Population pharmacokinetics of mycophenolic acid in children and young people undergoing blood or marrow and solid organ transplantation

Abstract

AIMS

To characterize the population pharmacokinetics of mycophenolic acid (MPA) and evaluate dose regimens using a simulation approach and accepted therapeutic drug monitoring targets in children and young people undergoing blood or marrow, kidney and liver transplantation.

METHODS

MPA concentration–time data were collected using an age specific sampling protocol over 12 h. Some patients provided randomly timed but accurately recorded blood samples. Total and unbound MPA were measured by HPLC. NONMEM was employed to analyze MPA pharmacokinetic data. Simulations (n = 1000) were conducted to assess the suitability of the MPA dose regimens to maintain total MPA AUC(0,12 h) within the range 30 and 60 mg l⁻¹ h associated with optimal outcome.

RESULTS

A two-compartment pharmacokinetic model with first-order elimination best described MPA concentration–time data. Population mean estimates of MPA clearance, inter-compartmental clearance, volumes of distribution in the central and peripheral compartments, absorption rate constant and bioavailability were 6.42 l h⁻¹, 3.74 l h⁻¹, 7.24 l, 16.8 l, 0.39 h⁻¹ and 0.48, respectively. Inclusion of bodyweight and concomitant ciclosporin reduced the inter-individual variability in CL from 54.3% to 31.6%. Children with a bodyweight of 10 kg receiving standard MPA dose regimens achieve an MPA AUC below the target range suggesting they may be at a greater risk of acute rejection.

CONCLUSIONS

The population pharmacokinetic model for MPA can be used to explore dosing guidelines for safe and effective immunotherapy in children and young people undergoing transplantation.

WHAT IS ALREADY KNOWN ABOUT THIS PROJECT?

• Mycophenolate mofetil (MMF) is an immunosuppressant drug used for the treatment and prevention of graft vs. host disease in blood or marrow transplantation and acute graft rejection in solid organ transplantation.

• Mycophenolic acid (MPA) pharmacokinetics have not been thoroughly studied in paediatric blood or marrow transplant recipients and guidance for optimal dosing of mycophenolic acid in children is lacking.

• Mycophenolic acid exhibits considerable inter- and intra-patient pharmacokinetic variability in adults and paediatric transplant recipients.

• The AUC of mycophenolic acid over a 12 h dose interval at steady-state is generally agreed to be the most reliable metric associated with the risk of acute rejection.

• Population pharmacokinetic analysis can utilize concentration information from both intensive sampling and sparse sampling to provide pharmacokinetic parameter estimates, estimates of inter- individual and intra-individual variability in these parameters and allows patient characteristics explaining inter-individual variability to be quantified.

THIS STUDY ADDS

• This study is one of the first investigations in which a population pharmacokinetic modelling approach was applied to assess the pharmacokinetics of both intravenous and oral MMF in children and young people undergoing blood or marrow and solid organ transplantation.

• Bodyweight and concomitant ciclosporin were found to influence MPA pharmacokinetics.

• This study evaluated current dosing strategies and found that they may be suboptimal for children weighing less than 10 kg.

Introduction

Mycophenolate mofetil (MMF) is an immunosuppressant drug used for the treatment and prevention of graft vs. host disease (GvHD) in blood or marrow transplantation [1] and acute graft rejection in solid organ transplantation [2]. MMF is the 2, 4-morpholinoethyl ester pro-drug of mycophenolic acid (MPA) which is a potent, selective and reversible inhibitor of inosine monophosphate dehydrogenase (IMPDH). MPA targets the de novo purine biosynthesis pathway by inhibiting of IMPDH, thus acting against the proliferation of T and B lymphocytes. It also blocks the synthesis of guanosine nucleotide, which is required as a building block for DNA and RNA synthesis [3].

MMF undergoes rapid and complete in vivo metabolism to MPA following intravenous infusion or oral administration [4]. The absolute bioavailability of MPA after oral administration of MMF was estimated to be 94% in healthy subjects [4]. MPA is highly bound (97–99%) to serum albumin in patients with normal renal and liver function [5, 6]. Only unbound MPA is capable of inhibiting IMPDH in vitro[7]. MPA is metabolized in the gastrointestinal tract, liver and kidney by uridine diphosphate glucuronosyltransferases (UGTs) [8] to form 7-O-MPA-glucuronide (MPAG) which is the major metabolite of MPA. MPAG also displays significant binding to serum albumin (approximately 82% in stable patients) [9] and is usually present in the plasma at 20- to 100-fold higher concentrations than MPA. However, this metabolite is not pharmacologically active [7]. MPAG is mainly excreted into the urine via active tubular secretion [9] and into the bile by the multi-drug resistance protein 2 (MRP-2) [10]. MPAG excreted into the bile may be de-conjugated back to MPA by bacteria found in the gastrointestinal tract and then reabsorbed into the systemic circulation in the lower bowel [4]. This enterohepatic recirculation process contributes approximately 40% (range 10–60%) to MPA exposure [5, 11].

There have been several studies on the pharmacokinetics of MPA in children with solid organ transplantation [12–14]. However, MPA pharmacokinetics have not been thoroughly studied in paediatric blood or marrow transplant recipients and guidance for optimal dosing of MPA in children is lacking [15].

