Background. We sought to develop a clinically useful subtyping system for the non-melancholic depressive disorders, and here we assess one weighted to central aetiological factors.

Methods. We studied 185 patients meeting DSM-III-R and/or clinical criteria for non-melancholic depression. Data were obtained by self-report, interview of patients and from corroborative witnesses. We developed a set of variables for class definition, assessing: (i) ‘P’, disordered personality as a vulnerability factor; (ii) ‘A’, meeting criteria for a lifetime anxiety disorder or positive on probe questions about trait anxiety characteristics, so assessing anxiety as a vulnerability factor; and (iii) ‘L’, psychiatrist and consensually-rated life event stress prior to depression onset.

Results. A latent class analysis generated a four-class solution for the P–A–L variables. Life event stressors had similar item probabilities across all four classes, and did not influence the four-class ‘P–A’ solution when deleted from the analysis, suggesting that life event stress may act more as a general provoking agent, rather than constituting any distinct ‘reactive’ or ‘situational’ depression class. Three classes generated clinically meaningful groupings, reflecting varying contributions of anxiety and disordered personality functioning, and with evidence of differential outcome over the following 12 months.

Conclusions: We suggest that a refined aetiologically-weighted model may assist definition of the non-melancholic depressive disorders, and provide the logic for exploring the comparative utility of differing treatments to identified vulnerability-based classes.