Kraepelin distinctions back in favor

Oct 11 2013

By Eleanor McDermid, Senior medwireNews Reporter

Modern statistical methods support Kraepelin’s original distinction between schizophrenia and bipolar disorder, but not the more recent schizoaffective disorder category, say researchers.

Roman Kotov (Stony Brook University, New York, USA) and colleagues used nonlinear modeling to identify natural distinctions produced by associations between symptoms over 4 years and outcome measures at 10 years in 413 patients with psychotic disorders.

The results for nonaffective psychotic symptoms did not support a linear model, and, on the whole, supported a kraepelinian two-category model over a DSM-IV three-category model.

Based on their findings, the team contends that continued use of the schizoaffective disorder diagnosis “is difficult to justify,” and suggests that psychosis and mood disorder, when both present, should be coded as separate comorbidities.

This would “align the nosology with the natural taxonomy of psychoses, simplify diagnosis, and improve its reliability,” write Kotov et al in JAMA Psychiatry.

However, editorialist Kenneth Kendler (Virginia Commonwealth University, Richmond, USA) adds a note of caution, saying that “the results do not, to my eye, permit as confident a distinction between [the models] as argued by the authors.”

The results do clearly exclude a linear relationship between nonaffective psychosis (psychosis occurring separately from a mood episode) and outcomes. When assessed using locally weighted scatterplot smoothing, outcomes worsened as nonaffective psychosis time climbed from 0% to 40%, with a particularly steep drop up to 20%, but higher percentages had no further effect.

Kendler says this supports the clinical intuition “that psychosis occurring within vs outside of major mood syndromes reflects distinct psychopathologic processes.”

When the researchers used spline regression modeling to examine the associations in more detail, they found that most of the four fit indices used (eg, the Bayesian information criterion) favored the kraepelinian model for functional outcomes; however, three indices supported the DSM-IV model for symptom outcomes.

The spline regression results also identified a cutoff of just 1.5% of time spent in nonaffective psychosis as distinguishing between mood disorders and schizophrenia, but Kendler says this does not appear to fit with the findings from locally weighted scatterplot smoothing. He agrees that the data do not generally support the schizoaffective disorder category, but neither do they completely exclude it.

Kendler says the findings “remind us that classification of psychiatric disorders needs to be viewed and evaluated from multiple perspectives.”

He adds: “Although one diagnostic formulation might best correlate with genetic findings, another might optimally predict outcome or treatment response.”

Licensed from medwireNews with permission from Springer Healthcare Ltd. ©Springer Healthcare Ltd. All rights reserved. Neither of these parties endorse or recommend any commercial products, services, or equipment.