MPA exhibits considerable inter- and intra-patient pharmacokinetic variability in adult and paediatric transplant recipients [9, 16–18]. The pharmacokinetics of MPA are complex with evidence that pharmacokinetic parameters for MPA change over time following transplantation and appear to be influenced by renal function, albumin concentrations and drug interactions [19]. A greater than 10-fold range in MPA area under the concentration–time curve (AUC) has been reported following a standard dose [5, 9]. The AUC of MPA over a 12 h dose interval at steady-state is generally agreed to be the most reliable metric associated with the risk of acute rejection [20, 21].

However, measuring 12 h MPA AUC in paediatric patients is impractical due to the ethical and logistical difficulties in obtaining adequate samples in children. Population pharmacokinetic analysis can utilize concentration information from both intensive sampling and sparse sampling to provide pharmacokinetic parameter estimates, estimates of inter- individual and intra-individual variability in these parameters and allows patient characteristics explaining inter-individual variability to be quantified [22].

The aims of this study were to characterize the population pharmacokinetics of MPA following intravenous infusion of MMF and oral administration of MMF in children and young people undergoing blood or marrow or solid organ transplantation, and evaluate currently recommended dose regimens using pharmacokinetic model-based simulations and the accepted therapeutic drug monitoring targets in children and young people.

Methods

Patients

Children and young people receiving mycophenolic acid after blood or marrow or solid organ transplantations were included in this prospective, single-centre observational investigation of the pharmacokinetics of MPA. This study was approved by the Human Research Ethics Committee at the Children's Hospital at Westmead and written informed consent was obtained from parents of all children.

Drug administration and blood sampling

MMF (CellCept® Powder for Infusion, Roche Products Pty. Ltd, Australia) was diluted in 5% glucose and administered as a 2 h intravenous infusion at a dose of between 10 to 15 mg kg⁻¹, twice or three times daily. MMF (CellCept® Tablet, Roche Products Pty. Ltd, Australia) was administered as an oral dose of between 10 to 15 mg kg⁻¹, twice or three times daily.

Children and young people who weighed more than 20 kg had a series of 13–14 blood samples (3 ml each) drawn over an 8 h or 12 h time period (depending on the dosing interval) following the intravenous or oral MMF dose. The blood samples were collected from the patient's central venous cannula prior to the dose, and at the end of the infusion, then 15, 30 and 45 min, 1, 1.5, 2, 3, 5, 6, 8, 10 and 12 h after intravenous administration of MMF, or at 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 6, 8, 10 and 12 h after oral administration of MMF. The last two samples were taken only in those patients who received MMF twice daily. Fewer samples were collected from children who weighed less than 20 kg. These typically included a sample prior to the dose, then at the end of the infusion, 0.5, 1, 2 and 4 h after the dose or at 0.5, 1, 3, 5 and 7 h after oral administration of MMF. The remaining children and young people had randomly timed but accurately recorded blood samples collected. All the blood samples were taken when steady-state had been achieved based on the dosing history for each patient. Plasma was separated by centrifugation at 1200 g for 10 min at 4°C (Beckman CS-15R; Beckman Instruments, Fullerton, CA, USA) and was stored at −40°C until analysis. All samples were analyzed within 3 months of collection.

MPA assay

MPA concentrations were measured in plasma samples using a previously established and validated high-performance liquid chromatography assay [23]. The extraction recovery was over 93% and the assay response was linear over the MPA concentration range of 0.07–50 mg l⁻¹ for total MPA. Intra- and inter-day assay reproducibility was less than 10% over this range. The limit of detection and limit of quantitation were 0.04 mg l⁻¹ and 0.07 mg l⁻¹, respectively. A comparison of numerous plasma samples from patients receiving a range of medicines revealed no interfering peaks in the MPA assay.

Population pharmacokinetic analysis

Data were analyzed using the nonlinear mixed-effects modelling program NONMEM version V, level 1.1 (Globomax, Hanover, USA). The program Wings for NONMEM version 405 (http://wfn.sourceforge.net/) was used as a front-end processor. The Compaq Visual Fortran version 6.1A (Compaq Computer Corporation, Texas, USA) was used to perform the Fortran compilation. Graphical outputs of NONMEM were undertaken using CrossGraphs version 2.3 (PPD Development, Cambridge MA, USA). The first-order conditional estimation with interaction method was used throughout the model building and evaluation procedure.

Model selection was based on goodness-of-fit in addition to criteria of statistical significance, model plausibility and stability. Models were compared statistically using a likelihood ratio test on the differences in the NONMEM derived objective function value (OFV). Statistical significance for base model building and individual covariate analysis was set at P < 0.05 (ΔOFV = 3.84), while statistical significance for forward addition and backward elimination was set at a higher criteria (P < 0.01, ΔOFV = 6.63). Plots of residuals and weighted residuals, standard errors of parameter estimates and changes in estimates of inter-individual, inter-occasion and residual variability were also examined.

Base model building

Preliminary analyses focused on the selection of the structural and statistical models without taking into account the effects of covariates. The plots of observed concentration–time data of MPA were initially examined. One- and two-compartment pharmacokinetic models with first-order elimination were compared to investigate the best fit of the concentration–time data. The NONMEM subroutines ADVAN2 TRANS2 and ADVAN4 TRANS4 were employed in the analyses. The use of a lag time for drug absorption and the absorption rate constant (k_a) were examined. An exponential error model was employed to model inter-individual variability (IIV) for the pharmacokinetic parameters, defined as follows: , where θ_i represents the pharmacokinetic parameter of the ith individual, is the typical value of pharmacokinetic parameter θ in the population (e.g. population mean), ηi quantifies the deviation of θ_i from with a normal distribution of (0, ω²). The difference between the jth observed concentration (Y) in the ith individual and its respective prediction () was modelled with an exponential error model such that , where ε₁ is random effect quantifying the errors between Y and with a normal distribution of (0, σ²). Inter-occasion variability (IOV) for the parameter θ was evaluated as an additional level of random effect and was expressed as follows: , where OCC_n has the value of 1 for the nth occasion and 0 otherwise; k_{1 …n}are random variables assumed to be normally distributed with a mean of 0, and variance is denoted by π². In the modelling analysis, each ‘occasion’ was defined as 7 days in patients who were administered MMF daily. The effect of IOV was tested on MPA clearance (CL) using NONMEM's BLOCK (1) SAME option, assuming IOV was the same for all occasions.

