DSM Issues: Incorporation of Biological Tests, Avoidance of Reification, and an Approach to the “Box Canyon Problem”

I chaired the Scientific Review Committee for DSM-5 (1). In that capacity, I obtained an in-depth exposure to the questions and controversies that arose during the development of DSM-5. Based on that experience, I would like to review three issues that I see confronting our nosology as we move forward and beyond DSM-5.

First, while we all want to move from a descriptive to an etiologically based nosology, it is not clear how best to accomplish this. This issue was reviewed by Charney et al. at the very beginning of the DSM-5 process (2). In 2002, they concluded that “the field of psychiatry has thus far failed to identify a single neurobiological phenotypic marker or gene that is useful in making a diagnosis of a major psychiatric disorder” (p. 33).

How should we approach this issue now, in 2014? One option is to await dramatic research breakthroughs in imaging or genetics that will permit a revolutionary change that will enable us to shift in one fell swoop from signs and symptoms to key DNA variants, imaging findings, or neurochemical markers that will define our disorders etiologically. I am skeptical of this approach because all the accumulating evidence suggests that we will not have one big discovery of the etiology of psychiatric disorders. Rather, for most, if not all of our disorders, we will have a series of small discoveries of important parts of complex multifactorial puzzles.

I would advocate a more modest evolutionary approach. In future revisions of our nosology, we should consider adding specific laboratory or biological tests. However, these measures would have to compete, in terms of their impact on diagnostic reliability and validity, with our current descriptive criteria. It should be an even playing field. We would add a new biological criterion for the same reasons as we might add a new symptom or sign. Furthermore, this approach avoids the deep problem of choosing a priori a single etiological level on which to focus (3). Rather, we can pragmatically consider individual findings from multiple levels including single or aggregate molecular genetic markers, biochemical findings, imaging findings, or neuropsychological findings. They just have to help improve the performance of our diagnoses.

This process has already started. In DSM-5, criterion B2 for narcolepsy is reduced levels of CSF hypocretin, and criterion B3 is specific findings from nocturnal sleep polysomnography. This year saw publication of a major advance in gene discovery in schizophrenia, with 108 independent loci (4). The effect sizes of all of them were quite small, but when combined into a genetic risk profile and applied to three new genotyped populations, the odds ratios for schizophrenia in the highest compared with the lowest decile ranged from 8 to 20. Would the addition of this measure to our other criteria for schizophrenia improve diagnostic validity?

There will be concerns about specificity, as a number of genetic and neurobiological markers for psychiatric disorders appear to have an impact on risk for several psychiatric disorders. But we have already confronted this problem at a symptom level. On its own, insomnia is nonspecific, but when it occurs with sad mood, reduced appetite, and guilt, it has proven to be a useful marker for depressive illness.

Finally, diagnostic approaches that utilize a combination of historical information, symptoms, signs, and laboratory tests of a variety of different kinds are common across the rest of medicine. If firmly based in empirical findings, we should welcome them into psychiatry.

The second issue I want to address is the “reification problem.” We all know that the current DSM diagnostic criteria have been selected because they usefully index our disorders. Many other symptoms and signs could have been used, but the ones selected, often by clinical tradition rather than careful scientific selection, have been shown to perform well. However, in many ways, we act as if the DSM criteria constitute our disorders. We teach our residents to focus their evaluations largely or solely on assessment of DSM criteria. Our clinical and research structured interviews assess only DSM criteria. Our treatment algorithms utilize DSM categories evaluated by DSM criteria. Our large epidemiological surveys typically assess only the diagnostic criteria for each DSM category. We act as if the DSM criteria are all that really matter when it comes to evaluating psychiatric illness.

In our teaching, we should consistently remind trainees that DSM criteria are a good place to start for a diagnostic evaluation, but that examining only these symptoms and signs would lead to an impoverished view of psychopathology. In our clinical work, we should remember to explore the diversity of the pathological experiences of our patients, with the DSM criteria being only a good jumping off point for further inquiry. For our research, we should insist on sampling beyond the DSM criteria to capture more of the wide domains of symptomatology for the disorders we are examining. After all, how can we possibly improve on our current criteria if our large data sets are restricted to these limited sets of signs and symptoms?

My third point is that we need to further ponder what might be called the “box canyon problem” in psychiatric nosology. An evolutionary approach to our nosology—trying to improve incrementally upon our current criteria and judging improvements against specified validators—has a good chance of working if the current criteria are in the right ball park (5). But what if they are not? What if they are so deeply flawed that small improvements cannot fix the problem—that is, they are stuck in a box canyon with no easy way out? In this case, real improvement of the criteria would require a deeper, more revolutionary change—a “reboot.” Some might advocate that we launch such a change by consensus-based conceptual analyses. That is, we can think our way out of the box canyon, developing a different and superior theoretical framework for our set of disorders. I oppose this approach. As we move toward a more empirically grounded nosological system, where small changes have to be vetted through validating measures, it is not sensible to use the older, more consensus-based approach to these bigger changes. While much more effortful, an empirically rigorous approach challenges researchers and clinicians to develop alternative formulations of their diagnoses that may differ radically from the DSM system in use. They then need to show their empirical superiority to the “boxed-in” DSM system before advocating for a wholesale transformation of the criteria.