OBJECTIVE The objective of this study was to explore the prevalence of anxious depression in an inpatient population, to describe its clinical and sociodemographic correlates, and to compare treatment outcomes between patients with anxious and nonanxious depression. Furthermore, the efficacy of algorithm-guided treatment versus treatment as usual in patients with anxious versus nonanxious depression was evaluated. METHODS Data were collected on 429 inpatients with the diagnosis of a depressive episode (according to ICD-10) and a score of ≥ or = 15 on the 21-item Hamilton Depression Rating Scale (HDRS-21). The German Algorithm Project, phase 3 (GAP3), was conducted between 2000 and 2005 in 10 psychiatric departments throughout Germany. A baseline HDRS-21 anxiety/somatization factor score of ≥ or = 7 was considered indicative of anxious depression. Remission was defined as an HDRS-21 score or ≤ = 9. To evaluate the efficacy of algorithm-guided treatment, patients were randomly assigned into 3 groups: 2 different treatment algorithms or treatment as usual. RESULTS The prevalence of anxious depression was 49%. Patients with anxious depression were more likely than those with nonanxious depression to be older (mean ± SD = 45.3 ± 12.8 vs 42.9 ± 12.0 years, odds ratio [OR] = 1.02 [95% CI, 1.00-1.03], P = .046), retired (70% vs 30%, OR = 3.09 [95% CI, 1.70-5.62], P = .000), without school qualification (74% vs 26%, OR = 3.11 [95% CI, 1.09-8.83], P = .035), more severely depressed (mean ± SD HDRS-21 score = 20.1 ± 5.0 vs 18.5 ± 4.4, OR = 1.08 [95% CI, 1.03-1.12], P = .001), and more likely to have a longer duration of the current episode (mean ± SD = 20.9 ± 26.2 vs 13.7 ± 14.3 weeks, OR = 1.02 [95% CI, 1.01-1.03], P = .011). Patients with anxious depression were more likely to display a variety of melancholic features. In patients with anxious depression compared to those with nonanxious depression, remission was less likely to be achieved (48.6% vs 61.5%, OR = 0.63 [95% CI, 0.42-0.92], P = .018) and took longer to occur (mean ± SD = 44 ± 3.4 vs 30 ± 2.8 days, HR = 0.65 [95% CI, 0.50-0.85], P = .001). There was no significant interaction with the treatment mode with regard to remission (Wald = 0.20, P = .890). CONCLUSIONS Anxious depression is common in patients diagnosed with depression. The poorer treatment outcome in patients with anxious depression demonstrates the need to address the issue of specific treatment strategies for this subgroup. However, anxious depression has no moderating effect on the efficacy of algorithm-guided treatment. BACKGROUND http://www.germanctr.de/ Identifier: DRKS00000161.