Categorical and/or continuous? Learning from vascular surgery

R. Krueger and his impressive international team of co‐authors offer a clear and comprehensive review of current issues in dimensional approaches to classifying psychopathology1. They make a challenging case for the advantages of this approach, as embodied in their Hierarchical Taxonomy of Psychopathology (HiTOP) model, over the categorical classifications widely adopted in psychiatry.

The issues as such are not new. They were widely discussed in the 1960s and 1970s. The British psychiatrist and epidemiologist R. Kendell covered much the same ground in his now classic book The Role of Diagnosis in Psychiatry 2. Then as now the question was whether psychopathology could be “cut at the joints” into distinct categories or whether it was better described quantitatively along one or more dimensions of continuous change. Then as now the answer depended in part on the kind of psychopathology in question and in part on the statistical methods adopted. Then as now opinion remained divided largely along professional lines, with psychiatrists favouring categorical and psychologists favouring dimensional approaches (Krueger, like a majority of his co‐authors, is a psychologist).

There are, certainly, as Krueger and his colleagues point out, new factors in play, some positive, others negative. On the positive side, there have been important methodological advances. Statistical methods have progressed dramatically with developments in computing science. Formal logic, too, has a novel role to play: the British philosopher and psychologist P. Koralus’ semantic modelling of decision making, for example, offers potentially exciting applications to psychopathology3. On the negative side, fifty years of experience with symptom‐based psychopathological categories have been disappointingly thin on aetiological insights. The promise of early 20th century advances (with discoveries such as neurosyphilis and Alzheimer's disease) remains, despite all the power of contemporary neuroscience, largely unfulfilled.

Should we then be persuaded by Krueger et al's case that categorical classifications of psychopathology should be abandoned in favour of dimensional description? Experience from other areas of medicine suggests that we should not.

Vascular surgery offers a case in point. As a relatively new specialty (the Vascular Surgical Society of Great Britain and Ireland was founded in 1966), vascular surgery adopted from the start an explicitly evidence‐based approach and remains strongly research‐led. In this respect, its predominantly categorical classification of disease entities has (as in most other areas of bodily medicine) served it well. Where psychiatry has suffered fifty years of frustration, vascular surgery has made significant and sustained progress in understanding the pathophysiology of a whole range of categorically‐defined disorders, ranging from aortic aneurysm to varicose veins, with corresponding advances in both surgical and non‐surgical management options.

So far so good then, it would seem, for traditional disease entities. However, closer inspection shows that, while the objects of scientific interest in vascular surgery are indeed categorically defined disorders, the science of vascular surgery has been in many instances dimensional in character. Progress in the treatment of aortic aneurysm, for example, has depended critically on quantitative studies of the relative risk of death respectively from vascular surgery and from aneurysm rupture. The key variable in these studies is the diameter of the aneurysm. The risk of rupture increases as the aneurysm expands. In most people this happens slowly, and international guidelines recommend annual monitoring until the diameter of the patient's aneurysm reaches five and a half centimeters, this being the point at which the risk of rupture within the next twelve months (5%) is sufficient to justify the risks of surgery4.

Vascular surgical science has thus made progress by combining categorical with dimensional approaches. Similar combined approaches continue to be adopted in ongoing research on the management of aortic aneurysm. The object of interest remains the categorically defined disease entity “aortic aneurysm”; the key variables remain the essentially dimensional variables of relative risk.

Psychopathology, it is true, is different from and in certain respects more complex than vascular pathology. There are, for example, no counterparts in vascular pathology of the conceptual challenges presented by comorbidity in psychopathology (reflected in the difficulties described by Krueger et al in establishing a stable hierarchical structure for their dimensional approach). Comorbidities are, of course, common in vascular pathology, but the requisite divisions and distinctions are largely unproblematic. Similarly unproblematic in vascular pathology are criteria of functioning. Descriptively similar experiences of voice hearing, for example, may be for one person functionally impairing and for another empowering5, 6. A swelling aorta, by contrast, is a functionally impaired aorta for anyone.

Such differences, though, make the example of vascular pathology more rather than less pertinent for psychopathology. If progress in vascular pathology has been achieved with a combined categorical and dimensional approach, it is at the very least likely that a similar approach will be needed if progress is to be made with the more complex challenges of psychopathology. The point, anyway, is general. All sciences make progress through quantification. But progress through quantification has usually been by way of addition, not substitution. This is evident throughout the medical and biological sciences. It is evident, too, in physics, surely the paradigm of a successful quantitative science (think of wave/particle dualism in quantum mechanics). So why should psychopathology be any different?

Krueger et al might reply: “because this is where the science leads”. In the opening paragraphs of their paper, they claim in support of their HiTOP model the high ground of empirical science, contrasting this with what they describe as the received authority of the DSM. But this is tendentious. The scientific basis specifically of DSM‐5 has indeed been widely criticized7. But the criticism is precisely that DSM‐5 has departed from the explicitly evidence‐based principles on which earlier revisions of the DSM (and ICD) were based. Notably, the Research Domain Criteria project, although bracketed by Krueger et al with DSM‐5, was in fact inspired by much the same aims as HiTOP for a return to empiricism in psychopathological research8.

We should thus welcome the advances in quantification of psychopathology described by Krueger et al. But we should welcome these advances as adding to rather than displacing categorical classifications as the basis of psychopathological science. More will be required for effective translation of psychopathological science into practice. In vascular surgery, translation has required teamwork rather than competition between professionals, and attention to values as well as evidence9. But, as to the science, the example of vascular surgery suggests that it is time for a change of conjunction. For fifty years the focus of debate in psychopathology has been “categorical or continuous”. The example of vascular surgery suggests that its time to think instead “categorical and continuous”.