European Journal of Clinical Pharmacology

https://doi.org/10.1007/s00228-018-2440-6

PHARMACOKINETICS AND DISPOSITION

Effect of CYP2C19, UGT1A8, and UGT2B7 on valproic acid clearance

in children with epilepsy: a population pharmacokinetic model
Shenghui Mei 1,2 & Weixing Feng 3,4 & Leting Zhu 1 & Xingang Li 1 & Yazhen Yu 4 & Weili Yang 4 & Baoqin Gao 4 &
Xiaojuan Wu 4 & Fang Fang 3 & Zhigang Zhao 1,2

Received: 7 December 2017 / Accepted: 2 March 2018

# Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
Purpose Valproic acid (VPA) is an important drug in seizure control with great inter-individual differences in metabolism and
treatment effect. This study aims to identify the effects of genetic variants on VPA clearance in a population pharmacokinetic
(popPK) model in children with epilepsy.
Methods A total of 325 VPA plasma concentrations from 290 children with epilepsy were used to develop the popPK model by
using the nonlinear mixed-effects modeling method. The one-compartment model was established to describe the pharmacoki-
netics of VPA. Twelve single nucleotide polymorphisms involved in the pharmacokinetics of VPA were identified by
MassARRAY system and their effects on VPA clearance were evaluated.
Results In the two final popPK models, inclusion of a combined genotype of four variants (rs1042597, rs28365062, rs4986893,
and rs4244285), total daily dose (TDD), and body surface area (BSA) significantly reduced inter-individual variability for
clearance over the base model. The inter-individual clearance equals to 0.73 × (TDD/628.92)0.59 × eUGT-CYP for TDD included
model and 0.70 × (BSA/0.99)0.57 × eUGT-CYP for BSA included model. The precision of all parameters were acceptable (relative
standard error < 32.81%). Bootstrap and visual predictive check results indicated that both two final popPK models were stable
with acceptable predictive ability.
Conclusion TDD, BSA, and genotype might affect VPA clearance in children. The popPK models may be useful for dosing
adjustment in children on VPA therapy.

Keywords Valproic acid . Uridine diphosphate glucuronosyltransferase . Cytochrome P450 family 2 subfamily C member 19 .
Clearance . Children . Population pharmacokinetic model . Nonlinear mixed-effects modeling

Introduction person ranged from $10,192 to $47,862 for general epilepsy

populations in the USA [1]. Valproic acid (VPA) is a first-line
Epilepsy is a global health problem with particularly high drug in treating patients with various kinds of seizures [2].
morbidity in children. The total direct healthcare costs per Due to its narrow therapeutic range and wide inter- and

Shenghui Mei and Weixing Feng are equal first authors.

Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s00228-018-2440-6) contains supplementary
material, which is available to authorized users.

* Fang Fang 2
Department of Clinical Pharmacology, College of Pharmaceutical
13910150389@163.com Sciences, Capital Medical University, Beijing 100045, People’s
Republic of China
* Zhigang Zhao
3
ttyyzzg1022@126.com Department of Neurology, Beijing Children’s Hospital, Capital
Medical University, 56 Nanlishi Road, Xicheng District,
1
Department of Pharmacy, Beijing Tiantan Hospital, Capital Medical Beijing 100045, People’s Republic of China
University, 6 Tiantan Xili, Dongcheng District, Beijing 100050, 4
Department of Pediatrics, Beijing Tiantan Hospital, Capital Medical
People’s Republic of China University, Beijing 100050, China

