The AAPS Journal, Vol. 18, No.

6, November 2016 ( # 2016)

DOI: 10.1208/s12248-016-9956-4

Research Article

A Physiologically Based Pharmacokinetic Model for Ganciclovir and Its Prodrug

Valganciclovir in Adults and Children

V. Lukacova,1 P. Goelzer,2 M. Reddy,3 G. Greig,4 B. Reigner,4 and N. Parrott5,6

Received 5 May 2016; accepted 7 July 2016; published online 22 July 2016

Abstract. A physiologically based pharmacokinetic (PBPK) model has been developed

for ganciclovir and its prodrug valganciclovir. Initial bottom-up modeling based on
physicochemical drug properties and measured in vitro inputs was verified in preclinical
animal species, and then, a clinical model was verified in a stepwise fashion with
pharmacokinetic data in adult, children, and neonatal patients. The final model incorporated
conversion of valganciclovir to ganciclovir through esterases and permeability-limited tissue
distribution of both drugs with active transport processes added in gut, liver, and kidney. A
PBPK model which accounted for known age-related tissue volumes, composition and blood
flows, and renal filtration clearance was able to simulate well the measured plasma exposures
in adults and pediatric patients. Overall, this work illustrates the stepwise development of
PBPK models which could be used to predict pharmacokinetics in infants and neonates,
thereby assisting drug development in a vulnerable patient population where clinical data are
challenging to obtain.
KEY WORDS: esterases; pediatrics; permeability limited distribution; physiologically based
pharmacokinetic models.

INTRODUCTION active tubular secretion [1]. Based on population pharmaco-

Ganciclovir is an antiviral product used intravenously
(IV) to treat cytomegalovirus (CMV) disease in immuno-
compromised patients and to prevent CMV disease in solid
organ transplant (SOT) recipients [1]. Valganciclovir [2] is a
valine ester prodrug of ganciclovir delivering improved oral
bioavailability [3] with established safety and efficacy in
prevention of CMV disease in adult transplant patients when
administered as a once-daily 900-mg dose [4–6].
Valganciclovir, an inactive prodrug, is metabolized to
form the active ganciclovir through the esterase
valacyclovirase [7], which is expressed in enterocytes, hepa-
tocytes, and kidney [8], while ganciclovir’s main route of
elimination is via the kidney through glomerular filtration and
Electronic supplementary material The online version of this article
(doi:10.1208/s12248-016-9956-4) contains supplementary material,
which is available to authorized users.
1
Simulations Plus, Inc., 42505 10th Street West, Lancaster, California
93534, USA.
2
Teva Pharmaceuticals, West Chester, Pennsylvania 19380, USA.
3
Array BioPharma, Boulder, Colorado 80301, USA.
4
Clinical Pharmacology, Roche Pharma Research and Early Devel-
opment, Roche Innovation Center Basel, Basel, Switzerland.
5 [14]. This is of particular importance in the development of
Pharmaceutical Sciences, Roche Pharma Research and Early
Development, Roche Innovation Center Basel, Basel, Switzerland.
6
To whom correspondence should be addressed. (e-mail:
neil_john.parrott@roche.com)
kinetic modeling of patient exposure to ganciclovir [9, 10], a
dosing algorithm has been developed for pediatric patients
that results in ganciclovir exposure comparable to that
achieved with the once-daily 900-mg regimen in adults [11].
This dosing algorithm is based on calculated creatinine
clearance (CrCLS) and patient body surface area (BSA)
(dose = 7 × BSA × CrCLS; see [11] for details); however, in
developing the algorithm, no transplant patients younger than
4 months were exposed to valganciclovir, and it was unknown
how differences in physiology might affect ganciclovir expo-
sure in such young children. To address this gap, a pharma-
cokinetic and safety study of valganciclovir in pediatric heart
transplant recipients aged <4 months (study NP22523) was
conducted [12]. As recruitment of this very vulnerable patient
population into the study was extremely challenging, a
comprehensive modeling and simulation strategy, including
physiologically based pharmacokinetic (PBPK) modeling,
was employed to optimally utilize all available data and
address various questions from health authorities throughout
the review process [13].
PBPK modeling is increasingly recognized as a way of
improving the efficiency of the drug development process
drugs for pediatric use, where patient numbers are limited
and optimal use of all relevant data is paramount [15]. During
development, changes in physiological factors such as body

1453 1550-7416/16/0600-1453/0 # 2016 American Association of Pharmaceutical Scientists

