The aim of this study was to replicate the findings of a 1994 study, in which a 30% response rate to venlafaxine was found in patients with treatment-refractory depression, as well as to examine for any predictors of such an outcome. The study was an 8-week, open-label, prospective investigation of venlafaxine in doses up to 300 mg in 312 patients fulfilling criteria for either "absolute" or "relative" treatment resistance. By week 8, 52.6% of the patients had responded, which was defined as a 50% reduction in scores on the Montgomery-Asberg Depression Rating Scale; 49% of those defined with "absolute resistance" demonstrated such an outcome. Forty-five percent of the patients with absolute resistance who had failed to respond to at least one tricyclic antidepressant responded to venlafaxine. Response rates were higher in those with an absence (57.5%) compared with the presence (31.0%) of any comorbid psychiatric disorder (p < 0.001), "marked" (60.3%) compared with "mild or moderate" (51.6%) or "severe" (43.4%) baseline ratings on the patient-rated Clinical Global Impressions Scale (p < 0.05), and "relative" (61%) compared with "absolute" resistance (49%) (p = 0.06). Furthermore, improvement in scores of 20% or 30% at weeks 1 or 2 was associated with higher rates of final response (p < 0.0005). After logistic regression, both comorbid psychiatric illness (p < 0.001) and early improvement (p < 0.0001) remained significant and independent predictors of final response.