Embed Size (px)
DESCRIPTION
vhvj
Citation preview
ORIGINAL RESEARCH ARTICLE
A Population Pharmacokinetic and Pharmacodynamic Studyof a Peripheral j-Opioid Receptor Agonist CR665 and Oxycodone
Anne E. Olesen • Kim Kristensen • Camilla Staahl • Sherron Kell •
Gilbert Y. Wong • Lars Arendt-Nielsen • Asbjørn M. Drewes
Published online: 5 December 2012
� Springer International Publishing Switzerland 2012
Abstract
Background Peripherally acting opioids, particularly
peripheral j-opioid agonists, may be effective for treating
visceral pain by activating receptors expressed on afferent
nerves within the gut.
Objective The objective of this study was to investigate
the pharmacokinetic/pharmacodynamic profile of a novel
peripherally selective j-opioid agonist, CR665 (JNJ-
38488502), and compare it to that of oxycodone, a non-
selective brain-penetrant opioid.
Methods In a randomized, placebo-controlled, double-
blind, three-way crossover study, healthy male volunteers
were administered CR665 (0.36 mg/kg, intravenous),
oxycodone (15 mg, oral) or placebo (intravenous and oral),
followed by assessment of visceral pain tolerance thres-
holds (VPTT) measured as volume of water (mL) in the
bag placed on an oesophageal probe. Plasma drug con-
centration data were used to generate pharmacokinetic
models, which were then used to fit the VPTT data using
NONMEM� VI to generate population pharmacokinetic/
pharmacodynamic models.
Results CR665 kinetics were optimally fitted with a
two-compartment model, while oxycodone kinetics were
best described by a one-compartment model with transit
compartment absorption feeding directly into the central
compartment. For both drugs, the plasma concentration
effects on VPTT were best fit by a direct linear model,
i.e. without the concentration–analgesia delay character-
istic of brain-penetrant opioids. The slope of oxycodone
(0.089 mL per ng/mL) was steeper than that of CR665
(0.0035 mL per ng/mL) for the plasma drug concentration
acting on the VPTT.
Conclusion The results are consistent with the peripheral
selectivity of CR665, as well as the possibility that
peripheral actions of oxycodone contribute to its visceral
analgesic efficacy.
1 Introduction
Visceral pain is prevalent in clinical practice and causes
challenges in pain management [1]. Currently, centrally
acting, non-selective opioid agonists are used in the treat-
ment of visceral pain. However, inadequate analgesia and
undesirable adverse effects often limit their usage [2].
A. E. Olesen (&) � A. M. Drewes
Mech-Sense, Department of Gastroenterology and Hepatology,
Aalborg Hospital, Aarhus University Hospital, Mølleparkvej 4,
9000 Aalborg, Denmark
e-mail: [email protected]
K. Kristensen
Insulin Pharmacology, Novo Nordisk A/S, Maløv, Denmark
C. Staahl
R & D, Grunenthal GmBH, Aachen, Germany
S. Kell
Research Development, Impax Pharmaceuticals,
Hayward, CA, USA
S. Kell
ALZA Corporation, Mountain View, CA, USA
G. Y. Wong
Pfizer, Inc., Worldwide Research and Development,
South San Francisco, CA, USA
L. Arendt-Nielsen � A. M. Drewes
Center for Sensory-Motor Interaction,
Department of Health Science and Technology,
Aalborg University, Aalborg, Denmark
Clin Pharmacokinet (2013) 52:125–137
DOI 10.1007/s40262-012-0023-8
Because dose-limiting adverse effects such as euphoria,
sedation, respiratory depression and nausea are mediated
by l-opioid receptors in the central nervous system (CNS),
there has been a focus on the development of peripherally
acting opioids (i.e. opioids that do not readily enter the
CNS), as well as receptor subtype-selective opioids (e.g.
j- and d-opioid agonists) that do not interact with the
classical l-opioid receptor [3, 4]. Visceral afferents
expressing j-opioid receptors in the gut are thought to have
an important role in the modulation of visceral pain and
are, therefore, a possible receptor target for novel drugs to
attenuate visceral pain [4, 5]. CR665 (JNJ-38488502) is a
novel peripherally selective j-opioid receptor peptide
agonist that has demonstrated efficacy in rodent models of
visceral pain, as well as an experimental visceral pain
model in humans [6–8]. Moreover, animal studies have
demonstrated analgesic effect of the main metabolite of
CR665 (CR665N-oxide) [unpublished data].
Opioid effects can be difficult to evaluate in a clinical
setting due to the many confounding factors affecting a
patient’s pain experience. Experimental pain models in
healthy volunteers provide less confounding pain mea-
sures, and may be more suitable for establishing pharma-
cokinetic/pharmacodynamic relationships for analgesic
effects [9–11]. In population pharmacokinetic/pharmaco-
dynamic modelling, mathematical models are used to
summarize the information on dose–plasma concentration–
effect relationships. The population method is able to iso-
late and characterize variability between subjects and it
allows characterizing of the unique time course of the pain
behaviour [12]. Thus, this technique offers advantages over
traditional statistical methods in situations when high inter-
and/or intra-subject variability can be expected, e.g. in
research on pain behaviour. Pharmacokinetic/pharmaco-
dynamic modelling has possibilities to enable the under-
standing of the underlying mechanisms of drug action [13].
Additionally, the population method has the potential of
using sparse and randomly collected data in contrast to
traditional analysis [14].
The rationale for the present study was that the phar-
macokinetic/pharmacodynamic profile of CR665 could
support evidence relating to the peripheral mechanism of
action in the management of visceral pain [15–18]. The
pharmacokinetic/pharmacodynamic relationship for orally
administered oxycodone (15 mg) has previously been
successfully described for different dynamic endpoints in
human experimental pain models [9, 11] and, therefore,
oxycodone was used as active comparator.
The aim of this study was to investigate the pharma-
cokinetic/pharmacodynamic profile of single doses of a
novel peripherally selective j-opioid receptor peptide
agonist CR665 and compare it to that of oxycodone in a
human experimental model of visceral pain.
2 Subjects and Methods
2.1 Subjects and Study Design
This pharmacokinetic/pharmacodynamic analysis was based
on data from a previous study investigating the analgesic
efficacy of the peripheral j-opioid receptor agonist CR665
compared to oxycodone in a multi-modal, multi-tissue
experimental human pain model [8]. Healthy male volun-
teers (aged 18–45 years) were recruited at a single site to
participate in this randomized, double-blind, placebo-con-
trolled, three-way crossover study. The study was conducted
in the laboratory at Mech-Sense, Department of Gastroen-
terology, Aalborg Hospital, Aalborg, Denmark. Inclusion
criteria were as follows: healthy male aged 18–45 years,
inclusive; body mass index (BMI) within 18–25 kg/m2,
inclusive, and stable bodyweight; bodyweight B90 kg; good
general health with no clinically relevant abnormalities as
assessed by the investigator; semi-recumbent blood pressure
(after resting for 5 min) in the ranges of 90–139 mmHg
systolic and 50–89 mmHg diastolic, inclusive; negative
urine drug test at screening and at check-in prior to each
treatment period; negative alcohol test at check-in prior to
each treatment period; and no known allergies to any of the
compounds used in the study. Exclusion criteria were as
follows: blood donor within 4 weeks prior to dosing; use of
any medication up until 30 days before study start; drinking
of alcohol-, grapefruit juice-, caffeine- or chinin-containing
products 48 h before study start; smoking; drinking more
than 5 cups of coffee, tea or 2.5 L of cola a day; and drinking
more than 14 units of alcohol per week.
