Paroxetine is characterized by large interindividual pharmacokinetic variability and heterogeneous response patterns. The present study investigates plasma concentration and therapeutic response to paroxetine for the influence of age, sex, and CYP2D6 and ABCB1 polymorphisms, the latter gene encoding for the permeability glycoprotein. Genotyping for CYP2D6 (alleles *3, *4, *5, *6, and *xN) and ABCB1 polymorphisms (61A>G, 2677G>T, and 3435C>T) was performed in 71 depressed patients who started 20 mg paroxetine per day and had plasma concentration measured after 2 weeks at a fixed dose. A dose increase to 30 mg per day was possible starting at week 2. For 63 patients, severity of depression (Montgomery-Asberg Depression Rating Scale) was assessed at weeks 0, 2, and 4 and every 2 weeks thereafter until discontinuation. Persistent response was defined as 50% improvement from baseline score sustained from the first occurrence to study end point. Paroxetine concentration significantly differed between female and male patients (median, 28 versus 16 ng/mL; P = 0.001). Differences were not significant between CYP2D6 heterozygous and homozygous extensive metabolizers (median, 27 versus 22 ng/mL; P = 0.074) and between ABCB1 genotypes (P > 0.10). When considered in a multivariate model, CYP2D6 heterozygous extensive metabolizer phenotype (P = 0.062) and female gender (P = 0.001) predicted 1.3-fold and 1.6-fold higher paroxetine concentration, respectively, but fraction of explained variability was modest (21%). Frequency of persistent response at study end point did not significantly differ according to CYP2D6 heterozygous extensive metabolizer versus homozygous extensive metabolizer phenotype and ABCB1 polymorphisms in univariate analyses. After adjusting for age, sex, paroxetine concentration at week 2, and daily dose at study end point, ABCB1 genotype contributed to improving the model significantly for 61A>G (P = 0.043), but not 2677G>T (P = 0.068) and 3435C>T (P = 0.11). None of two poor metabolizers and four ultrarapid metabolizers showed persistent response to paroxetine. The hypothesis that permeability glycoprotein activity might be a relevant predictor of therapeutic response deserves to be further investigated while controlling for pharmacokinetic variability.