A Recirculatory Model for Pharmacokinetics and

the Effects on Bispectral Index After Intravenous

Infusion of the Sedative and Anesthetic AZD3043
in Healthy Volunteers
Marcus A. Bjrnsson, MSc Pharm, PhD,* ke Norberg, MD, PhD, Sigridur Kalman, MD, PhD,
and Ulrika S. H. Simonsson, MSc Pharm, PhD

BACKGROUND: AZD3043 is a positive allosteric modulator of the -aminobutyric acid type

A receptor, with sedative and anesthetic properties. We describe a population pharmacoki-
netic (PK) model of arterial and venous concentrations of AZD3043 and the pharmacodynamic
effects on bispectral index (BIS) in healthy volunteers.
METHODS: Arterial and venous plasma concentrations of AZD3043 and BIS were measured
in 2 clinical studies in 125 healthy volunteers, where AZD3043 was given as a 1-minute bolus
(16 mg/kg), a 30-minute infusion (181 mg/kg/h), or 0.8 + 10, 1 + 15, 3 + 30, and 4 + 40
(mg/kg bolus + mg/kg/h infusion for 30 minutes). Population PK/pharmacodynamic analysis
was performed with NONMEM.
RESULTS: A recirculatory model, comprising a series of 5 compartments for the transit of drug
between venous and arterial plasma, 2 peripheral distribution compartments, and 1 compart-
ment for the nondistributive transit of drug from arterial to venous plasma, described the PK of
AZD3043. Systemic clearance was high (2.2 L/min; 95% confidence interval, 2.122.25), and
apparent volumes of distribution were low, leading to a short elimination half-life. The apparent
volumes of distribution of the arterial and peripheral compartments increased with increasing
administered dose, giving a total apparent volume of distribution of 15 L after the lowest dose
and 37 L after the greatest dose. A sigmoid maximum effect (Emax) model with an EC50 of 15.6
g/mL and a of 1.7 described the relationship between AZD3043 effect-site concentrations
and BIS. The between-subject variability in EC50 was 37%. An effect compartment model, with a
half-life of the equilibration rate constant ke0 of 1.1 min, described the delay in effect in relation
to the arterial plasma concentrations.
CONCLUSIONS: AZD3043 had a high clearance and a low apparent volume of distribution,
leading to a short half-life. However, the apparent volume of distribution was dose dependent
(P<0.001), leading to an increased half-life with increasing dose. The distribution to the effect
site was fast and together with the short plasma half-life led to a fast onset and offset of
effects. (Anesth Analg 2015;121:90413)

A
n easily controllable depth of anesthesia is an impor-
tant property for a new anesthetic compound. This
could be achieved with a short half-life, resulting
from rapid metabolism and a low volume of distribution,
and rapid equilibration between blood and the site of action.
Pharmacokinetic (PK)/pharmacodynamic (PD) modeling is
a useful tool to assess these properties.1 A quantitative mea-
sure of the depth of anesthesia, to be used for guiding the dos-
ing of the anesthetic, can be obtained by the bispectral index
(BIS), which is derived from the electroencephalogram.2
From the *Quantitative Clinical Pharmacology, AstraZeneca R&D,
Sdertlje, Sweden; Department of Pharmaceutical Biosciences, Uppsala
University, Uppsala, Sweden; Department of Clinical Science Intervention
and Technology, Karolinska Institutet, Stockholm, Sweden; and Department
of Anesthesiology and Intensive Care, Karolinska University Hospital,
Huddinge, Sweden.
Accepted for publication April 5, 2015.
Funding: This study was funded by AstraZeneca R&D.
Conflict of Interest: See Disclosures at the end of the article.
Reprints will not be available from the authors.
Address correspondence to Ulrika S. H. Simonsson, MSc Pharm, PhD, De-
partment of Pharmaceutical Biosciences, Uppsala University, Uppsala, Swe-
den. Address e-mail to ulrika.simonsson@farmbio.uu.se.
Copyright 2015 International Anesthesia Research Society
DOI: 10.1213/ANE.0000000000000814
AZD3043, originally called THRX-918661, is a posi-
tive allosteric modulator of the -aminobutyric acid type
A receptor, with sedative and anesthetic properties.3 It is
metabolized by esterases in plasma and the liver, leading
to a high systemic clearance and a short half-life. In 2 clini-
cal studies in healthy volunteers, AZD3043 induced seda-
tion and anesthesia in a dose-dependent manner, and the
subjects recovered rapidly after the end of the AZD3043
infusions.4,5
The aims of this analysis were to describe the population
PK of arterial and venous AZD3043 concentrations with a
recirculatory model and the relationship between AZD3043
concentrations and the effect on BIS in healthy volunteers,
including potential covariate (COV) effects of body weight,
sex, age, and esterase activity.
METHODS
Study Designs
This analysis was performed with data from the first 2 clini-
cal studies of AZD3043 in healthy volunteers (clinicaltrials.
gov identifiers NCT00918515 and NCT00984880).4,5 Both
studies were performed at Karolinska University Hospital,
Huddinge, Sweden, in accordance with the Declaration
of Helsinki and Good Clinical Practice. The studies were

