Pediatric Anesthesia ISSN 1155-5645

EDUCATIONAL REVIEW

An introduction to physiologically-based pharmacokinetic

models
Richard N. Upton, David J.R. Foster & Ahmad Y. Abuhewla
Australian Centre for Pharmacometrics and Sansom Institute, School of Pharmacy and Medical Sciences, University of South Australia,
Adelaide, SA, Australia

Keywords
pharmacokinetics; physiology; regional
blood flow; anesthesia; pediatrics; computer
simulation
Correspondence
Prof. Richard N. Upton, Australian Centre for
Pharmacometrics, School of Pharmacy and
Medical Sciences, University of South
Australia, Adelaide, SA 5000, Australia
Email: richard.upton@unisa.edu.au
Section Editor: Brian Anderson
Accepted 24 July 2016
doi:10.1111/pan.12995
Summary
Physiologically-based pharmacokinetic (PBPK) models represent drug
kinetics in one or more ‘real’ organs (and hence require submodels of organs/
tissues) and they describe ‘whole-body’ kinetics by joining together submod-
els with drug transport by blood flow as dictated by anatomy. They attempt
to reproduce ‘measureable’ physiological and/or pharmacokinetic processes
rather than more abstract rate constants and volumes. PBPK models may be
built using a ‘bottom-up’ approach, where parameters are chosen from first
principles, literature, or in vitro data as opposed to a ‘top-down’ approach,
where all parameters are estimated from data. The basic principles of PBPK
models are described, focusing on the equations for three individual organs: a
single flow-limited compartment describing distribution only, a membrane-
limited compartment describing distribution, and a single flow-limited com-
partment with elimination. These organ models are linked to make a basic
three-compartment physiological model of the whole body. PBPK models are
particularly suited to scaling kinetics across body size (e.g., adult to neonate)
and species (e.g., animal to first-in-man) as physiology and pharmacology can
be represented by independent parameters. Maturation models can be incor-
porated as for compartmental models. PBPK models are now available in
commercial software packages, and are perhaps now more accessible than
ever. Alternatively, even complex PBPK models can be represented in generic
differential equation-solving software using the simple principles described
here. The relative ease of constructing the code for PBPK models belies the
most difficult aspect of their implementation—collecting, collating, and justi-
fying the data used to parameterize the model.

does a hyper- or hypodynamic circulation mean for drug

Introduction
Anesthesiologists, as ‘applied pharmacologists’, have
long had an interest in pharmacokinetic (and pharmaco-
dynamic) models. The majority of pharmacokinetic
models in the literature are empirical mammillary com-
partmental (‘compartmental’) models. While these have
been reliable servants in many fields of pharmacology,
many anesthesiologists find the connection with patient
physiology (arguably at its most variable in the pre-,
peri- and postoperative setting) is not well represented.
For example, an anesthesiologist may ask why do arte-
rial and venous blood concentrations have different
time-courses, and what does the difference mean? What
disposition and effect? How should the dose regimen be
modified? These questions are not readily answered
using compartmental models.
There has been a long-standing interest, particularly
by anesthesiologists, in physiologically-based pharma-
cokinetic (PBPK) models. PBPK models represent drug
kinetics in one or more ‘real’ organs (and hence require
a submodel of the organ/tissue). They describe ‘whole-
body’ kinetics by joining together organ submodels.
Drug transport is by blood flow as dictated by anatomy.
They attempt to reproduce real, measureable physiologi-
cal and/or pharmacokinetic processes rather than more
abstract rate constants and volumes. Model parameters