1454
composition and hepatic and renal functions can influence
pharmacokinetics in young children, infants, and neonates via
effects on drug absorption, distribution, metabolism, and
elimination (ADME) [16–18]. A strength of the PBPK
approach is the ability to integrate such knowledge on
physiological development with drug-specific ADME knowl-
edge to simulate the expected changes in pharmacokinetics.
Development of PBPK models for valganciclovir and
ganciclovir is particularly challenging as in vitro data suggest
that active drug transporters play a significant role for both
drugs. Ganciclovir is a substrate for the human multidrug
resistance protein 4 (MRP4) efflux transporter [19], which is
expressed in the apical membrane in the kidney [20] and the
basolateral membrane in the liver [21]. The mRNA expres-
sion of MRP4 transporter was reported also in the human
small intestine [22], and based on the reported localization in
Caco2 cells [23], it is assumed to be expressed on the
basolateral membrane in the intestine. The multidrug and
toxin extrusion protein 1 (MATE1) and MATE2-K trans-
porters found in the liver and kidney, respectively, also
transport ganciclovir [24] out of the cells (into bile via
MATE1 and into urine via MATE2-K). Ganciclovir is also
transported into cells by the organic anion transporter 1
(OAT1), which was detected in the basolateral membrane in
the kidney. Valganciclovir is a substrate for the human
peptide transporter 1 (PepT1) influx transporter [25], which
is found in the apical membranes of the gut and the kidney
[26] and also in the liver [27]. A schematic representation of
the transporters involved in disposition of both compounds is
shown in Fig. 1.
Fig. 1. Schematic diagram of the valganciclovir and ganciclovir PBPK model including
more detailed views of three organs with summary of transporters pertinent to the liver,
gut, and kidney
Lukacova et al.
The objective of this work was to develop and verify
linked PBPK models for valganciclovir and ganciclovir, which
could be used to simulate pharmacokinetics in infants and
neonates, a vulnerable patient population where clinical data
are difficult to obtain. The models integrate in vitro ADME
data with optimized parameters based on clinical data.
Whenever parameter values were optimized, simulations
were afterward compared to an independent set of clinical
data. Simulations covered a range of patient populations and
different dosing regimens. An initial PBPK model describes
ganciclovir pharmacokinetics after IV and oral (PO) admin-
istrations and valganciclovir oral pharmacokinetics in adults.
The adult PBPK model is then adapted to describe the
pharmacokinetics in younger populations by accounting for
known physiological age dependencies. Comparison of simu-
lations to data from studies in 13 to 16, 9 to 12, 5 to 8, and 1-
to 3-year olds is performed. The verified model was
subsequently used to model pharmacokinetics and support
dosing in infants and neonates as described elsewhere [13].
METHODS
Software
All modeling and simulation were carried out using
GastroPlus™ version 9.0 (Simulations Plus Inc., Lancaster,
CA). Physiologies, including age- and body weight-matched
organ weights, volumes, and blood perfusion rates, were
generated by the program’s internal module for Population
Estimates for Age-Related (PEAR™) Physiology™. While
Pediatric PBPK Model for Valganciclovir and Ganciclovir
GastroPlus does include a virtual population simulator, we
did not use that feature here. For adult simulations, a single
PBPK simulation matching the age and body weight of the
average of the subjects in the study was used, while for
children, a PBPK model was generated for each individual
patient matching the reported age, body weight, and
creatinine clearance. The mixed multiple dose (MMD)
feature in GastroPlus is designed to set up complex dosing
scenarios including different dose levels, administration
routes, and irregular dosing intervals. As this feature also
allows changes to the PBPK physiology, it was used in this
work to account for physiological changes that happened
throughout the course of a study. In particular, this feature
was used to account for changes in glomerular filtration rates
in multiday studies in older children, where the renal function
was monitored and reported each day.
Preliminary Work in Animal Species
Initial understanding of the pharmacokinetics of ganci-
clovir and valganciclovir was gained by PBPK modeling and
simulation in animal species (mouse, rat, dog, and monkey).
Ganciclovir pharmacokinetic profiles after IV administration
of ganciclovir were used to parameterize the distribution and
renal elimination of ganciclovir. Ganciclovir and
valganciclovir pharmacokinetic profiles after IV administra-
tion of valganciclovir were used to parameterize the distribu-
tion of valganciclovir, its conversion to ganciclovir, and
ganciclovir biliary secretion. More details of those simulations
are outlined in Online Resource 1.
Input Data
The basic physicochemical and biopharmaceutical prop-
erties for ganciclovir and valganciclovir are listed in Table I.
Table I. Physicochemical and Biopharmaceutical Properties of Ganciclovir and Valganciclovir Used in the PBPK Models
Parameter
Molecular weight (g/mol)
LogPa
Human jejunal Peff (×10−4 cm/s)b
Measured solubility at reference pH (mg/mL)a
pKa
Solubility factorc
Mean-estimated precipitation time (s)c
Diffusion coefficientc, cm2/s × 105
Estimated drug particle density (g/mL)
Particle radius (μm)e
Human blood-to-plasma ratioc
Unbound percent in human plasma (%)d
LogP log of the octanol-water partition coefficient, Peff effective permeability, pKa acid dissociation constant
a
In vitro measurement made at Roche
b
These values were optimized based on clinical data
c
In silico value calculated with GastroPlus™ version 9.0
d
The protein binding of ganciclovir was measured in-house and was less than 2% bound. Protein binding for valganciclovir was not measured,
but it was assumed to be similar to that of ganciclovir because it is slightly more hydrophilic
e
Simulations for solid dosage forms of both ganciclovir and valganciclovir were not sensitive to dissolution rate, and so, particle radius was not
a sensitive parameter
1455
Clinical data used for model calibration and verification are
listed in Table II. Complete model parameter listings for both
ganciclovir and valganciclovir are listed in Online Resource 2
Tables I through III.
Incorporation of Permeability Limited Distribution and
Active Transport
Both ganciclovir and valganciclovir are hydrophilic,
poorly permeable molecules, and a permeability limitation
was needed when modeling their tissue distribution. This was
achieved using a permeability-surface-area product (PStc)
parameter for each tissue. To reduce the number of model
parameters to be estimated, a single value of PStc per cell
volume (specific PStc) was used for all tissues and was scaled
based on tissue cell volume. Full details on the permeability-
limited distribution model and the calculation of PStc values
are given in Online Resource 2.
In vitro studies have shown that active transport plays
an important role in the pharmacokinetics of both
valganciclovir and ganciclovir and have identified the
specific transporters involved (details provided in BGanciclovir
PBPK Model for Adults^ and BValganciclovir PBPK
Model for Adults^ sections). However, quantitative tissue
expression levels for these transporters are not available and
extrapolation from in vitro data to in vivo has not been
established. Therefore, active transport was added to the
model in those tissues, where evidence exists for transporter
expression and where a significant impact on the pharma-
cokinetics is expected. The transport parameters were then
optimized to match the clinical pharmacokinetic data. For
optimization of active renal secretion, the passive permeability
was assumed to be negligible and the activity of renal
transporters on both apical and basolateral membranes were
adjusted simultaneously based on plasma and urine
Ganciclovir Valganciclovir
255.2 354.4
−1.60 −2.01
0.027 0.117
3.7 at pH = 6.0 20 at pH = 8.91
Acidic 9.4a; basic 2.2a Acidic 10.84; basic 2.44, 7.6a
18.28 40.18
900 (default) 900 (default)
0.972 0.750
1.2 (default) 1.2 (default)
25 (default) 25 (default)
0.89 0.91
99% 99%
1456 Lukacova et al.

