Clearance of anticancer drugs in children is highly variable, leading to wide variability in the systemic exposure associated with each dosage escalation in Phase I studies. The purpose of this Phase I study was to escalate targeted systemic exposure to paclitaxel, rather than dose, to attenuate the impact of pharmacokinetic variability at each escalation level. Children with recurrent acute leukemias received paclitaxel as 24-h infusions at escalating levels of paclitaxel systemic exposure [i.e., area under the concentration-versus-time curve (AUC)]. Plasma paclitaxel concentrations were measured by high-performance liquid chromatography-UV detection. A Bayesian algorithm using a two-compartment model with saturable distribution and saturable elimination was used to estimate clearance during the first 8 h of infusion; the infusion rate was adjusted 12 h after the start of infusion to achieve the target AUC. Toxicity and evidence of activity were assessed after each course. Six boys and one girl received eight courses of paclitaxel. The target AUC ranged from 31.5 to 45 microM x h. Five of the seven evaluable courses had AUCs between 75% and 125% of the target after adjustment (median, 100.2%; range, 51-151%), whereas none of the seven courses was projected to be in the target range had no change in dose been made (P = 0.021). The ratio of the achieved AUC to the target AUC was inversely related to clearance (r = -0.857; P = 0.014). Mucositis was the exposure-limiting toxicity, at AUCs lower than were expected based on Phase I studies in children with solid tumors. Pharmacokinetically-guided dosing strategies reduced variability in systemic exposure. Alternatives to traditional Phase I studies may be important in the setting of childhood leukemias because these patients show poor tolerance to Phase I therapy.