The prediction of chemotherapeutic efficacy is complicated by "protocol dependencies" in dose-effect and dose-toxicity relationships. It has been proposed that pharmacokinetic-pharmacodynamic mathematical models may allow characterization of chemotherapeutic protocol dependencies, and may facilitate the prediction of chemotherapeutic efficacy; however, few demonstrations exist in the literature. The present study examines the pharmacokinetics and toxicodynamics of methotrexate (MTX), a commonly used anticancer agent, after intraperitoneal (i.p.) administration to mice. MTX was administered via bolus or infusion (24, 72, and 168 h), at doses of 2.5-1000 mg/kg. MTX plasma and peritoneal pharmacokinetics were characterized through standard noncompartmental and compartmental techniques. Body weight loss was used as a measure of MTX-induced toxicity. We found that MTX pharmacokinetics were independent of dose (over a range of 3-600 mg/kg) and independent of dosing mode (i.e., i.p. bolus vs. i.p. infusion). However, MTX-induced toxicity was shown to be highly dependent on the dosing protocol used. For example, the maximally tolerated dose (i.e., the dose related to a mean body weight loss of 10%) was 200-fold greater after bolus administration relative to that observed for 72-h infusion (760 mg/kg vs. 3.8 mg/kg). This profound protocol dependence in the relationship between MTX-induced toxicity and MTX exposure was characterized through the use of a time-dissociated pharmacokinetic-pharmacodynamic model (median prediction error: 3.9%).