Table II. Summary of Studies Used for PBPK Model Development and Evaluation

Study identifier Patient population Number Drug with route Age rangea Reference

Asano-Mori Hematopoietic stem cell transplantation recipients 12 IV ganciclovir 23–61 [28]

with normal and mildly impaired kidney function
Spector Subjects with HIV and stable CMV retinitis PO ganciclovir 38.5b [29]
Pescovitz Liver transplant recipients 28 PO valganciclovir 47.2c [3]
WP15347 HIV- and CMV-seropositive patients 39 PO valganciclovir 20–47 [30]
PV16000 SOT recipients (kidney, liver, liver-kidney, heart, 240 PO valganciclovir 14–71d Roche data on file
and kidney-pancreas)
WP16296 Pediatric kidney transplant recipients 26 IV ganciclovir, PO 3 months to 16 years Roche data on file
valganciclovir
WV16726 Pediatric SOT recipients 63 PO valganciclovir 3 months to 17 years Roche data on file
NP22523 Heart transplant recipients 17 PO valganciclovir <4 months Roche data on file
Acosta Symptomatic congenital CMV patients 24 IV ganciclovir, PO 8–33 days [31]
valganciclovir

CMV cytomegalovirus, IV intravenous, NA not applicable, PO oral

a
The age range specifies the range of actual patients included in the study
b
This is the median age of patients administered 1000-mg ganciclovir PO. A range was not provided
c
This is the mean. A range was not provided
d
This is the age range of patients administered valganciclovir. Although the study included some pediatric patients, the majority were adult,
with a median age of 48 years and a mean of 45.7 years

measurements. Thus, while the overall contribution of
secretion could be estimated, the specific Clint for efflux
on the apical membrane could not be uniquely identified.
Ganciclovir PBPK Model for Adults
Ganciclovir is a substrate for the human MRP4 trans-
porter [19], and based on the MRP4 tissue expression profile
[20, 21], active secretion from hepatocytes and enterocytes
into blood and from kidney cells to the urine was included in
the model. The effect of MATE1 transport in the liver [24]
was incorporated by adding transport through the apical
membrane of hepatocytes into bile. Furthermore, organic
anion transporter 1 (OAT1) transport occurs in vitro [8],
supporting the addition of active uptake from blood into the
kidney. All parameter values for active influx and efflux of
ganciclovir in the kidney, liver, and intestine are given in
Online Resource 1 Table II. The only routes of elimination
for ganciclovir were assumed to be into bile or into urine via
glomerular filtration or active tubular secretion.
Ganciclovir absorption into enterocytes and subsequent
transport into portal vein via MRP4 was simulated based on
intestinal permeability and MRP4 parameters, which were
optimized based on the clinical pharmacokinetic data

Table III. Simulated and Observed Pharmacokinetic Parameters for Ganciclovir After Intravenous Ganciclovir or Oral Valganciclovir in
Pediatric Kidney Transplant Recipients

Age range Observed mean (SD)a Simulated mean (SD) Average fold error (SD)

AUC (μg h/mL) Cmax (μg/mL) AUC (μg h/mL) Cmax (μg/mL) AUC Cmax

Ganciclovir after IV administration of ganciclovir

13–16 (N = 11) 44.42 (25.85) 11.66 (10.95) 37.42 (9.01) 10.03 (3.29) 1.37 (0.44) 1.57 (0.75)
9–12 (N = 7) 40.60 (8.23) 9.46 (2.03) 37.82 (7.71) 10.59 (4.03) 1.15 (0.10) 1.29 (0.19)
5–8 (N = 4) 30.62 (10.13) 9.59 (5.02) 33.15 (13.18) 11.21 (6.46) 1.51 (0.38) 1.71 (0.27)
1–3 (N = 3) 23.78 (8.58) 9.26 (1.95) 34.98 (13.04) 13.36 (0.39) 1.52 (0.62) 1.50 (0.38)
Ganciclovir after PO administration of valganciclovirb
13–16 (N = 11) 38.79 (12.33) 4.98 (2.13) 44.51 (8.75) 5.28 (1.51) 1.45 (0.28) 1.51 (0.46)
9–12 (N = 7) 39.45 (8.39) 5.44 (0.74) 47.98 (8.51) 6.65 (1.16) 1.35 (0.16) 1.30 (0.38)
5–8 (N = 4) 30.90 (22.09) 5.03 (3.67) 44.85 (18.11) 7.73 (1.96) 1.85 (0.94) 1.92 (0.64)
1–3 (N = 3) 17.77 (2.68) 4.62 (1.06) 35.65 (7.40) 7.30 (0.90) 2.04 (0.56) 1.62 (0.35)