intra-individual pharmacokinetic variability, the dosage ad-
justment is a challenge for clinicians because under dosing
might evoke epilepsy, whereas overdoing increased side ef-
fects, especially in pediatric patients [3]. Various factors in-
volved in VPA absorption (diet, soybean intake, and dosage
form) [4], distribution [body weight (BW), age, total daily
dose (TDD), and protein binding] [5–11], and metabolism
(gender, age, hepatic function, TDD, genotype of enzymes
related to VPA glucuronidation and oxidation, co-
administered enzyme inducers of antiepileptic drugs such as
phenobarbital, phenytoin, and carbamazepine) have signifi-
cant influence on VPA plasma concentration [3, 5, 6, 8,
10–24]. For example, cytochrome P450 family 2 subfamily
C member 9*3 (CYP2C9*3, rs1057910) was associated with a
decreased enzyme activity [25] and an increased VPA concen-
tration and clearance [3, 12, 18, 19].
The quantitative relation between VPA plasma concentra-
tion and these influence factors should be identified prior to
the individualized therapy. Therefore, various population
pharmacokinetic (popPK) models were constructed by using
the nonlinear mixed-effects modeling method (NONMEM)
[5–16, 22]. The pharmacokinetics of VPA could be best esti-
mated by one-compartment model with first-order absorption
and elimination [5, 6, 8–16]. Only two studies used two-
compartment model for data fitting [7, 22]. In previous pub-
lished studies, BW, age, and TDD could affect VPA distribu-
tion volume [6–11], while BW, age, gender, TDD, and co-
medications had significant influence on VPA clearance
[5–16]. Although various genetic polymorphisms were asso-
ciated with the increase or decrease of VPA plasma concen-
tration [3, 12, 17–21, 23, 24], only one study identified the
quantitative relationship between genetic polymorphisms [cy-
tochrome P450 family 2 subfamily C member 19 (CYP2C19)
and CYP2C9 genotype] and VPA clearance [12].
The aim of this study is to identify the influence of meta-
bolic enzyme genotypes and patients’ characteristics on VPA
pharmacokinetic parameters in a popPK model, which might
be useful for VPA dose adjustment in clinical practice.
Methods
Study design
This study was approved by the Ethics Committee of Beijing
Tiantan Hospital, Capital Medical University, Beijing, China.
Consent from the parents and assent from children were ob-
tained. All patients were Chinese origin. The inclusion and
exclusion criteria were the same as that mentioned in our
previous study [23].
A total of 290 children (325 plasma concentrations) on
VPA treatment in Beijing Tiantan Hospital were enrolled ac-
cording to the criteria described above. Patients’ demographic
Eur J Clin Pharmacol
data [gender, age, BW, and body surface area (BSA)], dosage
regimen (dose, dosing time, and frequency), and co-
administered antiepileptic drugs were recorded. After regular-
ly taking VPA for at least 6 days, the blood samples were
collected and the sampling time was recorded. The blood
plasma was used for VPA concentration analysis while the
white blood cells were used for genotyping [26].
Plasma concentration of VPA
The total plasma concentrations of VPA were evaluated by a
fluorescence polarization immunoassay (TDx, ABBOTT,
USA). Calibrators and quality control samples were routinely
performed according to the manufacturer’s instructions for
quality control.
Genotype identification
Based on previous studies and the PharmGKB database, 12
genetic variants were selected [3, 12, 17–21, 23, 24]. Patients’
DNA were purified by QIAamp DNA purification kit
(Qiagen, Hilden, Germany). Genotyping was identified in
Bio Miao Biological Technology (Beijing) by MassArray
method (Sequenom, USA) [26]. To control the quality of
genotyping, 5% of the whole patients were measured repeat-
edly, and the results were acceptable.
Grouping and combination of variants
Each of the selected variants was divided into three or two
groups by genotype. The group number of variants increased
with the decrease of enzyme activity or VPA clearance
(Table 1). Any two of the variants that were divided into three
groups were combined into a new variant by summation of
their group numbers. Then, the combined new variant was
separated into three or two groups according to the combina-
tion rules (Table 2) and the new group number increased with
the decrease of enzyme activity and VPA clearance.
Statistical analysis
Statistical analysis was carried out by PLINK software (ver-
sion 1.07, Shaun Purcell, Boston, USA). For all selected al-
leles, the minor allele frequency, genotype, and Hardy-
Weinberg Equilibrium (P value, chi-square test) were
calculated.
Population pharmacokinetic model development
To develop the popPK model, a nonlinear mixed-effects
modeling approach was performed by using the Phoenix®
NLME™ 7.0 (Certara, St. Louis, MO) software. The first-
order conditional estimation-extended least squares method,
Eur J Clin Pharmacol

Table 1 Grouping rules for

variants Group type Enzyme activity Wild-type Heterozygote Variant
changing by variants homozygote homozygote

Three groups Increased 3 2 1

Decreased 1 2 3
Two groups, rule 1 Increased 2 1 1
Decreased 1 2 2
Two groups, rule 2 Increased 2 2 1
Decreased 1 1 2

which was equivalent to the NONMEM first-order condition-
al estimation methodology with interaction, was used for
model construction [27]. VPC and bootstrap were performed
to test the predictive ability and stability of the final model,
respectively.
the absorption site and central compartment, respectively. Cc
represents the VPA concentration in the central compartment.
Population pharmacokinetic model

After the construction of the base model, the influence of

Base model
The one-compartment model with first-order absorption and
first-order elimination was used to describe the time course of
VPA in human plasma (Appendix 1) [5, 6, 8–16, 22]. The
model was parameterized by using the absorption rate con-
stant (Ka), apparent volume of central compartment (Vc/F),
and apparent total clearance (CL/F). Very few samples were
at the absorption phase, and the VPA formulations were the
same to a published study [6]. Therefore, Ka was fixed at 2.64
and 0.46 h−1 for immediate release tablets/solutions and con-
trolled release tablets, respectively [6]. The residual error was
characterized by the multiple model. The following equations
are used to describe the model:
dAa =dt ¼ −K a  Aa ð1Þ
dAc =dt ¼ K a  Aa −CLc  C c ð2Þ
C c ¼ Ac =V c ð3Þ
various covariates on VPA pharmacokinetic parameters
was evaluated. All of the continuous covariates were cen-
tered at their mean or median values. Between- and intra-
subject variability (η and ε) of pharmacokinetic parame-
ters were assumed to follow a normal distribution with a
mean of zero and a variance of ω2 and σ2, respectively. In
the present study, most patients had only one observation,
which might be associated with a higher value of shrink-
age [28]. Therefore, the differences in the objective func-
tion value (OFV) were used for model comparison [6]. In
covariate selection, compared to its previous model, an
intermediate model with an OFV decrease > 6.635
(P < 0.01, df = 1) was considered to be superior for for-
ward addition. After all covariates were added, the full
model was refined by removing the added covariates
using a tougher criterion: an increase in the OFV of >
10.828 (P < 0.001, df = 1) [5–7, 10–12, 29].

