OBJECTIVE: To investigate patients treated for depression with respect to steady-state concentration of venlafaxine enantiomers, to quantify within- and between-subject variability and to study the possible influence of individual characteristics such as gender and age. METHODS: Thirty-five inpatients received venlafaxine orally at a fixed 300-mg daily dose. Blood samples were taken on day 14 and day 28 for therapeutic drug monitoring purposes. All measurements reflected steady-state trough values. In a first stage, plasma concentrations of racemic venlafaxine (V) and O-desmethylvenlafaxine (ODV) were measured using a gas chromatography method. In a second stage, (+)/(-) enantiomeric ratios for V and ODV were determined using a stereoselective capillary electrophoresis method. RESULTS: Interindividual variability was 77% and 33% for concentrations of racemic V and ODV, respectively. Intraindividual variability was below 20% for both compounds. Enantiomeric ratios did not statistically differ from unity, with median (+)/(-) ratios of 1.14 for V and 0.97 for ODV. ODV/V metabolite formation ratios for the (+) and (-) enantiomers did not significantly differ from each other (median values 2.85 and 2.37, respectively). However, reduced ODV/V ratio for the (-) enantiomer was strongly associated with decreased (+)/(-) ratio for V (r(S)=0.71, P<0.001) and increased (+)/(-) ratio for ODV (r(S)=-0.79, P<0.001). In contrast, ODV/V ratio for the (+) enantiomer did not significantly correlate with parent compound (+)/(-) ratio and correlated only weakly with metabolite (+)/(-) ratio (r(S)=-0.38, P<0.05). When compared with males, females displayed a significantly lower ODV/V ratio for the (-) enantiomer (median values 1.42 vs 5.08 on day 14, P<0.05) but not for the (+) enantiomer (median values 2.36 vs 3.27, n.s.). Analysis did not reveal any significant association between ODV/V ratios and age, weight, height, creatinine clearance, smoking or co-medication. A pharmacokinetic model at steady state was developed that postulated two different enzyme systems to contribute to O-desmethylation. ODV(-) formation was supposed to largely depend on a single pathway, possibly impaired in a patient subpopulation. ODV(+) formation was postulated to rely on both pathways to a similar extent. Model predictions were in close agreement with observations in patients. CONCLUSION: Observations, together with model-based simulations, suggested that marked stereoselectivity in a patient subgroup may be related with impairment of O-desmethylation greater for (-) than (+) venlafaxine. This hypothesis requires testing against phenotypic and genotypic characteristics of patients.