The Kraepelinian dichotomy

McDonald et al (2005) investigated the Kraepelinian dichotomy of psychosis using brain imaging. They reported distinct grey matter volumetric deficits in patients with schizophrenia and those with psychotic bipolar I disorder but common white matter abnormalities in the two disorders.

Kraepelin distinguished dementia praecox and manic–depressive psychosis on the basis of symptomatology, course and outcome. He wrote that the basic disturbances in dementia praecox were the ‘impoverishment of those feelings and strivings which continually stoke the furnace of our will’ and ‘a loss of the internal integrity of comprehension, emotion and volition’. Furthermore, his description of manic–depressive psychosis included cases of ‘periodic and circular insanity, simple mania, melancholia and affective changes that could be regarded as rudiments of more severe disasters’ (Berner et al, 1992). This formulation is what we would today consider a spectrum concept of manic–depressive illness. A test of the Kraepelinian dichotomy would thus be better served by the use of patients with affective disorders rather than bipolar I disorder (with psychotic symptoms) as the comparator group.

The non-significant differences in grey matter between patients with bipolar I disorder and healthy volunteers could be a result of sampling bias. Recruitment of patients from voluntary support groups might have resulted in inclusion of those with less-severe illness. In addition, depression, anxiety, medical disorders (e.g. hypertension, diabetes mellitus) and seizures, which can give rise to structural abnormalities on magnetic resonance imaging, were not excluded in the ‘healthy volunteers’. The mean IQ and ethnicity of patient groups and the healthy volunteers were not given. These variables are important as they may contribute to differences in brain structure among groups (Thase, 2000). Similarly, the use of spoiled gradient recall echo sequence instead of inversion recovery sequence might have led to type 2 errors in comparisons of white matter volumes between patients with schizophrenia and those with bipolar I disorder (Karson & Renshaw, 2000).

The statistical analysis used the analysis of covariance (ANCOVA) model for differences between each patient group and the healthy volunteer group and differences between the two patient groups. Risk of type 1 errors would have been lower in a single ANCOVA (3 × 2) model.

Finally, it would be interesting to know whether ‘normalisation’ using the International Consortium for Brain Mapping data-set instead of the Talairach space would have made a difference to the results and whether some of the results were confirmed by the ‘region of interest’ methodology, which is known to be more accurate.