Ka and CLc/F represent the absorption rate and clearance of Goodness-of-fit and model evaluation
VPA, respectively. Aa and Ac represent the amount of VPA in
OFV and the scatter plots including observed concentra-
tions versus population predicted concentrations, condi-
Table 2 Combination rules for any two of variants tional weighted residuals (CWRES) versus population pre-
dicted concentrations, CWRES versus time after the last
Combined new group type Summation of two Combined new dose, and CWRES versus standard normal quantiles were
group numbers group number
used to evaluate the goodness-of-fit between base model
Three groups 2 1 and final model. A total of 1000 bootstrap and 2000 VPC
3 2 were performed to evaluate the stability and predictive
4, 5 3 ability of the final model, respectively [30]. The median,
Two groups, rule 1 2 1 the 2.5–97.5% intervals of estimated parameters, and the
3, 4, 5 2 5–95% prediction intervals of the simulated data were cal-
Two groups, rule 2 2, 3 1 culated. The prediction ability of the final popPK model is
4, 5 2 acceptable when 90% of the measured VPA concentrations
are within the 90% prediction interval.

Results
Demographic data and genotyping of enrolled
patients
A total of 325 VPA plasma concentrations were obtained from
290 children (108 females and 182 males). A total of 196
patients were treated by controlled release tablets of VPA,
while 94 children were treated by immediate release tablets
or solutions. Patients’ characteristics and dosage regimens are
shown in Appendix 2 (detail in Appendix 6). The base infor-
mation of the 12 identified variants is listed in Appendix 3.
The frequency of selected gene locus all conformed to the
Hardy-Weinberg equilibrium (P > 0.05).
Model construction and final popPK model
The covariates were added in descending order of reduction in
the OFV value (Appendix 4). Two final popPK models were
obtained by using TDD or BSA as covariates. TDD, BSA, and
uridine diphosphate glucuronosyltransferase and cytochrome
P450 (UGT-CYP) genotype have significant influence on
VPA clearance. Their quantitative relationships are listed
below:
For model with TDD,
CL=F ¼ 0:73  ðTDD=628:92Þ0:59  eUGT−CYP  eηCL ð4Þ
V c = F ¼ 22:12  eηV c ð5Þ
For model with BSA,
CL=F ¼ 0:70  ðBSA=0:99Þ0:57  eUGT−CYP  eηCL ð6Þ
ηV c
V c = F ¼ 18:36  e ð7Þ
where 0.73 and 0.70 (L/h) are the typical value of CL/F (L/h),
and 22.12 and 18.36 (L) are typical value of Vc/F (L). The
mean value of TDD is 628.92 mg/day, and the median value
of BSA is 0.99 m2. The estimated coefficients representing the
relationship between clearance and TDD or BSA are 0.59 and
0.57, respectively. In Eqs. 4 and 6, UGT-CYP is the genotype
of four combined variants (rs1042597, rs28365062,
rs4986893, and rs4244285), and UGT-CYP = 0 for patients
with group 1 genotype, otherwise UGT-CYP = − 0.22 for Eq.
4 and − 0.19 for Eq. 6. Table 3 lists the estimate, standard
error, 95% confidence interval, and inter-individual variability
of the parameters for the base model, two final models, and
bootstrap.
Goodness-of-fit and model evaluation
Four pairs of scatter plots are shown in Fig. 1 to evaluate the
goodness-of-fit of the final popPK models. In general, the two
final popPK models obviously improved data fitting
Eur J Clin Pharmacol
compared to the base model. The success rate (successful in
minimization) of both bootstrap analyses was 100%. The me-
dian value and the distribution of the popPK parameters ob-
tained from bootstrap were similar to the values observed in
the final popPK model, which indicated that the final popPK
model was robust (Table 3). For popPK models with TDD or
BSA as covariates, 89.2% (35/325) and 90.4% (31/325) of the
measured VPA plasma concentrations fell inside the 90% pre-
diction interval (Appendix 5). Stratified VPC analysis with
genotype indicated that both two popPK models did a better
job for genotype 2 versus genotype 1. Moreover, the predic-
tive ability of BSA-included popPK model was superior than
the TDD-included model.
Discussion
Model development
The one-compartment popPK model, which has been widely
used in previous studies [5, 6, 8–16], was employed to de-
scribe the pharmacokinetic property of VPA in our patients. In
one of the two final popPK models, TDD has an extremely
significant influence on VPA clearance, which has been re-
ported in four published popPK models [5, 10, 11, 13], and it
could be partly explained by the fact that patients with higher
VPA clearance need a higher dose to ensure the VPA concen-
tration within the therapeutic window (also known as thera-
peutic drug monitoring effect) [6]. The mechanism for this
phenomenon is due to protein binding being saturable in the
therapeutic range, increasing doses result in increased free
fraction and therefore higher clearance [5, 6]. TDD has a sig-
nificant influence on Vc/F in two popPK studies [10, 11]. In
the present study, TDD alone had a significant influence on
Vc/F, but this effect was removed when it was added after the
addition of TDD on clearance.
BSA, an important indicator, was used widely in popula-
tion pharmacokinetic analyses to describe inter-individual var-
iance in drug clearance [31]. Two VPA popPK studies chose
BSA as covariates, but neither of them successfully added
BSA on any of the parameters [13, 16]. In the final popPK
model without TDD, BSA increased with the increase of VPA
clearance, which could be explained by the following reasons:
VPA is excreted partly via kidney, and higher BSA is related
to a higher basal metabolic rate and glomerular filtration rate
[27]; in children, VPA dose was mainly depended on the BW,
which was highly correlated with BSA [31].
Glucuronidation and mitochondrial β-oxidation are two
major pathways for VPA metabolism in adults. About 30–
50% of VPA was metabolized into VPA-glucuronide conju-
gate by uridine diphosphate glucuronyl transferase family
members including uridine diphosphate glucuronosyltransfer-
ase family 1 member A3/4/6/8/9/10 (UGT1A3/4/6/8/9/10)
Eur J Clin Pharmacol

