The Corpus Callosum

To put this in context, the DSM is in the final stages of revision, from DSM-IV-TR, to DSM-V.  The draft to which he is referring is online, at www.dsm5.org.  Shorter begins with a rookie mistake.  You see, in order to have any hope of understanding the DSM (whatever version), you cannot merely read the lists of criteria; you have to read the text.  I know he has not read the text, because the text hasn't been written yet.  Only the draft criteria are out.  It is not a sign of academic rigor to pass judgment on a draft, much less a draft of the framework of the work itself.  It is like looking at the pictures in a chemistry textbook, and thinking you can discern something about the textbook itself.  Well, perhaps you can discern something, but not enough to write an article for a newspaper...even if it is a Rupert Murdoch paper. 

(OK, maybe I am wrong, maybe there is a draft of the entire text, that has not been announced or put online.  But I don't think so.  The APA has been making an effort to be publicly transparent in their process, and the website only mentions a draft of the criteria sets.)

The next error is more subtle, but still significant: "...stand a good chance of being diagnosed with a disease that doesn't exist..."  It is my impression that few psychiatrists think of the criterion sets as definitions of diseases.  They are lists of symptoms and items in a patient's history.  These define syndromes, not diseases.  He must know that.  Perhaps he is dumbing the article down, to make it appropriate for a mass audience.  But if that is the case, what point is he trying to make?

Then it gets worse:

In the 1950s and '60s, when psychiatry was still under the influence of the European scientific tradition, reasonably accurate diagnoses still sat at center stage. If you felt blue, uneasy and generally jumpy, "nerves" was a common diagnosis. For the psychotherapeutically oriented psychiatrists of the day, "psychoneurosis" was the equivalent of nerves. There was no point in breaking these terms down: clinicians and patients alike understood "a case of nerves," or a "nervous breakdown."

Sure, if you leave everything vague, then no one can prove anything to be incorrect, and you can say anything you want.  In would be very difficult to disprove a diagnosis of "nerves," for example. 

It is hard to know what to make of the statement that "there was no point in breaking these terms down."  He is a historian, he must have read works such as Fenichels' Psychoanalytic Theory of Neurosis (OK, it's from the 40's, but the point stands).  The psychiatrists of that era broke things down into excruciating detail.   (The problem was, that the way they broke things down did not help them come up with diangoses that were sufficiently specific as to be useful.)

Then it gets worse:

Our psychopathological lingo today offers little improvement on these sturdy terms. A patient with the same symptoms today might be told he has "social anxiety disorder" or "seasonal affective disorder." The increased specificity is spurious. There is little risk of misdiagnosis, because the new disorders all respond to the same drugs, so in terms of treatment, the differentiation is meaningless and of benefit mainly to pharmaceutical companies that market drugs for these niches.

See that?  Can you believe it?  In his opening paragraph, he states that the drugs do not work.  Now, he is telling us that they do work.  As for his point about there being no reason to differentiate between two conditions, if they both respond to the same drug, this is ridiculous.  It is like saying that there is no point in differentiating between two bacteria, if they both are killed by the same antibiotic.  You can't do research by throwing out data.  Social phobia is very different from seasonal affective disorder.  Sure, persons with these conditions might respond to the same drugs (SSRIs), but I've never seen anyone seriously propose that we treat social phobia with bright-light therapy in the AM.  They respond to the same drugs, but their responses to other therapies are different.  Cripes, what was he thinking???

What more can he muddle?

The first of the "tricyclic antidepressants" (because of their chemical structure) was launched in the U.S. in 1959, called imipramine generically and Tofranil by brand name. It remains today the single most effective antidepressant on the market for the immediate treatment of serious depression.

Imipramine is a great drug.  Nothing has been shown, convincingly, to be more effective, on a population basis (individual results vary, of course).  But to say it is the single most effective, implies that it -- alone -- is better than the rest.  This simply is not true. 

Dr. Shorter than goes on to imply that that the changes in diagnostic criteria, and the emergence of new classified syndromes, is all a plot to market drugs.  This is illustrated in the following, although he goes on at some length to make his point:

New drugs appeared to match the new diseases. In the late 1980s, the Prozac-type agents began to hit the market, the "SSRIs," or selective serotonin reuptake inhibitors, such as Zoloft, Paxil, Celexa and Lexapro.