All subjects were informed about the risks of the study,
gave their written informed consent prior to participating,
and were compensated for participating. The study was
conducted in accordance with the Declaration of Helsinki
on biomedical research involving human subjects and
guidelines on Good Clinical Practice were followed. The
study protocol was approved by the Regional Ethics Com-
mittee, Aalborg, Denmark (registration no. VN-20060021)
and the Danish Medicine Agency, Copenhagen, Denmark
(ref. no. 2612-3145).
2.2 Medication
Equal numbers of subjects were randomly assigned to one
of three treatment sequences (ABC, BCA or CAB). Intra-
venous infusions were administered in the right forearm
vein and oral solutions were administered via the medica-
tion port in the oesophageal tube. All subjects received the
following treatments:
• Treatment A: CR665 solution 1.1 mL vials containing
10 mg/mL of active drug, total dose of 0.36 mg/kg
126 A. E. Olesen et al.
mixed with an appropriate volume of isotonic saline
solution for injection to achieve a final volume of
25 mL and administered as an intravenous infusion
over 1 h at a rate of 25 mL/h, and an oral placebo
solution consisting of 15 mL orange juice solution.
• Treatment B: oxycodone 15 mg (1.5 mL of a 10 mg/mL
oral liquid solution of oxycodone hydrochloride in a
pure water vehicle) mixed with a 13.5 mL orange juice
solution used for the oral placebo solution in treatment
A, and a 1-h intravenous infusion of vehicle placebo
solution (i.e. isotonic saline for injection).
• Treatment C: 1-h intravenous infusion of vehicle
placebo solution and oral placebo solution, as used in
treatments A and B.
Subjects were admitted to the unit on the morning of
each dosing after overnight fasting and an oesophageal
tube and intravenous lines were placed prior to dosing.
Subjects were released from the clinic 24 h after the start of
dosing. Each treatment was followed by a 1- to 3-week
washout period. Prior to the pharmacokinetic analysis of
CR665 the individual doses were calculated by multiplying
the dose in mg/kg with individual bodyweights.
2.3 Pain Assessment
The visceral pain assessment parameter was measured
before each treatment, and at 30, 60 and 90 min after drug
administration.
The oesophageal probe has been described previously
[19]. The bag, attached to the probe, was placed 7 cm
proximal to the sphincter.
Water was infused to the bag by infusion channels
connected to a precision infusion–withdrawal pump (Type
111, Ole Dich Instrument Makers, Hvidovre, Denmark).
For distension, the oesophageal bag was filled with 37 �C
water at a constant infusion rate of 10 mL/min. As a pre-
viously validated measure of tolerated distension, the total
volumes of water (mL) in the oesophageal bag at pain
detection threshold and moderate pain were recorded.
Stimulation was stopped when subjects reported moderate
pain.
A single 10-point electronic visual analogue scale was
used to assess painful sensations in response to the dis-
tension of the bag in the oesophagus. A rating of 5 was
defined as the ‘pain detection threshold’ and a rating of 7 as
‘moderate pain’ [20].
2.4 Blood Samples
Blood samples for pharmacokinetic analysis were drawn
from an intravenous line in the left forearm pre-dosing and
then at 5, 15, 30, 60, 75 and 120 min after start of dosing.
Therefore, collection of blood samples was initiated during
the 1-h infusion phase. Blood was collected into pre-chilled
3 mL tripotassium (K3) EDTA-containing glass tubes
(Vacutainer�) and placed on ice for 1–2 h until centrifu-
gation. The samples were centrifuged at 1,0009g (about
2,500–3,000 rpm) for 10 min at approximately 4 �C. For
each sample, the separated plasma was transferred into two
polypropylene tubes and stored immediately at approxi-
mately –20 �C until analysis.
The analysis of the plasma samples for the determina-
tion of the concentrations of CR665 and its N-oxide
metabolite (CR665N-oxide) was carried out at Covance
Laboratories Limited (UK). Plasma concentrations were
determined using liquid chromatography with tandem mass
spectrometric detection (LC-MS/MS) with a lower limit of
quantification of 1 ng/mL.
The analysis of the plasma samples for the determina-
tion of the concentrations of oxycodone was carried out at
CEDRA Corporation (USA). Oxycodone concentrations
were determined by high-pressure liquid chromatography
(HPLC) with LC-MS/MS. The limits of quantification were
0.1 ng/mL for oxycodone. Values that were below the limit
of quantification were considered to be zero. The calibra-
tion range was 0.1–50 ng/mL.
Inter- and intra-day precision (% coefficient of varia-
tion) were below 10.2 and 10.5 %, respectively, across the
entire calibration range for all analyses.
2.5 General Population Modelling Methods
In total, 180 plasma concentration samples of CR665,
including 72 metabolite samples, were available for the
population pharmacokinetic analysis. For pharmacokinetic/
pharmacodynamic modelling an additional 140 pharma-
codynamic measurements were used. For the oxycodone
pharmacokinetic model, 91 plasma concentration samples
and 144 pharmacodynamic measurements were available
for the modelling. Both pharmacokinetic and pharmaco-
dynamic models were fitted to the data using non-linear
mixed–effects modelling software NONMEM� version VI,
level 1.0 (Globomax LLC, Elliott City, MD, USA), with
the first-order conditional estimation with interaction
approach. Unexplained inter-individual variability (IIV) in
structural model parameters was estimated using a log–
normal error model if not otherwise indicated (Eq. 1):
Pj ¼ TVP� egj ð1Þ
where Pj is the individual value of the parameter in the jth
individual, TVP is the typical value of the parameter for the
population and gj is an independent random variable with a
mean of zero and variance of x2.
Residual error was represented by a combined propor-
tional and/or additive error model. This was reduced to
PopPK/PD Study of CR665 and Oxycodone 127
either a proportional or an additive model by omitting
small residual error terms, as indicated.
The main tool for selection between hierarchical models
was the difference in objective function value (OFV)
between models, i.e. the likelihood ratio test. The OFV is
proportional to minus twice the log likelihood and the
difference in OFV for the two models is approximately
chi-square (v2)-distributed. A difference in OFV of greater
than 3.84 is significant at the 5 % level if the models differ
by one parameter. Corresponding values for p = 0.01 and
p = 0.001 are 6.63 and 10.83, respectively.
Goodness of fit was also assessed by visual inspection of
plots of observed and predicted values against time,
residual plots and the distribution of the residual error. The
standard errors of model parameters outputted by NON-
MEM� were compared with 1,000 bootstraps of these final
models from 1,000 models using Perl-speaks-NONMEM�
(PsN) version 3.4.2 [21, 22].
2.5.1 Pharmacokinetic Models
Using all available data, pharmacokinetic modelling of the
plasma drug concentrations was performed to make infer-
ences about the kinetics of intravenously administered
CR665 and orally administered oxycodone in these healthy
volunteers. The pharmacokinetics was then used to provide
input concentrations for separate subsequent pharmacody-
namic models. The CR665 and CR665N-oxide concentra-
tion data were fitted to one- and two-compartment
pharmacokinetic models, with zero-order infusion for
60 min. A zero-order reaction has a rate that is independent
of the concentration of the reactant and increasing the
concentration of the reacting species will not speed up the
rate of the reaction. Rate constants of the CR665 and
metabolite model were used to determine which process in
the model was the rate-limiting step in the pharmacokinetic
model.
The oxycodone concentration data were fitted to one-
and two-compartment pharmacokinetic models, with oral
absorption represented as either a first-order absorption
process with a lag term, or as one or two chains of transit
compartments [23].
2.5.2 Effect Models
Models were also tested for the placebo response encom-
passing any changes in oesophageal distension. These were
models with no systematic change from baseline value,
linear change with time and quadratic change with time
(the latter giving a curved shape to effect versus time).
