OBJECTIVE--A comprehensive meta-analysis of clinical trial data was performed to assess the possible association of fluoxetine and suicidality (suicidal acts and ideation). DESIGN--Retrospective analysis of pooled data from 17 double blind clinical trials in patients with major depressive disorder comparing fluoxetine (n = 1765) with a tricyclic antidepressant (n = 731) or placebo (n = 569), or both. MAIN OUTCOME MEASURES--Multiple data sources were searched to identify patients with suicidal acts. Suicidal ideation was assessed with item 3 of the Hamilton depression rating scale, which systematically rates suicidality. Emergence of substantial suicidal ideation was defined as a change in the rating of this item from 0 or 1 at baseline to 3 or 4 during double blind treatment; worsening was defined as any increase from baseline; improvement was defined as a decrease from baseline at the last visit during the treatment. RESULTS--Suicidal acts did not differ significantly in comparisons of fluoxetine with placebo (0.2% v 0.2%, p = 0.494, Mantel-Haenszel adjusted incidence difference) and with tricyclic antidepressants (0.7% v 0.4%, p = 0.419). The pooled incidence of suicidal acts was 0.3% for fluoxetine, 0.2% for placebo, and 0.4% for tricyclic antidepressants, and fluoxetine did not differ significantly from either placebo (p = 0.533, Pearson's chi 2) or tricyclic antidepressants (p = 0.789). Suicidal ideation emerged marginally significantly less often with fluoxetine than with placebo (0.9% v 2.6%, p = 0.094) and numerically less often than with tricyclic antidepressants (1.7% v 3.6%, p = 0.102). The pooled incidence of emergence of substantial suicidal ideation was 1.2% for fluoxetine, 2.6% for placebo, and 3.6% for tricyclic antidepressants. The incidence was significantly lower with fluoxetine than with placebo (p = 0.042) and tricyclic antidepressants (p = 0.001). Any degree of worsening of suicidal ideation was similar with fluoxetine and placebo (15.4% v 17.9%, p = 0.196) and with fluoxetine and tricyclic antidepressants (15.6% v 16.3%, p = 0.793). The pooled incidence of worsening of suicidal ideation was 15.3% for fluoxetine, 17.9% for placebo, and 16.3% for tricyclic antidepressants. The incidence did not differ significantly with fluoxetine and placebo (p = 0.141) or tricyclic antidepressants (p = 0.542). Suicidal ideation improved significantly more with fluoxetine than with placebo (72.0% v 54.8%, p less than 0.001) and was similar to the improvement with tricyclic antidepressants (72.5% v 69.8%, p = 0.294). The pooled incidence of improvement of suicidal ideation was 72.2% for fluoxetine, 54.8% for placebo, and 69.8% for tricyclic antidepressants. The incidence with fluoxetine was significantly greater than with placebo (p less than 0.001) and did not differ from that with tricyclic antidepressants (p = 0.296). CONCLUSIONS--Data from these trials do not show that fluoxetine is associated with an increased risk of suicidal acts or emergence of substantial suicidal thoughts among depressed patients.