Résumé : Objective: To investigate whether subtypes of depression predict differential outcomes of treatment with selective serotonin-reuptake inhibitor (SSRI) and a tricyclic antidepressant in major depression. Method: Among 811 adults with moderate-to-severe depression, melancholic, atypical, anxious and anxious-somatizing depression subtypes established at baseline were evaluated as predictors of outcome of treatment with flexible dosage of the SSRI escitalopram or the tricyclic antidepressant nortriptyline. The primary outcome measure was the Montgomery-Åsberg Depression Rating Scale (MADRS). Secondary outcome measures were the 17-item Hamilton Rating Scale for Depression (HRSD-17) and the Beck Depression Inventory (BDI). Results: Melancholic depression was associated with slightly worse outcomes among individuals treated with escitalopram, but did not affect outcome of treatment with nortriptyline. The interaction between melancholic depression and drug did not reach statistical significance for the primary outcome measure and significant results for secondary outcome measures were not robust in sensitivity analyses. Atypical depression was unrelated to outcome of treatment with either antidepressant. Anxious and anxious-somatizing depression did not predict outcome on the primary measure, but inconsistently predicted worse outcome in some secondary analyses. Limitations: Some participants were non-randomly allocated to drug. Therefore, drug-by-predictor interactions had to be validated in sensitivity analyses restricted to the 468 randomly allocated individuals. Conclusions: Melancholic, atypical or anxious depression, are not sufficiently robust differential predictors of outcome to help clinician choose between SSRI and tricyclic antidepressants. There is a need to investigate other predictors of outcome. © 2011 Elsevier B.V. All rights reserved.