904 www.anesthesia-analgesia.org October 2015 Volume 121 Number 4

Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 approved by the Regional Independent Ethics Committee
in Stockholm, Sweden, and the Swedish Medical Products
Agency, and written informed consent was obtained from
all subjects.
The first study4 was a dose-escalation study, where
30-minute constant rate IV infusions of 1, 3, 6, 12, 18, 27, 36,
54, or 81 mg/kg/h were given to 53 healthy male volun-
teers. There were 6 subjects per group except for the 1-mg
group, where there were 5 volunteers. All subjects had nor-
mal plasma butyrylcholinesterase activity, as measured with
the dibucaine number test.6,7 Subject demographics are pre-
sented in Table1. Arterial and venous plasma samples were
obtained before start of infusion and at 2, 5, 15, 29, 31, 32, 35,
37, 40, 45, 60, 75, 90, and 120 minutes after start of infusion.
In addition, venous plasma samples were obtained at 4.5, 8,
and 24 hours after start of infusion.
The second study5 was divided into 2 parts. In part A,
each subject received a 1-minute IV bolus of 1, 1.5, 2, 4, or
6 mg/kg. In part B, each subject received 0.8 + 10, 1 + 15, 3
+ 30, or 4 mg/kg bolus + 40 mg/kg/h 30-minute infusion.
Different subjects were included in the 2 parts, i.e., each
subject received only 1 treatment. Seventy-two healthy vol-
unteers (men or women of nonchildbearing potential, with
normal plasma butyrylcholinesterase activity), 8 subjects
per group, were included in the study (Table1). Arterial and
venous plasma samples were collected and analyzed for
AZD3043 as described previously, before start of infusion, at
50 to 55 seconds (before stop of 1-minute infusion); at 1, 4, 14,
and 29 minutes during the 30-minute infusion (part B); and
at 1, 4, 6, 9, 14, 29, 44, 59, and 119 minutes after the end of the
1-minute (part A) or 30-minute (part B) infusion. In addition,
venous samples were also collected at 4 hours after stop of
infusion and at 8 and 24 hours after start of infusion.
The number of patients included in these exploratory
studies was not determined based on any statistical con-
siderations but rather on previous experience with similar
studies. In a survey of 105 first-in-human studies published
between 1995 and 2004,8 the most common number of sub-
jects on active treatment per cohort was 6, as in the first
study of this work. In the second study included in this
work, the number of subjects was increased to 8.
All samples from both studies were collected in the
presence of stabilizers to avoid ester-induced hydrolysis of
AZD3043 and analyzed for total AZD3043 concentrations
by Huntingdon Life Sciences Ltd, Huntingdon, United
Kingdom, by the use of liquid chromatography with tan-
dem mass spectrometry. The method was validated over
the concentration range 0.01 to 20 g/mL, and the over-
all mean precision (percent coefficient of variation, or
%CV) and accuracy (%Bias) for the QC samples at 3 con-
centrations in plasma were 8.4% and 4.6%, respectively.
Table 1.Descriptive Statistics of Demographics and Covariates
n Age (y) Sex (M/F)
Study 1 53 25 (1842) 53/0 (100/0)
Study 2 70b 31 (2065) 62/8 (89/11)
Total 123b 28 (1865) 115/8 (93/7)
Data for sex are represented as n (%). Other data are represented as median (range).
BMI = body mass index; F = female; M = male.
a
Dibucaine numbers >70 were considered normal.7
b
Two subjects were excluded from the analysis because of interruptions in the infusions and are not included in this table.
October 2015 Volume 121 Number 4 www.anesthesia-analgesia.org 905
Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Unbound AZD3043 concentrations were measured in the
venous samples obtained at 29 and 40 minutes after start of
infusion in the first study. In both studies, BIS (BIS VISTA,
Aspect Medical Systems, Inc., Norwood, MA) was elec-
tronically captured every minute until the subjects were
awake according to the clinical signs, including Modified
Observers Assessment of Alertness/Sedation scale.9
Data Analysis
Nonlinear mixed effects modeling was performed with
NONMEM 7.2 (ICON Development Solutions, Ellicott
City, MD).10 The first-order conditional estimation method
with interaction was used for all estimations. Perl speaks
NONMEM (PsN)11 was used for automating and control-
ling runs, and Xpose 412 and R (http://www.r-project.org)
were used to produce goodness-of-fit graphics, including
visual predictive checks (VPCs).
The modeling was performed sequentially, starting with
the PK model, followed by the analysis of BIS. Models were
selected based on the goodness-of-fit plots, precision in
parameter estimates, and statistically using the objective
function value (OFV). A decrease in OFV by 6.63 (corre-
sponding to a P value <0.01) was required for a parameter
to be included in the model. Confidence intervals of the
parameters were obtained from nonparametric bootstraps
with 1000 samples, implemented in PsN. Because the
range of doses was large and every individual was dosed
based on their body weight, VPC was normalized by the
population predictions (PREDs) for each individual via the
PRED-corrected VPC (PC-VPC) in PsN.13 In a PC-VPC, each
observed or simulated value is divided by the correspond-
ing PRED for that individual at that time and multiplied
by the median PRED of all individuals, regardless of dose,
in that time-bin. That way all data are normalized to fit on
the same scale while the shape of the curve is maintained.
Log-likelihood profiles for the PK parameters were created
by fixing the parameters, 1 at a time, to different values and
re-estimating the other parameters. The difference in OFV
was then plotted versus the value of the fixed parameter.
Log-normal distribution of the parameters around the
typical value was assumed:
Pi = P exp(i )
where Pi is the value of the parameter in individual i, P is the
typical value of the parameter in the population, and i is
the normally distributed interindividual random variability
with mean 0 and variance 2. Additive, proportional, and
combined additive and proportional residual error models
were investigated.
Body weight (kg) BMI (kg/m2) Dibucaine numbera
76 (5796) 24 (1930) 134 (89192)
77 (5498) 24 (1930) 131 (77182)
77 (5498) 24 (1930) 133 (77192)
 Recirculatory Pharmacokinetics and Pharmacodynamics of AZD3043
Median prediction error (MDPE) and median absolute
performance error (MDAPE) were calculated for the final
models according to Varvel et al.14 Plasma concentrations
that were less than the limit of quantification (0.01 g/mL)
were excluded from the analysis. In case the concentra-
tions increased after observations below the limit of quan-
tification, these samples were also excluded. In 4 subjects
in study 1 and in 16 subjects in study 2, unexpected, and
sometimes large, increases in venous concentrations were
observed at late time points, likely due to sampling errors
(sample contamination by initial dose) because the corre-
sponding metabolite concentrations did not increase. These
samples were regarded as artifacts and were not included
in the analysis. In 2 subjects, there were BIS values that sud-
denly decreased to 0 at the exact times of several arterial
blood pressure measurements. These values were regarded
as artifacts and were excluded from the analysis. Two sub-
jects in the second study were excluded from all analyses
because there were interruptions in the infusions, and there-
fore, the exact dosing history was not known. This means
that the total number of subjects included in the analysis
was 123. In 2 subjects, sudden transient increases in BIS to
