Population pharmacokinetics and Bayesian estimation of mycophenolate mofetil in paediatric renal transplant recipients using a non-parametric approach.
Prémaud, A(1), Van Guilder M (2), Armstrong VW (3), Weber LT(3), Ghandi A (2), Oellerich M (3), Tonshoff B (3), Jelliffe R (2), Marquet P (1), Rousseau A (1)
(1) Faculty of Medicine, Limoges France, (2) Laboratory of Applied Pharmacokinetics, University of Southern California School of Medicine (2), Department of Clinical Chemistry, Georg-August University, Gottingen Germany
Introduction: Recent studies found that therapeutic drug monitoring of mycophenolic acid (MPA) could improve patient outcome. There is little information in the paediatric population and no population pharmacokinetic (popPK) analyses dedicated to children only.
Objectives: To investigate the popPK of MPA in paediatric renal transplant patients, using a non-parametric approach and to develop a Bayesian estimator able to reliably estimate individual PK parameters and AUC. Methods: Twenty eight patients aged 3.4 to 16.3 were investigated three and six months after renal transplantation. They received an immunosuppressive regimen including cyclosporine, prednisone and mycophenolate mofetil. The patients were randomly divided in two groups: an index group of 20 patients and a test group including 8 patients. In each patient, nine blood samples were collected (immediately prior to dosing, then at 0.33, 0.67, 1.25, 2, 4, 6 , 8 and 12 H after dosing) and months 1 and 3 post-grafting. The parametric Iterative Two-stage Bayesian Population Model program was used, followed by the nonparametric adaptive grid (NPAG) maximum likelihood program (implemented in BigWinpops software), to determine popPK parameter values of MPA in the index group. These population parameters were used as a priori information for Bayesian prediction in the test group using 3 or 4 blood samples. Prediction error (MPE) and root mean square prediction error (RMSE) were calculated for AUC.
Results: A 2-compartment model with first order absorption and a lag-time best described the pharmacokinetics of MPA. The standard deviation (SD) of the assay was 0.963 + 0.0227*C +0.0008*C² where C is the MPA concentration in mg/L. The following NPAG means were obtained: K12 = 3.43 h-1, K20= 0.99 h-1, K23 =0.95 h-1, K32=0.16 h-1, Vc/F=11.5 L, TLAG, 0.29 h. Inter-individual variability was 80%, 33%, 51%, 151%, 47% and 75%, respectively. Bayesian forecasting allowed prediction of MPA AUC0-12h using a combination of three sampling times (0;67, 1.25 and 4 hours post-dose), with a significant mean bias of 7.92 % (RMSE 28 %). Using a combination of 4 sampling times (0.67, 1.25, 2 and 4 hours post-dose) led to better results with a non-significant mean bias =-3.5 % (range -36 to 13%) and an acceptable precision (RMSE 16 %); only 2/14 AUC biases were out of the range 20%.
Conclusion: This study confirms both the high interindividual variability of MPA PK and the need for a lag-time to describe MPA absorption profiles. It shows the ability of the non-parametric approach to accurately describe MPA population pharmacokinetics. Interestingly, using a non-compartimental approach, E Jacqz-Aigrain et al (Pediatr Nephrol 2000) reported apparent clearance values in accordance with ours.