The non-linear plasma pharmacokinetics of paclitaxel in patients has been well established, however, the exact underlying mechanism remains to be elucidated. We have previously shown that the non-linear plasma pharmacokinetics of paclitaxel in mice results from Cremophor EL. To investigate whether Cremophor EL also plays a role in the non-linear pharmacokinetics of paclitaxel in patients, we have established its pharmacokinetics in patients receiving paclitaxel by 3-, 24- or 96-h intravenous infusion. The pharmacokinetics of Cremophor EL itself was non-linear as the clearance (Cl) in the 3-h schedules was significantly lower than when using the longer 24- or 96-h infusions (Cl175-3 h = 42.8+/-24.9 ml h(-1) m(-2); CI175-24 h = 79.7+/-24.3; P = 0.035 and Cl135-3 h = 44.1+/-21.8 ml h(-1) m(-1); Cl140-96 h = 211.8+/-32.0; P < 0.001). Consequently, the maximum plasma levels were much higher (0.62%) in the 3-h infusions than when using longer infusion durations. By using an in vitro equilibrium assay and determination in plasma ultrafiltrate we have established that the fraction of unbound paclitaxel in plasma is inversely related with the Cremophor EL level. Despite its relatively low molecular weight, no Cremophor EL was found in the ultrafiltrate fraction. Our results strongly suggest that entrapment of paclitaxel in plasma by Cremophor EL, probably by inclusion in micelles, is the cause of the apparent nonlinear plasma pharmacokinetics of paclitaxel. This mechanism of a (pseudo-)non-linearity contrasts previous postulations about saturable distribution and elimination kinetics and means that we must re-evaluate previous assumptions on pharmacokinetics-pharmacodynamics relationships.