This multinational, randomized, double-blind, flexible-dose study evaluated the short- and long-term antidepressant tolerability and efficacy of escitalopram and paroxetine. Tolerability was assessed by monitoring adverse events throughout the study, and discontinuation events during brief treatment interruption and tapered withdrawal. Discontinuation-emergent effects were evaluated in two separate double-blind periods. First, to mimic the consequences of non-compliance, patients were randomized to one of two treatment interruption periods (placebo-substitution for 3–5 days). Second, patients were randomized to a 1–2-week tapered withdrawal period randomly scheduled between weeks 28 and 31. The pre-specified primary efficacy endpoint was the mean change from baseline in total Montgomery–Åsberg Depression Rating Scale (MADRS) score at week 8, using the principle of last observation carried forward. A total of 323 patients entered 8 weeks of double-blind treatment and received at least one flexible dose of escitalopram (10–20 mg/day) or paroxetine (20–40 mg/day). Patients who demonstrated evidence of a significant clinical improvement (Clinical Global Impression–Improvement of 1 or 2) at week 8 entered a 19-week, double-blind maintenance period during which they were treated with the same dose they received at week 8, followed by a 1–2-week tapered withdrawal period. A total of 89 patients (28%) withdrew during the study; significantly (P<0.01) more patients withdrew from the paroxetine group (34%) than from the escitalopram group (21%), and significantly (P<0.05) more paroxetine patients withdrew due to lack of efficacy. The mean MADRS total score improved for both treatment groups from baseline to week 8, with no statistical difference between groups. In severely depressed patients (baseline MADRS total score ≥30), escitalopram was superior (P<0.05) to paroxetine at week 27 (end of maintenance treatment). There was a high prevalence of sexual dysfunction at baseline: the mean Arizona Sexual Experience Scale (ASEX) score was approximately 20 points in both treatment groups. Mean total ASEX scores increased slightly above baseline values during the acute period and declined slightly below baseline values towards the end of the maintenance period. During taper and cessation of treatment, patients in the paroxetine group demonstrated significantly more discontinuation symptoms relative to escitalopram based on the Discontinuation Emergent Signs and Symptoms scores.