Population pharmacokinetics of methotrexate (MTX) was evaluated from intravenous test-dose (TD) data (n = 20 corresponding to 174 measured samples). Bayesian prediction of MTX clearance from TD experiments combining population data with measured levels (at times 0.5 and 6 h) was found to be feasible in routine situations with good performance (root mean squared error : rmse (precision) = 1.14 1.h-1 (11.2%) and mean error : me (bias) = 0.06 1.h-1 (NS) relatively to weighted least-square estimates, n = 50). The precision of Bayesian prediction was comparable to that of the model independent which is used in routine practice and involves 9 measured levels over 30 h, (rmse = 1.35 1.h-1 (10.9%), n = 50). However, the routine method presented a significative bias (me = -0.81 1.h-1, n = 50).