predicting how his state changes based on treatment. In simpler terms, we need to know what "sick" and "healthy" mean in terms that can be measured, and how different treatments will affect the person.

Note that the concept of diagnosis has disappeared, and been replaced by the concept of correlation. The nature of the illness is irrelevant, because what matters is how results are correlated with the possible types of treatment.

The concept of correlation provides a completely different paradigm (organizing strategy) for how we think about illness and medicine. What we lose is the concept of a diagnosis, meaning an understanding of why we are ill. What we gain is a strategy for treating problems that cannot be diagnosed in any meaningful sense. (Is the loss a serious one? No, it isn't, because correlation methods are additional to strategies for discovering treatments, not a substitute for diagnostic strategies that work.)

One Correlation Method: Referenced EEG

A company named CNS Response has pioneered a correlation method based on electroencephalogram technology, called "Referenced EEG" (rEEG). At the heart of the rEEG method is a database of EEG recordings from thousands of healthy individuals, and individuals suffering from some common mental illnesses, both before and after taking various psychiatric medications. The company has created a predictive model based on statistical analysis of how different people respond to medications, which allows them to predict the response of a specific person to specific medications.

The claim is a bold one, and seems too good to be true. However, the company does not claim perfect success, only a better success rate than alternative strategies for selecting medications. For example, a study of patients with refractory depression (treatment-resistant depression) showed that six out of seven patients treated with the rEEG method improved, compared to one out of seven whose treatment was not guided by this method. The company claims an overall success rate of 70 - 95% in treating depression, attention-deficit disorder, eating disorders, and addiction.

These studies are sufficiently new that researchers in psychopharmacology have not confirmed or generally supported the results at this time. However, the relatively inexpensive and uninvasive nature of the technique, coupled with the reported success rates, make rEEG a technique definitely worth exploring.

Conclusion

While the concept of diagnosis remains valuable, it is no longer the sole paradigm for the selection of medical treatment. The new concept of correlation, while perhaps less intellectually satisfying, provides a very different approach to finding effective treatment for medical problems. Correlation methods hold out the prospect of successful treatments in areas where traditional diagnostic techniques have proven inadequate. Those who suffer from depression, and other types of mental illness, have reason to be glad that a new correlation method, rEEG, is now available to them.