© 2016 John Wiley & Sons Ltd 1

Introduction to PBPK models
may be ‘intrinsic’ covariates (e.g., weight) rather than a
posthoc explanation that improves model fit. In this
review, the basic principles of PBPK models are exam-
ined. The aim is to help anesthesiologists interpret and
critique the PBPK anesthetic literature, and perhaps
provide a catalyst for their wider use in anesthetic
research.
Modeling strategies
Compartmental models are traditionally developed using
a ‘top-down’ approach where all of the information for
the model comes from a given pharmacokinetic and
covariate data set. In contrast, PBPK models can be built
using a ‘bottom-up’ approach, where the information for
the model comes from a priori physiological and pharma-
cological mechanisms and in vitro data. At its most
extreme, this approach could be used to predict the kinet-
ics of a new chemical entity based only on physiology and
physicochemical properties (1). A ‘middle-out’ approach
may also be possible (and sometimes desirable) that mixes
new data, known physiology and pharmacological princi-
ples, as necessary, to construct a model.
Full PBPK models have representations of the major
organs of the body. In contrast, Semi-PBPK or ‘Recir-
culatory’ models (2) may use pooling or lumping where
different regions or organs of the body are represented
as a single kinetic system. This may be useful where a
particular organ of the body is of particular interest
(e.g., the brain for anesthetics (3)) but kinetics in the
remainder of the body can be represented as one, two,
or three lumped compartments. The lumping of model
compartments need not be ad hoc, but can potentially be
done using formal approaches (4).
PBPK vs compartment models
Physiologically-based pharmacokinetic and compart-
mental models are both made from ‘compartments’, but
differ in the way the compartments are joined together.
A compartment is a region assumed to have uniform
distribution, such that a sample from the compartment
is representative of the rest of the compartment. The
analytical solutions of compartmental models become
intractable once more than three compartments are
used, and hence, models with more than three compart-
ments are rare. Population pharmacokinetic methods
(5) have allowed compartmental models to describe
complex clinical populations. However, the limitations
of compartmental models have been widely discussed:
they are not necessarily identifiable (more than one
model can describe the same data set equally well). What
do the compartments actually represent? Why are some
2 © 2016 John Wiley & Sons Ltd
R.N. Upton et al.
patients described by a one-compartment model and
some by a two-compartment model for the same drug?
How do physiological processes relate to the model?
Addressing these limitations has seen PBPK models
evolve alongside compartmental models.
History and applications of PBPK models
Some of the earliest PBPK models appear in the anes-
thetic literature, including the work of Eger (6) and
Mapleson (7) on volatile anesthetics. Price (8) used a
PBPK model to argue that recovery from thiopental
was due to redistribution rather than elimination.
Benowitz et al. (9,10) used a PBPK model to describe
the effect of hemorrhage on the disposition of lido-
caine in monkeys. Davis and Mapleson showed that
the kinetics of pethidine in organs could be estimated
from first principles (11). More recently, recirculatory
PBPK models have been used to describe the disposi-
tion of propofol (12) and fentanyl (13) in clinical pop-
ulations.
Commercially available physiologically-based
pharmacokinetic models
Commercially available PBPK models have increased
the accessibility of PBPK methods. Software packages
include GastroPlusTM (Simulation Plus, Inc., Lancaster,
CA), SimcypÒ simulator (Simcyp Ltd, Sheffield, United
Kingdom), and PK-SimÒ (Bayer Technology Services,
Leverkusen, Germany). For orally administered drugs,
commercially available PBPK models have been used
predict food effects (14), formulation effects (15), and to
identify causes of poor oral bioavailability (16). PBPK
models were also used to study the effect of organ dys-
function (17), and drug–drug interactions (18). Pediatric
modules are also available in these software using
known information on age-related changes in blood
flows, organ volumes, binding protein concentrations,
and drug metabolism capacity.
GastroPlusTM was the first commercially available
whole-body PBPK software initially based on the Com-
partmental Absorption and Transit (CAT) model (19),
which was further expanded by Simulation Plus
(www.simulations-plus.com) to the Advanced CAT
model (ACAT) (16). The ACAT model divides the GI
tract into nine compartments in sequence representing
stomach, duodenum, jejunum (two compartments),
ileum (three compartments), cecum, and ascending
colon. It accounts for drug dissolution, precipitation,
efflux or influx transporters, chemical stability, and first-
pass metabolism. Formulation and drug-specific input
data, obtained from in vitro, in vivo, or in silico estimates
R.N. Upton et al.
are fed into the dissolution and the absorption model.
Physiological parameters and characteristics for each
gastrointestinal (GI) compartment are built in the model
based on literature data or user input. The cumulative
absorption is estimated and then convoluted with the
disposition kinetics of the drug to predict plasma con-
centration-time profiles.
SimcypÒ was initially designed for mechanistic pre-
diction of drug metabolism then expanded to predict
pharmacokinetics including drug absorption (20). It
uses the Advanced Dissolution, Absorption, and Meta-
bolism (ADAM) model itself structured on the CAT
model; however, it uses a more sophisticated model to
describe dissolution (21). SimcypÒ incorporates a priori
estimates of interindividual variability (IIV) on physio-
logic parameters. For example, it considers a popula-
tion mean and IIV of GI pH and bile acid
concentrations in the fed and fasted states and models
the effects of these factors on the drug’s solubility and
dissolution rate (20).
The PK-SimÒ absorption model was based on the dis-
persion model (22) which considers the GI tract as a
one-dimensional cylindrical tube with spatially different
properties (e.g., pH, length, surface area) rather than a
series of transit compartments.
Getting pharmacokinetic data
Compartmental models are usually parameterized by fit-
ting plasma concentration data from a single site in the
body (e.g., arm venous blood). However, such a data set
is generally not sufficient to estimate all parameters of
PBPK models, and kinetic data for other sites in the
body are needed. Whether obtained by new experimen-
tation or from the literature, data sources could include:
Tissue concentrations
Direct measurement of drug concentrations in tissue
homogenates is possible (23) but this method is generally
only available for animals, and tissue samples are most
often taken postmortem. Therefore, only one time-point
is available per animal (but with the collection of many
tissues at that time). Defining the time-course of tissue
concentrations requires the sacrifice of a number of ani-
mals at different time points. Steady-state tissue:plasma
partition coefficients can be estimated and may be suffi-
cient if the mechanisms of tissue:plasma equilibrium are
well understood (e.g., perfusion limited). Tissue homoge-
nate data may not distinguish between intravascular,
extracellular, and intracellular locations of the drug.
Microdialysis provides an alternative method for estimat-
ing free extracellular concentrations in tissues (24,25),
© 2016 John Wiley & Sons Ltd
Introduction to PBPK models
but the time-course of tissue equilibration is required to
estimate the distribution volume of the tissue.
Regional venous concentrations
The kinetics of drugs in individual organs can also be
inferred from measurements of concurrent arterial and
regional venous plasma drug concentrations (i.e., the
drug concentration in blood entering and leaving the
organ). Cannulae can be placed acutely in man (26) or
chronically in animals (27). This method potentially
allows the collection of dense data with many time
points per subject and many sites per subject, even in
man (28).
In silico
Extensive experimentation has allowed the development