Table 3 Parameter estimates and bootstrap results of valproic acid population pharmacokinetic model in children with epilepsy
Parameters Base model Final model-TDD Bootstrap-TDD Final model-BSA Bootstrap-BSA
Estimate 95% CI IIV Estimate 95% CI IIV Median 95% CI Estimate 95% CI IIV Median 95% CI
(% SE) (CV%) (% SE) (CV%) (% SE) (% SE) (CV%) (% SE)
Ka (h 1)
ˉ
2.64 or 0.46* 2.64 or 0.46* 2.64 or 2.64 or 0.46* 2.64 or 0.46*
0.46*
CL (L/h) 0.61 (0.021) 0.57 to 0.65 38.11 0.73 (0.038) 0.66 to 0.81 25.86 0.73 (5.56) 0.65 to 0.81 0.70 (0.039) 0.63 to 0.78 32.84 0.70 (0.037) 0.63 to 0.78
V (L) 24.93 (4.14) 16.78 to 33.08 6.42 22.12 (7.26) 7.84 to 36.40 6.47 21.83 (7.25) 13.30 to 41.75 18.36 (4.15) 10.20 to 26.52 6.49 17.46 (4.22) 12.36 to 28.91
TDD on CL (L/h) - - 0.59 (0.072) 0.45 to 0.73 0.60 (0.072) 0.46 to 0.75 0.57 (0.084) 0.41 to 0.74 0.57 (0.083) 0.42 to 0.73
UGT-CYP - - ˉ0.22 ˉ0.33 to ˉ ˉ0.22 ˉ0.33 to ˉ ˉ0.19 ˉ0.31 to ˉ ˉ0.19 ˉ0.32 to ˉ
genotype (0.056) 0.11 (0.056) 0.11 (0.061) 0.072 (0.061) 0.077
on CL (L/h)
σ (multiple, 0.35 (0.023) 0.30 to 0.39 0.33 (0.017) 0.29 to 0.36 0.32 (0.020) 0.28 to 0.36
mg/L) 0.32 (0.026) 0.27 to 0.37 0.31 (0.039) 0.22 to 0.35
TDD total daily dose, BSA body surface area, RSE relative standard error, IIV inter-individual variability, CV coefficient of variation; 95%CI 95%
confidence interval, Ka absorption rate, CL clearance, V distribution volume, UGT-CYP uridine diphosphate glucuronosyltransferase and cytochrome
P450 family, σ coefficient variation of intra-individual variability
*Ka was fixed at 2.64 h−1 for immediate release tablets/solutions, and 0.46 h−1 for controlled release tablets

and uridine diphosphate glucuronosyltransferase family 2
member B7/15 (UGT2B7/15) [32]. The influence of 12
selected variants on VPA clearance was evaluated.
However, none of them had significant influence
(P < 0.001) on VPA clearance after the addition of TDD
or BSA on clearance. This could be explained by two rea-
sons: the extensive metabolism of VPA via various kinds of
enzymes, and the limited effect of a single nucleotide poly-
morphism on enzyme activity and VPA metabolism.
Therefore, combination of the variants was performed. In
our two final popPK models, the combined genotype
(rs1042597, rs28365062, rs4986893, and rs4244285) has sig-
nificant influence on VPA clearance, and it was partly in agree-
ment with the results in the published popPK model, which
indicated that rs4986893, rs4244285, and rs1057910 had sig-
nificant influence on VPA clearance [12]. Interestingly, when
we combined rs1057910 with the combined variant of the four
variants, the OFV value increased by 4.56. The effect of
rs4986893 and rs4244285 on VPA clearance could be ex-
plained by the fact that both CYP2C19*2 (rs4244285) and
CYP2C19*3 (rs4986893) resulted in nonfunctional protein ex-
pression [33, 34] and subsequently reduced VPA metabolism.
When added on CL/F alone, rs1042597 (C > G, UGT1A8) in-
creased VPA clearance, this finding was in accordance with its
effect on raloxifene glucuronidation [35]. rs28365062 (A > G,
UGT2B7) decreased the VPA clearance, and the result demon-
strated that this variant was associated with a decreased enzyme
activity, which was in agreement with its effect on efavirenz
metabolism [36].
Cytochrome P450 family 2 subfamily C member 9
(CYP2C9), cytochrome P450 family 2 subfamily A
member 6 (CYP2A6), and cytochrome P450 family 2

