Purpose: PSC 833 (valspodar) is a second-generation P-glycoprotein (Pgp) antagonist developed to reverse multidrug resistance. We conducted a phase I study of a 7-day oral administration of PSC 833 in combination with paclitaxel, administered as a 96-hour continuous infusion. Patients and Methods: Fifty patients with advanced cancer were enrolled onto the trial. PSC 833 was administered orally for 7 days, beginning 72 hours before the start of the paclitaxel infusion. Paclitaxel dose reductions were planned because of the pharmacokinetic interactions known to occur with PSC 833. Results: In combination with PSC 833, maximum-tolerated doses were defined as paclitaxel 13.1 mg/m2/d continuous intravenous infusion (CIVI) for 4 days without filgrastim, and paclitaxel 17.5 mg/m2/d CIVI for 4 days with filgrastim support. Dose-limiting toxicity for the combination was neutropenia. Statistical analysis of cohorts revealed similar mean steady-state concentrations (Cpss) and areas under the concentration-versus-time curve (AUCs) when patients received paclitaxel doses of 13.1 or 17.5 mg/m2/d for 4 days with PSC 833, as when they received a paclitaxel dose of 35 mg/m2/d for 4 days without PSC 833. However, the effect of PSC 833 on paclitaxel pharmacokinetics varied greatly among individual patients, although a surrogate assay using CD56+ cells suggested inhibition of Pgp was complete or nearly complete at low concentrations of PSC 833. Responses occurred in three of four patients with non-small-cell lung cancer, and clinical benefit occurred in five of 10 patients with ovarian carcinoma. Conclusion: PSC 833 in combination with paclitaxel can be administered safely to patients provided the paclitaxel dose is reduced to compensate for the pharmacokinetic interaction. Surrogate studies with CD56+ cells indicate that the maximum-tolerated dose for PSC 833 gives serum levels much higher than those required to block Pgp. The variability in paclitaxel pharmacokinetics, despite complete inhibition of Pgp in the surrogate assay, suggests that other mechanisms, most likely related to P450, contribute to the pharmacokinetic interaction. Future development of combinations such as this should include strategies to predict pharmacokinetics of the chemotherapeutic agent. This in turn will facilitate dosing to achieve comparable CPss and AUCs.