of in silico methods for estimating the steady-state tis-
sue:plasma partition coefficients of drugs based on their
physicochemical properties, plasma binding, and the
lipid and water content of the tissue in question (29,30).
These partition coefficients can be used in ‘flow-limited’
PBPK models.
Getting blood-flow data
The anatomical structure of PBPK models dictates that
drugs are transported to and from organs by blood flow;
hence the knowledge of blood flow values is integral to
PBPK models and can be obtained from a number of
sources.
Databases and publications
Documented values for blood flows (and organ size) first
came from the idea of a ‘Standard Man’ in radiation biol-
ogy (http://en.wikipedia.org/wiki/Standard_person) who
is 20–30 years of age, 70 kg and 170 cm tall. These values
are useful starting points, but do not account for the
between subject variability in body size and composition.
More usefully, Price et al. (31) extracted data from the
NHANES database and developed software that can
return the organ size and blood flows for ‘virtual’ popula-
tions (http://www.thelifelinegroup.org/p3m). Data for
animals (albeit not accounting for variability in size) are
available for rats, mice, dogs, monkeys, sheep, and pigs
(32–34).
Direct measurement
In animals, microspheres (35), flow probes (36), or tracer
kinetics (37) can be used to measure blood flow. In man,
3
Introduction to PBPK models
tracer kinetics, ultrasound methods (e.g., Transcranial
Doppler (26)) and imaging methods are possible (38). It
is not yet clear under what circumstances measuring
blood flow is warranted over using estimated or litera-
ture blood flow values. It is almost certainly important
for drugs or diseases associated with acute changes in
blood flow.
The basic maths
Consider the transport of drug in a blood vessel (e.g.,
the blood vessels entering or leaving an organ). The rate
(or flux) of drug transport rate (J) in the vessel is a func-
tion of the blood flow (Q) and drug concentration in the
blood (C):
J ¼ Q:C ð1Þ
Alternatively, if drug is added to the flowing blood at
a constant rate (e.g. as a constant rate infusion), the con-
centration achieved in the blood is proportional to the
infusion rate (J) and inversely proportional to the blood
flow:
J may be determined experimentally in vitro from tissue
C¼ ð2Þ
Q homogenates if a ‘bottom-up’ approach is employed, or
As an inverse relationship, low blood flows are associ-
ated with relatively higher blood concentrations for a
given infusion rate—a fundamental concept underlying
the relationship between pharmacokinetics and the cir-
culation.
Organs as a differential equations
Ultimately a PBPK model connects representations of
kinetics in organs in the body, but these can be consid-
ered in isolation.
Flow-limited distribution
The simplest model of an organ considers the case where
a drug is constantly infused into a blood vessel which
perfuses a single noneliminating organ, and where there
is no restriction to the distribution of the drug within
the organ (Figure 1). The amount of drug in the organ
at any time ‘t’ is equal to the integral of the difference
between the rate of entry and the rate of exit from the
organ:
Z t
amountðtÞ ¼ ðJin  Jout Þ ð3Þ
0 capillary wall may present a substantial barrier. In
Substitution of Equation 1 into Equation 3 then
allows for a description in terms of concentrations and
blood flows:
4 © 2016 John Wiley & Sons Ltd
R.N. Upton et al.
Z t
amountðtÞ ¼ ðQ:Cart  Q:Cven Þ ð4Þ
0
The rate of change in the amount (A) of drug in the
tissue can be obtained by differentiation with respect to
time:
dA
¼ Q:Cart  Q:Cven ð5Þ
dt
Which can then be further expressed as the rate of
change in the venous (effluent) concentration by divid-
ing both sides by the apparent volume of the tissue (V):
dCven Q:Cart  Q:Cven
¼ ð6Þ
dt V
The apparent volume of the tissue is, in turn, the pro-
duct of the real physical volume of the tissue (Vreal) and
the partition coefficient of the drug between blood and
tissue (represented by convention as Kp):
V ¼ Vreal :Kp ð7Þ
Thus, the real physical volume of the tissue may be
based upon known values, while the partition coefficient
as an estimated parameter when fitting to blood and tis-
sue concentration-time data. As can be seen in Equa-
tion 6, during a constant infusion of drug into the
arterial blood, the venous drug concentrations will
increase to a steady-state value equal to that in the arte-
rial blood. Importantly, the rate of change in concentra-
tion will be affected by both the apparent volume of the
organ and the blood flow, as illustrated in Figure 2.
Membrane-limited distribution
The next most complex situation for distribution only
occurs when the drug does not distribute within the tis-
sue instantaneously due to a diffusion barrier which
restricts the drug’s movement into a component of the
tissue (Figure 1). In this case, the tissue can be divided
into two (or more) ‘compartments’ with a permeability
term (represented by convention as PS) used to describe
the movement of drug across the barrier separating the
initial (V1) and the ‘deeper’ (V2) tissue spaces.
The diffusion barrier is typically some form of mem-
brane, although the nature of the ‘membrane’ will be
specific to the drug and tissue, as will any physical inter-
pretation of the two compartments. For example, the
which case the initial volume (V1) may represent the
vascular space of the tissue, while the ‘deeper’ tissue vol-
ume (Vtissue) may represent the total of extracellular and
R.N. Upton et al.
Figure 1 Simple models of individual organs. Upper: flow-limited
distribution of a drug into a single tissue perfused by a single blood
vessel. Middle: membrane-limited distribution in a single tissue per-
fused by a single blood vessel. Lower: flow-limited distribution with
elimination of a drug from a single tissue perfused by a single blood
vessel. Cart and Cven are the arterial and venous blood concentrations
of drug, respectively, Q is the blood flow through the tissue and V is
the apparent volume of the tissue, CL is the clearance of the drug by
the tissue. For the membrane-limited model, C2 is the deeper tissue
concentration of the drug, and PS is the permeability term.
intracellular spaces. Similarly, if the cell membrane pre-
sents a barrier to drug distribution, then V1 would
approach extracellular volume. There may be other bar-
riers to drug distribution such as the nuclear or other
unknown membranes. Assuming that the distribution
across the membrane is reversible, then a mathematical
description of this system is described below:
dCven Q:Cart  Q:Cven  PS:Cven þ PS:Ctissue
¼ ð8Þ
dt V1
dCtissue PS:Cven  PS:Ctissue
¼ ð9Þ
dt Vtissue
While the term ‘membrane-limited’ implies that only
permeability across the membrane is a limiting factor,
this is not necessarily the case and more generally it is a
mixture of both flow and membrane limitation
(Figure 3). The division of a tissue into more than one
© 2016 John Wiley & Sons Ltd
Introduction to PBPK models
Figure 2 Flow-limited distribution of a drug into a single tissue per-
fused by a single blood vessel. The impact of changes in blood flow
(upper panel) and apparent tissue volume (lower panel) on the con-
centration-time profile resulting from a 10 min 5 mgmin1 continu-
ous arterial infusion. Arterial concentrations are shown in red and
venous concentrations are shown in blue, with line type grouping the
data. In the upper panel, the apparent volume of the tissue remained
constant at 10 l while flow was changed. In the lower panel, flow
remained constant at 5 lmin1 while apparent tissue volume was
changed, and so arterial concentrations are identical for all three con-
ditions.
compartment based on distribution rate can be an
empirical decision equivalent to the way that compart-
mental model compartments are identified, or based on
an understanding of physicochemical properties (e.g.,
lipid solubility).
Flow-limited with elimination
In the case where an organ has a capacity to eliminate
the drug, this can be accommodated by incorporating a
clearance (CL) term. For simplicity consider a simple
flow-limited tissue with clearance added (Figure 1). For
this model CL should be linked to the arterial concen-
trations (39) and cannot exceed flow (Q) for obvious
reasons.
5
Introduction to PBPK models R.N. Upton et al.

