This study explored the variability of zidovudine concentrations with computer simulations and measured concentrations. A one-compartment oral absorption model was selected to characterize zidovudine disposition. Mean (+/-standard deviation) values for the pharmacokinetic parameters were taken from the literature. Five different Monte Carlo simulations (50 each) were performed of zidovudine concentrations following repetitive administration of 100-mg oral doses 6 times/day in patients weighing 45-85 kg. A sixth simulation considered a weight-adjusted regimen. Predicted concentrations were compared with those measured in 30 HIV-infected persons receiving 100 mg/dose. Predicted concentrations 1 hour after 100 mg was administered fell in the range of 0.52-5.18 microM; measured values in 30 patients were 0.54-3.07 microM. This study confirms substantial variability in zidovudine serum concentrations. The simulation study of a weight-adjusted regimen suggests one possibility to reduce this variability. These observations provide a basis to explore dosing strategies that control for pharmacokinetic and perhaps pharmacodynamic sources of interpatient variability.