Covariate analysis

Covariates screened for their possible influence on MPA pharmacokinetic parameters included sex, bodyweight, weight with an allometric scaling function (with the exponent fixed to 0.75), height, age, body surface area (BSA), body mass index (BMI), time after transplantation, days of therapy, transplantation type (blood or marrow, kidney or liver), concomitant ciclosporin (note: tacrolimus was co-administered to patients who did not receive ciclosporin), concomitant acyclovir, MPA unbound fraction, glomerular filtration rate (GFR), serum albumin concentration and estimated creatinine clearance (CL_CR). The covariates were implemented in the model using two different approaches: TVCL = θ_CL× (1 +θ_COV× covariate) and TVCL = θ_CL× covariate. Serum albumin concentrations were determined by a colorimetric method using bromocresol green dye (by the Department of Biochemistry, Children's Hospital at Westmead). Creatinine clearance for children and young people was estimated using the Counahan formula [24] as follows:

The serum creatinine concentration used in the analysis was the median value observed during the sample collection period. The influence of individual covariates on pharmacokinetic parameters was examined by plotting the empirical Bayesian estimates of individual parameters (derived from the base model) against covariates. Covariates identified as potentially influential were then included in the population pharmacokinetic model for MPA.

Following initial screening of individual covariates, in a forward stepwise fashion covariates were added cumulatively to the model in the order of their contribution to the reduction in the OFV in the preliminary analysis until there was no further reduction in the OFV. Next, a backward elimination step was performed in which the influence of each covariate was removed from the model in descending order of their contribution to the change in the OFV and the remaining parameters in the model were re-estimated.

It is plausible that changes in MPA unbound fraction due to changes in albumin concentrations will influence the pharmacokinetic parameters (except for the drug absorption rate constant) estimated from the total MPA concentrations [25]. In addition to including serum albumin concentration as a covariate for individual pharmacokinetic parameters, the effect of its incorporation into the best base model with an albumin-dependent scaling of concentration was also examined using the following equation:

where Y represents the total MPA concentration and U is the unbound ligand; D is a factor in the relationship between serum albumin and plasma concentrations; ALB is serum albumin concentration in g l⁻¹.

Model evaluation

A non-parametric bootstrap re-sampling method was performed to evaluate the accuracy and stability of the final population pharmacokinetic model. A total of 1000 minimization successful bootstrap samples were generated from the original data set and the final population model was fitted to the bootstrap data sets using the NONMEM program. Parameter estimates of each of the 1000 data sets were re-estimated. For each parameter, the resulting estimates were sorted, and the 2.5th and 97.5th percentiles were obtained as the lower and upper boundaries for the 95% confidence interval (95% CI) of the parameter. The mean parameter estimates obtained by bootstrapping were compared with those obtained by the population pharmacokinetic model to assess the reliability of the final model estimates.

Simulated dose regimens

The final population pharmacokinetic model was used to simulate MPA clearance data sets for 1000 hypothetical patients within the range of this group of children and young people using eight different dosing scenarios (intravenous infusion and oral MMF 10 mg kg⁻¹ with ciclosporin or tacrolimus or intravenous infusion and oral MMF 15 mg kg⁻¹ with ciclosporin or tacrolimus). The proportions of the hypothetical patients with different weights and concomitant ciclosporin (or tacrolimus) who achieved a total MPA AUC(0,24 h) of 60–120 mg l⁻¹ h were determined. The proportions of patients whose total MPA AUC(0,24 h) was below or above the target range were also calculated. This target range was derived from a total MPA AUC(0,12 h) of 30–60 mg l⁻¹ h below which a previously published report suggested that there may be a significant risk of developing acute rejection [26]. A total MPA AUC(0,12 h) of approximately 30–60 mg l⁻¹ h is also a reasonable target for kidney transplantation in both adults [20] and children [18]. A similar therapeutic range for mycophenolic acid was also used in liver transplantation for acute rejection [21, 26] and blood or marrow transplantation for GvHD [27].

Results

Pharmacokinetics

Thirty-eight children and young people aged between 4.8 months and 19.9 years who were receiving MPA after blood or marrow (n = 23), kidney (n = 5) and liver (n = 10) transplantations were recruited into this study. The characteristics of the children and young people who participated in this study are summarized in Table 1. All patients were given MMF via intravenous infusion or oral administration (Table 1). A total of 859 concentration–time measurements were collected from 38 children and young people (Table 1). The number of MPA concentrations per patient ranged from 2 to 49 samples, with a median of 23 samples per patient; 13 patients had intensive sampling (median 34, range 12 to 49) and 25 had sparse or random sampling (median 12, range 2 to 43). A frequency histogram of blood samples collected at specific times during the dose interval is shown in Figure 1. MPA concentrations ranged from 0.07 mg l⁻¹ to 11.1 mg l⁻¹ with a mean of 1.3 ± 1.8 mg l⁻¹ after intravenous dosing and 0.07 mg l⁻¹ to 10.6 mg l⁻¹ with a mean of 1.6 ± 1.9 mg l⁻¹ after oral dosing of MMF.

