Objective: To predict the clearance and toxicity of high-dose methotrexate (HDMTX) based on serum concentration of methotrexate (MTX) at different time points in children with acute lymphoblastic leukemia (ALL). Methods: Blood samples were collected from 198 children with acute lymphoblastic leukemia treated with HDMTX (303 infusions) at different time points. MTX levels at different time points were determined by modified enzyme amplification immunoassay. Correlation between MTX levels and toxicity was evaluated by receiver operating characteristic (ROC) curve. The clinical data of another 22 children were extracted to verify the accuracy and specificity of the prediction model. Results: The upper limit of the optimal concentration of 42 h MTX level for predicting delayed clearance was 1.50 μmol/L

with a sensitivity of 82.17%; the upper limit of the optimal concentration for predicting the safety threshold was 0.76 μmol/L

with a sensitivity of 90.78%. The upper limit of the optimal concentration of 66 h MTX level for predicting delayed clearance was 0.50 μmol/L

with a sensitivity of 89.09%; the upper limit of the optimal concentration for predicting non-toxicity was 0.10 μmol/L

with a sensitivity of 92.73%. The accuracy of 42 h MTX level in predicting delayed clearance was 80.60% in 22 children with model validation. The accuracy of predicting the safety threshold was 85.78% and the specificity was 91.50%. The accuracy of 66 h MTX level in predicting toxicity was 89.18%. The accuracy and specificity for predicting non-toxicity were respectively 87.24% and 92.20%. Conclusion: MTX levels can be accurately measured at various time points up to 66 h post-infusion

with 42 h MTX level being used for prediction of delayed clearance and 66 h MTX level being used for prediction of toxicity.