Open-label studies and randomized clinical trials have suggested that mifepristone may be effective for the treatment of major depression with psychotic features (psychotic depression). A recent study reported a correlation between mifepristone plasma concentration and clinical response. The current study aimed to evaluate the safety and efficacy of mifepristone and, secondarily, to test whether response was significantly greater among patients with mifepristone plasma concentrations above an a priori hypothesized threshold. A total of 433 patients who met criteria for psychotic depression were randomly assigned to receive 7 days of either mifepristone (300, 600, or 1200 mg) or placebo. Response was defined as a 50% reduction in psychotic symptoms on both days 7 and 56. Cochran-Mantel-Haenszel tests compared (1) the proportion of responders among patients assigned mifepristone versus placebo and (2) the proportion of responders among the subset of patients with plasma concentrations greater than 1660 ng/mL versus placebo. Mifepristone was well tolerated at all 3 doses. The proportion of responders randomized to mifepristone did not statistically differ from placebo. Patients with trough mifepristone plasma concentrations greater than 1660 ng/mL were significantly more likely to have a rapid and sustained reduction in psychotic symptoms than those who received placebo. The study failed to demonstrate efficacy on its primary end point. However, the replication of a statistically significant linear association between mifepristone plasma concentration and clinical response indicates that mifepristone at sufficient plasma levels may potentially be effective in rapidly and durably reducing the psychotic symptoms of patients with psychotic depression.