high values were seen during infusion while the subjects
were deeply anesthetized. These increased values were
regarded as artifacts and were excluded from the analysis.
PK Analysis
A recirculatory PK model was developed and fitted to arte-
rial and venous concentrations simultaneously. The model
consisted of an arterial and a venous compartment, periph-
eral distribution compartments, a tanks-in-series transit
representing the transport of drug through the heart and
lungs, and a nondistributive pathway between arterial
and venous plasma (Fig.1). First, the structural model was
identified by increasing the number of compartments in the
central circulation, the nondistributive pathway, and the
peripheral distribution until no further improvement of the
906
www.anesthesia-analgesia.org
Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
fit was found. Drug elimination was assumed to occur from
the arterial compartment, but elimination from the venous
compartment also was assessed. Thereafter, the effect of
dose, accumulated administered dose, and concentration
on the clearance and apparent volume of distribution were
evaluated. The influence of body weight on clearance and
volume parameters was then assessed by having them not
scaled to body weight, directly proportional to body weight,
or scaled allometrically to body weight (normalized to the
median value 77 kg), where the volume parameters were
assumed to be directly proportional to body weight, while
the elimination and intercompartmental clearance param-
eters were related to body weight raised to the power 0.75.15
Finally, the effects of esterase activity, sex, and age on clear-
ance were investigated. COVs were modeled as linear rela-
tions, centered at the median value of the COV:
P = 1 (1 + 2 ( COV COVmedian ) )
where 1 is the typical value of parameter P for an indi-
vidual with the median value of a COV (COVmedian), and 2
is the fractional change in the parameter for each unit the
COV differs from the median. For the categorical COV, sex,
the effect on clearance was modeled as a fractional change
between male and female volunteers.
PD Analysis
When developing the BIS model, the PK parameters (both
fixed and random effects) were fixed to their estimates from
the final PK model and the PK data were retained in the
dataset.16 This approach conditions the PD analysis not only
on the PK parameters but also on the PK data, accounting for
uncertainty in the individual predictions of the PK param-
eters. An effect-compartment model was used to describe
the delay of effects on BIS in relation to the arterial plasma
concentrations of AZD3043. The rate constant of the effect
Figure 1. Schematic representation of
the final recirculatory pharmacokinetic
model. CL = elimination clearance;
ke0 = rate constant for equilibration
between plasma and effect-site concentra-
tion; QCO = cardiac output (plasma flow);
QND = plasma flow of nondistributive
pathway; QP1 = plasma flow through
peripheral compartment 1; QP2 = plasma
flow through peripheral compartment
2; VA = arterial volume; VV = venous vol-
ume; VC = volume of central circulation;
VP1 = volume of peripheral compartment 1;
VP2 = volume of peripheral compartment 2.
anesthesia & analgesia
 delay was described by the parameter ke0. A 2-compartment From the arterial plasma, with the volume VA, the drug
effect-site model17 also was tested, in which distribution is transferred to the venous plasma, with the volume VV,
was assumed to occur from the effect-site compartment to through 2 peripheral distribution compartments with the
peripheral parts of the brain. The effect-site concentrations apparent volumes VP1 and VP2 and 1 compartment for the
were related to BIS by a linear model, a maximum effect nondistributive transit. The plasma flow through the non-
(Emax) model, and a sigmoid Emax model, according to the distributive and distribution compartments is represented
following equation: by QND, QP1, and QP2. In the final model, elimination clear-
ance occurred from the arterial compartment. Plasma clear-
E Ce ance was high (2.2 L/min), which was as much as 76% of the
BIS = Baseline 1 max
estimated cardiac output (plasma flow), with low between-
EC50 + Ce
subject variability (14%). The effect of body weight on the
parameters was best described by an allometric model. No
where Baseline is the BIS before start of drug administra- significant effects (P > 0.1) of dose, esterase activity, sex, or
tion, Emax is the maximum effect (in this analysis fixed to 1 age on clearance were found. The arterial and the peripheral
as very low BIS values were observed), EC50 is the effect- apparent volumes of distribution increased with increasing
site concentration needed to achieve 50% of the maximum administered dose. Total apparent volume of distribution
effect, Ce is the concentration at the effect site, and is a (the sum of the distribution volumes in the arterial, venous,
shape factor. The effects of sex, age, and body weight on central, and peripheral compartments at the end of infu-
EC50 were investigated. The criteria for including a COV in sion) was 15 L in the lowest dose group (receiving a total
the model were the same as for the PK analysis. dose of 0.5 mg/kg) and increased with increasing infused
dose up to 37 L in the greatest dose group (40.5 mg/kg).
RESULTS Shrinkage was <30% for all PK parameters except VV (31%)
Pharmacokinetics and QCO (49%). A high shrinkage implies that there is lim-
Observed and predicted arterial and venous concentrations ited information of a parameter in some individuals, and
of AZD3043, stratified by the different dose groups, are therefore, their individual estimates of the parameters are
presented in Figures2 and 3, respectively. During the infu- shrunk toward the population mean. For example, in the
sions, arterial concentrations were greater than venous con- case of oral absorption, a patient with only samples late in
centrations, whereas after the end of the infusions, venous the elimination phase will provide no information on the
concentrations were greater than arterial concentrations absorption parameters and that patients individual predic-
(Fig.4). The unbound fraction of AZD3043 increased with tions of the absorption parameters will be the population
increasing concentrations, from 5% to 12% in the lowest mean. The population parameters for absorption may still
dose group to 18% to 30% in the greatest dose group. be well estimated if other individuals provide data on the
A recirculatory model was used to fit arterial and venous absorption phase. If shrinkage is high, goodness-of-fit plots
plasma concentrations of AZD3043 simultaneously (Fig.1). using individual predictions may be less useful.18
The final model consisted of a series of a 5 tanks-in-series Key modeling steps are listed in Table2. The parameter
transit of drug from venous plasma, where the drug was estimates for the final PK model are presented in Table 3.
administered, to arterial plasma. This represents the central The PC-VPC for the final PK model, stratified for arterial
circulation, described by the cardiac output (QCO) and a cen- and venous plasma and the 3 different dosing regimens,
tral apparent volume of distribution (VC). The cardiac out- is shown in Figure5, and the plots of observed divided by
put was not measured in the studies but rather estimated as population predicted concentrations over time are shown
a plasma flow based on the limited sampling of AZD3043. in Figure6. MDPE and MDAPE are summarized in Table4.

