In randomized controlled trials, maintenance treatment for relapse prevention has been proven to be efficacious in patients responding in acute treatment, its efficacy in long-term outcome in "real-world patients" has yet to be proven. Three-year long-term data from a large naturalistic multisite follow-up were presented. Severe relapse was defined as suicide, severe suicide attempt, or rehospitalization. Next to relapse rates, possible risk factors including antidepressant medication were identified using univariate generalized log-rank tests and multivariate Cox proportional hazards model for time to severe relapse. Overall data of 458 patients were available for analysis. Of all patients, 155 (33.6%) experienced at least one severe relapse during the 3-year follow-up. The following variables were associated with a shorter time to a severe relapse in univariate and multivariate analyses: multiple hospitalizations, presence of avoidant personality disorder, continuing antipsychotic medication, and no further antidepressant treatment. In comparison with other studies, the observed rate of severe relapse during 3-year period is rather low. This is one of the first reports demonstrating a beneficial effect of long-term antidepressant medication on severe relapse rates in naturalistic patients. Concomitant antipsychotic medication may be a proxy marker for treatment resistant and psychotic depression.