AUC area under the plasma concentration-time curve, Cmax maximum plasma concentration, GFR glomerular filtration rate, IV intravenous,
PO oral, SD standard deviation
Kidney transporter expression levels per milligram of tissue were left as the optimized values for adult transplant recipients
a
The data are from study WP16296
b
The model predicted concentrations of both valganciclovir and ganciclovir after valganciclovir PO administration. However, valganciclovir
concentrations were not measured and so only ganciclovir concentrations are shown in the table
Pediatric PBPK Model for Valganciclovir and Ganciclovir
described by Spector et al. 1995 [29] (Table II). The MRP4
distribution along the intestinal tract in human is not known,
and so, a uniform distribution was assumed.
Valganciclovir PBPK Model for Adults
It was assumed that conversion of valganciclovir to
ganciclovir through the esterase valacyclovirase [7] occurred
only in the gut, liver, and kidney. The only other route of
elimination of valganciclovir was renal elimination via
glomerular filtration.
Valganciclovir is a substrate for the human PepT1 influx
transporter [25], which is found in the apical membranes of the
gut and the kidney [26] and also in the basolateral membrane of
hepatocytes [27]. Therefore, active uptake of valganciclovir into
the liver from the blood and into kidney cells from urine through
PepT1 was included in the model. The parameter values for
active influx of valganciclovir in the kidney, liver, and intestine
are given in Online Resource 1 Table II.
The absorption of valganciclovir was simulated based on
intestinal permeability (Table I) and uptake by PepT1, which
were optimized based on clinical pharmacokinetic data
obtained after PO administration of 450, 875, 1750, and
2625 mg in study WP15347 (Table II). The PepT1 distribution
along the human intestinal tract was defined as previously
reported in literature [32], and degradation in the gastroin-
testinal tract was included based on measured in vitro stability
of valganciclovir at a range of pH values [33].
Pediatric PBPK Models
For simulations in children, the PBPK models were
adjusted to account for known age dependencies in physio-
logical parameters which are built-in to PEAR ™ in
GastroPlus 9.0 [34]. These models account for reported age-
dependent values for tissue sizes, blood flows, tissue compo-
sitions, total body water, extracellular tissue water, plasma
protein concentration, and hematocrit. The development of
renal function with age is also taken into account, and where
clinical data were available, glomerular filtration rate (GFR)
was estimated based on measured serum creatinine levels.
A stepwise procedure was followed to scale the adult
model for use in children. Model simulations using the default
age dependencies were first compared to observed plasma
concentrations in age-matched groups of older children
keeping the expression levels of enzymes and transporters
per milligram of tissue (or for intestinal transporters per
square centimeter of intestinal surface area) at the adult
levels. Only if needed for older children, were specific model
adjustments made and then used for predictions in younger
children and infants.
The model was subsequently extended for predictions in
neonates and infants as described in [13] and briefly in Online
Resource 5.
RESULTS
Modeling in Animal Species
Modeling in animal species was used to gain understand-
ing of the pharmacokinetics of gangciclovir and
1457
valganciclovir. Comparison of simulations to the measured
plasma concentrations after intravenous doses revealed that
the simulated volumes of distribution and profile shapes were
best matched when using permeability-limited tissue models
for both drugs. Furthermore, for both ganciclovir and
valganciclovir, it was found that the same specific PStc value
could be applied for all animals. It was found that enzymatic
conversion of valganciclovir to ganciclovir in liver alone was
not sufficient to account for valganciclovir clearance and
glomerular filtration alone was insufficient to match the
observed clearance of ganciclovir. Thus, additional metabo-
lism of valganciclovir in kidney or blood and transporter-
mediated renal tubular secretion of ganciclovir were included
as well as elimination of ganciclovir into the bile. The detailed
results of simulations in animal species are outlined in Online
Resource 1.
Adult Ganciclovir Model
The initial model for ganciclovir disposition based on the
work in animal species was optimized to match the plasma
pharmacokinetics measured in HIV patients without renal
impairment [29]. As already identified in the animal models,
clearance by glomerular filtration alone was insufficient to
match the observed renal clearance and so transporter-
mediated secretion (MRP4, OAT1) was added. Additional
elimination via MATE1 secretion into the bile was also
included in the model. Similar to simulations in animal
species, the contribution of renal and biliary secretion to
ganciclovir elimination was dependent on whether ganciclovir
or valganciclovir was administered. After administration of
ganciclovir, the majority of the drug was eliminated in urine
(99%). However, after oral administration of valganciclovir,
about 30% of the ganciclovir were eliminated in the bile and
70% in urine. This difference is explained by the formation of
higher levels of hepatic ganciclovir after valganciclovir enters
the hepatocyte. To match the simulated profile shape and the
VSS estimated by non-compartmental analysis of both
observed and simulated data (Fig. 2a, b), the specific PStc
value was also adjusted from 0.00031 mL/s/mL-cell volume in
animals to 0.0002 mL/s/mL-cell volume in humans.
Ganciclovir clearance (CL) is lower in organ transplant
recipients than in healthy subjects or HIV patients [28]. The
exact reason for the lower clearance is not clear, but in the
current model, a 60% decrease in the kidney transporter levels
(0.75 vs. 0.3 mg/s/g-tissue for Vmax) was fitted to the in vivo IV
pharmacokinetic data in transplant recipients [3] (Fig. 2e, f).
Simulations of ganciclovir oral absorption were verified
by comparison to steady state plasma levels following oral
administration of 1000 mg every 8 h in HIV patients
(Fig. 2c, d) [29]. Oral simulations in organ transplant
recipients were also consistent with observed data on day 1
following oral ganciclovir administration every 6 h at a dose
of 1000 mg (Fig. 2g, h) [3]. The simulated bioavailability was
8% in both groups of patients, which is in line with values
reported independently [35].
Adult Valganciclovir Model
The conversion of valganciclovir to ganciclovir by
valacyclovirase as well as valganciclovir transport by PepT1
1458 Lukacova et al.