1 mg/kg/h 3 mg/kg/h 6 mg/kg/h 12 mg/kg/h 18 mg/kg/h 27 mg/kg/h

100
Arterial Concentration (ug/mL)

10
1
0.1
36 mg/kg/h 54 mg/kg/h 81 mg/kg/h 1 mg/kg bolus 1.5 mg/kg bolus 2 mg/kg bolus

100
10
Figure 2. Individual observed (blue)
1
and typical predicted (red) arterial
0.1 plasma concentrations of AZD3043
4 mg/kg bolus 6 mg/kg bolus 0.8 mg/kg + 10 mg/kg/h 1 mg/kg + 15 mg/kg/h 3 mg/kg + 30 mg/kg/h 4 mg/kg + 40 mg/kg/h versus time, stratified for different
100 dosing regimen.
10
1
0.1

0 30 60 90120 0 30 60 90120 0 30 60 90120 0 30 60 90120 0 30 60 90120 0 30 60 90120

Time (min)

October 2015 Volume 121 Number 4 www.anesthesia-analgesia.org 907

Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 Recirculatory Pharmacokinetics and Pharmacodynamics of AZD3043

1 mg/kg/h 3 mg/kg/h 6 mg/kg/h 12 mg/kg/h 18 mg/kg/h 27 mg/kg/h

100
Venous Concentration (ug/mL)

10
1
0.1
36 mg/kg/h 54 mg/kg/h 81 mg/kg/h 1 mg/kg bolus 1.5 mg/kg bolus 2 mg/kg bolus

100
10
Figure 3. Individual observed (blue)
1
and typical predicted (red) venous
0.1 plasma concentrations of AZD3043
4 mg/kg bolus 6 mg/kg bolus 0.8 mg/kg + 10 mg/kg/h 1 mg/kg + 15 mg/kg/h 3 mg/kg + 30 mg/kg/h 4 mg/kg + 40 mg/kg/h versus time, stratified for different
100 dosing regimens.
10
1
0.1

0 30 60 90120 0 30 60 90120 0 30 60 90120 0 30 60 90120 0 30 60 90120 0 30 60 90120

Time (min)

the final PK-BIS model, stratified for the 3 different dosing

regimens, is shown in Figure8. Plots of observed divided by
population-predicted concentrations over time are shown
in Figure9. MDPE and MDAPE are summarized in Table4.
Concentration (g/mL)

10
1 pated liver blood flow, suggesting esterases not only in the
0 10
Figure 4. Arterial (red) and venous (blue) concentrations for a typical
individual after a 30-minute infusion of 36 mg/kg/h.
Pharmacodynamics
The observations of BIS ranged from 4 to 98. A sigmoid
Emax model, where Emax was fixed to 1, i.e., a 100% decrease
in BIS from baseline, described the relationship between
AZD3043 concentrations and BIS. EC50 was 15.6 g/mL,
with an interindividual variability of 37% (range, 4.138.3
g/mL), and the Hill coefficient () was 1.7. No interindi-
vidual variability was estimated in . No significant effects
(P > 0.1) of body weight, age, or sex on EC50 were found.
The rate constant for the effect delay, ke0, was 0.64/min
(range, 0.04910.8), corresponding to a half-life of the effect
delay of 1.08 minutes (range, 0.06414.1). A 2-compart-
ment effect-site model,16 with a distribution compartment
off from the effect site, allowing for different onset and off-
set rates in relation to plasma concentrations, did not sig-
nificantly improve the fit (P > 0.1). Shrinkage was 15%
in all PD parameters. The parameter estimates of the final
PK-BIS model are presented in Table5. Key modeling steps
are listed in Table2. Observed and predicted BIS, stratified
by dose group, are presented in Figure7. The PC-VPC for
DISCUSSION
A recirculatory model, with 2 peripheral distribution com-
partments, a 5 tanks-in-series transit from venous to arte-
rial plasma, and 1 transit compartment from arterial to
venous plasma, adequately described the PK of AZD3043
in the healthy volunteers in the 2 studies. The elimination
of AZD3043 was fast, with an estimated plasma clearance
of AZD3043 of 2.2 L/min, which is >75% of the estimated
cardiac output (plasma flow) and greater than the antici-
20 30 40 50 60
liver but also in blood to be involved in the metabolism.
Time (min) This finding is also supported by in vitro data.3 However,
no significant relationship between esterase activity and
clearance was found in this analysis, which could have
been because of the limited range in esterase activity in
this healthy volunteer population, where normal esterase
activity was an inclusion criterion for participation in the
study. Disease conditions or genetic polymorphism influ-
encing the esterase levels or coadministration of esterase
inhibitors might still influence the clearance of AZD3043,
and caution should be taken when investigating such
patients or patients with unknown esterase activity. The
data did not support inclusion of sex or age on clearance,
but because the study was not designed to detect such dif-
ferences and the population was rather small and homog-
enous, no firm conclusions could be made regarding a
broader population. An allometric model was the best
descriptor of the relationship of body weight and clear-
ance parameters, whereas those for volumes were directly
proportional to weight alone. In this homogenous popula-
tion, the difference between allometrically scaled param-
eters and parameters directly proportional to body weight
is small, but the allometric model could be an advantage
when scaling to children or obese patients. However, fur-
ther studies in a wider range of body weights are needed
to draw any firm conclusions.