Details of the equations used and their purpose were
described previously by our group [11].
The combined data for the placebo and CR665 or oxy-
codone response for each subject were modelled together in
this stage of the analysis. The placebo response (if any) was
assumed to be additive with any drug response (that is, the
placebo data established what would have happened if the
drug dose was zero). Linear and maximum concentration–
effect relationships were examined as described in Table 1.
For linear models, the concentration–effect relationship was
described by the term ‘slopeD’ (D for drug). Positive values
indicated concentration-dependent analgesia. The models
were characterized by the maximum effect (Emax) and the
concentration producing 50 % of the maximum effect
(EC50). Pain measures were related to either the measured
venous plasma concentration of drug (C) or concentration of
the drug in the hypothetical effect compartment (Ceff). The
Ceff was delayed relative to the plasma concentration by an
equilibrium rate constant between plasma and the effect
compartment (ke0). If the pain measures were not delayed
relative to the opioid plasma concentration (i.e. the elimi-
nation half-life for effect–site equilibrium [t�,ke0] = 0), the
data would support a pharmacodynamic model without a
separate effect compartment. If the analgesic effects were
delayed relative to the opioid plasma concentration, e.g. due
to a requirement for transport across the blood–brain barrier,
the data would support a pharmacodynamic model with an
effect compartment, with a delay quantified by a positive,
non-zero value of t�,ke0.
2.6 Statistical Analysis
The parameter estimates obtained using NONMEM� were
expressed as the population average value and the popu-
lation parameter variability (its variability across the
population).
2.6.1 Model Evaluation
A predictive model check was used to validate the final
model using PsN version 3.4.2 [21, 22]. The final popula-
tion pharmacokinetic and pharmacokinetic/pharmacody-
namic model, including the final fixed and random effects,
were used to simulate a dataset 1,000 times for the subjects
studied. The 95 % confidence intervals for the simulated
plasma concentration–time profile or effect–time profile
were plotted against the observed data [24, 25].
3 Results
Eighteen healthy, non-smoking, opioid-naıve male volun-
teers (17 Caucasian and 1 Asian; ages 19–43 years, median
25 years; bodyweight 62.4–94.5 kg, median 80.0 kg; BMI
20.5–27.4 kg/m2, median 23.6 kg/m2) completed the study.
128 A. E. Olesen et al.
The mean plasma concentrations of CR665, the
N-oxide metabolite of CR665 and oxycodone for all 18
healthy subjects are shown in Fig. 1. The subjects’
tolerances to pain induced by oesophageal distensions
measured as visceral pain tolerance threshold (VPTT)
versus time following placebo and drug treatment are
shown in Fig. 2.
3.1 Pharmacokinetics
3.1.1 CR665
A two-compartment and one-compartment distribution
model was fitted to CR665 and N-oxide metabolite data,
respectively. Distribution of CR665 was controlled by
rate constants k12 (first-order rate constant of drug
transfer for the central compartment to the peripheral
compartment) and k21 (first-order rate of drug transfer for
the peripheral compartment to the central compartment).
The elimination of CR665 was described by a first-order
formation rate constant (k13) from the central compart-
ment of CR665 to the N-oxide compartment, reflecting
the known conversion of the parent compound to this
active metabolite. Due to lack of information during the
elimination of the metabolite, a first-order elimination
rate constant for this metabolite (k30), equal to the for-
mation rate constant of metabolite (i.e. k30 = k13), best
described the pharmacokinetics of the N-oxide following
intravenous infusion of CR665 (a schematic model is
depicted in Fig. 3).
10 1
Time (h)
20.5 1.5
10
100
a 1,000
Pla
sma
conc
entr
atio
n (n
g/m
L)
0.10 1
Time (h)
20.5 1.5
1
10
b 100
Pla
sma
conc
entr
atio
n (n
g/m
L)
CR665N-oxide metabolite
Fig. 1 Spaghetti plots describing individual time courses of the
plasma concentration of CR665 and its N-oxide metabolite following
intravenous infusion over 1 h of 0.36 mg/kg CR665 (a) and oral
administration of 15 mg oxycodone (b)
Ta
ble
1T
este
dm
od
els
of
the
CR
66
5an
do
xy
cod
on
eco
nce
ntr
atio
n–
effe
ctre
lati
on
ship
s
Model
Des
crip
tion
Equat
ion
Purp
ose
1P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
linea
rphar
mac
odynam
icm
odel
E=
bas
elin
e?
slopeD
9C
Tes
ted
the
assu
mpti
on
that
the
dat
aco
uld
be
des
crib
edby
ali
nea
rm
odel
wit
hno
effe
ct
del
ay
2P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
IIV
inbas
elin
ean
dsl
ope
wit
h
Cfi
xed
atze
ro
E=
bas
elin
e?
slopeD
9C
Tes
ted
the
assu
mpti
on
that
dru
gco
nce
ntr
atio
ns
did
not
infl
uen
ceth
eef
fect
3P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
log-l
inea
rphar
mac
odynam
ic
model
E=
m9
ln(C
?bas
elin
e)T
este
dth
eas
sum
pti
on
that
the
dat
aco
uld
be
des
crib
edby
ali
nea
rm
odel
wit
hno
effe
ct
del
ay
4P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
non-l
inea
rphar
mac
odynam
ic
model
;a
refe
rence
non-l
inea
rm
odel
wit
hno
effe
ctco
mpar
tmen
t
E=
bas
elin
e?
(Em
ax
9C
)/
(EC
50
?C
)
Tes
ted
the
assu
mpti
on
that
the
dat
aco
uld
be
des
crib
edby
anon-l
inea
rm
odel
wit
hno
effe
ctdel
ay;
hig
hE
C50
val
ues
mad
eth
ism
odel
rever
tto
ali
nea
rm
odel
5P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
linea
rphar
mac
odynam
icm
odel
wit
hef
fect
del
ay;
IIV
inbas
elin
e,sl
ope
and
ke0
Ceff0
=k
e0
9(C
-C
eff)
As
for
model
1but
wit
ha
ke0
term
E=
bas
elin
e?
slopeD
9C
eff
6P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
non-l
inea
rphar
mac
odynam
ic
model
wit
hef
fect
del
ay
Ceff0
=k
e0
9(C
-C
eff)
As
for
model
3but
wit
ha
ke0
term
E=
m9
ln(C
eff
?bas
elin
e)
7P
Km
odel
inte
rpola
tion
of
pla
sma
conce
ntr
atio
ns;
non-l
inea
rphar
mac
odynam
ic
model
wit
hef
fect
del
ay
Ceff0
=k
e0
9(C
-C
eff)
As
for
model
4but
wit
ha
ke0
term
E=
bas
elin
e?