Fig. 1 Goodness-of-fit plot for the base model and the final model. a The time after the last dose. d The conditional weighted residuals versus
observed concentrations versus population predicted concentrations. b standard normal quantiles. Red lines represent LOESS smoothing. TDD
The conditional weighted residuals (CWRES) versus population total daily dose, BSA body surface area
predicted concentrations. c The conditional weighted residuals versus

subfamily B member 6 (CYP2B6) account for 15–20%
of VPA metabolism [19, 37]. However, the lower expres-
sion of uridine 5′-diphosphate-glucuronyl transferases
[38, 39] and the higher cytochrome P450 enzyme activ-
ities in children as compared to adults [40, 41], and the
inhibition of mitochondrial β-oxidation by VPA and its
metabolites [42, 43] indicated that cytochrome P450 en-
zymes might be more important for children than for
adults in VPA metabolism [3]. CYP2C9*3 (1075 A > C,
rs1057910) was associated with a decreased enzyme ac-
tivity [25] and an increased VPA plasma concentrations
[3, 18, 19], and it has been used to adjust VPA clearance
in the popPK model [12]. Moreover, CYP2C9*3 was
used in clinical practice for dose adjustment to avoid
adverse drug reactions [3]. However, the influence of
CYP2C9*3 on VPA clearance was not observed in our
study due to its extremely low frequency (0.038 in our
patients). The enzyme activity (represented by the
bupropion clearance) in CYP2B6*4 (rs2279343) carriers
was 1.66-fold higher than wild-type carriers [44], while
CYP2A6*9 (rs28399433) resulted in a reduced level of
mRNA and protein expression [45, 46]. However, neither
CYP2B6*4 nor CYP2A6*9 has significant influence on
VPA clearance in our final popPK model.
In three popPK models of children, age had significant
influence on VPA clearance and Vc/F [6–8]. The effect of
age on VPA clearance could be explained by its effect on
glucuronidation, a major pathway in VPA metabolism.
Generally, the hepatic glucuronidation activities are low
in infants, particularly in children younger than 2 years
old, and reach the adult levels after 10–15 years old [38,
39]. The increase of Vc/F with age could be understood as
the increase of BW with age [7]. In our popPK model, age
had significant influence on VPA clearance and Vc/F when
they were added alone, but both of the two effects were
gone when they were added after the addition of TDD or
BSA on clearance [5, 10, 13].
BW has significant influence on VPA clearance and Vc/
F in four popPK models in children [6–8, 13]. The influ-
ence of BW on Vc/F is easy to understand when patients
with a larger BW have a higher distribution volume, and
the effect of BW on VPA clearance in children may be
explained by the correlation between BW and age, which
was associated with VPA clearance in two popPK studies
in children [6, 8]. The influence of BW on VPA clearance
and Vc/F was observed in our model when each of them
was added alone. However, after the addition of TDD on
clearance, these effects were gone. The TDD of VPA usu-
ally increases with BW. In our children, TDD, BW, and age
were collinear. TDD was correlated with BW (r = 0.49) and
age (r = 0.53), and BW was strongly correlated with age
(r = 0.906). Moreover, clearance and distribution volume
has mathematical relationship (CL = kV). When TDD was
Eur J Clin Pharmacol
added on clearance with a reduction of OFV value by 87.2,
the influence of age and BW on clearance or Vc/F was
disappeared.
Enzyme inducers of co-medicated antiepileptic drugs
could increase VPA clearance [5, 6, 8, 10, 11, 13, 14,
16]. In our patients, only 12 patients were co-medicated
with at least one of the three enzyme inducers (carbamaz-
epine, phenobarbital, and phenytoin), and their influences
on VPA clearance were not observed in our study as well
as in other models [7, 9, 12, 15]. Females have lower VPA
clearance than males in three VPA popPK models [10, 11,
14]; however, this finding was neither confirmed by other
VPA popPK models [5–9, 12, 13, 15, 16] nor by ours. VPA
clearance increased with TDD:BW ratio in two published
popPK models [6, 16]; however, it was not confirmed in
our models.
Deficiencies of the study
Limitations of the study are listed below: (1) the sample size