908
www.anesthesia-analgesia.org anesthesia & analgesia
Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 Table 2.Key Modeling Steps in the Pharmacokinetic and Pharmacodynamic Analysis
Description Compared with Additional degrees of freedom OFV*
Pharmacokinetic model
Fifth compartment in the central circulation 4 compartments 0 14.4
Sixth compartment in the central circulation 5 compartments 0 18.1
Second compartment in the nondistributive pathway 1 compartment 0 22.5
Concentration-dependent VP1 Constant VP1 1 0.4
Concentration-dependent VP2 Constant VP2 1 4.0
Dose-dependent VP1 Constant VP1 1 17.9
Dose-dependent VP2 Constant VP2 1 28.7
VP1 dependent on accumulated dose Constant VP1 1 44.7
VP2 dependent on accumulated dose Constant VP2 1 21.6
VA dependent on accumulated dose Constant VA 1 52.2
Dose-dependent CL Constant CL 1 0.3
Parameters proportional to body weight No relation with body weight 0 41.7
Parameters allometrically scaled to body weight No relation with body weight 0 48.7
Pharmacodynamic model
Sigmoid Emax model Emax model 1 696.4
CL = plasma clearance; Emax = maximum effect; OFV = objective function value; VA = volume of arterial compartment; VP1 = volume of peripheral compartment 1;
VP2 = volume of peripheral compartment 2.
*P 0.05 for OFV 3.84 at 1 degrees of freedom; P 0.01 for OFV 6.64 at 1 degrees of freedom; P 0.001 for OFV 10.83 at 1 degrees of freedom.

Table 3.Parameter Estimates of the Final Recirculatory PK Model for AZD3043

Parameter Estimate (95% CI) IIV (95% CI) Description
CL (L/min) 2.20 (2.122.25) 14 (1216) Plasma elimination clearance
VA (L) 3.07 (2.343.96) 52 (3565) Volume of arterial compartment
Dose-VA (%/mg) 0.146 (0.1200.204) Effect of accumulated dose on VA
VV (L) 1.77 (1.142.13) 117 (95146) Volume of venous compartment
VC (L) 1.03 (0.681.44) 75 (6197) Volume of central circulation
VP1 (L) 1.56 (0.991.79) 97 (88113) Intercept for volume of peripheral compartment 1
Dose-VP1 (%/mg) 0.0252 (0.001300.0380) Effect of accumulated dose on VP1
VP2 (L) 7.21 (6.778.15) 29 (2132) Volume of peripheral compartment 2
Dose-VP2 (%/mg) 0.0338 (0.02400.0480) Effect of accumulated dose on VP2
QCO (L/min) 2.90 (2.503.20) 8.5 (414) Cardiac output (plasma flow)
QP1 (L/min) 1.41 (1.081.62) 24 (1932) Intercompartmental clearance 1
QP2 (L/min) 0.829 (0.7530.963) 29 (2133) Intercompartmental clearance 2
QND (L/min) 0.0235 (0.02040.0268) Intercompartmental clearance for nondistributive pathway
A,prop (%) 16.2 (14.417.9) Proportional residual variability for arterial concentrations
A,add (g/mL) 0.00287 (0.0007770.00391) Additive residual variability for arterial concentrations
V,prop (%) 25.4 (23.027.3) Proportional residual variability for venous concentrations
V,add (g/mL) 0.00432 (0.002510.00636) Additive residual variability for venous concentrations
All volumes and clearance terms refer to plasma where applicable. 95% CIs are obtained from a nonparametric bootstrap with sample size 1000. Parameter
estimates are for a typical individual of median weight (77 kg).
CL = plasma clearance; CI = confidence interval; IIV = interindividual variability expressed as percent coefficient of variation (%CV); PK = pharmacokinetic;
QCO = cardiac output (plasma flow); QND = plasma flow of nondistributive pathway; QP1 = plasma flow through peripheral compartment 1; QP2 = plasma flow through
peripheral compartment 2; VA = volume of arterial compartment; VV = venous volume; VC = volume of central circulation; VP1 = volume of peripheral compartment 1;
VP2 = volume of peripheral compartment 2.

