A preliminary methotrexate (MTX) kinetic evaluation following administration of an IV bolus (test-dose) allowed individualization of high-dose infusions (HD-MTX). This approach combined with therapeutic drug monitoring was found to have good performance over a large scale of predicted steady-state levels (Css) (10(-5) to 10(-4) M over 24 to 36 h) corresponding to 17 to 650 mg/h deliveries (root mean squared error : rmse (precision) = 1.54 X 10(-5) M (21.4%) and mean error : me (bias) = 0.043 X 10(-5) M (NS)). However a significative (p less than 0.05) but rather low over-estimation of dosage (me = 7.38 X 10(-5) M (14.8%)) associated to a decrease in the prediction precision (rmse = 13.3 X 10(-5) M (26.6%)) occurred in 5 X 10(-4) M predicted Css over 8 h (970 to 1970 mg/h deliveries). However in a number of cases (6 out of 29) important deviations from predicted Css occurred, implying the need to stop the infusion before 8 h. These results indicated that MTX pharmacokinetics was linear from low test-dose bolus injections to high-dose infusions. This allowed dosage predictions based upon preliminary estimation of MTX clearance and associated to therapeutic drug monitoring during and following infusion.