Limited sampling formulas and bayesian estimation for mycophenolic acid 12 hours Area Under the concentration-time Curve prediction in Stable renal transplant recipients co-medicated with cyclosporine or sirolimus
F T Musuamba (1), J L Bosmans(2), J J Senessael (3), J Cumps (1), P Wallemacq (4) and R K Verbeeck (1)
(1)School of Pharmacy, Catholic University of Louvain, Brussels, Belgium, (2) Department of Nephrology and Hypertension, University Hospital, University of Antwerp, Edegem, Belgium, (3) Renal Unit, Academic Hospital, Free University of Brussels, Brussels, Belgium, (4) Laboratory of Clinical Biochemistry, University Hospital St. Luc, Catholic University of Louvain, Brussels, Belgium.
Background: Mycophenolate mofetil (MMF), the prodrug of mycophenolic acid (MPA), is an immunosuppressive agent used in combination with corticosteroids, calcineurin inhibitors or sirolimus for the prevention of acute rejection after solid organ transplantation. Mycphenolic acid glucuronide (MPAG), the major metabolite of MPA is subject to enterohepatic recycling. Controversy remains about the interaction between MMF and other immunosuppressive drugs. To date, MPA pharmacokinetic (PK) analysis of stable transplant recipients treated with sirolimus as co-meditation has only been reported on two clinical trials including 12 and 11 patients.
MPA area under the MPA plasma concentration-time profile during one dosing interval (AUC0-12), rather than trough concentrations is being considered as the best exposure marker. To estimate an individual patient's AUC0-12 without measuring the full MPA plasma concentration-time profile two different methods can be used. The limited sampling formulas (LSF) based on multiple linear regression models using a small number of blood samples, preferably obtained in the early post-dose period, to predict the full AUC0-12. This approach requires strict adherence to the time of blood sample collection. Maximum a priori (MAP) Bayesian estimation of AUC0-12 for each individual patient is also based on a limited number of plasma concentration measurements in the early post-dose period and is more flexible in blood sample timing, but in addition requires population pharmacokinetic data being available for the drug .
The objectives of the present study were
- to identify and model the effect of demographic and routine biochemistry factors and of the immunosuppressive co-drugs on MPA PK variability by using nonlinear mixed-effect modelling techniques,
- to predict MPA AUC0-12 by using multiple linear regression models (limited sampling formula) and MAP Bayesian estimation methods ,
- to assess the robustness of various previously reported limited sampling strategies 3 (LSS) by testing them on our patient sample.
Methods:
Data from 40 stable adult renal allograft recipients, transplanted in one of two Belgian university hospitals (Free University of Brussels and University of Antwerp) were included in this study. All patients received mycophenolate mofetil (MMF) (0.75 g b.i.d.), cyclosporine and steroids, all per os, during the initial post transplantation period. At 7.4 ± 1.4 months, cyclosporine was replaced by sirolimus while continuing MMF (0.75 g b.i.d.) and steroid treatment. Full pharmacokinetic profiles for MMF during one dosing interval were determined on three different occasions: A) the day before switching from cyclosporine to sirolimus at 7.4 ± 1.4 months (N=40), B) at 60 days after the switch (N=39 ), and C) at 270 days after the switch (N=37 ).
AUC0-12 was estimated by using the linear trapezoidal method (Noncompartmental Analysis, WinNonlin® version 5.01, Pharsight, Mountainview CA, USA).
Nonlinear mixed effects modelling was performed by using NONMEM Version VI and VNM a Windows®-based interface to NONMEM containing graphical and statistical tools. The sample (N=40) was randomly split in two groups: 1) a model building subgroup comprising 27 patients, and 2) a validation subgroup of the remaining 13 patients which was also used for Bayesian estimation. The first-order conditional estimation (FOCE) approach with interaction between parameters was used throughout the entire modelling process. Between- and within-patient variability was modelled with exponential error models. The difference between observed serum concentrations and the corresponding model-predicted serum concentrations was estimated with a mixed error model. Various models were tested:They were first fitted separetely to MPA and MPAG plasma concentrations and in a second time simultaneously to MPA and MPAG plasma concentrations. Bayesian estimation on the validation group by the NONMEM "posthoc" subroutine was performed by using the final model based on different combinations of 3 MPA concentration-time points sampled within 2 hours following MMF dosing.
Limited sampling formulas were developed to predict MPA AUC0-12 by using various combinations of three MPA serum concentrations determined during the 2-hour interval following MMF dosing. Multiple linear regression analyses were performed using JMPTM software to correlate predicted MPA AUC0-12 values with MPA AUC0-12 values calculated by using the full pharmacokinetic profiles . Repeated cross-validation was used to evaluate each LSS as described by Pawinski 1.
Predicted AUC0-12 from each model was compared to the observed AUC0-12 by linear regression to evaluate the strength of the relationship between the AUC0-12 values predicted by the various LSS and the observed AUC0-12 values. Predictive performance of the various LSS and agreement between predicted and observed AUC0-12 were assessed as described by Sheiner and Beal 2.
Results:
The data were best fitted by a four-compartment model fitting MPA and MPAG plasma concentrations and was significantly improved by introduction of a rate constant describing transfer from the fourth to the first compartment and describing the enterohepatic cycle. Glomerular filtration rate as described by Nankivell significantly influenced the MPAG elimination constant whereas hepatic transaminases significantly influenced the transfer rate constant from the MPA central to the MPAG central compartment. Weight was significantly correlated to MPA central compartment volume of distribution.
AUC0-12 was best predicted by MAP using different combinations of patients samples within the 2-hours following MMF intake. The best LSF involved samples drawn 0.66, 1.25 and 3 hours after drug intake.
None of the previously published models acceptably fitted our data
Conclusions:
-These are the first results on population pharmacokinetic analysis of MPA/MPAG in patients co-medicated with sirolimus.
- These results show that there is still a large inter-individual and inter-occasion variability in MPA pharmacokinetics and thus, a need of TDM long term (15 months) after transplantation.
- LSS using Bayesian estimation were better performing than those using MLR due in part to the fact that the population model used are more physiological and take into account the large variability in absorption and the enterohepatic recycling.
References:
[1] Sheiner, L.B. & Beal, S.L. Some suggestions for measuring predictive performance. J. Pharmacokinet. Biopharm. 9, 503-512 (1981).
[2] Pawinski, T. et al. Limited sampling strategy for the estimation of mycophenolic acid area under the curve in adult renal transplant patients treated with concomitant tacrolimus. Clin. Chem. 48, 1497-1504 (2002).
[3] Staatz, C.E. and Tett S.E. Clinical Pharmacokinetics and Pharmacodynamics of Mycophenolate in solid organ transplant recipients. Clin. Pharmacokinet. 46 (1) 13-58 (2007).