The results from this analysis were consistent with the
results from noncompartmental analyses performed in the
2 studies. A low but dose-dependent total apparent volume
of distribution, ranging from 15 L after the lowest dose
to 37 L after the greatest dose in a typical individual was
estimated. The reason for dose-dependent distribution is
unknown, but because AZD3043 is a lipophilic compound
administered as an emulsion, the distribution of the lipids
in the emulsion used in the drug formulation could poten-
tially influence the distribution of AZD3043. Similarly, infu-
sion of a lipid emulsion is used as treatment in situations of
overdoses of lipophilic local anesthetics19,20 where the local
anesthetics distribute to the infused lipids.
Litonius et al.21 showed that plasma concentrations
of bupivacaine decreased and volume of distribution
increased when patients were treated with an IV lipid
emulsion. However, it could be questioned whether such
a large increase in apparent volume of distribution could
have been because of lipids alone. Another hypothesis is
that cardiovascular effects could have influenced the dis-
tribution, because tachycardia was seen after greater doses.
It is also possible that blood flows to different tissues are
altered during anesthesia, which could possibly alter the
distribution.2224 The dose-dependent apparent volume of
distribution does not influence the area under the plasma
concentration-time curve or the steady-state concentrations,
because they are dependent on clearance only, but the half-
life is affected, meaning that the greater the dose, the lon-
ger time it takes for the concentrations to decrease and for
a subject to recover after anesthesia. This could potentially
be of importance if the apparent volume of distribution
continues to increase after infusions >30 minutes studied.
The maximum concentrations after a short infusion are also
influenced by the apparent volume of distribution, leading

October 2015 Volume 121 Number 4 www.anesthesia-analgesia.org 909

Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 Recirculatory Pharmacokinetics and Pharmacodynamics of AZD3043

Figure 5. PRED-corrected visual predictive check for the arterial (top row) and venous (bottom row) concentrations versus time based on the
final pharmacokinetic model. Solid line is the PRED-corrected observed median, and the dashed lines are the 5th and 95th percentiles of
the PRED-corrected observations. The blue- and red-shaded areas represent the 95% confidence interval for the median and 5th and 95th
percentiles, respectively, of the simulated data (n = 1000). Left, 30-minute infusions; middle, bolus; right, bolus + 30-minute infusions. As
data were PRED corrected, all doses are shown on the same scale, and the PRED-corrected observations are not the same as the uncorrected
observations. PRED = population predictions.

10 10 10
DV/PRED (arterial)

DV/PRED (arterial)

DV/PRED (arterial)

1 1 1

0.1 0.1 0.1

0 30 60 90 120 150 0 30 60 90 120 150 0 30 60 90 120 150

Time (min) Time (min) Time (min)

10 10 10
DV/PRED (venous)

DV/PRED (venous)

DV/PRED (venous)

1 1 1

0.1 0.1 0.1

0 30 60 90 120 150 0 30 60 90 120 150 0 30 60 90 120 150

Time (min) Time (min) Time (min)

Figure 6. Observed divided by population-predicted concentrations over time. Top row, arterial concentrations; bottom row, venous concentra-
tions; left, 30-minute infusion; middle, bolus; right, bolus + 30-minute infusion.

to less than proportional increases in maximum concentra-
tions with increased dose.
The short terminal half-life, because of the high clear-
ance and relatively low apparent volume of distribution,
provides potential for a rapid recovery after anesthesia. In a
study with similar sampling, propofol was found to have a
clearance of 1.6 L/min and a steady-state apparent volume
of distribution of 91 L,16 compared with 2.2 L/min and 15
to 37 L for AZD3043. The high ke0 and the limited or rapid
distribution within the brain, as suggested by the lack of
improvement of fit when applying a 2-compartment effect-
site model, are consistent with a rapid onset and offset of
anesthesia. However, because of increasing volume of dis-
tribution with greater doses, the terminal half-life could

910
www.anesthesia-analgesia.org anesthesia & analgesia
Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 potentially be prolonged after long infusions. More stud-
Table 4.Median Prediction Error and Median
Absolute Performance Error of Observed Data in ies are needed to investigate whether this could affect the
Comparison with the Final Pharmacokinetic and recovery after long-term dosing.
Pharmacodynamic Models The recirculatory model contains a large number of
MDPE (%) MDAPE (%) parameters to obtain a physiologically plausible model.
Pharmacokinetic model The model has a tendency toward overparametrization,
Arterial concentrations 3.9 (21 to 30) 20 (10 to 37) resulting in the minimization easily ending up in local
Venous concentrations 1.7 (26 to 32) 26 (12 to 53) minima or not converging successfully. This is seen in the
Pharmacodynamic model log-likelihood profiling (Fig. 10), where the profiles are
BIS 0.75 (6.9 to 3.2) 3.9 (2.8 to 8.1)
not always smooth parabolas. However, simplifications of
Data are represented by median (1090 percentiles). the model would have the drawback of losing physiologic
BIS = bispectral index; MDPE = median prediction error; MDAPE = median
absolute performance error. interpretation.

Table 5.Parameter Estimates of the Final Pharmacodynamic Model for AZD3043

Parameter Estimate (95% CI) IIV (95% CI) Description
Baseline 93.9 (93.394.4) 3.3 (2.73.9) Baseline BIS
EC50 (g/mL) 15.6 (14.718.7) 37 (2745) Effect-site concentration needed to reach 50% of Emax
1.7 (1.531.89) Shape factor
ke0 (per min) 0.640 (0.5530.834) 91 (67114) Rate constant for delay of effect
4.93 (4.665.20) Additive residual variability
95% CIs are obtained from a nonparametric bootstrap with a sample size of 1000.
BIS = bispectral index; CI = confidence interval; IIV = interindividual variability expressed as percent coefficient of variation (%CV).