(Em
ax
9C
eff)/
(EC
50
?C
eff)
Cpla
sma
dru
gco
nce
ntr
atio
n,
Ceff
obse
rved
conce
ntr
atio
nin
the
effe
ctco
mpar
tmen
t,C
eff0
esti
mat
edco
nce
ntr
atio
nin
the
effe
ctco
mpar
tmen
t,E
tim
eco
urs
eof
the
effe
ctof
the
oxyco
done
dose
,E
C50
pla
sma
dru
g
conce
ntr
atio
nat
whic
hth
ere
was
ahal
f-re
duct
ion
inth
epai
nsc
ore
,E
max
pla
sma
dru
gco
nce
ntr
atio
nat
whic
hth
ere
was
am
axim
um
reduct
ion
inth
epai
nsc
ore
,II
Vin
ter-
indiv
idual
var
iabil
ity,
k e0
equil
ibri
um
rate
const
ant
bet
wee
npla
sma
and
the
effe
ctco
mpar
tmen
t,w
hic
hw
asal
soa
popula
tion
par
amet
er,
msl
ope
of
the
appar
ent
linea
rse
gm
ent
of
the
log-t
ransf
orm
edco
nce
ntr
atio
nver
sus
resp
onse
curv
e,P
Kphar
mac
okin
etic
,
slopeD
conce
ntr
atio
n–ef
fect
rela
tionsh
ip
PopPK/PD Study of CR665 and Oxycodone 129
Covariate bodyweight (WT) was incorporated into the
models according to Holford [26], so the volumes of
distribution were scaled with the quantity (WT/median
bodyweight), and rate constants with the quantity (WT/
median bodyweight)-0.25. IIV was found for the apparent
volume of distribution of the central compartment for the
parent compound (V1) and metabolite (V3) as well as for
k12. Various one- and two-compartment distribution models
with similar and different elimination rate constants were
tested, with IIV in different parameters estimated using an
exponential error structure, and residual error modelled as a
proportional error structure. The model described above
was judged based on convergence of the estimation and
covariance routines, goodness-of-fit plots, precision of the
population pharmacokinetic parameters, and minimum
value of the objective function, a statistical criterion for the
addition of a single parameter, as the model best suited for
further pharmacokinetic/pharmacodynamic modelling of
CR665. The parameters for this model are summarized in
Table 2 and visual predicted check plots are shown in
Fig. 4. The model seems to describe the data of both CR665
and metabolite well as the visual predicted check plots
shows that the 2.5th, 50th and 97.5th percentiles of the
observed data are closely followed by the by the simulated
2.5th, 50th and 97.5th percentiles. In addition, the areas of
95 % prediction intervals around the 2.5th, 50th and 97.5th
percentiles of the simulated data overlap the 2.5th, 50th and
97.5th percentiles of the observed data.
3.1.2 Oxycodone
For the observed oxycodone plasma concentrations, a one-
compartment model with a single transit chain representing
CR665OxycodonePlacebo
b
0
Time (h)
10
20
30
50
40
60
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
c
0
Time (h)
0 1.0 2.00.5 1.5 0 1.0 2.00.5 1.5
10
20
30
50
40
60
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
151.00
Time (h)
2.01.50.5
20
25
a30
CR665Placebo
OxycodonePlacebo
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
Fig. 2 a Oesophageal distension measured as visceral pain tolerance
threshold (mean ± standard error of the mean) vs. nominal time
following CR665, oxycodone and placebo treatment. b, c Spaghetti
plots describing individual time courses of the visceral pain tolerance
threshold after 1 h intravenous infusion of CR665 and placebo (b) and
oral administration of 15 mg oxycodone and placebo (c)
ktr ktr ktr ktr ktrAmountA
Vd/F
CL/F
TransA1
TransAn −1
TransAn
MTT
IV infusion
CR665V1
V2
k12
k21
k13 k30N-oxideV3
a
b
Fig. 3 a Schematic representation of the CR665 pharmacokinetic
model. b Schematic representation of the oxycodone pharmacokinetic
model. A, A1, An–1, An represent the amount of oxycodone in each
compartment, CL/F apparent total oral clearance, IV intravenous, k12
first-order rate constant of drug transfer for the central compartment
to the peripheral compartment, k13 first-order rate constant of
metabolite formation, k30 elimination rate constant of N-oxide (k13
and k30 are equal in the final model), k21 first-order rate constant of
drug transfer for the peripheral compartment to the central compart-
ment, ktr rate constants for oxycodone transfer through transit
compartment, MTT mean transit time for oxycodone movement
through n transit compartments, Trans transit compartment, V1 central
volume of distribution for CR665, V2 peripheral volume of distribu-
tion, V3 central volume of distribution for CR665N-oxide, Vd/Fvolume of distribution after oral administration of the drug
130 A. E. Olesen et al.
gastrointestinal absorption after oral administration pro-
vided the best fit to the data (the model is depicted in
Fig. 3). The parameters for this model are summarized in
Table 3. The absorption process was characterized by a
mean transit time of 0.295 h (17.7 min), with a typical
value of N (number of compartments in the transit
chain) = 4.88. Visual predicted check plots performed for
the oxycodone pharmacokinetic model are shown in Fig. 4.
The model seems to describe the data of oxycodone well as
the visual predicted check plots shows that the 2.5th, 50th
and 97.5th percentiles of the observed data are closely
followed by the simulated 2.5th, 50th and 97.5th percen-
tiles. In addition, the areas of 95 % prediction intervals
around the 2.5th, 50th and 97.5th percentiles of the simu-
lated data overlap the 2.5th, 50th and 97.5th percentiles of
the observed data.
3.2 Pharmacodynamics
3.2.1 Placebo
A baseline model was selected for the placebo measured
effect. As the study was a three-way crossover study,
the same placebo data were used for both the CR665 or
oxycodone pharmacodynamic modelling. In the final CR665
or oxycodone models similar estimates of the population
mean baseline of 21.7 and 21.0 mL as well as similar base-
line IIV of 0.175 and 0.180 mL, respectively, were found
(see Table 4).
3.2.2 Cr665
Due to a large difference in concentrations of parent
compound and metabolite (Fig. 1), parent compound was
considered the driver for the dynamics and only parent
compound was included in the pharmacokinetic/pharma-
codynamic modelling. The model with the best fit of the
data from the CR665 arm of the study was Model 1,
describing a linear relationship between the concentration
and the effect, with IIV in the baseline–effect relationship
and a proportional error model. There was a clear
advantage of Model 1 over the baseline model, Model 2
(OFV difference of [5, see Table 5), demonstrating a
significant analgesic effect of CR665 on the oesophageal
distension VPTT. This finding was supported by the
positive value of slopeD (concentration–effect relation-
ship) for this model (Table 4). Model 3 had a higher OFV
than Model 1. Models 4, 5 and 6 all had a significant
drop in the OFV but EC50 values for Model 4 were very
low (0.05 ng/mL), and the ke0 of Models 5 and 6 was low
and high, respectively. None of Models 4, 5 and 6 passed
the covariance step of NONMEM�. The non-linear model
Emax Model 7 in Table 5 with an effect delay ke0 passed
the covariance step with an OFV difference of [10
compared to Model 1. The estimation of the parameters
were baseline 20.0 mL (95 % CI 14.5–25.5), ke0 3.61 h
(95 % CI 1.14–6.51), Emax 3.65 mL (95 % CI 2.07–5.23)
and EC50 11.2 ng/mL (95% CI –2.52 to 24.9). However,
considering the study design with a relatively sparse
number of pharmacodynamic samples and use of only
one dose level, the maximum effect of CR665 might not
have been obtained and, thus, Model 1 was selected as
the final pharmacokinetic/pharmacodynamic model for
CR665.