The implication is that psychiatrists created a new diagnostic category, so that drug companies could sell their drugs.  Now, I am very much aware of the fact that there are many shady practices in the marketing of these products.  But this is not one of them.  You see, in order for a drug company to sell a drug in the USA, it has to have what the FDA calls an "indication."  That is, it has to be show to be safe and effective for a particular condition.  Thus, ALL drugs marketed in the USA will have an indication.  The indication will be drawn from whatever the standard diagnostic scheme is, pertaining to the condition at hand.  As it happens, oncologists reclassify tumors from time to time.  Is he going to argue that oncologists are defining new cancers, just so the drug companies can sell their drugs?  Is he going to argue that Pfizer would not have had a market for Zoloft, if Major Depression had not been defined?  This simply does not make any sense. I am reasonably confident that pharmaceutical companies would do just fine, if not better, if we still used his favored terms. 

Examples: Mellaril (an antipsychotic) once was marketed for the treatment of "anxiety, tension, and agitation."  Taractan, another antipsychotic, was indicated for "moderate to severe emotional disorders, especially agitated states associated with neuroses, depression, or schizophrenia."  (Those are quotes from the FDA-approved product labeling, as published in the July-August 1965 issue of Psychosomatic Medicine.)  Is he seriously saying that this way of describing drug indications was superior?  That it led to more precisely-targeted drug prescribing?  He is a historian.  He should take a look at some old documents to see if they support the points he is trying to make.  Unfortunately, they do not.

Because Dr. Shorter is a serious scholar, and obviously quite knowledgeable, I was curious as to how he could make such a fundamental error.  As it happens, he recently published a paper in the Canadian Journal of Psychiatry that sheds some light on this.  The paper is in a debate format, structured as a debate with another scholar.  (The link opens a PDF of the paper.)

Disease Versus Dimension in Diagnosis
Edward Shorter, PhD, FRSC; Herman M van Praag, MD, PhD
Can J Psychiatry. 2010;55(2):59-64.

Dr. Shorter mentions five disease states that match his conceptualization of a real, definable disease: melancholia, catatonia, atypical depression, panic, and ADHD.

And that is it. Those are the disease entities in psychiatry that, to date, have been relatively well characterized. Others will doubtless follow as the concept of biological markers again gains traction (after the premature discarding of the DST).

Ah, the DST (dexamethasone suppression test) reals its lochnessian head again!  I'm not sure why he says it was discarded prematurely.  It was studied exhaustively as for its potential to provide a reliable diagnosis of major depression.  Nobody could figure out how to make it clinically useful.  Although I cannot prove that nobody ever will, I must say that it seems exceedingly doubtful.

Anyway, the impression I get from Dr. Shorter's part of the article, is that he seems to feel that diagnoses need to correspond to his conceptualization of a disease, in order to be useful.  He adapts these criteria to form his idea of a disease:

If we are to regard psychiatric diagnoses as medical diagnoses, how might they be constructed? Building on Guze, Max Fink and Michael Alan Taylor have proposed a 3-fold basis for carving out discrete psychiatric illnesses: describing an illness entity in well-circumscribed psychopathologic terms; verifying its existence with laboratory measures; and, validating it with a distinctive response to treatment.

To be fair, I should say that it would be nice if it were so.  That is, it would be nice if all psychiatric diagnoses had corresponding biological markers, differential response to specific treatments, and statistically cohesive symptom sets.  Alas, this is not the case; moreover, it is not clear whether it ever will be.  So we have choice to make.  We can sit around and do nothing, while awaiting scientific advances that may never come, or we can work with the existing -- albeit inadequate -- models that we already have. 

Let's put this in perspective.  The human brain is the most complex organ that we know.  Indeed, it is the most complex entity that we know of, in the entire Universe.  Sometimes it does not work right.  This is no surprise.  Often, we have difficultly coming up with a good model to describe these dysfunctions.  This also is no surprise.  The models are false.  We know this, because we know that all models are false, even if they are modeling something relatively simple.  In my view, the question is not whether the model is true, but: is it useful? 

There are many uses for a diagnosis.  The most important one, is to provide a link between research and treatment.  Evidence-based medicine is practices by gathering up a bunch of studies that seem to relate to the condition that a specific patient has.  Then, the practitioner assesses how likely it is that the outcomes described in the studies reflect the outcome that is likely to occur if a specific treatment is applied to that patient.  In order to do that, we need to have some way to link the research to the patient.  The diagnosis is that link. 

If the diagnostic scheme fulfills that purpose, then it is useful.  This usefulness transcends any quaint notions about reality.  We all would feel better if we knew that the diagnosis represented something real.  However, the point is not for us to feel better, the point is to find effective ways of linking research to specific patients.  Whether the diagnosis represents something real, is actually beside the point. 

So when Dr. Shorter rants about "a disease that doesn't exist," he is talking about something that we know is true (it is true that the model is false), but which is immaterial (what we care about is whether the model works).