dCven Q:Cart  Q:Cven  CL:Cart

¼ ð10Þ
dt V
Alternatively, elimination can be represented as an
extraction ratio (E) which takes values ranging from 0
to 1. In this case, the clearance of the tissue is the pro-
duct of flow and extraction ratio:
CL ¼ Q:E ð11Þ
Extraction ratio can be calculated from the arterial
and venous concentrations and is a function of blood
flow and intrinsic clearance (CLint) of the organ:
Cart  Cven CLint
E¼ ¼ ð12Þ
Cart Q þ CLint
Extraction ratio, rather than CL, may be used in a
model and determined from perfused tissue experiments
if a ‘bottom-up’ approach is employed, or as an esti-
mated parameter when fitting to blood and tissue con-
centration-time data. Importantly changes in CL do not
affect the rate of equilibration of the tissue. However,
during a constant infusion of drug into the arterial
blood, the venous drug concentrations of the eliminating
organ will increase to a steady-state equal to (1-E) of
those in the arterial blood (Figure 4).

Figure 3 Membrane-limited distribution of a drug into a single tissue

perfused by a single blood vessel. The impact of changes in perme-
ability across a membrane barrier on the concentration-time profile
resulting from a 10-min 5 mgmin1 continuous arterial infusion.
Figure 4 Flow-limited distribution with elimination of a drug in a sin-
Arterial concentrations are shown in red, venous concentrations are
gle tissue perfused by a single blood vessel following an intravenous
shown in blue, and tissue concentrations are shown in green, with
infusion. The impact of changes in clearance on the concentration-
line type grouping the data. In all panels, the apparent initial volume,
time profile resulting from a 10-min 5 mgmin1 continuous arterial
deeper tissue volume and flow remained constant at 10 l, 50 l and
infusion in a simple flow-limited tissue. Arterial concentrations are
5 lmin1, respectively, while the permeability was changed from
shown in red and venous concentrations are shown in blue, with line
100 to 10 to 1 lmin1. In the lower panel, flow remained constant at
type grouping the data. The apparent volume of the tissue and blood
5 lmin1 while apparent tissue volume was changed, and so arterial
flow (Q) remained constant at 10 l and 5 lmin1, respectively, while
concentrations were identical for all three conditions. Note the mark-
clearance (CL) was changed. The extraction ratio (E) is equal to CL/Q.
edly slower equilibration of the tissue concentrations as permeability
Note that the venous concentrations approach a steady-state equal
decreases.
to (1-E) multiplied by the arterial concentration.

6 © 2016 John Wiley & Sons Ltd

R.N. Upton et al. Introduction to PBPK models

More complicated organ models

The three organ models described above are the simplest
conceivable, and could describe the kinetics of many
drugs. More complex models such as the well-stirred
hepatic model (40) may be needed for orally adminis-
tered drugs, where first-pass effects are important and
there is a more complex interplay between protein bind-
ing, elimination capacity, and blood flow (41). The well-
stirred model also allows useful descriptions of drug
interactions (42).