Table 1.

Characteristics of children and young people receiving MMF by intravenous infusion or oral administration

Characteristics	Median	Range
Number of children and young people (MPA concentrations)	38 (859)
0 < Age (years) < 2	7 (155)
2 < Age (years) ≤ 5	4 (75)
5 < Age (years) ≤ 12	17 (376)
12 < Age (years) < 20	10 (253)
Number who had intravenous dosing	13
Number who had oral dosing	18
Number who had both intravenous and oral dosing	7
Number who had intensive sampling	13
Number who had sparse sampling	7
Number who had random sampling	18
Number who had enterohepatic recirculation	9
Transplantation type
Blood or marrow transplant	23
Kidney transplant	5
Liver transplant	10
Concomitant immunosuppressant
With ciclosporin	23
With tacrolimus	15
Concomitant antiviral
With acyclovir	9
Without acyclovir	29
Unbound fraction	0.0116	0.0049–0.06
Sex (male/female)	20/18
Age (years)	8.4	0.4–19.9
Weight (kg)	27.9	3.4–87.7
Body surface area (m2)	1.0	0.22–2.0
Body mass index (kg m−2)	17.2	11.9–31.8
Height (cm)	128.6	53.5–169.0
Creatinine clearance (l h−1 kg−1)*	0.12	0.01–0.31
Albumin (g l−1)†	31.0	22.0–44.0

Body surface area, BSA [45];

^*Creatinine clearance, CL_CR was estimated using the Counahan formula [24].

^†Serum albumin was determined using bromocresol green dye (a colorimetric method).

Figure 1.

Frequency histogram of blood samples collected at specific times during the dose interval

Base model building

A two-compartment pharmacokinetic model with first-order elimination (estimating parameters clearance (CL), inter-compartmental clearance (Q), volume of distribution in the central (V₁) and peripheral (V₂) compartments, absorption rate constant (k_a) and bioavailability (F), with IIV on CL, k_a and F, and IOV on CL) was found to provide the best description of the MPA concentration–time data. This model included an exponential error model to estimate residual error. The model with inclusion of IIV on the pharmacokinetic parameters (including V₁, V₂ and Q) and IOV on V₁ was not supported by the data yielding large standard errors for these parameters. The evaluation of a lag time parameter was tested in the present model as the MPA concentration-time plot revealed evidence of a lag time in drug absorption in some population pharmacokinetic studies [28–31]. However, this could not be reliability estimated by the model possibly due to insufficient MPA concentration–time data collected between 0 to 0.5 h after an oral dose.

Covariate analysis

Individual posterior Bayesian estimates of pharmacokinetic parameters generated from the base model were plotted against each covariate. Four covariates were identified as being of interest (Figures 2 and 3). Estimates of mycophenolic acid CL values increased with increasing patient weight (r² = 0.36), decreased with increasing CL_CR (r² = 0.1), and was greater in patients receiving concomitant ciclosporin when compared with those patients receiving tacrolimus (P < 0.001, Mann–Whitney Test). There was also a trend towards increased MPA CL with reducing serum albumin concentration (r² = 0.10). Inclusion of these covariates individually significantly reduced OFV by the pre-defined cut-off value of 3.84 (P < 0.05) when tested against the base model. Covariates were not tested on V₁, as IIV for V₁ was not evaluated in the base model.

Figure 2.

Scatterplots of empirical Bayesian estimates vs. covariates: (A) CL vs. bodyweight, (B) CL vs. CL_CR, (C) CL vs. serum albumin concentration

Figure 3.

Empirical Bayesian CL estimates (median value for each individual) vs. concomitant immunosuppressive therapy (ciclosporin vs. tacrolimus)

The inclusion of total bodyweight as a covariate on CL was found to be statistically significant (ΔOFV = −17.40), with a reduction of IIV in CL from 54.3% to 36.9%. While the effect of bodyweight scaled using an allometric function in which the exponent was fixed to 0.75 was found to reduce the IIV in CL from 54.3% to 46.8% (ΔOFV = −4.92), this was less significant that the effect of total body weight. The inclusion of concomitant ciclosporin on mycophenolic acid CL was found to reduce the IIV in CL from 54.3% to 45.3% (ΔOFV = −18.24). The consideration of estimated CL_CR and serum albumin concentration on CL were also found to be statistically significant, with a decrease in OFV by 5.17 and 7.86 points, respectively, and the IIV in CL was reduced to 44.7% and 50.5%, respectively.

In the forward stepwise modelling building approach, the cumulative inclusion of bodyweight and concomitant ciclosporin reduced the OFV by 39.78 and IIV in CL reduced from 54.3% to 31.6%, without increasing variability in other parameter estimates. Inclusion of CL_CR and/or serum albumin concentration in this covariate model did not result in a further statistically significant reduction in OFV or IIV in CL. Finally, in the backward elimination step, removal of bodyweight and concomitant ciclosporin in order resulted in an increase in the OFV of 21.54 and 22.38, respectively. The model-building process is summarized in Table 2.

Table 2.

