The relationship between zidovudine phosphorylation inside mononuclear cells and plasma zidovudine pharmacokinetic was assessed in six subjects. Plasma and intracellular concentrations were measured by radioimmunoassay over an 8-h period after administration of 100 or 200 mg of zidovudine. Plasma pharmacokinetics followed expected patterns, with considerable interpatient variability in area under the concentration-versus-time curve (AUC), and a terminal half-life of 1.5 h. Intracellular AUC was even more variable than plasma AUC, but the data suggested a crude linear relationship between these parameters. The intracellular half-life of 3.5 h was consistently longer than the plasma half-life, and varied little between patients. The prolonged intracellular half-life suggested that total phosphorylated zidovudine, as measured by the method described, is not greatly dominated by the 5'-monophosphate as predicted from the in vitro studies reported in the literature. Plasma concentrations of zidovudine have shown little correlation with clinical effect. Study of the relationship between phosphorylated zidovudine and clinical outcome could lead to a more effective management of therapy.