was relatively small (290 patients with 325 observations); (2)
the Ka was fixed because very few samples were in the ab-
sorption phase [6]; (3) despite their influence on VPA metab-
olism, variants such as CYP2C9*2 were not chosen for anal-
ysis for their low frequency (minor allele frequency < 0.05)
[12, 17, 47]; (4) the external evaluation of the model was not
performed because data with genetic information could not be
obtained from other study groups; (5) the influence of food on
VPA metabolism was not considered [4]. However, this influ-
ence might be very limited because most of the observations
were trough concentrations; (6) the analytical method used in
the present study could not measure 5 of active VPA metabo-
lites; therefore, the influence of these genetic polymorphisms
on the concentration of VPA metabolites could not be evalu-
ated; (7) the developed popPK model was not performed in
actual clinical practice. However, Lin et al. (2015) used their
popPK model in case examples for VPA dose adjustment with
satisfactory results [5]. Moreover, Budi et al. (2015) nicely
present real utility of CYP2C9 genotype guided VPA dosing
to avoid the misdosing-induced side effects [3].
Conclusion
Two popPK models for VPA in children with epilepsy have
been successfully developed. TDD, BSA, and UGT-CYP ge-
notype have significant influence on VPA clearance.
Bootstrap and VPC results indicated that the two final
popPK models were stable with acceptable predictive ability.
The popPK models may be useful for dosing adjustment in
children on VPA therapy. Further studies are warranted to
confirm the results.
Eur J Clin Pharmacol
Acknowledgements Thanks are given to our patients.
Funding Author Weixing Feng was supported by the National Natural
Science Foundation of China (No. 81301118).
Compliance with ethical standards
All procedures performed in studies involving human participants were in
accordance with the ethical standards of the institutional and/or national
research committee and with the 1964 Declaration of Helsinki and its later
amendments or comparable ethical standards.
Conflict of interest The authors declare that they have no conflict of
interest.
References
1. Begley CE, Durgin TL (2015) The direct cost of epilepsy in the
United States: a systematic review of estimates. Epilepsia 56:1376–
1387. https://doi.org/10.1111/epi.13084
2. Loscher W (2002) Basic pharmacology of valproate: a review after
35 years of clinical use for the treatment of epilepsy. CNS Drugs 16:
669–694
3. Budi T, Toth K, Nagy A et al (2015) Clinical significance of
CYP2C9-status guided valproic acid therapy in children.
Epilepsia 56:849–855. https://doi.org/10.1111/epi.13011
4. Marahatta A, Bhandary B, Jeong SK, Kim HR, Chae HJ (2014)
Soybean greatly reduces valproic acid plasma concentrations: a
food-drug interaction study. Sci Rep 4:4362. https://doi.org/10.
1038/srep04362
5. Lin WW, Jiao Z, Wang CL, Wang HY, Ma CL, Huang PF, Guo XZ,
Liu YW (2015) Population pharmacokinetics of valproic acid in
adult Chinese epileptic patients and its application in an individu-
alized dosage regimen. Ther Drug Monit 37:76–83. https://doi.org/
10.1097/ftd.0000000000000100
6. Ding J, Wang Y, Lin W, Wang C, Zhao L, Li X, Zhao Z, Miao L,
Jiao Z (2015) A population pharmacokinetic model of valproic acid
in pediatric patients with epilepsy: a non-linear pharmacokinetic
model based on protein-binding saturation. Clin Pharmacokinet
54:305–317. https://doi.org/10.1007/s40262-014-0212-8
7. Williams JH, Jayaraman B, Swoboda KJ, Barrett JS (2012)
Population pharmacokinetics of valproic acid in pediatric patients
with epilepsy: considerations for dosing spinal muscular atrophy
patients. J Clin Pharmacol 52:1676–1688. https://doi.org/10.1177/
0091270011428138
8. Jiang DC, Wang L, Wang YQ et al (2007) Population pharmacoki-
netics of valproate in Chinese children with epilepsy. Acta
Pharmacol Sin 28:1677–1684. https://doi.org/10.1111/j.1745-
7254.2007.00704.x
9. Park HM, Kang SS, Lee YB, Shin DJ, Kim ON, Lee SB, Yim DS
(2002) Population pharmacokinetics of intravenous valproic acid in
Korean patients. J Clin Pharm Ther 27:419–425
10. Nakashima H, Oniki K, Nishimura M, Ogusu N, Shimomasuda M,
Ono T, Matsuda K, Yasui-Furukori N, Nakagawa K, Ishitsu T,