Joining organs together

The complexity of the whole-body model will be dic-
tated by the availability of organ data in a top-down
approach, the organs of particular interest or impor-
tance in a bottom-up approach, or a combination of
both in a middle-out approach. It is also common to
‘pool’ or ‘lump’ organs which, although anatomically
distinct, may be considered as a single kinetic unit.
For example, lumping all muscle tissue, bone, fat, vis-
cera, or skin is often employed, and quite often these
may also be lumped together and termed the ‘body’ or
Figure 5 A simple whole-body physiologically-based pharmacoki-
‘carcass’. The appropriateness and nature of such netic model with hepatic elimination. Cart and Cven are the arterial and
lumping will be drug-dependent and should be guided venous blood concentrations of the drug, respectively, QCO, Qhep are
by careful consideration of the known physicochemical cardiac output and hepatic blood flow, respectively. Qbody is equal to
and pharmacological properties of the drug. Most QCO minus Qhep. Vlung, Vhep, and Vbody are the apparent volume of
organs are connected in parallel with several notable the lung, liver, and body, respectively. CLhep is the clearance of the
exceptions. The lung is connected in series with all drug by the liver.

other organs, and is arguably a key organ to include

in any whole-body model, particularly for intravenous
administration where it can be an important first-pass
organ. Similarly, blood flow from the gut, spleen, and
pancreas are in parallel and lead to the liver via the
portal vein. Bjorkman et al. (23,43,44) provide an
excellent example for the opioid analgesics, fentanyl
and alfentanil. In order to construct a model of a
whole body, models describing the inputs and outputs
of the individual organs were anatomically connected
via blood flows, with blood flow through any lumped
organs equaling cardiac output minus the sum of all
blood flows elsewhere defined.
A simple whole-body model
A simple case of a whole-body, semi-PBPK model incor-
porates three compartments: the lungs, the liver as an
eliminating organ, and the rest of the body (Figure 5).
In this example all organs are assumed to display flow-
limited distribution with the drug administered as an
intravenous infusion. The equations that describe this
system are outlined below:
dClung Rateinf  QCO :Cart þ Qhep :Chep þ Qbody :Cbody
¼
dt Vlung
ð13Þ
dChep Qhep :Cart  Qhep :Chep  CLhep :Cart
¼ ð14Þ
dt Vhep
dCbody Qbody :Cart  Qbody :Cbody
¼ ð15Þ
dt Vbody
Figure 6 illustrates the concentration-time course in
the arterial blood (Cart), body (Cbody), and liver tissue
(Chep) during a constant infusion of drug (Rateinf) into
the arterial blood at a rate of 100 mgmin1 for
200 min, with cardiac output (QCO) at varying values.
Checking the coding of the PBPK model is important.
For example, in Figure 6, the arterial concentrations
reach an eventual steady-state of 100 mgl1, corre-
sponding to the expected value calculated as dose rate/
clearance (100 mgmin1/1 lh1). Also note the slowing
of the rise in concentrations as cardiac output decreases

© 2016 John Wiley & Sons Ltd 7

Introduction to PBPK models
Figure 6 A simple whole-body physiologically-based pharmacoki-
netic model with hepatic elimination following an intravenous infu-
sion. The impact of changes in cardiac output (3, 6, and 9 lmin1) on
the concentration-time profile resulting from a 200 min
100 mgmin1 continuous infusion in a simple three-compartment,
whole-body model (see Figure 5). Arterial concentrations are shown
in red and concentrations in the body and liver are in blue and green,
respectively. The apparent volume of the body, liver, and lungs
remained constant at 25 l, 2 l, and 1 l, respectively, liver blood flow
and hepatic clearance also remained constant at 6 lmin1 and
1 lmin1, respectively.
with no gradient at steady state between the body and
arterial blood. In contrast, the hepatic extraction ratio,
illustrated by steady-state liver concentrations compared
8 © 2016 John Wiley & Sons Ltd
R.N. Upton et al.
to arterial concentrations, remained constant at 50%
regardless of cardiac output as neither hepatic clearance
nor liver blood flow were changed, which is in agreement
with the ratio of CLhep/Qhep; an alternative expression
of extraction ratio.
Adjusting for size and body composition
Physiologically-based pharmacokinetic models are often
structured so that the physiological parameters (e.g.,
organ size and blood flow) and pharmacological param-
eters (e.g., free fraction, extraction ratio, intrinsic clear-
ance) are distinct. This allows the physiological
parameters to be adjusted for differences in body size or
composition, both within and across species.
Within species
Allowing for differences in body size within a species
(e.g., with age) follows the standard principles of allome-
try. The general allometric equation is y = a + log(wt)
*b where b is the allometric coefficient. Organ size can
be scaled as a fixed fraction of body weight (b = 1),
while blood flow can be scaled with an allometric coeffi-
cient of 0.75 (45). The later value (implying higher rela-
tive perfusion (flow per g) in smaller subjects), reflects
the geometry of oxygen delivery to cells (46). The con-
cept of maturation models (45,47,48) accounting for the
early development of metabolic and renal capacity can
be implemented directly in PBPK models as per com-
partmental models.
Across species
The PBPK models are amenable to scaling kinetics
across species, where they have found a useful role in
predicting ‘first in man’ kinetics from animal data (49).
The task is to account not only for the difference in
body size between the species (via allometric principles)
but also account for the sometimes substantial differ-
ence in body composition between species (34). Allow-
ing for any differences in plasma protein binding is also
important (13).
Software
Any general-purpose software platform that can solve
systems of differential equations can implement a
bespoke PBPK model. Software can be classified by
whether they can be used for simulation only
[STELLAÒ(Cognitus Ltd., North Yorkshire, United
Kingdom), Berkeley MadonnaTM (www.berkeley-
madonna.com)], for fitting single subject data (R,
R.N. Upton et al. Introduction to PBPK models

