The link between virological response and exposure to zidovudine was studied in 40 HIV-infected patients of the protocol ANRS 01. During this 45-day trial, the patients received only oral zidovudine in six treatment groups. Our objectives were: to analyze and model the pharmacokinetics of zidovudine and the decrease of P24 antigenemia; to study the links between exposure and efficacy. For the pharmacokinetic study, 12 blood samples were collected from 0.16 to 24 h after the first dose and a compartmental model was used. For the pharmacodynamic study of P24 antigenemia, blood samples were collected before treatment and every 3 days until day 45; an exponantial decay model was used. The pharmacokinetic and pharmacodynamic parameters were estimated for each patient by nonlinear regression. The correlations between efficacy parameters and exposure parameters, were then studied in the 40 patients. The mean (+/- SD) apparent volume of distribution and clearance were 151 L (+/- 94) and 184 L/h (+/- 72), respectively. The mean initial antigen level was 472 pg/mL (+/- 409), the coefficient of reduction of antigenemia was 0.27 (+/- 0.21) and the rate of decrease was 0.27/day (+/- 0.16). The coefficient of P24 reduction was found to be significantly correlated to the daily area under the curve (P < 0.0014). This relationship was adequately described by an Imax model and the daily area under the curve, leading to 50% of antigenemia decrease, was estimated to be 2.32 mg x h/L (+/- 0.33). In conclusion, a significant relationship between exposure to zidovudine at day 1, and decrease of P24 antigenemia was found. It was estimated that the average steady-state concentration, which corresponds to 70% of maximal efficacy, was 0.22 mg/L. Together with the large interpatient variability of zidovudine pharmacokinetics, these findings confirmed that zidovudine should be monitored and a clinical target concentration was defined.