Saruwatari J (2015) Determination of the optimal concentration of
valproic acid in patients with epilepsy: a population
pharmacokinetic-pharmacodynamic analysis. PLoS One 10:
e0141266. https://doi.org/10.1371/journal.pone.0141266
11. Ogusu N, Saruwatari J, Nakashima H, Noai M, Nishimura M,
Deguchi M, Oniki K, Yasui-Furukori N, Kaneko S, Ishitsu T,
Nakagaswa K (2014) Impact of the superoxide dismutase 2
Val16Ala polymorphism on the relationship between valproic acid
exposure and elevation of gamma-glutamyltransferase in patients
with epilepsy: a population pharmacokinetic-pharmacodynamic
analysis. PLoS One 9:e111066. https://doi.org/10.1371/journal.
pone.0111066
12. Jiang D, Bai X, Zhang Q, Lu W, Wang Y, Li L, Müller M (2009)
Effects of CYP2C19 and CYP2C9 genotypes on pharmacokinetic
variability of valproic acid in Chinese epileptic patients: nonlinear
mixed-effect modeling. Eur J Clin Pharmacol 65:1187–1193.
https://doi.org/10.1007/s00228-009-0712-x
13. Correa T, Rodriguez I, Romano S (2008) Population pharmacoki-
netics of valproate in Mexican children with epilepsy. Biopharm
Drug Dispos 29:511–520. https://doi.org/10.1002/bdd.636
14. Birnbaum AK, Ahn JE, Brundage RC, Hardie NA, Conway JM,
Leppik IE (2007) Population pharmacokinetics of valproic acid
concentrations in elderly nursing home residents. Ther Drug
Monit 29:571–575. https://doi.org/10.1097/FTD.
0b013e31811f3296
15. Bondareva IB, Jelliffe RW, Sokolov AV, Tischenkova IF (2004)
Nonparametric population modeling of valproate pharmacokinetics
in epileptic patients using routine serum monitoring data: implica-
tions for dosage. J Clin Pharm Ther 29:105–120. https://doi.org/10.
1111/j.1365-2710.2003.00538.x
16. Serrano BB, Garcia Sanchez MJ, Otero MJ et al (1999) Valproate
population pharmacokinetics in children. J Clin Pharm Ther 24:73–
80
17. Guo Y, Hu C, He X, Qiu F, Zhao L (2012) Effects of UGT1A6,
UGT2B7, and CYP2C9 genotypes on plasma concentrations of
valproic acid in Chinese children with epilepsy. Drug Metab
Pharmacokinet 27:536–542
18. Tan L, Yu JT, Sun YP, Ou JR, Song JH, Yu Y (2010) The influence
of cytochrome oxidase CYP2A6, CYP2B6, and CYP2C9 polymor-
phisms on the plasma concentrations of valproic acid in epileptic
patients. Clin Neurol Neurosurg 112:320–323. https://doi.org/10.
1016/j.clineuro.2010.01.002
19. Kiang TK, Ho PC, Anari MR et al (2006) Contribution of CYP2C9,
CYP2A6, and CYP2B6 to valproic acid metabolism in hepatic
microsomes from individuals with the CYP2C9*1/*1 genotype.
Toxicol Sci 94:261–271. https://doi.org/10.1093/toxsci/kfl096
20. Chu XM, Zhang LF, Wang GJ, Zhang SN, Zhou JH, Hao HP
(2012) Influence of UDP-glucuronosyltransferase polymorphisms
on valproic acid pharmacokinetics in Chinese epilepsy patients. Eur
J Clin Pharmacol 68:1395–1401. https://doi.org/10.1007/s00228-
012-1277-7
21. Hung CC, Ho JL, Chang WL, Tai JJ, Hsieh TJ, Hsieh YW, Liou HH
(2011) Association of genetic variants in six candidate genes with
valproic acid therapy optimization. Pharmacogenomics 12:1107–
1117. https://doi.org/10.2217/pgs.11.64
22. Jawien W, Wilimowska J, Klys M et al (2017) Population pharma-
cokinetic modelling of valproic acid and its selected metabolites in
acute VPA poisoning. Pharmacol Rep 69:340–349. https://doi.org/
10.1016/j.pharep.2016.12.003
23. Mei S, Feng W, Zhu L, Yu Y, Yang W, Gao B, Wu X, Zhao Z, Fang
F (2017) Genetic polymorphisms and valproic acid plasma concen-
tration in children with epilepsy on valproic acid monotherapy.
Seizure 51:22–26. https://doi.org/10.1016/j.seizure.2017.07.005
24. Feng W, Mei S, Zhu L, Yu Y, Yang W, Gao B, Wu X, Zhao Z, Fang
F (2016) Effects of UGT1A6 and GABRA1 on standardized
valproic acid plasma concentrations and treatment effect in children
with epilepsy in China. Ther Drug Monit 38:738–743. https://doi.
org/10.1097/ftd.0000000000000337
25. Shintani M, Ieiri I, Inoue K, Mamiya K, Ninomiya H, Tashiro N,
Higuchi S, Otsubo K (2001) Genetic polymorphisms and functional
characterization of the 5′-flanking region of the human CYP2C9
gene: in vitro and in vivo studies. Clin Pharmacol Ther 70:175–182.
https://doi.org/10.1067/mcp.2001.117367
26. Buetow KH, Edmonson M, MacDonald R et al (2001) High-
throughput development and characterization of a genomewide