MATLAB) or fitting population data (NONMEM, S-
Adapt, Monolix, WinBugs, Stan). Examples of fitted
population PBPK are uncommon as the size of the mod-
els make fit times intractable, but examples do appear in
the literature (13,41,50–53).
Pharmacodynamic models
The general principles for linking of pharmacodynamic
models to PBPK models are the same as those for mam-
millary PK models (54). For PBPK models, drug con-
centrations in organs or specific sites in the vasculature
(e.g., arterial or venous) can be linked to drug effect
rather than linking to the more ambiguously defined
central compartment concentration of compartmental
PK models. This can be advantageous from some classes
of drugs (such as anesthetics) where the site of drug
action is well understood (e.g., the CNS). Using drug
concentrations in the target organ to drive pharmacody-
namic effects may obviate the need for an empirical
effect compartment (13), with the advantage of an ‘in-
built’ representation of the factors (such as organ blood
flow) affecting target organ equilibration.
Conclusions
The PBPK models come in a variety of shapes and sizes.
Most are built using compartments and differential equa-
tions and as such a few basic building blocks such as
those outlined here can be used to build relatively large
and sophisticated models. It is prudent and time-efficient
to adapt the level of complexity of a PBPK model to the
requirements of the project at hand. Models can expand
or develop more complexity as more data or understand-
ing becomes available, or as it becomes necessary for the
model to address more complex questions. Despite the
growing acceptance and commercial support of PBPK,
we suspect the greatest potential of the PBPK approach
will be realized when there is increased acceptance of
adapting commonly used mammillary models with mod-
els with physiological features as needed. The relative
ease of constructing the code for PBPK models belies the
most difficult aspect of their implementation—collecting,
collating, and justifying the data used to parameterize the
model. Furthermore, for complex PBPK models the task
of validating that the model is ‘fit for purpose’ is not triv-
ial (49). Recent developments in data archiving and pub-
lishing (55) and the standardization of model
terminology (http://www.ddmore.eu/) will surely make
these tasks easier in the future.
How have PBPK models helped address some of the
questions posed at the beginning of the review? Why do
arterial and venous blood concentrations have different
time-courses? PBPK-based analysis has shown that
venous blood represents systemic kinetics ‘convolved’
with the kinetics in the tissue drained by the vein in
question. Maximum concentrations in venous blood
after intravenous bolus administration can be lower and
later than in arterial blood, and this has the potential to
confound PD analyses in anesthetic settings (13). What
does a hyper- or hypodynamic circulation mean for drug
disposition and effect? PBPK-based analysis has shown
that for intravenous anesthetics, peak concentrations
after intravenous bolus injection are inversely related to
cardiac output, and potentially dangerous high peak
concentrations are possible in the hypodynamic patient,
while efficacy may be reduced in the hyperdynamic
patient (27,56,57). While these analyses have only begun
to explore the potential uses of PBPK models in anes-
thetic research, we are very confident that the trend for
increasing use of PBPK models will allow valuable
insight into optimal drug use in pediatric anesthesia.
Ethical approval
Not applicable.
Funding
The study received no external funding.
Conflict of interest
The authors report no conflicts of interest.

References
1 Edginton AN. Knowledge-driven
approaches for the guidance of first-in-
children dosing. Pediatr Anesth 2011; 21:
206–213.
2 Krejcie TC, Avram MJ. Recirculatory phar-
macokinetic modeling: what goes around,
comes around. Anesth Analg 2012; 115:
223–226.
3 Upton RN, Ludbrook GL. A model of the
kinetics and dynamics of induction of anaes-
thesia in sheep: variable estimation for
thiopental and comparison with propofol.
Br J Anaesth 1999; 82: 890–899.
4 Gulati A, Isbister GK, Duffull SB. Scale
reduction of a systems coagulation model
with an application to modeling
pharmacokinetic-pharmacodynamic data.
CPT Pharmacomet Syst Pharmacol 2014; 3:
e90.
5 Mould DR, Upton RN. Basic concepts in
population modeling, simulation, and
model-based drug development. CPT
Pharmacomet Syst Pharmacol 2012; 1:
e6.