Fig. 2. Mean observed (symbols) and simulated (lines) ganciclovir plasma concentrations following a 1-h IV
infusion of ganciclovir at a dose of 5 mg/kg in a, b 66 adult HIV patients of average age 38 years [29] and e, f 28 liver
transplant recipients of average age 47 years [3]. c, d Profiles on day 14 following 1000-mg oral ganciclovir
administration three times a day in 13 adult HIV patients average age 38 years [29]. g, h Day 1 profiles for oral
ganciclovir administration every 6 h at a dose of 1000 mg in 28 liver transplant recipients of average age 47 years [3].
Concentrations are shown on linear a, c, e, g as well as log scale b, d, f, h. The only model adjustment made to fit the
simulations to the data in liver transplant recipients was a 60% lower expression of kidney transporters

in enterocytes, hepatocytes, and kidney were optimized to
match plasma pharmacokinetic measurements obtained in
HIV- and CMV-seropositive subjects (Fig. 3a–d). The passive
diffusion of valganciclovir through tissue membranes was
modeled using the same specific PStc value as fitted in animal
species. This model was then verified by predicting pharma-
cokinetics in liver transplant recipients (Fig. 3e, f). While
ganciclovir concentrations are simulated reasonably well, the
valganciclovir concentrations are over-predicted for both
doses in the second study. However, the reported observed
valganciclovir concentrations show about a twofold difference
for similar doses from these two studies, suggesting that the
apparent mis-prediction may be a result of variability
between the subject populations in the two studies, which
may in turn be related to the different diseases in the two
patient populations.
Pediatric PBPK Model for Valganciclovir and Ganciclovir 1459

Fig. 3. a–d Mean-observed (symbols) and mean-simulated (lines) plasma concentrations for valganciclovir a, c and
ganciclovir b, d following oral administration of 450 (blue diamond), 875 (red square), (green triangle), or 2625
(purple circle) mg valganciclovir in 28 fasted c, d and fed a, b HIV- and CMV-seropositive adult patients on the third
day of daily dosing [30]. Plots e and f show the observed (symbols) and simulated (lines) plasma concentrations for
valganciclovir e and ganciclovir f following a single oral dose of 450 (blue diamond) or 900 (red square) mg
valganciclovir in 16 adult liver transplant recipients [3]. Note that the simulations in liver transplant recipients used a
model with reduced kidney transporter expressions as optimized for ganciclovir PK reported in organ transplant
recipients

Model simulations also closely agreed with the data from
high-risk adult SOT recipients (study PV16000; Table II), as
shown in Fig. 4 and Online Resource 3. Simulations indicated
that virtually, all valganciclovir are eliminated by conversion to
ganciclovir and the contribution of urinary elimination is
negligible. Simulated bioavailability of valganciclovir after oral
administration of valganciclovir was 8%, while that of ganciclo-
vir was 60–70%. The majority (>98%) of the valganciclovir-to-
ganciclovir conversion was simulated in liver, and less than 1%
was simulated in the gut or kidney. These simulated low
contributions of gut and kidney were based on fitting simula-
tions to the measured valganciclovir pharmacokinetic profiles.
Thus, the high conversion in liver and the low contribution of
kidney were supported by the low observed valganciclovir
bioavailability, while the low contribution of gut compared to
liver was supported by matching the profile shape since gut
metabolism mainly affected the bioavailability through first pass
while liver contributed also to systemic clearance.
Pediatric Model in Children
Simulations in children were compared to data from a
pharmacokinetics and safety study WP16296 (Table II) in 25
pediatric kidney transplant recipients at risk of developing
CMV disease. This study included patients aged from 1 to
16 years. Doses were based on the adult dose adjusted for
BSA and renal function [11]. On study days 1 and 2, all
patients received IV ganciclovir. On day 3, all patients
received valganciclovir oral solution at a dose equivalent to
half the adult dose, and on day 4, the dose was increased to
be equivalent to the full adult dose. The simulated and
observed pharmacokinetic parameters for children of differ-
ent age ranges are compared in Table III and Fig. 5a. Overall
agreement is good, although a tendency to simulate higher
exposures than are observed seems to be more pronounced in
the younger age groups.
1460 Lukacova et al.

Fig. 4. Observed (symbols) and simulated (lines) ganciclovir plasma concentrations on four different days following
daily administration of 900 mg valganciclovir. a The complete time course and b–e the periods with observed data.
The symbols represent the sparse samples taken from individual adult solid organ transplant recipients in study
PV16000 (Table II). The simulation used a model with reduced kidney transporter expressions as optimized for
ganciclovir PK reported in solid organ transplant recipients. The simulation results for 450-mg valganciclovir dose
are reported in Online Resource 3

The model was further verified against data from 13
heart transplant recipients at risk of CMV disease from study
WV16726 with ages ranging from 3 months to 17 years. These
children were administered valganciclovir once daily for up to
100 days with doses adjusted based on BSA and renal
function. The model predictions obtained were consistent
with the observed pharmacokinetic data as shown in Table IV
and Fig. 5b.
Simulations for both studies used physiologies generated
using built-in algorithms in PEAR™ matching reported age,
body weight, and GFR for individual subjects. Kidney
transporter levels were decreased by 60% as this had been
shown to be appropriate for adult transplant recipients.
Simulated and observed ganciclovir pharmacokinetic profiles
for individual subjects from both studies are available in
Online Resource 4.
The model verified against data in children aged 3 months
to 17 years (Tables III and IV, Fig. 5a, b, and Online
Resource 4) was subsequently used to predict ganciclovir
pharmacokinetics after IV ganciclovir and PO valganciclovir
administration in neonates and infants less than 4 months old
as described in [13]. After adjusting for known physiological
changes based on the gestational age, postnatal age, gender,
and body weight of each subject, the model simulations were
in reasonable agreement to observations. Although a slight
bias to under-prediction was noted in the very youngest
subjects, this was not strong enough to justify a dose
adjustment for neonates. These results are shown in Fig. 5c, d
and briefly reported in Online Resource 5.
DISCUSSION
This manuscript describes the stepwise development and
verification of a pediatric PBPK model for ganciclovir and its
prodrug valganciclovir. This work enhances our understanding of
the processes determining pharmacokinetics within this vulnera-
ble patient population, where clinical data are difficult to obtain.
Firstly, a PBPK model was developed to describe
ganciclovir pharmacokinetics after IV and oral administration
to adult patients. This was then extended to valganciclovir
oral pharmacokinetics. This human model was informed by
prior work done to develop a PBPK model for animal species
and took into account drug physicochemical properties,
metabolic biotransformation, and active transport. Key model
Pediatric PBPK Model for Valganciclovir and Ganciclovir 1461

