The efficacy and safety of fluoxetine and desipramine were compared in a 6-week double-blind, parallel group study of patients with major depression. Twenty-five were studied while hospitalized for treatment, and 33 were studied as outpatients. Improvement on the Hamilton Rating Scale for Depression was significant for both treatments from week 1 through the end of the study and did not differ between the two treatments at any week. Overall, 64% of fluoxetine-treated patients and 68% of desipramine-treated patients had at least a 50% reduction in Hamilton Depression score. We assessed whether improvement relatively early in treatment was predictive of categorical response at 6 weeks. Among fluoxetine-treated patients, but not desipramine-treated patients, the week 3 change in the Hamilton Depression mood item was significantly predictive of the response at 6 weeks. Patients treated with fluoxetine had significantly fewer side effects than those treated with desipramine. Desipramine, but not fluoxetine, caused a persistent increase in heart rate. The results suggest that early signs of response to fluoxetine are not dependent on achieving steady-state levels of the drug.