© 2016 John Wiley & Sons Ltd 9

Introduction to PBPK models
6 Eger EII. A mathematical model of uptake
and distribution. In: Papper EM, Kitz RJ,
eds. Uptake and Distribution of Anesthetic
Agents. New York: McGraw-Hill, 1963: 72–
87.
7 Mapleson WW. Uptake and distribution of
anesthetic agents. In: Papper E, Kitz RJ, eds.
Uptake and Distribution of Anesthetic
Agents. New York: McGraw-Hill, 1963:
104–119.
8 Price HL. A dynamic concept of the distribu-
tion of thiopental in the human body. Anes-
thesiology 1960; 21: 40–45.
9 Benowitz N, Forsyth FP, Melmon KL et al.
Lidocaine disposition kinetics in monkey
and man. I. Prediction by a perfusion model.
Clin Pharmacol Ther 1974; 16: 87–98.
10 Benowitz N, Forsyth RP, Melmon KL et al.
Lidocaine disposition kinetics in monkey and
man. II. Effects of hemorrhage and sympath-
omimetic drug administration. Clin Pharma-
col Ther 1974; 16: 99–109.
11 Davis NR, Mapleson WW. Structure and
quantification of a physiological model of the
distribution of injected agents and inhaled
anaesthetics. Br J Anaesth 1981; 53: 399–405.
12 Masui K, Upton RN, Doufas AG et al. The
performance of compartmental and physio-
logically based recirculatory pharmacokinetic
models for propofol: a comparison using
bolus, continuous, and target-controlled
infusion data. Anesth Analg 2010; 111: 368–
379.
13 Upton RN, Foster DJR, Christrup LL et al.
A physiologically-based recirculatory meta-
model for nasal fentanyl in man. J Pharma-
cokinet Pharmacodyn 2012; 39: 561–576.
14 Parrott N, Lukacova V, Fraczkiewicz G
et al. Predicting pharmacokinetics of drugs
using physiologically based modeling–appli-
cation to food effects. AAPS J 2009; 11: 45–
53.
15 Lukacova V, Woltosz WS, Bolger MB. Pre-
diction of modified release pharmacokinetics
and pharmacodynamics from in vitro, imme-
diate release, and intravenous data. AAPS J
2009; 11: 323–334.
16 Agoram B, Woltosz WS, Bolger MB. Pre-
dicting the impact of physiological and bio-
chemical processes on oral drug
bioavailability. Adv Drug Deliv Rev 2001; 1
(50 Suppl 1): S41–S67.
17 Edginton AN, Willmann S. Physiology-based
simulations of a pathological condition: pre-
diction of pharmacokinetics in patients with
liver cirrhosis. Clin Pharmacokinet 2008; 47:
743–752.
18 Yeo KR, Jamei M, Rostami-Hodjegan A.
Predicting drug-drug interactions: applica-
tion of physiologically based pharmacoki-
netic models under a systems biology
approach. Expert Rev Clin Pharmacol 2013;
6: 143–157.
10
19 Yu LX, Lipka E, Crison JR et al. Transport
approaches to the biopharmaceutical design
of oral drug delivery systems: prediction of
intestinal absorption. Adv Drug Deliv Rev
1996; 19: 359–376.
20 Jamei M, Dickinson GL, Rostami-Hodjegan
A. A framework for assessing inter-indivi-
dual variability in pharmacokinetics using
virtual human populations and integrating
general knowledge of physical chemistry,
biology, anatomy, physiology and genetics: a
tale of ‘bottom-up’ vs ‘top-down’ recognition
of covariates. Drug Metab Pharmacokinet
2009; 24: 53–75.
21 Kostewicz ES, Aarons L, Bergstrand M et al.
PBPK models for the prediction of in vivo
performance of oral dosage forms. Eur J
Pharm Sci Off J Eur Fed Pharm Sci 2014; 16:
300–321.
22 Ni PF, Ho NFH, Fox JL et al. Theoretical
model studies of intestinal drug absorption
V. Non-steady-state fluid flow and absorp-
tion. Int J Pharm 1980; 5: 33–47.
23 Bj€orkman S, Stanski DR, Verotta D et al.
Comparative tissue concentration profiles of
fentanyl and alfentanil in humans predicted
from tissue/blood partition data obtained in
rats. Anesthesiology 1990; 72: 865–873.
24 Bouw R, Ederoth P, Lundberg J et al.
Increased blood-brain barrier permeability of
morphine in a patient with severe brain
lesions as determined by microdialysis. Acta
Anaesthesiol Scand 2001; 45: 390–392.
25 Ederoth P, Tunblad K, Bouw R et al. Blood-
brain barrier transport of morphine in
patients with severe brain trauma. Br J Clin
Pharmacol 2004; 57: 427–435.
26 Ludbrook GL, Visco E, Lam AM. Propofol:
relation between brain concentrations, elec-
troencephalogram, middle cerebral artery
blood flow velocity, and cerebral oxygen
extraction during induction of anesthesia.
Anesthesiology 2002; 97: 1363–1370.
27 Upton RN, Ludbrook GL. A physiological
model of induction of anaesthesia with propo-
fol in sheep. 1. Structure and estimation of
variables. Br J Anaesth 1997; 79: 497–504.
28 Mather LE, Runciman WB, Ilsley AH et al.
Direct measurement of chlormethiazole
extraction by liver, lung and kidney in man.
Br J Clin Pharmacol 1981; 12: 319–325.
29 Small H, Gardner I, Jones HM et al. Mea-
surement of binding of basic drugs to acidic
phospholipids using surface plasmon reso-
nance and incorporation of the data into
mechanistic tissue composition equations to
predict steady-state volume of distribution.
Drug Metab Dispos Biol Fate Chem 2011; 39:
1789–1793.
30 Rodgers T, Rowland M. Mechanistic
approaches to volume of distribution predic-
tions: understanding the processes. Pharm
