Predictors of relapse in a prospective study of fluoxetine treatment of major depression.

Loss of response to a previously effective antidepressant is a common clinical problem. Retrospective analyses have shown that the pattern of response during antidepressant treatment (late onset and persistent versus other patterns) can be used to predict relapse during continuation and maintenance treatment and possibly to identify placebo responses to treatment. This study was designed to test the predictive value of response pattern prospectively and to examine the data for other predictors of relapse. Five hundred seventy persons with major depressive disorder were treated with fluoxetine for 12 weeks and their pattern of response was determined. Those who responded (N=292) underwent random assignment, under double-blind conditions, to continue taking fluoxetine or to switch to placebo for 52 weeks or until relapse. Survival analysis was used to examine the effect of covariates on relapse. Although fluoxetine was significantly more effective than placebo during maintenance treatment, this chronically ill group had a high rate of relapse. Contrary to previous findings, a pattern of acute response was not predictive of relapse. Chronicity, symptom severity, a neurovegetative symptom pattern, and female gender were all associated with a significantly greater risk of relapse, with no difference observed between fluoxetine and placebo. The pattern of response to acute treatment appears to be inconsistently predictive of relapse. There is a high rate of relapse with both active medication and placebo in patients with chronic depression. Illness characteristics predict loss of response both to fluoxetine and to placebo. No variable examined was predictive of differential relapse rates between fluoxetine and placebo.