Utility without validity is useless

A. Jablensky's paper1 raises important questions of many kinds. In this commentary, due to space constraints, I will not consider the evidence regarding clinical course as a diagnostic validator in schizophrenia or the overlap between psychosis and mood conditions, nor the nosological views of K. Jaspers, except to note that other interpretations exist that would not agree with Jablensky's perspective. The main focus here will be instead on whether the concept of utility can or should be the basis of psychiatric nosology.

The central assumption in Jablensky's paper is the statement, made in passing, that medical classifications have as their primary purpose pragmatic needs, and only secondarily generation of new knowledge. Here is the heart of the validity versus utility debate. However, there is another way of thinking about the matter. Almost a century ago, A. Lewis2 noted: “Classifications may be useful for the wrong ends… The clinician may never come to see how vicious are the uses to which he has been, contentedly, putting his classification”. Reversing the DSM/ICD view, Lewis held that nosology had to be “valid and useful”. If invalid, a nosology is not useful. He concluded: “A valid classification is one which is not only useful, but useful for sound medical and scientific ends”. Put another way, the primary source for diagnostic classification should be our best scientific knowledge, i.e., classification should be valid scientifically, first and foremost, and also clinically relevant. Only secondarily, in rare cases, can purely utilitarian diagnosis be justified when there is compelling clinical need but zero scientific evidence. DSM/ICD reverses the terms, with hundreds of scientifically unjustified utilitarian diagnoses, versus only a dozen or two with some scientific bases.

This is the kernel of the problem: should validity be central to the diagnostic process, or can we just give up on it, and happily celebrate utility?

To answer this question, let's go back for a history lesson. The original justification for the radical changes of DSM‐III in 1980 was that it represented a common language, providing “reliability” and utility. This was not the final goal, though. The claim was made repeatedly that this reliability/utility would be a way‐station to validity3. In other words, we would get to validity more effectively by having a reliable common language. We would change this language with further scientific research, each revision of DSM moving gradually closer to validity. However, as Jablensky admits, the DSM project has failed to achieve validity. And now we are told that we should change our goal to pure utility, an attempt to make a virtue out of defeat.

Recent debates around DSM‐5 have exposed some ideas which previously were expressed mainly behind closed doors. We learned that our DSM leaders have important post‐modernist assumptions: they have given up entirely on the whole concept of validity4.

Contrary to initial DSM‐III claims about achieving gradual validity in the future, we now have 40 years of the converse experience. The DSM‐IV and 5 leadership stated very explicitly to their task forces that they should make as few changes as possible4. This is an anti‐scientific attitude. Scientists do not make and test hypotheses by saying to themselves: “Now, let's make as few changes to prior beliefs as possible”. DSM classification is now a pure paean to utility, entirely “pragmatic”, in the worst meaning of the term: an extremist utilitarianism that has no purpose other than to reflect the wishes and beliefs of the American Psychiatric Association or DSM leadership, or the loudest interest groups. This statement is documented by historians who have reviewed internal DSM documents5, 6.

Besides its basic anti‐scientific attitude, DSM revisions have used higher and higher thresholds for making changes based on research, making it harder to move toward empirically‐based validity. Call it the “Sisyphus problem”: researchers obtain data, rolling the boulder of knowledge up the hill of ignorance; then DSM leaders say it is not good enough. Another generation of researchers adds to that knowledge, and, if their results pass the DSM task force itself, they are vetoed in the American Psychiatric Association by the Scientific Review Committee, or the Board of Trustees.