Fig. 5. Simulated and observed ganciclovir plasma concentrations after IV ganciclovir and PO valganciclovir
administration in different pediatric studies. a Study WP16296 in 1- to 16-year-old kidney transplant recipients
(overall RMSE = 2.21, AFE = 1.91), b study WV16726 in 3-month to 17-year-old heart transplant recipients (overall
RMSE = 2.95, AFE = 2.26), c study NP22523 in 1- to 4-month-old heart transplant recipients [13] (overall RMSE =
2.89, AFE = 2.21), and d Acosta study [31] in neonates up to 1 month old [13] (overall RMSE = 3.49, AFE = 6.0 for
simulation with adult expression levels; RMSE = 2.50, AFE = 5.80 for simulation with adjusted expression levels).
Different age groups are shown in different colors: in a, black (1 to 3 years old), green (5 to 8 years old), red (9 to
12 years old), and blue (13 to 16 years old); in b, black (0.3 to 2 years old), green (4 to 6 years old), and blue (15 to
17 years old); and in c, d, the simulations with adult and adjusted expression levels of renal transporters are shown
in blue and red, respectively. The simulations in organ transplant recipients a, b, and c used a model with reduced
kidney transporter expressions as optimized for ganciclovir reported PK in adult organ transplant recipients. RMSE
root-mean-square error across individual concentrations, AFE average fold error across individual concentrations

assumptions were (i) ganciclovir was actively transported kidney tubules, as well as from the liver and gut into the
from the blood into the kidney and from the kidney to the blood; (ii) ganciclovir clearance was renal by glomerular

Table IV. Simulated and Observed Pharmacokinetic Parameters for Ganciclovir After Oral Administration of Valganciclovir in Pediatric
Heart Transplant Recipients

Age range, years Mean observed (SD)a Mean predicted (SD) Average fold error (SD)

AUCb (μg h/mL) Cmax (μg/mL) AUCb (μg h/mL) Cmax (μg/mL) AUC Cmax

0.3–2, N = 6 54.6 (19.4) 11.3 (5.35) 63.6 (4.9) 12.8 (1.83) 1.48 (0.33) 1.63 (0.68)
4–6, N = 3 43.2 (13.2) 8.85 (3.83) 44.2 (22.0) 10.7 (6.23) 1.54 (0.49) 1.28 (0.10)
15–17, N = 4 60.2 (34.3) 9.98 (5.01) 41.0 (7.70) 6.46 (0.58) 1.70 (0.48) 1.75 (0.68)

