An End to Kraepelinian Nosology?

Kraepelin's dichotomy between manic-depressive illness and dementia praecox has formed the foundation of our current nosology of the functional psychoses, but its validity has been controversial for many years.¹ What light do neuropsychiatric data shed?

Kraepelin himself had little doubt that schizophrenia (his dementia praecox) was a brain disease: “We are concerned here with a palpable pathological process in the brain”² (p. 154). Moreover, he believed that he was initiating the replacement of descriptive psychiatry by a scientific nosology of disease entities, in accord with the accepted notions of clinical-pathological correlation that had established the foundations of modern medicine earlier in the 19th century.³ He fully expected that subsequent research would reveal the nature of the pathological processes, and he used disease course as a proxy.⁴ Decades after his systematic textbook descriptions of dementia praecox and manic-depressive illness, he recognized that “it is becoming increasingly clear that we cannot distinguish satisfactorily between these two illnesses and this brings home the suspicion that our formulation of the problem may be incorrect.” Yet even then he insisted that “we must at all costs adhere to the basic difference between the disease processes concerned”⁵ (p. 28).

But are dementia praecox and manic-depressive illness distinct disease processes? It must first be noted that psychiatric diagnosis in general developed under the pressure of social factors unrelated to neurobiology.⁶ Severe forms of mental disorder were geographically and professionally segregated from milder forms because of their social consequences, with the result of obscuring possible continuities in underlying biology and alternative ways of clustering cases. Prognosis and troublesomeness were crucial for asylum decision-makers, neurobiology of no practical importance.⁷

Discrimination between two forms of functional psychosis cannot be soundly based on symptom clusters, as Kendell and Gourlay showed many years ago.⁸ Recent work confirms that the phenomenology of psychosis simply does not sort itself out into schizophrenia on the one hand and mood disorder on the other.⁹ Phenomenological data are supplemented by genetic epidemiological studies, which have been extensive and diverse. A former partisan of the Kraepelinian dichotomy concluded from the recent evidence: “From a familial/genetic perspective, there is no sharp division between schizophrenia and other psychotic disorders”¹⁰ (p. 54). Berrettini¹¹ showed that for several of the loci linked by compelling evidence to one of the functional psychoses, strong evidence exists for linkage to the other; he concluded that “both family and molecular studies of these disorders suggest shared genetic susceptibility” (p. 245).

Neurodevelopmental precursors to both schizophrenic and affective psychosis include prenatal famine¹² and exposure to influenza,¹³ winter birth,¹⁴ obstetrical complications,¹⁵ minor physical anomalies,¹⁶ dermatoglyphic anomalies,¹⁷ and anomalous social and cognitive development.¹⁸ The evidence from pathophysiology and neuropathology is limited by the absence of psychotic control subjects from many studies, but some abnormalities may be shared between supposedly distinct groups.^19,20 The neuroimaging data suggestive of shared pathology, along with the neurodevelopmental and genetic data, are discussed in the Neuropsychiatric Practice and Opinion article by Curtis, van Os, and Murray²¹ that appears in this issue of the Journal. Although the data supporting shared risk are not all replicated and compelling, evidence for specificity of any biological factor is scant.

The issue is not whether Kraepelin described clinical syndromes with exquisite clarity. This he accomplished, even if the evidentiary basis of his research is questionable and the reasons for its domination of nosology for a century are the subject of little historical investigation.^22,23 The question is whether these syndromes represent biological natural kinds, disease entities with specific pathological correlates, such that the century-long search for the “causes of schizophrenia” is a sensible project. If not, no shame accrues to Kraepelin, who was operating in the absence of modern statistical and technological methods and of current neuroscientific knowledge. In my opinion, current data, including the neuropsychiatric data, indicate that the psychotic disorders are best understood not categorically but dimensionally,^24–27 comprising parameters that vary continuously and can appear in various combinations—a “mix-and-match” nosology. From these combinations can certainly emerge the classic clinical pictures described by Kraepelin and seen in daily practice. The claim made here is not that schizophrenia and psychotic mood disorder are the same, with the same antecedents, or that they differ unidimensionally on a spectrum of severity. To the contrary, the analysis of specific symptomatic elements distinguishing various clinical states will be critical to advances in understanding of pathophysiology. But clinging to the idea, outmoded by biological data, that the classic clinical pictures represent distinct disease entities has been and continues to be a brake on progress. It is past time to abandon it.

A lesson from the classification of other neuropsychiatric illnesses is that only biological understanding—not clinical syndromes and not even neuropathology—can ultimately organize the data. For example, a single mutation of the prion protein gene can produce varying clinical pictures and variable pathological findings within the same family.²⁸ Similar findings arise from the study of mutations of the tau gene producing frontotemporal dementia.²⁹ Obviously, such biological understanding is not yet available to guide a classification of the functional psychoses. Research to bring it into existence must focus broadly on psychotic disorders;³⁰ no study of the neurobiology of schizophrenia is complete without psychotic controls. Clinicians must move away from traditional but empty debates about whether patients “really” have schizophrenia or mood disorder—debates that often amount to trying to fit a peg with both round and square features into a square or a round hole, at great cost in clinical richness. The challenge is to devise diagnostic dimensions that have heuristic value for neuroscientific research and that can also guide clinical understanding and intervention.