Early improvement of depression severity is considered an important therapeutic goal, predictive of later remission. The present study aimed at testing the hypothesis that plasma concentration might influence the time course of response to paroxetine. Eighty-four patients with a severe depressive episode started paroxetine 20 mg/day, with a possible dose adjustment to 30 mg/day after 2 weeks. Severity of depression (Montgomery-Asberg Depression Rating Scale) was assessed at weeks 0, 2 and 4 for all patients, and every 2 weeks thereafter until discontinuation. Median duration on paroxetine was 6 weeks (range 4-18 weeks). Plasma concentration was measured at steady-state after 2 weeks at 20 mg/day. In a first stage, pattern analysis led to distinguish patients with non-response, non-persistent response, early persistent response (obtained at week 2) and delayed persistent response (week 4 or later). Comparison of patients with (n=29, 34.5%) and without persistent response (n=55, 65.5%) did not reveal any significant difference, whereas focus on patients with persistent response indicated that shorter time to response was significantly associated with shorter duration of current episode (r(S)=0.54, p=0.003) and higher plasma level (r(S)=-0.47, p=0.011). In a second stage, a sigmoid mixed effects model was developed that adequately fitted depression severity versus time profiles among patients with persistent response (n=157 data for 29 patients). Estimated median time to response was 3.2 weeks (range 0.9-6.6). Higher paroxetine concentration, younger age and shorter episode duration were confirmed as significant determinants of a shorter time to response (likelihood ratio tests, p<0.005). The present study supports the hypothesis that higher paroxetine concentration might contribute to hasten relief of depressive symptoms in severely depressed patients.