Successful base and covariate population pharmacokinetic models for MPA

Model number	Covariate	Model	Compared against	OFV	ΔOFV	P
One compartment
1	Base model	CL = θ1× EXP(IIV + IOV)		−207.92
		V = θ2× EXP(IIV)
Two compartment
2	Base model	CL = θ1× EXP(IIV + IOV)	1	−457.85	−249.93	<0.001
		V1 = θ2× EXP(IIV)
		Q = θ3× EXP(IIV)
		V2 = θ4× EXP(IIV)
3	Bodyweight on CL	CL = θ1× (1 +θ7× WT/27.9) × EXP(IIV + IOV)	2	−475.25	−17.40	<0.001
		V1 = θ2× EXP(IIV)
		Q = θ3× EXP(IIV)
		V2 = θ4× EXP(IIV)
4	Concomitant ciclosporin on CL	CL = θ1× (1 +θ7× WT/27.9) × (1 +θ8× CYTA) × EXP(IIV + IOV)	3	−497.63	−22.38	<0.001
		V1 = θ2× EXP(IIV)
		Q = θ3× EXP(IIV)
		V2 = θ4× EXP(IIV)

OFV, NONMEM derived objective function value; WT, bodyweight; CYTA, concomitant ciclosporin (CYTA = 0 if the patient taking ciclosporin, CYTA = 1 if without ciclosporin while taking tacrolimus); CL, clearance; Q, inter-compartmental clearance; V₁, volume of distribution in the central compartment; V₂, volume of distribution in the peripheral compartment. The population CL term was standardized to 27.9 kg, which represents the median value of weight in this study group.

As an alternative to modelling the effect of serum albumin concentration as a covariate on CL, its incorporation as an albumin-dependent scaling of concentration was also examined. This resulted in an OFV reduction of 24.99 points. This model yielded a typical (SE%) population CL of 44.4 l h⁻¹ (288%), Q of 14.7 l h⁻¹ (259%), V₁ of 30.6 l (228%), V₂ of 69.3 l (251%), k_a of 0.322 (18%), F of 0.618 (27%) and albumin factor of 0.129 (233%). These values were estimated with a much higher degree of uncertainty and the model had very wide 95% confidence intervals for CL, Q, V₁ and V₂ during bootstrapping. For this reason this model was not evaluated further.

The final covariate model for MPA CL was:

The population pharmacokinetic parameters estimated from the base and final models are presented in Table 3. The population mean estimates of CL, Q, V₁, V₂, k_a, F, θ₇ and θ₈ from the final model were 6.42 l h⁻¹ (14.7 l h⁻¹ with ciclosporin, 5.5 l h⁻¹ with tacrolimus), 3.74 l h⁻¹, 7.24 l, 16.80 l, 0.39 h⁻¹, 0.48, 1.09 and −0.60, respectively, and estimates of IIV were 31.6% for CL, 59.1% for k_a, 34.6% for F, respectively. The IOV estimate was 5.8% for CL. Figure 4 displays the diagnostic plots for the base and final models. The residual variability in the final model was reduced compared with the base model. The population-predicted concentrations were symmetrically distributed around the line of identity, indicating that the model adequately describes the concentration–time profile of MPA at the doses studied. The representative observed and model-predicted concentration–time profiles for two typical patients treated with an intravenous infusion of MMF at a dose of 10 mg kg⁻¹ or oral MMF dose of 15 mg kg⁻¹, respectively, are shown in Figure 5 and show a satisfactory model fit.

Table 3.

Population pharmacokinetic parameter estimates derived from the base and the final models with their relative standard error (RSE%) and the parameter estimates from 1000 successful bootstrap replicates with 95% confidence interval (CI)

	Base model		Final model		1000 bootstrap runs
Parameter	Mean	RSE %	Mean	RSE %	Mean	95% CI*
Fixed effects
CL (l h−1)	11.3	33	6.42 (14.74†/5.51‡)	34	6.93	3.64, 12.8
Q (l h−1)	3.93	47	3.74	46	3.89	1.66, 7.21
V1 (l)	7.86	55	7.24	55	8.20	2.90, 18.5
V2 (l)	18.10	54	16.80	52	17.8	5.79, 34.9
ka (h−1)	0.33	20	0.39	20	0.42	0.27, 0.71
F	0.64	36	0.48	32	0.50	0.27, 0.82
Weight factor	–	–	1.09	30	1.06	0.42, 1.80
Concomitant ciclosporin factor	–	–	−0.60	10	−0.61	−0.73, −0.48
Interindividual variability
ωCL (CV%)	54.3	24	31.6	26	30.0	20.8, 38.3
ωQ (CV%)	–	–	–	–	–	–
ωV1 (CV%)	–	–	–	–	–	–
ωV2 (CV%)	–	–	–	–	–	–
ωka (CV%)	61.1	87	59.1	79	57.0	9.5, 116.6
ωF (CV%)	49.0	54	34.6	59	33.0	10.2, 58.7
Interoccasion variability
πCL (CV%)	5.9	112	5.8	105	5.0	0.1, 9.1
πV1 (CV%)	–	–	–	–	–	–
Random residual variability
σ1	0.48	14	0.48	14	0.47	0.40, 0.53

RSE, relative standard error; CL, clearance; Q, inter-compartmental clearance; V₁ volume of distribution in central compartment; V₂ volume of distribution in peripheral compartment; k_a absorption rate constant; F bioavailability; σ₁ residual variability.

^*Percentile bootstrap 95% CI corresponding to parameter estimates at the 2.5th and 97.5th percentiles of bootstrap runs.

^†Concomitant ciclosporin

^‡Concomitant tacrolimus.