collection of gene-based single nucleotide polymorphism markers
by chip-based matrix-assisted laser desorption/ionization time-of-
flight mass spectrometry. Proc Natl Acad Sci U S A 98:581–584.
https://doi.org/10.1073/pnas.021506298
27. Mei S, Li X, Jiang X, Yu K, Lin S, Zhao Z (2018) Population
pharmacokinetics of high-dose methotrexate in patients with prima-
ry central nervous system lymphoma. J Pharm Sci. https://doi.org/
10.1016/j.xphs.2018.01.004
28. Savic RM, Karlsson MO (2009) Importance of shrinkage in empir-
ical bayes estimates for diagnostics: problems and solutions. AAPS
J 11:558–569. https://doi.org/10.1208/s12248-009-9133-0
29. Yano Y, Beal SL, Sheiner LB (2001) Evaluating pharmacokinetic/
pharmacodynamic models using the posterior predictive check. J
Pharmacokinet Pharmacodyn 28:171–192
30. Ette EI, Williams PJ, Kim YH, Lane JR, Liu MJ, Capparelli EV
(2003) Model appropriateness and population pharmacokinetic
modeling. J Clin Pharmacol 43:610–623
31. McLeay SC, Morrish GA, Kirkpatrick CM et al (2012) The rela-
tionship between drug clearance and body size: systematic review
and meta-analysis of the literature published from 2000 to 2007.
Clin Pharmacokinet 51:319–330. https://doi.org/10.2165/
11598930-000000000-00000
32. Ghodke-Puranik Y, Thorn CF, Lamba JK, Leeder JS, Song W,
Birnbaum AK, Altman RB, Klein TE (2013) Valproic acid path-
way: pharmacokinetics and pharmacodynamics. Pharmacogenet
Genomics 23:236– 241. https://doi.org/10.109 7/FPC .
0b013e32835ea0b2
33. de Morais SM, Wilkinson GR, Blaisdell J, Nakamura K, Meyer
UA, Goldstein JA (1994) The major genetic defect responsible for
the polymorphism of S-mephenytoin metabolism in humans. J Biol
Chem 269:15419–15422
34. De Morais SM, Wilkinson GR, Blaisdell J et al (1994)
Identification of a new genetic defect responsible for the polymor-
phism of (S)-mephenytoin metabolism in Japanese. Mol
Pharmacol 46:594–598
35. Sun D, Jones NR, Manni A, Lazarus P (2013) Characterization of
raloxifene glucuronidation: potential role of UGT1A8 genotype on
raloxifene metabolism in vivo. Cancer Prev Res 6:719–730. https://
doi.org/10.1158/1940-6207.capr-12-0448
36. Haas DW, Kwara A, Richardson DM, Baker P, Papageorgiou I,
Acosta EP, Morse GD, Court MH (2014) Secondary metabolism
pathway polymorphisms and plasma efavirenz concentrations in
HIV-infected adults with CYP2B6 slow metabolizer genotypes. J
Antimicrob Chemother 69:2175–2182. https://doi.org/10.1093/jac/
dku110
37. Ho PC, Abbott FS, Zanger UM, Chang TKH (2003) Influence of
CYP2C9 genotypes on the formation of a hepatotoxic metabolite of
Eur J Clin Pharmacol
valproic acid in human liver microsomes. Pharmacogenomics J 3:
335–342. https://doi.org/10.1038/sj.tpj.6500210
38. Strassburg CP, Strassburg A, Kneip S, Barut A, Tukey RH, Rodeck
B, Manns MP (2002) Developmental aspects of human hepatic
drug glucuronidation in young children and adults. Gut 50:259–265
39. Reith DM, Andrews J, Parker-Scott S, Eadie MJ (2000) Urinary
excretion of valproate metabolites in children and adolescents.
Biopharm Drug Dispos 21:327–330
40. Ginsberg G, Hattis D, Sonawane B, Russ A, Banati P, Kozlak M,
Smolenski S, Goble R (2002) Evaluation of child/adult pharmaco-
kinetic differences from a database derived from the therapeutic
drug literature. Toxicol Sci 66:185–200
41. Stewart CF, Hampton EM (1987) Effect of maturation on drug
disposition in pediatric patients. Clin Pharm 6:548–564
42. Silva MF, Aires CC, Luis PB et al (2008) Valproic acid metabolism
and its effects on mitochondrial fatty acid oxidation: a review. J
Inherit Metab Dis 31:205–216. https://doi.org/10.1007/s10545-
008-0841-x
43. Ponchaut S, van Hoof F, Veitch K (1992) In vitro effects of
valproate and valproate metabolites on mitochondrial oxidations.
Relevance of CoA sequestration to the observed inhibitions.
Biochem Pharmacol 43:2435–2442
44. Kirchheiner J, Klein C, Meineke I, Sasse J, Zanger UM, Mürdter
TE, Roots I, Brockmöller J (2003) Bupropion and 4-OH-bupropion
pharmacokinetics in relation to genetic polymorphisms in CYP2B6.
Pharmacogenetics 13:619–626. https://doi.org/10.1097/01.fpc.
0000054125.14659.d0
45. Kiyotani K, Yamazaki H, Fujieda M, Iwano S, Matsumura K,
Satarug S, Ujjin P, Shimada T, Guengerich FP, Parkinson A,
Honda G, Nakagawa K, Ishizaki T, Kamataki T (2003) Decreased
coumarin 7-hydroxylase activities and CYP2A6 expression levels
in humans caused by genetic polymorphism in CYP2A6 promoter
region (CYP2A6*9). Pharmacogenetics 13:689–695. https://doi.
org/10.1097/01.fpc.0000054136.14659.be
46. Yoshida R, Nakajima M, Nishimura K, Tokudome S, Kwon JT,
Yokoi T (2003) Effects of polymorphism in promoter region of
human CYP2A6 gene (CYP2A6*9) on expression level of messen-
ger ribonucleic acid and enzymatic activity in vivo and in vitro. Clin
Pharmacol Ther 74:69–76. https://doi.org/10.1016/s0009-9236(03)
00090-0
47. Wang C, Wang P, Yang LP, Pan J, Yang X, Ma HY (2017)
Association of CYP2C9, CYP2A6, ACSM2A, and CPT1A gene
polymorphisms with adverse effects of valproic acid in Chinese
patients with epilepsy. Epilepsy Res 132:64–69. https://doi.org/
10.1016/j.eplepsyres.2017.02.015