3.2.3 Oxycodone
The model with the best fit of the data from the oxycodone
arm of the study was Model 1, describing a linear rela-
tionship between the concentration and the effect, with IIV
Table 2 Parameters of the best pharmacokinetic model for CR665
and its N-oxidea
Parameter Population-
typical value
[%RSE]
Median parameter estimates of
1,000 bootstrap replicates
[2.5–97.5 % range]
V1 (L) 8.08 [8.33] 8.11 [6.93–9.60]
k13 ( = k30) [h-1] 0.501 [15.6] 0.505 [0.367–0.670]
k12 (h-1) 4.85 [9.65] 4.85 [3.79–5.70]
k21 (h-1) 0.0778 [13.1] 0.078 [0.056–0.099]
V3 (L) 488 [21.1] 493 [315–741]
Inter-individual variability (x)
xV1 0.0166 [65.1] 0.0155 [0.0012–0.0621]
xk12 0.0331 [57.4] 0.0314 [0.0041–0.0906]
xV3 0.164 [35.9] 0.145 [0.054–0.274]
Residual variability (r)
Proportional
error—CR665
(%CV)b
0.113 [12.9] 0.109 [0.083–0.143]
Proportional
error—N-oxide
(%CV)b
0.0311 [17.2] 0.030 [0.018–0.041]
CL clearance, CV coefficient of variation, k12 first-order rate constant
of drug transfer from the central compartment to the peripheral
compartment, k13 first-order rate constant of metabolite formation, k21
first-order rate constant of drug transfer from the peripheral com-
partment to the central compartment, k30 first-order rate constant of
metabolite elimination, RSE relative standard error, V1 apparent
volume of the central compartment for the parent compound, V3
apparent volume of the central compartment for the metabolite, WTbodyweighta Pharmacokinetic parameter values for CR665 following adminis-
tration of 0.36 mg/kg as intravenous infusion. All parameters were
bodyweight normalized so the volumes were scaled with WT/median
bodyweight (80 kg), and rate constants with (WT/median bodyweight
[80 kg])-0.25
b The residual errors (proportional) represented the variability in the
data that was unexplained by the model
PopPK/PD Study of CR665 and Oxycodone 131
in the baseline–effect relationship and a mixed structural
error model, i.e. a combined proportional and additive error
model. There was a clear advantage of Model 1 over the
baseline model (Model 2) as the OFV difference was
greater than 20 (see Table 6), thus demonstrating a sig-
nificant analgesic effect of oxycodone on the oesophageal
distension VPTT. This result was also supported by the
positive value of slopeD for this model (Table 4).
Models 3 and 4 also converged and passed the covari-
ance step with a non-significant drop in OFV and a sig-
nificant drop in OFV [6, respectively. The estimation of
the parameters of direct Emax for Model 3 were baseline
21.1 mL (95 % CI 16.8–25.4), Emax 13.2 mL (95 % CI
10.0–16.4) and EC50 166 ng/mL (95 % CI –69.2 to 401).
However, considering the study design with a relatively
sparse number of pharmacodynamic samples and use of
a
Time (h)
CR665
2.00.0 1.00.5 1.5
500
1000
Pla
sma
conc
entr
atio
n (n
g/m
L)b
Time (h)
CR665 N-oxide
5
10
Pla
sma
conc
entr
atio
n (n
g/m
L)
c
Time (h)
Oxycodone
10
20
Pla
sma
conc
entr
atio
n (n
g/m
L)
30
40
50
60
2.00.0 1.00.5 1.5 2.00.0 1.00.5 1.5
Fig. 4 Visual predictive check of pharmacokinetics of CR665 (a),
CR665N-oxide (b) and oxycodone (c) including 95 % prediction
intervals. On each plot: observed data (circles), median (solid red line),
2.5th and 97.5th percentiles (dashed red lines) of the observations. The
2.5th, 50th and 97.5th percentiles of the simulated data are plotted as
dashed black lines, solid black lines and dashed black lines, respec-
tively. The 95 % predicted confidence intervals around 2.5th, 50th and
97.5th percentiles of the simulated data are shown as coloured areas:
the 95 % confidence interval for the median (pink areas) and the 2.5th
and 97.5th percentiles (blue areas) of the simulated data
Table 3 Parameters of the best pharmacokinetic model for oxycodonea
Parameter Population typical value [%RSE] Median parameter estimates
of 1,000 bootstrap replicates [2.5–97.5 % range]
CL/F (L/h) 201 [5.12] 201 [185–221]
V1/F (L) 356 [6.43] 355 [310–401]
MTT (h) 0.295 [11.3] 0.300 [0.240–0.380]
NN -4.06 [9.29] –4.24 [–4.76 to –3.73]
Nb 4.88 [2.61–8.31] NC
Inter-individual variability (x)
xV1/F 0.0527 [42.1] 0.0420 [0.009–0.086]
xMTT 0.192 [31.6] 0.190 [0.07–0.32]
xNNc 0.516 [52.3] 0.490 [0.097–1.235]
Residual variability (r)
Proportional error (%CV)d 0.0638 [41.4] 0.0700 [0.0200–0.1300]
Additive error (nmol/L)d 0.131 [63.8] 0.098 [0.022–0.295]
CL/F apparent clearance after oral administration, CV coefficient of variation, MTT mean transit time for absorption, N number of compartments
in the transit chain (untransformed), NC not calculated, NN number of compartments in the transit chain (logit transformed), RSE relative
standard error, V1/F apparent volume of the central compartment for the parent compound after oral administration of the druga Pharmacokinetic parameter values for immediate-release oxycodone 15 mg, administered orallyb The logit transform constrained N to be between 1 and 200 via the equation N = 199 9 EXP(NN)/[1 ? EXP(NN)] ? 1c Additived The residual errors (additive and proportional) represented the variability in the data that was unexplained by the model. The additive error was
the component of the total error that was the same regardless of the opioid concentration. The proportional error was the component of the total
error that was proportional to the concentration
132 A. E. Olesen et al.
Ta
ble
4E
ssen
tial
feat
ure
so
fth
eb
est
mo
del
so
fth
ep
ain
mea
sure
Co
mpo
un
dan
dst
imu
lus
mo
dal
ity
Bas
elin
ep
ain
val
ue
[95
%C
I]B
asel
ine
IIV
(SE
[95
%C
I])
Conce
ntr
atio
n–ef
fect
rela
tionsh
ip(s
lopeD
)(p
ain
un
its
per
ng
/mL
of
com
po
un
d[9
5%
CI]
)
Slo
peD
IIV
(SE
[95
%C
I])
Res
idual
erro
r(%
[95
%C
I])
CR
665
VP
TT
(mL
)a
Po
pu
lati
on
esti
mat
eb2
1.7
[17
.5–
25
.9]
0.1
75
[0.0
81
5–
0.2
68]
0.0
03
5[0
.00
08
7–
0.0
061
]N
A0
.04
07
[0.0
276
–0
.05
38
]c
Bo
ots
trap
esti
mat
ed2
1.6
[17
.8–
26
.5]
0.1
59
[0.0
79
8–
0.2
67]
0.0
03
5[0
.00
08
4–
0.0
060
]N
A0
.04
00
[0.0
290
–0
.05
39
]
Ox
yco
do
ne
VP
TT
(mL
)a
Po
pu
lati
on
esti
mat
eb2
1.0
[16
.7–
25
.3]
0.1
8[0
.08
75
–0
.27
3]
0.0
79
3[0
.01
34
–0
.14
5]
1.3
4[–
0.0
96
7to
2.7
8]
0.1
51
[0.0
02
67
–0.0
275
]c;
4.5
2[–
1.0
9to
10
.1]e
Bo
ots
trap
esti
mat
ed2
1.3
[17
.6–
25
.8]
0.1
71
[0.0
91
–0
.26
7]
0.0
79
7[0
.02
32
–0
.15
2]
1.2
5[0
.15
9–
6.7
4]
0.0
13
5[0
.00
58
4–
0.0
26
6]c
;4
.53
[0.5
48–
10
.8]e
BIC
Bay
esia
nin
form
atio
ncr
iter
ion
,II
Vin
ter-
ind
ivid
ual
var
iab
ilit
y,
NA
no
tap
pli
cab
le,
SE
stan
dar
der
ror,
slop
eDco
nce
ntr
atio
n–
effe
ctre
lati
on
ship
,V
PT
Tv
isce
ral
pai
nto
lera
nce
thre
sho
lda
Mea
sure
das
the
vo
lum
eo
fw
ater
inth
eb
agp
lace
do
nan
oes
oph
agea
lp
rob
eb
Th
e9
5%
CIs
of
the
par
amet
eres
tim
ates
wer
eb
ased
on
the
asy
mp
toti
cst
and
ard
erro
rfo
rth
ep
aram
eter
retu
rned
by
NO
NM
EM
�
cP
rop
ort
ion
aler
ror
dM
edia
np
aram
eter
esti
mat
eso
f1
,00
0b
oo
tstr
apre
pli
cate
s[2
.5–
97
.5%
ran
ge]
eA
dd
itiv
eer
ror
(th
ep
aram
eter
was
no
tsi
gn
ifica
nt
ifth
eco
nfi
den
cein
terv
alin
clu
ded
zero
);h
ow
ever
,th
isad
dit
ive
erro
rst
abil
ized
the
mod
elto
con
ver
ge
Ta
ble
5M
od
el-b
uil
din
gst
eps
for
the
CR
66
5p
har
mac
ok
inet
ic/p
har
mac
od
yn
amic
mo
del
and
chan
ges
inth
eo
bje
ctiv
efu
nct
ion
val
ue
Mo
del
Eq
uat
ion
NO
PD
OF
V
1E
=b
asel
ine
?sl
op
eD9
C,
IIV
on
bas
elin
e1
4
2A
sM
od
el1
wit
hsl
op
eDfi
xed
toze
ro1
35
.30
2
3E
=m
9ln
(C?
