Sertraline and fluoxetine have different pharmacologic and pharmacokinetic profiles which may be of clinical relevance in the determination of response in different subtypes of depression. A randomized, double-blind, 6-week study comparing sertraline (50-100 mg/day) with fluoxetine (20-40 mg/day) in 286 outpatients with major depression, who had demonstrated comparable efficacy and tolerability for the two drugs, was analysed by subgroups of patients at baseline with melancholia, severe depression, single depressive episode, multiple depressive episodes, high anxiety, low anxiety, psychomotor retardation and psychomotor agitation. Multiple logistic regression with regressors including treatment-by-subgroup variables revealed that, within certain subgroups, the efficacy might differ substantially from that of the whole treatment group. However, the only treatment-by-subgroup interaction term that was significant was anxiety (P < 0.05). There was no evidence of interaction in single or recurrent episode subgroups, and these were not included in subsequent analyses. Subsequent two-sample statistical comparison tests of response (i.e. Hamilton Depression Scale reduction > or = 50%) rates at study endpoint between treatment groups demonstrated that patients with melancholic depression and those with symptoms of psychomotor agitation yielded a significantly greater proportion of responders with sertraline compared to fluoxetine (P < 0.05). Response rates in sertraline- and fluoxetine-treated patients, respectively, were: overall study 59%, 51%; melancholia 59%, 44%; severe depression 59%, 41%; low anxiety 71%, 55%; high anxiety 47%, 48%; psychomotor retardation, 48%, 46%; and psychomotor agitation 62%, 39%. Multiple logistic regression adjusting for possible confounding factors, that included a treatment by anxiety interaction term, also led to similar findings. In particular, the analysis showed that significant differences existed in favour of sertraline in patients with low anxiety in the melancholia and severe depression subgroups (P < 0.05), indicating that these characteristics predicted a superior response to 6 weeks of treatment with sertraline relative to fluoxetine. Sertraline also demonstrated advantages over fluoxetine on parameters such as sleep and weight disturbance in severely depressed patients, and sleep disturbance, weight, cognitive disturbance and retardation in melancholic patients.