The effect of fluoxetine (FLU) in posttraumatic stress disorder was studied in a one-year trial. Subjects received open-label treatment for 6 months, followed by double-blind randomized treatment with FLU or placebo (PBO) for 6 months. Rates of relapse were compared using the Clinical Global Impressions of Improvement. One hundred twenty-three subjects entered open-label treatment, of whom 114 returned at least once. Sixty-two subjects were randomized to receive FLU or PBO, of whom 57 returned at least once and were analyzed. The dose of FLU ranged from 10 to 60 mg/d; at randomization, mean doses were 48.6 and 42.1 mg for FLU and PBO groups. Rates of relapse were 22% for FLU versus 50% for PBO (P = 0.02), and time to relapse on FLU was longer than for PBO (P = 0.02, log-rank statistic). The odds ratio for relapse on PBO relative to FLU was 3.50. No significant differences were found on other measures. Fluoxetine was well tolerated during double-blind treatment.