The present study was conducted to compare the pharmacokinetics and pharmacodynamics (PD) of paclitaxel between Phase I trials of 3- and 24-h infusions and to determine the most informative pharmacokinetic parameter to describe the PD. Twenty-seven patients were treated in a Phase I study of paclitaxel by a 3-h infusion at one of six doses: 105, 135, 180, 210, 240, and 270 mg/m2. Pharmacokinetic data were obtained from all patients. Paclitaxel concentrations were measured in the plasma and urine using HPLC. The pharmacokinetics and PD were compared with those of a Phase I trial of paclitaxel by a 24-h schedule previously performed. The maximum tolerated dose of paclitaxel by a 3-h infusion was determined to be 240 mg/m2. The major toxicities were granulocytopenia, neuromuscular toxicities, and hypotension. Apparent differences in pharmacodynamic relationships for the change in granulocytes with dose, peak concentration, and areas under the concentration versus time curve were observed between the 3- and 24-h schedules. However, the relationship between the duration of plasma concentration above 0.05 microM and the change in granulocytes could be fitted to the same sigmoid maximum effect model in either schedule (P < 0.01). There were no clear relationships between peripheral neuropathy or hypotension and any pharmacokinetic parameters. The pharmacokinetics and PD of paclitaxel were schedule dependent. The duration of plasma concentration above 0.05 microM could be a common pharmacokinetic parameter predicting granulocytopenia for both schedules.