AUC area under the plasma concentration-time curve, Cmax maximum plasma concentration, SD standard deviation
Kidney transporter expression levels were set to adult values optimized against data from transplant recipients. Valganciclovir concentrations
were not measured in this clinical study
a
The data are for heart transplant recipients from study WV16726. Observed values are reported as mean (SD)
b
AUCs over 12 h were calculated only for patients where at least four data points were reported. AUCs for one patient from group 0.3–2 years
and patient from group 15–17 years were not calculated
1462
filtration and active secretion and biliary secretion; (iii)
valganciclovir undergoes active uptake into enterocytes, liver,
and kidney; (iv) valganciclovir metabolism to form ganciclovir
occurred in enterocytes as well as in the liver and kidney; and
(v) valganciclovir clearance was by conversion to ganciclovir
and passive renal filtration.
Both prodrug and active drug are poorly permeable, and
active transport is crucial in determining their pharmacoki-
netics. In vitro data have characterized the specific trans-
porters involved for both molecules as well as the tissues
where these transporters are expressed and the Km values for
interaction of both compounds with the relevant transporters.
However, quantitative translation of transporter kinetics from
in vitro to in vivo is not currently possible, and therefore,
model parameters for non-saturable transport in the relevant
tissues were optimized so that simulations matched the
available adult clinical pharmacokinetic data. After each
optimization of model parameters, the model was verified
by comparison of simulations to observed data from an
independent study.
This work was based on clinical data from patient groups
suffering from a variety of conditions and diseases including
stem cell transplantation, HIV and stable CMV retinitis, and
organ transplant (Table II). Given that drug pharmacokinet-
ics is known to be affected by disease as well as by the
associated co-administered medications, a high inter-study
variability is to be expected. Indeed, we found a lower
ganciclovir clearance in organ transplant recipients than in
healthy subjects or HIV patients and also noted that
valganciclovir concentrations were higher in liver transplant
recipients than in HIV- and CMV-seropositive subjects.
Neither the effect of co-administered drugs (such as steroids
which are known to be inducers in SOT patients) nor the
effect of disease on physiological factors which might affect
gangciclovir or valgangciclovir pharmacokinetics has been
accounted for mechanistically in our PBPK modeling, which
is a limitation of this work.
The adult PBPK model was then adapted to describe
pharmacokinetics in children. Firstly, the simulation results
were compared to the pharmacokinetic parameters deter-
mined in pediatric kidney transplant recipients administered
IV ganciclovir or oral valganciclovir (study WP16296). A
second study with children of similar ages was used to further
validate the model (study WV16726 in pediatric SOT
recipients), and good predictions were obtained for all age
groups without further model modifications. Our attempt to
bridge the ganciclovir pharmacokinetics in neonates using
PBPK is certainly challenging as the drug and its prodrug
valganciclovir are substrates of many transporters, for which
the expression in children is unknown. As there is no
ontogeny information for the intestinal and hepatic trans-
porters involved in the disposition of VGCV and GCV, the
expression levels of intestinal and hepatic transporters in the
model were kept at the adult values per square centimeter of
intestinal surface area and per gram of liver tissue. Thus, only
changes in tissue size affected the total activity of intestinal
and hepatic transporters. Although no data exist on the
expression in pediatric tissues of transporters relevant for
gangciclovir, a very recent paper [36] does support the
assumption made in our model that the density of PepT1
Lukacova et al.
expression per square centimeter of intestinal surface area is
similar in neonates and adults.
Success in simulating these pediatric populations in-
creased confidence that the pharmacokinetics in infants could
be predicted. The PBPK model for infants and neonates
accounted for age-related changes in tissue compartment
volumes and blood flows as well as tissue composition, while
the renal filtration clearance was estimated from measured
creatinine levels for each infant. In older infants in study
NP22523, the PBPK model provided simulations consistent
with the available pharmacokinetic data. In the youngest
patients, particularly in the study in neonates of 35 days old
or less with congenital CMV [31], a slight bias to under-
prediction was noted, although the under-prediction was not
strong enough to justify a dose adjustment for neonates [13].
Model predictions in the youngest infants were improved
when an ontogeny function for the expression of kidney
transporters in infants of less than 42 days postnatal age was
introduced (Online Resource 5) [13].
CONCLUSIONS
The work described in this manuscript illustrates the use of
PBPK modeling for pediatric drug development. Specifically,
this approach enabled the integration of clinical data in children
from a very limited patient population with age-related changes
in physiology, tissue volumes and composition, and relevant
in vitro data. We found that the default pediatric-PBPK models
accounting for known age-related changes were able to
accurately simulate plasma concentrations in pediatric patients
down to infants. For the neonatal population, there was a slight
under-prediction of exposures which might be related to the
immaturity of renal transporters.
REFERENCES
1. McGavin JK, Goa KL. Ganciclovir: an update of its use in the
prevention of cytomegalovirus infection and disease in transplant
recipients. Drugs. 2001;61(8):1153–83.
2. Cvetkovic RS, Wellington K. Valganciclovir: a review of its use in
the management of CMV infection and disease in immunocom-
promised patients. Drugs. 2005;65(6):859–78.
3. Pescovitz MD, Rabkin J, Merion RM, Paya CV, Pirsch J,
Freeman RB, et al. Valganciclovir results in improved oral
absorption of ganciclovir in liver transplant recipients.
Antimicrob Agents Chemother. 2000;44(10):2811–5.
4. Humar A, Snydman D. Cytomegalovirus in solid organ trans-
plant recipients. Am J Transplant. 2009;9:S78–86.
5. Vaziri S, Pezhman Z, Sayyad B, Mansouri F, Janbakhsh A,
Afsharian M, et al. Efficacy of valganciclovir and ganciclovir for
cytomegalovirus disease in solid organ transplants: a meta-
analysis. J Res Med Sci Off J Isfahan Univ Med Sci.
2014;19(12):1185–92.
6. Paya C, Humar A, Dominguez E, Washburn K, Blumberg E,
