In a previous study, of 41 depressed patients who had not responded to fluoxetine 20 mg/day, 53% were treated with high-dose fluoxetine (40-60 mg/ day) and responded (i.e., their 17-item Hamilton Rating Scale for Depression [HAM-D-17] score was <7) versus 29% and 25% of patients treated with fluoxetine plus lithium (300-600 mg/day) or fluoxetine plus desipramine (25-50 mg/day), respectively. We wanted to assess whether these findings could be replicated in a larger sample of depressed outpatients. We identified 101 outpatients with major depressive disorder (52 men and 49 women; mean age, 41.6 + 10.6 years) who were either partial responders (n = 49) or nonresponders (n = 52) to 8 weeks of treatment with fluoxetine 20 mg/ day. These patients were randomized to 4 weeks of double-blind treatment with high-dose fluoxetine (40-60 mg/day), fluoxetine plus lithium (300-600 mg/day), or fluoxetine plus desipramine (25-50 mg/day). In the overall group of patients (N = 101), there was no significant difference in response rates across the three treatment groups (high-dose fluoxetine, 42.4%; fluoxetine plus desipramine, 29.4%; fluoxetine plus lithium, 23.5%). Dropout rates were also comparable, ranging from 9.1% (high-dose fluoxetine) to 14.7% (fluoxetine plus desipramine and fluoxetine plus lithium). There were also no significant differences in response rates across the three treatment groups among partial responders (high-dose fluoxetine, 50.0%; fluoxetine plus desipramine, 33.3%; fluoxetine plus lithium, 33.3%) and nonresponders (high-dose fluoxetine, 35.3%; fluoxetine plus desipramine, 26.3%; fluoxetine plus lithium, 12.5%). At the end of the study, the mean lithium level was 0.37 + 0.15 mEq/L (n = 27; range, 0.1-0.8 mEq/L) among lithium-treated patients, and the mean desipramine level was 104.7 + 58.8 ng/mL (n = 22; range, 25-257 ng/mL). There were no significant relationships between lithium or desipramine blood levels and degree of improvement (as measured by the change in HAM-D-17 score). We found no significant differences in efficacy among these three treatment strategies among patients who had failed to respond adequately to 8 weeks of treatment with fluoxetine 20 mg/day, although the high-fluoxetine group was associated with nonsignificantly higher response rates in both partial responders and nonresponders.