bas
elin
e),
IIV
on
C0
14
7.4
5
4E
=b
asel
ine
?(E
max
9C
)/(E
C50
?C
),II
Vo
nb
asel
ine
and
Em
ax
16
–8
.15
1
5C
eff0
=k
e0
9(C
-C
eff),
E=
bas
elin
e?
slo
peD
9C
eff
,II
Vo
nb
asel
ine,
ke0
and
Slo
peD
17
–6
.39
4
6C
eff0
=k
e0
9(C
-C
eff),
E=
m9
ln(C
eff
?b
asel
ine)
,II
Vo
nb
asel
ine
and
ma
17
–1
4.1
94
7C
eff0
=k
e0
9(C
-C
eff),
E=
bas
elin
e?
(Em
ax
9C
eff
)/(E
C50
?C
eff
),II
Vo
nb
asel
ine
and
ke0
17
–1
3.6
26
Res
idu
aler
ror
mo
del
su
sed
for
the
ph
arm
aco
dy
nam
icp
art
wer
ep
rop
ort
ion
erro
rm
od
els
un
less
spec
ified
oth
erw
ise
Cp
lasm
ad
rug
con
cen
trat
ion
,C
0in
itia
l(fi
ctiv
e)o
rb
ack
-ex
trap
ola
ted
pla
sma
dru
gco
nce
ntr
atio
nat
tim
eze
rofo
llo
win
gb
olu
sin
trav
eno
us
inje
ctio
n,
Ceff
ob
serv
edco
nce
ntr
atio
nin
the
effe
ct
com
par
tmen
t,C
eff0
esti
mat
edco
nce
ntr
atio
nin
the
effe
ctco
mp
artm
ent,
Eti
me
cou
rse
of
the
effe
cto
fth
eo
xy
cod
on
ed
ose
,E
C50
pla
sma
dru
gco
nce
ntr
atio
nat
wh
ich
ther
ew
asa
hal
f-re
du
ctio
nin
the
pai
nsc
ore
,E
max
pla
sma
dru
gco
nce
ntr
atio
nat
wh
ich
ther
ew
asa
max
imu
mre
du
ctio
nin
the
pai
nsc
ore
,II
Vin
ter-
ind
ivid
ual
var
iab
ilit
y,
k e0
equ
ilib
riu
mra
teco
nst
ant
bet
wee
np
lasm
aan
dth
e
effe
ctco
mp
artm
ent,
wh
ich
was
also
ap
op
ula
tio
np
aram
eter
,m
slo
pe
of
the
app
aren
tli
nea
rse
gm
ent
of
the
log
-tra
nsf
orm
edco
nce
ntr
atio
nv
ersu
sre
spo
nse
curv
e,N
OP
nu
mb
ero
fes
tim
ated
par
amet
ers
inth
em
od
elin
clu
din
gp
har
mac
ok
inet
ics
par
amet
ers,
OF
Vo
bje
ctiv
efu
nct
ion
val
ue,
slo
peD
con
cen
trat
ion
–ef
fect
rela
tio
nsh
ipa
Mix
eder
ror
mo
del
use
d
PopPK/PD Study of CR665 and Oxycodone 133
only one dose level, the maximum effect of oxycodone
might not have been obtained and, thus, Model 1 was
selected as the final pharmacokinetic/pharmacodynamic
model for oxycodone. Model 5 passed the covariance step
of NONMEM�, but with an increase in OFV compared
with Model 1. The drop in OFV was significant for Model
6, but not for the later Model 7 when considering the
increased number of parameters compared with Model 1.
For Model 6 the ke0 was very small (\0.02 h) and, in
addition, neither of Models 6 or 7 passed the covariance
step of NONMEM�. An overview of the best effect model
fits for the various effect measures and the slopes of the
functions is given in Table 4.
Visual predicted check plots of effect observation and
population predictions of the model, as a summary of the
model predictions for the placebo, CR665 and oxycodone
arms of the study, are illustrated in Fig. 5. The models
seem to describe the pharmacodynamic data of placebo,
CR665 and oxycodone well as the visual predicted check
plots show that the 2.5th, 50th and 97.5th percentiles of the
observed data are closely followed by the by the simulated
2.5th, 50th and 97.5th percentiles. In addition, the areas of
95 % prediction intervals around the 2.5th, 50th and 97.5th
percentiles of the simulated data overlap with the 2.5th,
50th and 97.5th percentiles of the observed data.
3.2.4 CR665 Versus Oxycodone
Comparing the CR665 and oxycodone pharmacodynamic
models, the data fitted the same models and the baseline
values were similar, despite the small differences described
in Sects. 3.2.2 and 3.2.3. However, the slopes of the direct
linear pharmacodynamic model were found to be 0.0035
and 0.0793 VPTT mL per ng/mL for CR665 and oxyco-
done, respectively.
4 Discussion
A two-compartment and one-compartment distribution
model described the CR665 and CR665N-oxide metabolite
data, respectively, following a 1-h intravenous infusion.
For the observed oxycodone plasma concentrations, a one-
compartment model with a single transit chain representing
oral absorption provided the best fit for the data. The effect
of plasma concentration on VPTT for both CR665 and
oxycodone was best described by a direct linear model.
4.1 Methodological Considerations
More frequent pharmacodynamic evaluations and a longer
blood sampling period was not feasible in the present study
as the full study was a comprehensive pain study using a
variety of pain and safety measures [8]. This comprehen-
sive set-up in experimental pain studies is highly different
from a study in which the pharmacodynamic endpoint is,
for example, blood pressure, serum glucose, corrected QT
interval, hormone concentrations, temperature or circulat-
ing white cells. However, pharmacokinetic/pharmacody-
namic modelling can be performed sufficiently with
intervals of 30 min between pharmacodynamic evaluations
[9].