Figure 4.

Observed MPA concentrations vs. population predicted concentrations using the base and final population pharmacokinetic models (the line is the regression line)

Figure 5.

Concentration–time profile of MPA predicted by the final model in two representative patients following intravenous and oral administration of MMF

Model evaluation

From the original data set, 1000 minimization successful data sets (72%) were generated using a bootstrapping approach. The mean values for all fixed effects parameters and the mean values for random effects parameters were within ±15% of those obtained by the NONMEM final model, indicating good reliability of this model. Table 3 presents the results of the bootstrap procedure with parameters, presented as the mean and 95% CI.

Simulation

The results of the simulated dose regimens are shown in Tables 4 and 5. In the ‘simulated’ children who received MMF with concomitant ciclosporin, orally administered MMF (10 mg kg⁻¹ or 15 mg kg⁻¹) three times daily was associated with the lowest proportion of children achieving the target total MPA AUC range 60–120 mg l⁻¹ h (less than 24%). A large proportion (more than 93% of children when having an intravenous infusion MMF and more than 97% of children when having oral MMF) of light children who weighed 10 kg achieved total MPA AUC(0,24 h) at steady-state of less than 60 mg l⁻¹ h suggesting that dose increases may be beneficial. The children who weigh 50 kg may also benefit from dose (based on mg kg⁻¹) increases when receiving oral MMF three times daily, since more than 70% of children achieved total MPA AUC(0,24 h) at steady-state of less than 60 mg l⁻¹ h.

Table 4.

The percentage of children who achieved total MPA AUC(0,24 h) values below, within and above the daily target range of a 60–120 mg l⁻¹ h, calculated by 1000 simulations for children with different bodyweights and concomitant ciclosporin (tacrolimus) when given 10 mg kg⁻¹ intravenous infusion or oral doses of MMF

Intravenous infusion or oral MMF 10 mg kg−1
Weight	Concomitant treatment with ciclosporin
	Intravenous infusion MMF three times daily			Oral MMF three times daily
	Total MPA AUC(0,24 h) (mg l−1 h)			Total MPA AUC(0,24 h) (mg l−1 h)
	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h
10 kg	99.7%	0.3%	0.0%	99.4%	0.5%	0.1%
25 kg	84.0%	15.8%	0.2%	94.8%	4.7%	0.5%
50 kg	52.1%	46.3%	1.6%	87.3%	11.4%	1.3%
	Concomitant treatment with tacrolimus
	Intravenous infusion MMF twice daily			Oral MMF twice daily
	Total MPA AUC(0,24 h) (mg l−1 h)			Total MPA AUC(0,24 h) (mg l−1 h)
	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h
10 kg	87.4%	12.5%	0.1%	95.4%	4.2%	0.4%
25 kg	24.6%	69.9%	6.5%	77.9%	19.1%	3.0%
50 kg	6.0%	66.7%	27.3%	62.7%	30.2%	7.1%

Note:Target ranges for total MPA AUC(0,12 h) are 30–60 mg l⁻¹ h [21, 26]. A derived target range for total MPA AUC(0,24 h), 60–120 mg l⁻¹ h was used in the present study as the simulated doses were administered twice or three times daily.

Table 5.

The percentage of children who achieved total MPA AUC(0,24 h) values below, within and above the daily target range of a 60–120 mg l⁻¹ h, calculated by 1000 simulations for children with different bodyweights and concomitant ciclosporin (tacrolimus) when given 15 mg kg⁻¹ intravenous infusion or oral MMF

Intravenous infusion or oral MMF 15 mg kg−1
Weight	Concomitant treatment with cyclosporine
	Intravenous infusion MMF three times daily			Oral MMF three times daily
	Total MPA AUC(0,24 h) (mg l−1 h)			Total MPA AUC(0,24 h) (mg l−1 h)
	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h
10 kg	92.9%	7.1%	0.0%	96.7%	3.2%	0.1%
25 kg	38.1%	58.8%	3.1%	83.5%	14.6%	1.9%
50 kg	12.0%	71.7%	16.3%	70.4%	24.3%	5.3%
	Concomitant treatment with tacrolimus
	Intravenous infusion MMF twice daily			Oral MMF twice daily
	Total MPA AUC(0,24 h) (mg l−1 h)			Total MPA AUC(0,24 h) (mg l−1 h)
	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h	<60 mg l−1 h	60–120 mg l−1 h	>120 mg l−1 h
10 kg	43.4%	54.0%	2.6%	84.9%	13.4%	1.7%
25 kg	2.8%	56.1%	41.1%	55.7%	33.7%	10.6%
50 kg	0.4%	23.8%	75.8%	40.1%	37.5%	22.4%

Note:Target ranges for total MPA AUC(0,12 h) are 30–60 mg l⁻¹ h [21, 26]. A derived target range for total MPA AUC(0,24 h), 60–120 mg l⁻¹ h was used in the present study as the simulated doses were administered twice or three times daily.

In the ‘simulated’ children who had MMF with concomitant tacrolimus, orally administered MMF (10 mg kg⁻¹ or 15 mg kg⁻¹) twice daily was associated with a greater proportion of the children achieving the target total MPA AUC range 60–120 mg l⁻¹ h when compared with the oral dose regimen of MMF co-administered with ciclosporin, however, there was still a tendency for light children who weighed 10 kg to have lower MPA exposure (more than 85%), suggesting that dose increases may be required for these patients to achieve the target exposure. Total MPA AUC(0,24 h) values at steady-state of greater than 120 mg l⁻¹ h were achieved by a large proportion (76%) of children in weighing more than 50 kg, receiving an intravenous infusion MMF (15 mg kg⁻¹) twice daily with concomitant tacrolimus, which may increase the risk of MPA related toxicity [32].