We have an unimpeachable example of this Sisyphus problem in the work of the great psychiatric researcher J. Angst. For a century, ever since Kraepelin, the standard view in world psychiatry was that it did not matter if patients were manic (bipolar) or depressed (unipolar), but rather that all mood episodes reflected the same single manic‐depressive illness. Angst's Zürich cohort, collected in the early 1960s, suggested that bipolar and unipolar groups differentiated on diagnostic validators of course and genetics7. Hence the underappreciated radical anti‐Kraepelinian change in DSM‐III: the creation of bipolar disorder and major depressive disorder out of Kraepelin's concept of manic‐depressive illness. In the intervening decades, with over 40 years of more data, Angst now finds that his Zürich cohort does not differentiate well into bipolar and unipolar based on course and other diagnostic validators8. The same Zürich dataset, now even more valid with complete prospective follow‐up of the entire lifetime of its subjects, is rejected by the DSM‐5 task force. What was considered acceptable to make very radical changes in the 1970s is now rejected decades later for even minor changes (like duration of hypomania or definition of mixed states). Much more radical changes in the past were made with much less science.

There is not even a utilitarian justification for this resistance. DSM‐5 field trials now indicate that, after four decades, major depressive disorder has poor reliability9, even worse than in the past. Our current nosology of major depression is both false and useless.

Angst, being a true scientist, falsifies his own hypotheses, something the DSM/ICD leadership has been unwilling to do, which brings us to the most baneful consequence of the rejection of science/validity in favor of pragmatism/utility: because of DSM/ICD, all research, both clinical and biological, is doomed to failure. This self‐fulfilling prophecy is then used by DSM advocates of pragmatism/utility to justify further their rejection of science‐based classification. We reach a dead end in obtaining further new knowledge precisely because obtaining new knowledge is “secondary” to the pragmatism that ensures that no new knowledge will be achieved. Psychiatric progress never occurs, because it cannot occur with these anti‐scientific attitudes.

To state it otherwise: DSM/ICD is a “social construction”. That's what the concept of utility means. It is created for social – professional, insurance, forensic, economic, ideological, political, cultural – purposes. It is not, as admitted by Jablensky, primarily based on scientific research. The fact that DSM/ICD is a social construction reflects its underlying philosophy, post‐modernism10.

If we create diagnostic categories based on social, economic and political considerations, why should genes correlate with those categories? Why should neuroanatomy correlate with wishes for insurance reimbursement? When DSM/ICD phenotypes for biological studies are purely social constructions, it should be no surprise that hardly any major genes/biomarkers for DSM/ICD diagnoses are identified. Four decades of failure in DSM‐based research are hard to ignore. Recent change in the U.S. National Institute of Mental Health (NIMH) policy, such that DSM criteria are no longer acceptable for research11, is an institutional verification that an emphasis on utility actually prevents ever achieving validity.

Because DSM failed, one should not conclude, as the NIMH leadership does, that the whole clinical research project failed. In fact, because of DSM pragmatism, clinical research has not been the main basis of our diagnostic system for 40 years. Let us now not draw the false conclusion that clinical research into psychiatric diagnosis has failed, when instead it has been ignored.

Nor will it do to resort to prayer – wishing for a gene, or a brain circuit, that will someday, somehow, split the Red Sea. The gene/biological marker miracle will never happen as long as DSM/ICD fails to put science first12.

The explicitly vague term “disorder” reflects post‐modernist cynicism about the disease concept13. The attempt to base “disorder” definitions on functional impairment and severity of symptoms is not conceptually, biologically, or scientifically sound. There are many medical diseases that do not cause functional impairment (such as silent cancers), or involve mild rather than severe symptoms. Some medical diseases even are associated with some benefits, rather than only harms (e.g., decreased malaria risk with sickle cell trait). The extremist DSM/ICD ideology of rejecting mild symptoms does not solve the “false positives problem” nor improve predictive values of diagnosis14. Instead, it feeds into, and perhaps reflects, stigma against mental illness, an ironic result of DSM/ICD “pragmatism”, understandable as another baneful effect of cultural post‐modernism.

In sum, my main critique is that a primarily utilitarian approach, in the end, is not useful, because it matters – cultural post‐modernist assumptions notwithstanding – whether we are really right or wrong, i.e., whether our diagnoses are valid. In clinical medicine, where lives are in the balance and where scientific values are accepted, any other view is difficult to defend.

S. Nassir Ghaemi
Mood Disorders Program, Tufts Medical Center, Boston, MA, USA