Res 2007; 24: 918–933.
R.N. Upton et al.
31 Price PS, Conolly RB, Chaisson CF et al.
Modeling interindividual variation in physio-
logical factors used in PBPK models of
humans. Crit Rev Toxicol 2003; 33: 469–503.
32 Brown RP, Delp MD, Lindstedt SL et al.
Physiological parameter values for physio-
logically based pharmacokinetic models.
Toxicol Ind Health 1997; 13: 407–484.
33 Clewell RA, Clewell HJ. Development and
specification of physiologically based phar-
macokinetic models for use in risk assess-
ment. Regul Toxicol Pharmacol RTP 2008;
50: 129–143.
34 Upton RN. Organ weights and blood flows
of sheep and pig for physiological pharma-
cokinetic modelling. J Pharmacol Toxicol
Methods 2008; 58: 198–205.
35 Robertson HT, Hlastala MP. Microsphere
maps of regional blood flow and regional
ventilation. J Appl Physiol Bethesda Md 1985
2007; 102: 1265–1272.
36 Upton R, Grant C, Ludbrook G. An ultra-
sonic Doppler venous outflow method for
the continuous measurement of cerebral
blood flow in conscious sheep. J Cereb Blood
Flow Metab Off J Int Soc Cereb Blood Flow
Metab 1994; 14: 680–688.
37 Doolette DJ, Upton RN, Grant C. Agree-
ment between ultrasonic Doppler venous
outflow and Kety and Schmidt estimates of
cerebral blood flow. Clin Exp Pharmacol
Physiol 1999; 26: 736–740.
38 Cosgrove D, Eckersley R, Blomley M et al.
Quantification of blood flow. Eur Radiol
2001; 11: 1338–1344.
39 Upton RN, Foster DJR. Representing elimi-
nating organs in physiological pharmacoki-
netic models. World Congress of
Pharmacometrics. In Seoul, Korea; 2012.
40 Baker M, Parton T. Kinetic determinants of
hepatic clearance: plasma protein binding
and hepatic uptake. Xenobiotica Fate Foreign
Compd Biol Syst 2007; 37: 1110–1134.
41 Hopkins AM, Wiese MD, Proudman SM
et al. Semiphysiologically based pharmacoki-
netic model of leflunomide disposition in
rheumatoid arthritis patients. CPT Pharma-
comet Syst Pharmacol 2015; 4: 362–371.
42 Zhang D, Zhu M, Humphreys WG. Drug
Metabolism in Drug Design and Develop-
ment. Hoboken, NJ, USA: John Wiley &
Sons; 2007. 634 p.
43 Bj€
orkman S, Stanski DR, Harashima H et al.
Tissue distribution of fentanyl and alfentanil
in the rat cannot be described by a blood flow
limited model. J Pharmacokinet Biopharm
1993; 21: 255–279.
44 Bj€
orkman S, Wada DR, Stanski DR. Appli-
cation of physiologic models to predict the
influence of changes in body composition
and blood flows on the pharmacokinetics of
fentanyl and alfentanil in patients. Anesthesi-
ology 1998; 88: 657–667.
© 2016 John Wiley & Sons Ltd
R.N. Upton et al.
45 Anderson BJ, Holford NHG. Tips and traps
analyzing pediatric PK data. Pediatr Anesth
2011; 21: 222–237.
46 West GB, Brown JH. The origin of allomet-
ric scaling laws in biology from genomes to
ecosystems: towards a quantitative unifying
theory of biological structure and organiza-
tion. J Exp Biol 2005; 208: 1575–1592.
47 Anderson BJ, Holford NHG. Mechanism-
based concepts of size and maturity in phar-
macokinetics. Annu Rev Pharmacol Toxicol
2008; 48: 303–332.
48 Rhodin MM, Anderson BJ, Peters AM et al.
Human renal function maturation: a quantita-
tive description using weight and postmenstrual
age. Pediatr Nephrol Berl Ger 2009; 24: 67–76.
49 Rowland M, Lesko LJ, Rostami-Hodjegan
A. Physiologically based pharmacokinetics is
impacting drug development and regulatory
decision making. CPT Pharmacomet Syst
Pharmacol 2015; 4: 313–315.
© 2016 John Wiley & Sons Ltd
50 Langdon G, Gueorguieva I, Aarons L
et al. Linking preclinical and clinical
whole-body physiologically based pharma-
cokinetic models with prior distributions in
NONMEM. Eur J Clin Pharmacol 2007;
63: 485–498.
51 Gueorguieva I, Aarons L, Rowland M. Dia-
zepam pharamacokinetics from preclinical to
phase I using a Bayesian population physio-
logically based pharmacokinetic model with
informative prior distributions in WinBUGS.
J Pharmacokinet Pharmacodyn 2006; 33:
571–594.
52 Wendling T, Dumitras S, Ogungbenro K
et al. Application of a Bayesian approach to
physiological modelling of mavoglurant pop-
ulation pharmacokinetics. J Pharmacokinet
Pharmacodyn 2015; 42: 639–657.
53 Wendling T, Tsamandouras N, Dumitras
S et al. Reduction of a whole-body
physiologically based pharmacokinetic
Introduction to PBPK models
model to stabilise the Bayesian analysis
of clinical data. AAPS J 2016; 18:
196–209.
54 Upton RN, Mould DR. Basic concepts in
population modeling, simulation, and model-
based drug development: part 3-introduction
to pharmacodynamic modeling methods.
CPT Pharmacomet Syst Pharmacol 2014; 3:
e88.
55 Kratz J, Strasser C. Data publication consen-
sus and controversies. F1000Research 2014;
3: 94.
56 Ludbrook GL, Upton RN. A physiological
model of induction of anaesthesia with
propofol in sheep. 2. Model analysis and
implications for dose requirements. Br J
Anaesth 1997; 79: 505–513.
57 Upton RN, Ludbrook G. A physiologically
based, recirculatory model of the kinetics
and dynamics of propofol in man. Anesthesi-
ology 2005; 103: 344–352.
11