1 mg/kg/h 3 mg/kg/h 6 mg/kg/h 12 mg/kg/h 18 mg/kg/h 27 mg/kg/h

100
80
60
40
20
0 36 mg/kg/h 54 mg/kg/h 81 mg/kg/h 1 mg/kg bolus 1.5 mg/kg bolus 2 mg/kg bolus
100
80
60
BIS

Figure 7. Individual observed (blue)

40 and typical predicted (red) BIS versus
20 time, stratified for different dosing
0 4 mg/kg bolus 6 mg/kg bolus 0.8 mg/kg + 10 mg/kg/h 1 mg/kg + 15 mg/kg/h 3 mg/kg + 30 mg/kg/h 4 mg/kg + 40 mg/kg/h regimens. BIS = bispectral index.
100
80
60
40
20
0
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60

Time (min)

Figure 8. PRED-corrected visual predictive check for the BIS model. Solid black line is the PRED-corrected median, and the dashed black lines
are the 5th and 95th percentiles of the PRED-corrected observations. The blue- and red-shaded areas represent the 95% confidence interval
for the median and 5th and 95th percentiles, respectively, of the simulated data (n = 1000). Left, 30-minute infusions; middle, bolus; right,
bolus + 30-minute infusions. As data were PRED corrected, all doses are shown on the same scale, and the PRED-corrected observations are
not the same as the uncorrected observations. BIS = bispectral index; PRED = population predictions.

October 2015 Volume 121 Number 4 www.anesthesia-analgesia.org 911

Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 Recirculatory Pharmacokinetics and Pharmacodynamics of AZD3043

40 40 40

DVPRED (BIS)

DVPRED (BIS)

DVPRED (BIS)
20 20 20

0 0 0

20 20 20

40 40 40
0 30 60 0 30 60 0 30 60

Time (min) Time (min) Time (min)

Figure 9. Observed dependent variable (DV) minus population-predicted (PRED) BIS over time. Left, 30-minute infusion; middle, bolus; right,
bolus + 30-minute infusion. BIS = bispectral index.

CL QP1 QP2 QCO QND

6 6 6 6 6
dOFV

dOFV

dOFV

dOFV

dOFV
4 4 4 4 4

2 2 2 2 2

0 0 0 0 0
2.102.152.202.252.30 1.1 1.2 1.3 1.4 1.5 0.75 0.85 0.95 2.62.72.82.93.03.1 0.020 0.024

Estimate Estimate Estimate Estimate Estimate

VA VV VP1 VP2 VC

6 6 6 6 6
dOFV

dOFV

dOFV

dOFV

dOFV
4 4 4 4 4

2 2 2 2 2

0 0 0 0 0
2.5 3.0 3.5 4.0 1.0 1.5 2.0 1.0 1.2 1.4 1.6 1.8 6.5 7.0 7.5 8.0 0.15 0.20 0.25

Estimate Estimate Estimate Estimate Estimate

DoseVA DoseVP1 DoseVP2

6 6 6
dOFV

dOFV

dOFV

4 4 4

2 2 2

0 0 0
0.00100.00140.0018 0e+00 3e04 6e04 1e04 4e04 7e04

Estimate Estimate Estimate

Figure 10. Log-likelihood profiles of the pharmacokinetic parameters. CL = elimination clearance; Dose-VA, Dose-VP1, and Dose-VP2 = effect
of accumulated dose on VA, VP1, and VP2, respectively; QCO = cardiac output (plasma flow); QND = plasma flow of nondistributive pathway;
QP1 = plasma flow through peripheral compartment 1; QP2 = plasma flow through peripheral compartment 2; VA = arterial volume; VV = venous
volume; VC = volume of central circulation; VP1 = volume of peripheral compartment 1; VP2 = volume of peripheral compartment 2.

The large range of ke0 was because of a few outlying sub- titration, which is commonly used in anesthesia practice. In
jects with very high or low values. The subjects with the 2 subjects in the greatest dose group, sudden and transient
lowest ke0 (longest equilibration time) received low doses increases in BIS were seen during the infusion, despite the
with a small effect on BIS, and it is possible that lying still subjects being deeply anesthetized as judged by the inves-
in a quiet room also could have influenced BIS in these sub- tigators. The reason for this is unknown, but could possi-
jects. was estimated to be 1.7, implying a steeper concen- bly be because of other electroencephalogram signals that
trationeffect relationship than for an Emax model. The is in distort the transformation into BIS. One subject receiving
line with what was found for propofol in a similar setting.16 the greatest dose had a large outlying decrease in BIS from
The PK and PD properties of AZD3043, with its fast onset approximately 15 to 30 minutes after start of infusion. The
and offset of effect, suggest it may be suitable for individual reason for this is unknown.