4.2 Pharmacokinetics
Together with the number of samples within the timeframe,
a two-compartmental distribution model for CR665 was
superior to a one-compartment model based on the drop in
the NONMEM� OFV of approximately 60. Distribution of
the CR665 and N-oxide model in the final pharmacokinetic
model resulted in different distributions of parent and
metabolite, which might be due to the relatively short
Table 6 Model-building steps for the oxycodone and changes in the objective function value
Model Equation NOP DOFV
1 E = baseline ? slopeD 9 C, IIV on baseline and SlopeD 15
2 As Model 1 with slopeD fixed to zero 13 21.972
3 E = m 9 ln(C ? baseline), IIV on m 15 0.459
4 E = baseline ? (Emax 9 C)/(EC50 ? C), IIV on baseline and EC50 16 –6.498
5 Ceff0 = ke0 9 (C - Ceff), E = baseline ? slopeD 9 Ceff, IIV on baseline and SlopeD 16 4.828
6 Ceff0 = ke0 9 (C - Ceff), E = m 9 ln(Ceff ? baseline), IIV on baseline and m 15 –5.279
7 Ceff0 = ke0 9 (C - Ceff), E = baseline ? (Emax 9 Ceff)/(EC50 ? Ceff), IIV on baseline, ke0, Emax and EC50 19 –6.498
Residual error models used for the pharmacodynamic part were mixed error models
C plasma drug concentration, Ceff observed concentration in the effect compartment, Ceff0 estimated concentration in the effect compartment,
E time course of the effect of the oxycodone dose, EC50 plasma drug concentration at which there was a half-reduction in the pain score, Emax
plasma drug concentration at which there was a maximum reduction in the pain score, IIV inter-individual variability, ke0 equilibrium rate
constant between plasma and the effect compartment, which was also a population parameter, m slope of the apparent linear segment of the
log-transformed concentration versus response curve, NOP number of estimated parameters in the model including pharmacokinetics parameters,
OFV objective function value, slopeD concentration–effect relationship
134 A. E. Olesen et al.
blood sampling collection period of 2 h in which the
N-oxide metabolite only seems to be formed and distri-
bution/elimination of the metabolite has not been captured
(see Fig. 1). This in turn caused the slightly empirical
model for CR665 and the metabolite where k13 = k30
along with the different distribution of the two compounds.
Based on the current data, it might be difficult to conclude
whether the formation rate of the metabolite is rate limiting
for its elimination or if the N-oxide elimination is con-
trolled by its own elimination. Thus, studies including
longer sampling periods are warranted for a more extensive
pharmacokinetic description of CR665 and the N-oxide.
The resulting one-compartment distribution model with
a single chain of transit compartments for oxycodone was
also supported by previous studies with longer sampling
periods [27, 28]. The final pharmacokinetic model and
parameter estimates for oxycodone were also comparable
with previous findings from a study with a short pharmaco-
kinetic sampling period [11]. This earlier study identified a
similar one-compartment model with a single transit chain
a
Time (h)
Placebo
20
40
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
60
80
b
Time (h)
CR665
20
40
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
60
80
c
Time (h)
Placebo
20
40
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
60
80
d
Time (h)
Oxycodone
20
40
Vis
cera
l pai
n to
lera
nce
thre
shol
d (m
L)
60
80
2.00.0 1.00.5 1.5 2.00.0 1.00.5 1.5
2.00.0 1.00.5 1.5 2.00.0 1.00.5 1.5
Fig. 5 Visual predictive check of pharmacodynamics depicted as
effect on visceral pain tolerance threshold vs. time. Placebo (a),
CR665 (b), placebo (c) and oxycodone (d) with 95 % prediction
intervals. On each plot: observed data (circles), median (solid redline), 2.5th and 97.5th percentiles (dashed red lines) of the
observations. The 2.5th, 50th and 97.5th percentiles of the simulated
data are plotted as dashed black lines, solid black lines and dashedblack lines, respectively. The 95 % predicted confidence intervals
around 2.5th, 50th and 97.5th percentiles of the simulated data are
shown as coloured areas: the 95 % confidence interval for the median
(pink areas) and the 2.5th and 97.5th percentiles (blue areas) of the
simulated data
PopPK/PD Study of CR665 and Oxycodone 135
representing the absorption phase, with a mean transit time
of 26 min, which is comparable with the result obtained
from the current study (0.295 h, corresponding to 18 min).
The number of transit compartments of 4.88 (range
2.6–8.3) resulting from the Stirling approximation inherent
in the model [29] seems to be in a comparable range with
that of a previous finding of 2.66 (2.2–7.2) after oral
administration of oxycodone [11].
4.3 Pharmacodynamics
The pharmacokinetic/pharmacodynamic relationship in the
present study revealed identical models for CR665 and
oxycodone with similar baseline model and parameter
estimates for the placebo treatment. A direct linear effect of
the plasma drug concentration acting on the VPTT
described by the slope of the models indicated a steeper
slope of oxycodone of 0.0793 VPTT (mL per ng/mL)
[balloon volume per plasma concentration] than that of
0.0035 (mL per ng/mL) for CR665. This observation could
indicate that oxycodone is more potent than CR665 in
the VPTT model, e.g. due to its additional effects on
l-receptors in the CNS, or simply that oxycodone and
CR665 were not given in equivalent doses relative to their
intrinsic efficacy. Such dose equivalencies can be estab-
lished readily between two compounds if they share an
efficacy biomarker response, e.g. miosis for CNS-acting
l-opioids, but this has not yet been possible for oxycodone
and a peripherally selective j-opioid such as CR665.
4.4 Peripheral Versus Central Effect?
Clinically used opioids are generally believed to exert their
main analgesic effects via l-opioid receptors in the CNS.
Consistent with this view, there is evidence for a delay in
the time course of the pharmacodynamic effect (i.e. relief
of somatic pain) with respect to the time course of the
plasma drug concentration following opioid administration
[9, 30]. This delay in the analgesic response relative to
the pharmacokinetic profile, described as ‘hysteresis’, is
usually considered a consequence of rate-limiting transport
across the blood–brain barrier, in addition to other factors.
In principle, both the efficacy and the immediacy of vis-
ceral pain relief reported here could be explained by the
involvement of peripheral j-opioid receptors on visceral
afferents as rat studies have demonstrated that j-opioids
are more efficient than l-opioids in reducing visceral pain
[31]. However, from the present study, it is not possible to
make a clear evaluation of the effect–delay component and
therefore other explanations for the current observations
cannot at present be excluded, e.g. a very rapid uptake of
oxycodone into a CNS effect compartment.
The development of a peripherally selective j-opioid
remains a powerful approach to creating selective, rationally
targeted visceral pain analgesics. Substantial evidence has
emerged for the significance of peripherally located j-opioid
receptors as a target for treating visceral pain [4]. FE200041
(an earlier analogue of CR665) and CR665 have been
developed with the idea that these peptides should be less
likely to cross the blood–brain barrier and to cause centrally
mediated adverse effects [6, 7]. Evaluation of CR665 dem-
onstrated peripheral j-opioid selectivity and antinociceptive
activity in a wide range of preclinical visceral pain models
[6, 7]. However, in the present study, some CNS effect was
seen in the form of paraesthesia, which was reported as an
adverse event in 61.1 % of the healthy volunteers receiving
CR665 [8]. It cannot be excluded that CR665 also exerted
analgesic effects in the CNS.
Oxycodone metabolites were not included in this phar-
macokinetic/pharmacodynamic analysis, as it has previ-
ously been demonstrated that the pharmacodynamic data
are driven by the plasma concentration of the parent
compound [11]. This was again supported by successful
fitting of the pharmacodynamic data to the plasma con-
centration of the parent compound.
In the future, the integration of population-based phar-
macokinetic/pharmacodynamic models can be used for a
clinical trial simulation where trial design and execution
can be explored relative to performance and outcomes [32].
5 Conclusion
Pharmacokinetic/pharmacodynamic modelling provided
results consistent with a peripheral visceral analgesic effect
of both CR665 and oxycodone, which could be mediated
by opioid receptors on visceral afferent nerves. The mod-
elling showed a direct linear plasma drug concentration–
effect relationship for the plasma drug concentration acting
on the VPTT.
