Separation of subtypes of depression using discriminant analysis. I. Separation of unipolar endogenous depression from non-endogenous depression.

We derived a discriminant function based on clinical features to classify patients with endogenous depression ('melancholia') and non-endogenous ('neurotic') depression. The difference between the groups was not one of overall severity of illness alone. Bipolar melancholic patients were classified less well than were unipolars, supporting previous findings of clinical differences between these groups. The discriminant function (DF) was reduced to a discriminant index (DI) which classified a separate group of unipolar melancholic and non-endogenous patients with comparable accuracy. Approximately 80 per cent of all cases received a definite classification by the DI. The agreement between the 105 definite DI classifications and the clinical diagnoses was 90 per cent when results from the derivation and validation groups were combined. The DI classification was then validated against an objective biological marker, the dexamethasone suppression test (DST). The diagnostic discriminant index predicted the DST result with the same accuracy as the clinical diagnoses. The discriminant index can serve as an operational definition of the patients diagnosed as endogenous or nonendogenous unipolar depression in future studies by ourselves and other groups of investigators.