The relative efficacy of the various classes of antidepressants has not been established. Observational studies in naturalistic settings are important in evaluating treatment outcomes with antidepressants, since controlled clinical trials include only a minority of patients present in clinical practice. This study sought to evaluate in a naturalistic setting the treatment outcomes of dosulepin and venlafaxine for patients with depressive episodes. At the university hospital in Copenhagen, Denmark, between 1998 and early 2001, the first-line treatment for psychiatric inpatients with depression was dosulepin; after that time, venlafaxine was the first-line medication. We compared the treatment outcomes among inpatients during the respective periods. There was no significant difference in the primary outcome parameters between the two groups. A tendency in favour of dosulepin confirmed by a post-hoc analysis suggested that the failure to achieve significant difference was related to a type 2 error. However, missing data and possible confounders related to the different treatment periods weaken the results. This naturalistic study showed a non-significant trend for poorer treatment outcomes (probably because of an underpowered design) after replacing dosulepin with venlafaxine as first-line drug for depression in a naturalistic inpatient setting.