Duloxetine (DLX) is a dual serotonin and norepinephrine reuptake inhibitor that has been recently approved for the treatment of major depressive disorder (MDD). However, little is known about the relationship between DLX plasma levels and clinical response. The aims of this open-label study were 1) to assess clinical outcome and tolerability of DLX by means of clinician and patient assessments and 2) to evaluate the value of plasma DLX levels as predictors of clinical response and tolerability. This was a naturalistic, open-label study of 45 outpatients affected with MDD (16 men and 29 women), who received DLX at doses of 30-120 mg/day and were evaluated at baseline (T0) and after 2, 4 and 12 weeks (T1-3). The assessments included the Hamilton Rating Scales for Depression (HRSD) and Anxiety (HRSA), Clinical Global Impression-Severity (CGI-S), Beck's Depression Inventory (BDI) and a mood visual analogue scale (VAS). Compared with T0, there were significant improvements in HRSD at T1, T2 and T3 (P < 0.001), in HRSA, CGI-S and the self-administered BDI at T2 and T3 (P < 0.001), and in the VAS scores shown at T3 (P = 0.01). DLX treatment was safe and well tolerated. Plasma DLX levels at T2 ranged from 5 to 135 ng/mL (mean +/- SD = 53.56 +/- 39.45) and correlated almost significantly with the DLX dose (r = 0.35; P = 0.069). There was a significant curvilinear quadratic relationship between the improvement of HRSA scores and plasma DLX levels (R(2) = 0.27; P = 0.02). The incidence of anxiety or irritability was associated with the highest plasma levels. Our findings suggest that monitoring plasma DLX levels may be helpful in predicting better treatment responses and tolerability. The present data seem to suggest an optimal anxiolytic efficacy of DLX at intermediate plasma levels.