Discussion

This study is one of the first investigations in which a population pharmacokinetic modelling approach was applied to assess the pharmacokinetics of both intravenous and oral MMF in children and young people undergoing blood or marrow and solid organ transplantation. Recruitment of patients was a challenging aspect of this study, so data from three groups of patients (blood or marrow/kidney/liver transplant recipients) were combined to increase the power to detect significant covariate effects and to better estimate the population pharmacokinetic parameters. The largest subgroup were the blood or marrow transplant recipients (n = 23), but this number was still insufficient to provide meaningful results when analyzed and reported separately. In this study considerable inter-individual variability in the pharmacokinetics of MPA was observed. After systematically evaluating a range of patient characteristics and clinical factors, it was found that bodyweight and concomitant ciclosporin had a significant influence on the MPA CL. Inclusion of these factors, which described IIV in these parameters, significantly improved the population pharmacokinetic model performance based on the likelihood ratio test.

In children, MPA CL has been correlated with weight and MMF dosage recommendations in children and young people are based on weight. In the present study children and young people had a broad weight range (3.4–87.7 kg), and there was a clear trend towards increased MPA CL with increasing bodyweight. The inclusion of weight into the population model resulted in a large reduction in IIV in CL from 54.3% to 36.9%.

Concomitant ciclosporin did prove to be a significant covariate in the final population pharmacokinetic model. Ciclosporin inhibits the transport of MPAG from hepatocytes into the bile resulting in decreased enterohepatic circulation [10, 33]. The study performed by van Gelder et al. [33] showed that MPA concentrations during MMF treatment alone were not significantly different from MPA concentrations during tacrolimus co-administered treatment. In this study, co-administered ciclosporin as opposed to tacrolimus resulted in a mean increase in MPA CL of 63%.

Renal excretion is reported to be one of the elimination pathways for MPA [9] and impaired renal function has been reported to influence MPA disposition [34]. Poor renal function can reduce the renal excretion of MPAG that is believed to compete with MPA for protein binding sites [6]. In the present study, a trend towards a lower normalized CL_CR with an increased MPA CL was evident. The physiological plausibility of this observation was unclear but may relate to other clinical effects associated with renal impairment (and renal excretion) impacting on MPA pharmacokinetics. However, this covariate did not reach statistical significance and was not included in the final pharmacokinetic model. It is likely that the effect of renal impairment on MPA pharmacokinetics could not be reliably reported in this group of children and young people because of lack of patients with poor renal function.

MPA and MPAG are highly bound to human serum albumin [6] and may be altered in patients with severe renal impairment, liver disease and hypoalbuminaemia [32]. In the present study, a weak trend towards a decreased MPA CL for patients with higher serum albumin concentrations was identified. However, this relationship did not reach statistical significance in the final population pharmacokinetic model. MPA unbound fraction was also tested as a potential covariate, but it was not significant in the individual covariate screen. In theory, changes in MPA unbound fraction due to changes in albumin concentration have the potential to influence all pharmacokinetic parameters except those only involved with drug absorption [25]. The inclusion of albumin concentration as an albumin-dependent scaling of MPA concentration, rather than as a covariate on CL, resulted in parameter values with a high degree of uncertainty. Concentration–time data collected in this study may not be sufficient to characterize the influence of serum albumin concentration on all pharmacokinetic parameters simultaneously.

This study also evaluated the possible impact of concomitant administration of acyclovir on MPA CL but found no significant effects. It is consistent with previous reports that no statistically significant difference in the pharmacokinetic parameters of MPA was observed with co-administration of acyclovir [9, 35].

Transplantation type (blood or marrow, kidney and liver transplantation) was also evaluated as a possible covariate. However, the number of each transplant type was not high enough to demonstrate statistical significance.

The previous studies reported the high variability of MMF bioavailability in different populations including healthy volunteers, kidney transplant recipients, liver transplant recipients and non-myeloablative haematopoietic cell transplant recipients [4, 36–38]. The possible reasons for high variability include preparative regimen-induced gastrointestinal damage, underlying poor absorptive processes due to previous standard chemotherapy and chronic illness, and/or reduced enterohepatic recirculation (e.g. concomitant ciclosporin use) [37, 38]. Interestingly, MMF bioavailability was 48% in this study cohort when estimated using the final model. This suggests that higher oral MMF doses may be required to provide equivalent exposure as intravenous doses.

The occurrence of enterohepatic recirculation could only be studied in those 13 patients that underwent intensive sampling and was found in nine of these patients. The extent of enterohepatic recirculation of MMF may be dependent on each individual's post-surgical state and co-administered medication, but could not be modelled further in this study given the limited number of patients. This has the potential to lead to the inflation of the residual variability in the final model.