Alexander B, et al. Efficacy and safety of valganciclovir vs. oral
ganciclovir for prevention of cytomegalovirus disease in solid
organ transplant recipients. Am J Transplant. 2004;4(4):611–20.
7. Kim I, Chu XY, Kim S, Provoda CJ, Lee KD, Amidon GL.
Identification of a human valacyclovirase: biphenyl hydrolase-
like protein as valacyclovir hydrolase. J Biol Chem.
2003;278(28):25348–56.
8. Puente XS, Lopez-Otin C. Cloning and expression analysis of a
novel human serine hydrolase with sequence similarity to
Pediatric PBPK Model for Valganciclovir and Ganciclovir
prokaryotic enzymes involved in the degradation of aromatic
compounds. J Biol Chem. 1995;270(21):12926–32.
9. Welker H, Farhan M, Humar A, Washington C. Ganciclovir
pharmacokinetic parameters do not change when extending
valganciclovir cytomegalovirus prophylaxis from 100 to 200 days.
Transplantation. 2010;90(12):1414–9.
10. Caldes A, Colom H, Armendariz Y, Garrido MJ, Troconiz IF,
Gil-Vernet S, et al. Population pharmacokinetics of ganciclovir
after intravenous ganciclovir and oral valganciclovir administra-
tion in solid organ transplant patients infected with cytomegalo-
virus. Antimicrob Agents Chemother. 2009;53(11):4816–24.
11. Vaudry W, Ettenger R, Jara P, Varela-Fascinetto G, Bouw MR,
Ives J, et al. Valganciclovir dosing according to body surface area
and renal function in pediatric solid organ transplant recipients.
Am J Transplant. 2009;9(3):636–43.
12. Bradley D, Moreira S, Subramoney V, Chin C, Ives J, Wang K.
Pharmacokinetics and safety of valgancyclovir in pediatric heart
transplant recipients aged <4 months submitted for publication. 2015.
13. Jorga K, Chavanne C, Frey N, Lave T, Lukacova V, Parrott N,
et al. Bottom-up meets top-down: complementary PBPK and
popPK modeling for regulatory approval of dosing algorithm of
valganciclovir in very young children. Clinical Pharmacology &
Therapeutics (submitted). 2016.
14. Rowland M, Peck C, Tucker G. Physiologically-based pharma-
cokinetics in drug development and regulatory science. Annu
Rev Pharmacol Toxicol. 2011;51:45–73.
15. Leong R, Vieira M, Zhao P, Mulugeta Y, Lee C, Huang S-M,
et al. Regulatory experience with physiologically based pharma-
cokinetic modeling for pediatric drug trials. Clin Pharmacol
Ther. 2012;91(5):926–31.
16. Abdel-Rahman SM, Amidon GL, Kaul A, Lukacova V, Vinks
AA, Knipp GT. Summary of the National Institute of Child
Health and Human Development—best pharmaceuticals for
Children Act Pediatric Formulation Initiatives
Workshop—Pediatric Biopharmaceutics Classification System
Working Group. Clin Ther. 2012;34(11):S11–24.
17. Kearns GL. Developmental pharmacology—drug disposition,
action, and therapy in infants and children. N Engl J Med. 2003.
18. Tayman C, Rayyan M, Allegaert K. Neonatal pharmacology:
extensive interindividual variability despite limited size. J
Pediatric Pharmacol Ther. 2011;16(3):170–84.
19. Adachi M, Sampath J, Lan LB, Sun D, Hargrove P, Flatley R,
et al. Expression of MRP4 confers resistance to ganciclovir and
c o m p r o m i s e s b y s t a n d e r c e l l k i l l i n g . J B i o l C he m .
2002;277(41):38998–9004.
20. Lee W, Kim RB. Transporters and renal drug elimination. Annu
Rev Pharmacol Toxicol. 2004;44:137–66.
21. International Transporter C, Giacomini KM, Huang SM,
Tweedie DJ, Benet LZ, Brouwer KL, et al. Membrane trans-
porters in drug development. Nat Rev Drug Discov.
2010;9(3):215–36.
22. Maubon N, Le Vee M, Fossati L, Audry M, Le Ferrec E, Bolze
S, et al. Analysis of drug transporter expression in human
intestinal Caco-2 cells by real-time PCR. Fundam Clin
Pharmacol. 2007;21:659–63.
23. Prime-Chapman H, Fearn R, Cooper A, Moore V, Hirst B.
Differential multidrug resistance-associated protein 1 through 6
1463
isoform expression and function in human intestinal epithelial
Caco-2 cells. J Pharm Exp Ther. 2004;311:476–84.
24. Tanihara Y, Masuda S, Sato T, Katsura T, Ogawa O, Inui K.
Substrate specificity of MATE1 and MATE2-K, human multi-
drug and toxin extrusions/H(+)-organic cation antiporters.
Biochem Pharmacol. 2007;74(2):359–71.
25. Sugawara M, Huang W, Fei YJ, Leibach FH, Ganapathy V,
Ganapathy ME. Transport of valganciclovir, a ganciclovir
prodrug, via peptide transporters PEPT1 and PEPT2. J Pharm
Sci. 2000;89(6):781–9.
26. Ito K, Suzuki H, Horie T, Sugiyama Y. Apical/basolateral surface
expression of drug transporters and its role in vectorial drug
transport. Pharm Res. 2005;22(10):1559–77.
27. Hilgendorf C, Ahlin G, Seithel A, Artursson P, Ungell A-L,
Karlsson J. Expression of thirty-six drug transporter genes in
human intestine, liver, kidney, and organotypic cell lines. Drug
Metab Dispos. 2007;35(8):1333–40.
28. Asano-Mori Y, Kanda Y, Oshima K, Watanabe T, Shoda E,
Motokura T, et al. Pharmacokinetics of ganciclovir in
haematopoietic stem cell transplantation recipients with or
without renal impairment. J Antimicrob Chemother.
2006;57(5):1004–7.
29. Spector SA, Busch DF, Follansbee S, Squires K, Lalezari JP,
Jacobson MA, et al. Pharmacokinetic, safety, and antiviral
profiles of oral ganciclovir in persons infected with human
immunodeficiency virus: a phase I/II study. AIDS Clinical Trials
Group, and Cytomegalovirus Cooperative Study Group. J Infect
Dis. 1995;171(6):1431–7.
30. Brown F, Banken L, Saywell K, Arum I. Pharmacokinetics of
valganciclovir and ganciclovir following multiple oral dosages of
valganciclovir in HIV- and CMV-seropositive volunteers. Clin
Pharmacokinet. 1999;37(2):167–76.
31. Acosta EP, Brundage RC, King JR, Sanchez PJ, Sood S,
Agrawal V, et al. Ganciclovir population pharmacokinetics in
neonates following intravenous administration of ganciclovir and
oral administration of a liquid valganciclovir formulation. Clin
Pharmacol Ther. 2007;81(6):867–72.
32. Herrera-Ruiz D, Wang Q, Cook T, Knipp G, Gudmundsson O,
Smith R, et al. Spatial expression patterns of peptide transporters
in the human and rat gastrointestinal tracts, Caco-2 in vitro cell
culture model, and multiple human tissues. AAPS J.
2001;3(1):100–11.
33. Stefanidis D, Brandl M. Reactivity of valganciclovir in aqueous
solution. Drug Dev Ind Pharm. 2005;31(9):879–84.
34. Simulations Plus, Inc., GastroPlus user manual v9.0. Lancaster,
California 93534, 2015.
35. Anderson RD, Griffy KG, Jung D, Dorr A, Hulse JD, Smith RB.
D J, A D, JD H, RB S. Ganciclovir absolute bioavailability and
steady-state pharmacokinetics after oral administration of two
3000-mg/d dosing regimens in human immunodeficiency virus-
and cytomegalovirus-seropositive patients. Clin Ther.
1995;17(3):425–32.
36. Mooij MG, de Koning BAE, Lindenbergh-Kortleve DJ,
Simons-Oosterhuis Y, van Groen BD, Tibboel D, et al. Human
intestinal PEPT1 transporter expression and localization in
preterm and term infants. DRUG METABOLISM AND
DISPOSITION. 2016.