Site variability in treatment outcome in antidepressant trials

Objective: Site-specific differences in treatment outcome during multisite antidepressant drug trials may contribute to a negative or failed clinical trial. As part of a five-site, long-tern, double-blind, placebo-controlled, relapse prevention trial with fluoxetine in major depression, the authors examined site-specific variability in outcome ratings. Methods: Data from 390 patients with major depression who participated in a 64-week, placebo-controlled trial were retrospectively analyzed. chi(2) Analyses and Kaplan-Meier survival estimates were used to examine site-specific differences in relapse rates during 14 weeks of maintenance treatment following randomization to fluoxetine or placebo after remission with fluoxetine treatment. Results: Results from chi(2) analysis (P<.001) and Kaplan-Meier survival rates (P<.001) following randomization to placebo after 12 weeks of fluoxetine treatment showed a fluoxetine superiority over placebo among all five sites combined. Individually, however, only three of the five sites (60%) were able to distinguish fluoxetine superiority. In contrast, there was no fluoxetine versus placebo difference observed among all five sites after randomization following 26 weeks of fluoxetine treatment (P=.11, Fisher's exact test). However, one site (20%) individually could still distinguish a drug versus placebo difference (P<.05). Limitations: Analyses were performed retrospectively, with individual sites not specifically powered to distinguish a drug-placebo difference. Conclusion: Substantial site differences were observed in the ability to distinguish drug superiority over placebo, and this variability may contribute to a negative or failed clinical drug trial. (C) 2002 Elsevier Science Inc. All rights reserved.