Acknowledgments Stephen Hwang, PhD, and Sarita Khanna, PhD,
are acknowledged for participating in the study design.
Conflicts of interest The study was sponsored by Johnson and
Johnson (Mountain View, CA, USA). Sherron Kell is affiliated with
the sponsor and took part in study design, review and approval of the
manuscript. Gilbert Y. Wong was affiliated with the sponsor at the
time the study was conducted, and took part in study design, review
and approval of the manuscript. Sherron Kell owns stock in Johnson
& Johnson. Asbjørn Mohr Drewes has received unrestricted research
grants from Mundipharma, AstraZeneca, Lundbeck and Pfizer and
served as a Consultant/Advisory Board member for Mundipharma,
AstraZeneca and Shire. Anne Estrup Olesen has received honorarium
from Mundipharma. Camilla Staahl is currently employed by Gru-
nenthal GmBH, but was affiliated with Mech-Sense during conduct of
study and preparation of manuscript.
136 A. E. Olesen et al.
References
1. Konig HH, Bernert S, Angermeyer MC. Health status of the
German population: results of a representative survey using the
EuroQol questionnaire. Gesundheitswesen. 2005;67(3):173–82.
2. Bell TJ, Panchal SJ, Miaskowski C, et al. The prevalence,
severity, and impact of opioid-induced bowel dysfunction: results
of a US and European Patient Survey (PROBE 1). Pain Med.
2009;10(1):35–42.
3. Stein C, Schafer M, Machelska H. Attacking pain at its source:
new perspectives on opioids. Nat Med. 2003;9(8):1003–8.
4. Riviere PJ. Peripheral kappa-opioid agonists for visceral pain. Br
J Pharmacol. 2004;141(8):1331–4.
5. Burton MB, Gebhart GF. Effects of kappa-opioid receptor ago-
nists on responses to colorectal distension in rats with and without
acute colonic inflammation. J Pharmacol Exp Ther. 1998;285(2):
707–15.
6. Binder W, Machelska H, Mousa S, et al. Analgesic and antiin-
flammatory effects of two novel kappa-opioid peptides. Anes-
thesiology. 2001;94(6):1034–44.
7. Vanderah TW, Largent-Milnes T, Lai J, et al. Novel D-amino
acid tetrapeptides produce potent antinociception by selectively
acting at peripheral kappa-opioid receptors. Eur J Pharmacol.
2008;583(1):62–72.
8. Arendt-Nielsen L, Olesen AE, Staahl C, et al. Analgesic efficacy
of peripheral kappa opioid receptor agonist CR665 compared to
oxycodone in a multi-modal, multi-tissue experimental human
pain model: selective effect on visceral pain. Anesthesiology.
2009;111:616–24.
9. Staahl C, Upton R, Foster DJ, et al. Pharmacokinetic-pharma-
codynamic modeling of morphine and oxycodone concentrations
and analgesic effect in a multimodal experimental pain model.
J Clin Pharmacol. 2008;48(5):619–31.
10. Andresen T, Upton RN, Foster DJ, et al. Pharmacokinetic/phar-
macodynamic relationships of transdermal buprenorphine and
fentanyl in experimental human pain models. Basic Clin Phar-
macol Toxicol. 2011;108(4):274–84.
11. Olesen AE, Upton R, Foster DJ, et al. A pharmacokinetic and
pharmacodynamic study of oral oxycodone in a human experi-
mental pain model of hyperalgesia. Clin Pharmacokinet.
2010;49(12):817–27.
12. Shafer SL, Struys MM. Mixed effect modeling in analgesia trials.
Anesth Analg. 2008;107(1):9–10.
13. Csajka C, Verotta D. Pharmacokinetic-pharmacodynamic mod-
elling: history and perspectives. J Pharmacokinet Pharmacodyn.
2006;33(3):227–79.
14. Rosenbaum SE, Carter AA, Dudley MN. Population pharmaco-
kinetics: fundamentals, methods and applications. Drug Dev Ind
Pharm. 1995;21(9):1115–41.
15. Holtman JR, Wala EP. Characterization of the antinociceptive
effect of oxycodone in male and female rats. Pharmacol Biochem
Behav. 2006;83(1):100–8.
16. Nielsen CK, Ross FB, Lotfipour S, et al. Oxycodone and mor-
phine have distinctly different pharmacological profiles:
radioligand binding and behavioural studies in two rat models of
neuropathic pain. Pain. 2007;132(3):289–300.
17. Nozaki C, Saitoh A, Kamei J. Characterization of the antinoci-
ceptive effects of oxycodone in diabetic mice. Eur J Pharmacol.
2006;535(1–3):145–51.
18. Ross FB, Smith MT. The intrinsic antinociceptive effects of
oxycodone appear to be kappa-opioid receptor mediated. Pain.
1997;73(2):151–7.
19. Drewes AM, Gregersen H. Multimodal pain stimulation of the
gastrointestinal tract. World J Gastroenterol. 2006;12(16):
2477–86.
20. Drewes AM, Schipper KP, Dimcevski G, et al. Multimodal
assessment of pain in the esophagus: a new experimental model.
Am J Physiol Gastrointest Liver Physiol. 2002;283(1):G95–103.
21. Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit: a collection of
computer intensive statistical methods for non-linear mixed effect
modeling using NONMEM. Comput Methods Programs Biomed.
2005;79(3):241–57.
22. Lindbom L, Ribbing J, Jonsson EN. Perl-speaks-NONMEM
(PsN): a Perl module for NONMEM related programming.
Comput Methods Programs Biomed. 2004;75(2):85–94.
23. Savic RM, Jonker DM, Kerbusch T, et al. Implementation of a
transit compartment model for describing drug absorption in
pharmacokinetic studies. J Pharmacokinet Pharmacodyn. 2007;
34(5):711–26.
24. Fang L, Holford NH, Hinkle G, et al. Population pharmacoki-
netics of humanized monoclonal antibody HuCC49deltaCH2 and
murine antibody CC49 in colorectal cancer patients. J Clin
Pharmacol. 2007;47(2):227–37.
25. Yano Y, Beal SL, Sheiner LB. Evaluating pharmacokinetic/
pharmacodynamic models using the posterior predictive check.
J Pharmacokinet Pharmacodyn. 2001;28(2):171–92.
26. Holford NH. A size standard for pharmacokinetics. Clin Phar-
macokinet. 1996;30(5):329–32.
27. Lalovic B, Kharasch E, Hoffer C, et al. Pharmacokinetics and
pharmacodynamics of oral oxycodone in healthy human subjects:
role of circulating active metabolites. Clin Pharmacol Ther.
2006;79(5):461–79.
28. Mandema JW, Kaiko RF, Oshlack B, et al. Characterization and
validation of a pharmacokinetic model for controlled-release
oxycodone. Br J Clin Pharmacol. 1996;42(6):747–56.
29. Kass R, Raftery A. Bayes factors. J Am Stat Assoc. 1995;90:
773–95.
30. Lotsch J. Pharmacokinetic-pharmacodynamic modeling of opi-
oids. J Pain Symptom Manage. 2005;29(5):S90–103.
31. Ozaki N, Sengupta JN, Gebhart GF. Differential effects of mu-,
delta-, and kappa-opioid receptor agonists on mechanosensitive
gastric vagal afferent fibers in the rat. J Neurophysiol.
2000;83(4):2209–16.
32. Barrett JS, Fossler MJ, Cadieu KD, et al. Pharmacometrics: a
multidisciplinary field to facilitate critical thinking in drug
development and translational research settings. J Clin Pharma-
col. 2008;48(5):632–49.
PopPK/PD Study of CR665 and Oxycodone 137
Reproduced with permission of the copyright owner. Further reproduction prohibited withoutpermission.