912
www.anesthesia-analgesia.org anesthesia & analgesia
Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
 In conclusion, a recirculatory model, with 2 peripheral
distribution compartments, a 5 tanks-in-series transit delay
from venous to arterial plasma, and a single compartment
transit delay from arterial to venous plasma, described the
PK of AZD3043. Clearance was high and volume of distri-
bution was low, resulting in a short elimination half-life.
However, the volume of distribution was increasing with
increasing dose. The distribution of the drug to the effect
site was rapid, and together with the short half-life, the
onset and offset of effect was fast. The effect on BIS was
described using a sigmoidal Emax model, with the delay in
the effect described by a single-effect compartment. E
DISCLOSURES
Name: Marcus A. Bjrnsson, MSc Pharm, PhD.
Contribution: This author helped design the studies, analyze
the data, and write the manuscript.
Attestation: Marcus A. Bjrnsson has seen the original study
data, reviewed the analysis of the data, and approved the final
manuscript.
Conflicts of Interest: Marcus A. Bjrnsson was employed by
AstraZeneca and has equity interest in AstraZeneca.
Name: ke Norberg, MD, PhD.
Contribution: This author helped design the studies, conduct
the studies, analyze the data, and write the manuscript.
Attestation: ke Norberg has seen the original study data,
reviewed the analysis of the data, and approved the final
manuscript.
Conflicts of Interest: ke Norberg received an honorarium
from AstraZeneca 3 years ago for participating in an advisory
board on AZD3043.
Name: Sigridur Kalman, MD, PhD.
Contribution: This author helped design the studies, conduct
the studies, analyze the data, and write the manuscript.
Attestation: Sigridur Kalman approved the final manuscript.
Conflicts of Interest: Sigridur Kalman received an honorarium
from AstraZeneca 3 years ago for participating in an advisory
board on AZD3043.
Name: Ulrika S. H. Simonsson, MSc Pharm, PhD.
Contribution: This author helped analyze the data and write
the manuscript.
Attestation: Ulrika S. H. Simonsson approved the final manuscript.
Conflicts of Interest: Ulrika S. H. Simonsson has equity interest
in AstraZeneca.
This manuscript was handled by: Steven L. Shafer, MD.
REFERENCES
1. Minto CF, Schnider TW. Contributions of PK/PD modeling to
intravenous anesthesia. Clin Pharmacol Ther 2008;84:2738
2. Sigl JC, Chamoun NG. An introduction to bispectral analysis for
the electroencephalogram. J Clin Monitoring 1994;10:392404
3. Egan TD, Obara S, Jenkins TE, Jaw-Tsai SS, Amagasu S, Cook
DR, Steffensen SC, Beattie DT. AZD-3043: a novel, metaboli-
cally labile sedative-hypnotic agent with rapid and predictable
emergence from hypnosis. Anesthesiology 2012;116:126777
4. Kalman S, Koch P, Ahln K, Kanes SJ, Barassin S, Bjrnsson MA,
Norberg A. First human study of the investigational sedative
and anesthetic drug AZD3043: a dose-escalation trial to assess
the safety, pharmacokinetics, and efficacy of a 30-minute infu-
sion in healthy male volunteers. Anesth Analg 2015;121:88593
October 2015 Volume 121 Number 4 www.anesthesia-analgesia.org 913
Copyright 2015 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
5. Norberg A, Koch P, Kanes SJ, Bjrnsson MA, Barassin S, Ahlen
K, Kalman S. A bolus and bolus followed by infusion study of
AZD3043, an investigational intravenous drug for sedation and
anesthesia: safety and pharmacodynamics in healthy male and
female volunteers. Anesth Analg 2015;121:894903
6. Kalow W, Genest K. A method for the detection of atypical
forms of human serum cholinesterase; determination of dibu-
caine numbers. Can J Biochem 1957;35:33946
7. Pantuck EJ. Plasma cholinesterase: gene and variations. Anesth
Analg 1993;77:3806
8. Buoen C, Bjerrum OJ, Thomsen MS. How first-time-in-human
studies are being performed: a survey of phase I dose-escala-
tion trials in healthy volunteers published between 1995 and
2004. J Clin Pharmacol 2005;45:112336
9. Chernik DA, Gillings D, Laine H, Hendler J, Silver JM,
Davidson AB, Schwam EM, Siegel JL. Validity and reliabil-
ity of the observers assessment of alertness/sedation scale:
study with intravenous midazolam. J Clin Psychopharmacol
1990;10:24451
10. Beal S, Sheiner LB, Boeckmann A, Bauer RJ, eds. NONMEM
Users Guides (19892009). Ellicott City: ICON Development
Solutions, 2009
11. Lindbom L, Pihlgren P, Jonsson EN. PsN-toolkita collection
of computer intensive statistical methods for non-linear mixed
effect modeling using NONMEM. Comput Methods Programs
Biomed 2005;79:24157
12. Jonsson EN, Karlsson MO. Xposean S-PLUS based popu-
lation pharmacokinetic/pharmacodynamic model build-
ing aid for NONMEM. Comput Methods Programs Biomed
1999;58:5164
13. Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Prediction-
corrected visual predictive checks for diagnosing nonlinear
mixed-effects models. AAPS J 2011;13:14351
14. Varvel JR, Donoho DL, Shafer SL. Measuring the predic-
tive performance of computer-controlled infusion pumps.
J Pharmacokinet Biopharm 1992;20:6394
15. Holford NH. A size standard for pharmacokinetics. Clin
Pharmacokinet 1996;30:32932
16. Zhang L, Beal SL, Sheiner LB. Simultaneous vs. sequential
analysis for population PK/PD data I: best-case performance.
J Pharmacokinet Pharmacodyn 2003;30:387404
17. Bjrnsson MA, Norberg A, Kalman S, Karlsson MO, Simonsson
US. A two-compartment effect site model describes the
bispectral index after different rates of propofol infusion.
J Pharmacokinet Pharmacodyn 2010;37:24355
18. Savic RM, Karlsson MO. Importance of shrinkage in empirical
Bayes estimates for diagnostics: problems and solutions. AAPS
J 2009;11:55869
19. Ciechanowicz S, Patil V. Lipid emulsion for local anesthetic sys-
temic toxicity. Anesthesiol Res Pract 2012;2012:131784
20. Weinberg GL. Lipid emulsion infusion: resuscitation for
local anesthetic and other drug overdose. Anesthesiology
2012;117:1807
21. Litonius E, Tarkkila P, Neuvonen PJ, Rosenberg PH. Effect of
intravenous lipid emulsion on bupivacaine plasma concentra-
tion in humans. Anaesthesia 2012;67:6005
22. Avram MJ, Krejcie TC, Niemann CU, Enders-Klein C,
Shanks CA, Henthorn TK. Isoflurane alters the recirculatory
pharmacokinetics of physiologic markers. Anesthesiology
2000;92:175768
23. Niemann CU, Henthorn TK, Krejcie TC, Shanks CA, Enders-
Klein C, Avram MJ. Indocyanine green kinetics characterize
blood volume and flow distribution and their alteration by pro-
pranolol. Clin Pharmacol Ther 2000;67:34250
24. Ludbrook GL, Visco E, Lam AM. Propofol: relation between
brain concentrations, electroencephalogram, middle
cerebral artery blood flow velocity, and cerebral oxygen
extraction during induction of anesthesia. Anesthesiology
2002;97:136370