A multisite trial of mifepristone for the treatment of psychotic depression: A site-by-treatment interaction

Major Depression with Psychotic Features (psychotic depression) is a common, debilitating psychiatric disease. We hypothesized that mifepristone, a cortisol receptor (GRII) antagonist, would significantly reduce psychotic symptoms in psychotic depression. Two hundred fifty-eight patients with psychotic depression enrolled at 29 sites were randomized to mifepristone or placebo for 7 days. The primary outcome was rapid and sustained response, defined as a 50% or greater decrease in Brief Psychiatric Rating Scale Positive Symptom Subscale scores at the end of treatment (day7) and 49 days later (day 56). Cochran-Mantel-Haenszel compared proportions of responders to mifepristone versus placebo adjusting for site. Exploratory analyses compared response of patients with mifepristone plasma concentrations of >= 1800 ng/ml to placebo. The primary endpoint was not statistically significant. However, the Breslow-Day test indicated a statistically significant site-by-treatment interaction. Mifepristone produced significantly higher response among the twenty sites who participated from the trial onset (p<.05), whereas no difference was observed at the nine sites added late in the trial. Patients with mifepristone plasma levels >= 1800 ng/ml were significantly more likely to respond than placebo patients (Intent-to-Treat: OR=2.4, p=.03; Initial 20 sites: OR=4.1, p=.002). The results of this trial are instructive in two respects. First, while statistical adjustment site are common in multisite clinical trials, this study reminds trialists to formally evaluate the interaction of site-by-treatment. Second, the association between increased mifepristone plasma concentration levels and greater clinical response, detected despite the site-by-treatment interaction, suggests that higher plasma levels may be needed for maximizing the probability of a positive response. (C) 2009 Elsevier Inc. All rights reserved.