To date, seven population pharmacokinetic analyses (using NONMEM software) of MPA have been published. Five of these studies were performed in adult kidney transplant recipients [25, 29, 30, 39, 40], one in paediatric and adolescent renal transplant recipients [28] and another in healthy volunteers [31]. Previous population pharmacokinetic studies involved only oral administration of MMF. A two-compartment pharmacokinetic model was selected to describe MPA pharmacokinetics. Mean MPA CL/F estimates ranged from 14.1 to 34.9 l h⁻¹ in adults receiving concomitant cyclosporine and from 11.9 to 25.4 l h⁻¹ in adults receiving co-administration of tacrolimus. Bodyweight, serum albumin concentration and concomitant ciclosporine were the covariates mostly identified. Complex absorption and enterohepatic recirculation of MPA has made population modelling challenging. The only previously reported paediatric population pharmacokinetic analysis of MPA was performed in 41 children and adolescents (2–21 years) following oral administration of MMF [28]. In the study by Payen et al. bodyweight was found to have significant effects on the apparent volume of distribution of MPA in the central compartment [28]. The present study performed in 38 children and young people after intravenous infusion and oral doses of MMF showed comparable estimates of pharmacokinetic parameters to the study by Payen et al. [28]. Bodyweight and concomitant ciclosporin were identified as influential covariates. The bioavailability was also able to be evaluated in the present study.

As the pharmacokinetics of MPA exhibits wide inter-individual variability and concomitant ciclosporine has an influence on MPA pharmacokinetics, it is possible that the clinical outcomes of MPA could be improved with therapeutic drug monitoring (TDM) [26]. A total MPA AUC(0,12 h) of approximately 30–60 mg l⁻¹ h was reported to be a reasonable target for the early renal post-transplant period in both adults [20] and children [18]. Although inter-individual variability of MPA exposure in patients undergoing liver transplantation is at least as pronounced as that seen after kidney transplantation [19], there are limited data to provide guidelines for TDM of MPA after liver transplantation [41, 42] and more investigation is needed. For now, it is recommended that similar targets are used to those in kidney transplantation [21, 26].

Target values for MPA TDM cannot be definitely recommended in blood or marrow transplant recipients until more data are published from clinical studies that provide definitive guidelines in this patient population. Haentzschel et al. [27] individualized MMF dose regimens for GvHD prophylaxis in 29 patients after allogeneic blood stem cell transplantation using similar MPA concentration targets as in kidney transplantation. It is suggested that targeting of individual MMF doses is feasible in the early phase after allogeneic blood stem cell transplantation. Given the lack of an established TDM concentration target range for this group of patients, we did not believe it appropriate to suggest dosing guidelines. Instead, we used a simulation approach to evaluate current dosing schedules according to agreed MPA AUC target ranges.

In this study, children and young people were given a 10–15 mg kg⁻¹ intravenous infusion or an oral MMF dose. To investigate the suitability of MMF dosing regimens in children and young people a simulation study was employed using the final validated population pharmacokinetic model. The simulation results showed that the dosing regimens may not be optimal for achieving MPA TDM targets associated with the prevention of acute rejection or GvHD in all children and young people according to these criteria (bodyweight and concomitant ciclosporin). Children who weigh below 10 kg are relatively under-dosed and may be at a greater risk of acute rejection [43] while heavier children (who weigh over 50 kg) may be at greater risk of developing adverse reactions [32] after intravenous infusion of MMF co-administered with tacrolimus. Filler et al. [43] found a significant correlation between the dose-normalized exposure and age when studying 27 children and young people (age 1.8–20.7 years) receiving a combination therapy of MMF and tacrolimus. The data showed that young children under 2 years of age required twice the amount of MMF in combination with tacrolimus when compared with adolescents. Overall, rational dosing strategies for MMF need be considered for children and young people on the basis of bodyweight and concomitant therapy.

The data about the therapeutic drug monitoring of unbound MPA AUC are limited in solid organ transplant recipients and therapeutic unbound targets are undefined, although data suggest that an unbound AUC(0,12 h) greater than 400 ng ml⁻¹ h is associated with greater haematologic toxicity in paediatric kidney transplant recipients [12]. A relationship between MPA unbound AUC and the clinical outcome after hematopoietic cell transplantation has been defined recently by Jacobson et al. [44] in a pharmacokinetic and pharmacodynamic study involving 87 adult patients undergoing non-myeloablative-related and non-myeloablative-unrelated hematopoietic cell transplantation who were receiving 1 g MMF orally or intravenously every 12 h with concomitant ciclosporin. The patients with an unbound MPA AUC(0,6 h) of less than 150 ng ml⁻¹ h had a higher cumulative incidence of grade II-IV acute GvHD than the patients with a greater AUC (68% vs. 40%, P = 0.02). An unbound AUC(0,12 h) of less than 300 ng ml⁻¹ h was also associated with more frequent acute GvHD (58% vs. 35%, P = 0.05). The unbound MPA population pharmacokinetics and the unbound therapeutic targets will be explored in this group of children and young people in future studies.

In conclusion, a two-compartment pharmacokinetic model describing the concentration–time profile of MPA after intravenous infusion and oral dosing of MMF has been developed and evaluated in children and young people undergoing blood or marrow and solid organ transplantation. Bodyweight and concomitant ciclosporin were found to influence MPA clearance. The current dose regimens may not be optimal for achieving MPA TDM targets associated with the prevention of acute rejection or GvHD in all children and young people. It is possible to use this population pharmacokinetic model in conjunction with MPA efficacy data to develop an optimal dosing strategy to ensure that children and young people achieve the maximal immunosuppressant efficacy of MPA.

Competing interests

There are no competing interests to declare.

Lihua Zeng is supported, in part, by a NHMRC Project Grant 396702. Dr Christa Nath is supported by the Leukaemia Research Support Fund of The Children's Hospital Westmead. The authors thank the nurses and allied clinical staff in the Children's Hospital at Westmead for taking blood samples and for patient care and the patients and their families for taking part in this study.