REVIEW

published: 03 July 2020

doi: 10.3389/fphar.2020.00997

Application of Pharmacokinetic-
Pharmacodynamic Modeling in Drug
Delivery: Development and
Challenges
Huixi Zou , Parikshit Banerjee , Sharon Shui Yee Leung and Xiaoyu Yan *

School of Pharmacy, Faculty of Medicine, The Chinese University of Hong Kong, Hong Kong, Hong Kong

With the advancement of technology, drug delivery systems and molecules with more
complex architecture are developed. As a result, the drug absorption and disposition
processes after administration of these drug delivery systems and engineered molecules
become exceedingly complex. As the pharmacokinetic and pharmacodynamic (PK-PD)
modeling allows for the separation of the drug-, carrier- and pharmacological system-
specific parameters, it has been widely used to improve understanding of the in vivo
behavior of these complex delivery systems and help their development. In this review, we
Edited by:
Defang Ouyang,
summarized the basic PK-PD modeling theory in drug delivery and demonstrated how it
University of Macau, China had been applied to help the development of new delivery systems and modified large
Reviewed by: molecules. The linkage between PK and PD was highlighted. In particular, we exemplified
Han Qiao, the application of PK-PD modeling in the development of extended-release formulations,
Shanghai Jiao Tong University, China
Xiaoqiang Xiang, liposomal drugs, modified proteins, and antibody-drug conjugates. Furthermore, the
Fudan University, China model-based simulation using primary PD models for direct and indirect PD responses
*Correspondence: was conducted to explain the assertion of hypothetical minimal effective concentration or
Xiaoyu Yan
xiaoyuyan@[Link]
threshold in the exposure-response relationship of many drugs and its misconception.
The limitations and challenges of the mechanism-based PK-PD model were
Specialty section: also discussed.
This article was submitted to
Translational Pharmacology, Keywords: drug delivery, modified protein, pharmacokinetic modeling, pharmacodynamic modeling, mechanism-
a section of the journal based PK-PD modeling, the minimal effective concentration
Frontiers in Pharmacology

Received: 15 March 2020

Accepted: 19 June 2020
Published: 03 July 2020
INTRODUCTION
Citation: The research field of drug delivery focuses on the development of new techniques to manipulate
Zou H, Banerjee P, Leung SSY and drug absorption and disposition to achieve a desirable effect (Anselmo and Mitragotri, 2014; Asiri
Yan X (2020) Application of
and Mohammad, 2018). With the advancement of technology, drug delivery systems with more
Pharmacokinetic-Pharmacodynamic
Modeling in Drug Delivery:
complex architectures are developed. As a result, the drug absorption and disposition processes after
Development and Challenges. administration of these drug delivery systems become exceedingly complex. The lack of understanding
Front. Pharmacol. 11:997. of the in vivo behavior of these delivery systems may limit their successful translation into clinics.
doi: 10.3389/fphar.2020.00997 Mechanism-based pharmacokinetic-pharmacodynamic (PK-PD) modeling could be used to untangle

Frontiers in Pharmacology | [Link] 1 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

these complexities and improve the understanding of the in vivo predict the dosing regimen in humans. Once the clinical PK-PD
behavior of these drug delivery systems, consequently informing data is available, they can be incorporated into the PK-PD
their preclinical-to-clinical translation and clinical development. models to further optimize their design. The PK-PD modeling
PK-PD modeling, an indispensable component of drug can also evolve together with the clinical development to support
discovery and development, is a mathematical approach to the final approval.
study pharmacokinetics (PK), pharmacodynamics (PD), and Currently, modeling technique is commonly applied in the
their relationship (Peck et al., 1992; Danhof et al., 2005). As drug delivery system and modified large molecules. In the classic
Figure 1 shows, the mechanism-based PK-PD model can be drug delivery system, modeling has been widely utilized in aiding
incorporated into multiple stages in drug development. the formulation design based on preclinical studies, such as
Explicitly, PK modeling quantitatively describes the process of liposome, nanoparticle, and nanoemulsion (Soininen et al., 2016;
absorption and disposition of drug in the body. PD modeling Benchimol et al., 2019; Kadakia et al., 2019). As for the modification
evaluates the time course of the pharmacological effects of drugs, of large molecules related to drug delivery, such as PEGylated
with the consideration of the mechanism of drug action and protein, Fc-modified mAbs and antibody-drug conjugate (ADC),
major rate-limiting steps in the biology of the system (Mager et al., modeling technique has been widely used in both preclinical studies
2003). The PK and PD modeling can quantify the relationship of and clinical trials, providing valuable information for the animal-to-
drug exposure and response, and further characterize the influences human translation and dose regimen selection in clinical trials
of drug-specific, delivery system-specific, physiological and (Mager et al., 2005; Zheng et al., 2011; Krzyzanski et al., 2013; Ait-
pathological system-specific parameters on this relationship Oudhia et al., 2017; McSweeney et al., 2018). There are also many
(Agoram et al., 2007; Danhof et al., 2007). Drug-specific review papers and book chapters on the recent advancement of
parameters (e.g., drug clearance and receptor binding affinity) modeling in drug delivery, while those publications focused more
illustrate the interaction between the drug and the biological on pharmacokinetics (Yamashita and Hashida, 2013; Ait-Oudhia
system. The drug delivery system-specific parameters represent et al., 2014; Diao and Meibohm, 2015; Singh et al., 2015; Hedrich
the properties of carriers, such as the clearance, release rate, and the et al., 2018; Rodallec et al., 2018; Singh and Shah, 2018; Glassman
internalization rate of the carrier. The physiological system-specific and Muzykantov, 2019; He et al., 2019; Park, 2019). On the contrary,
parameters represent physiological values such as blood flow, life- in this review, the linkage between PK and PD is highlighted. We
span of cells, expression of enzymes, and transporters (Danhof et al., introduce the basic theory of PK-PD modeling and its application
2005; Danhof et al., 2007; Sager et al., 2015). relevant to drug delivery. The theory of PK focuses on the modeling
Through the separation of drug-specific and system-specific of absorption and deconvolution, which is a technique used to
parameters in PK-PD modeling, the influences of various identify an appropriate model structure for describing complex
properties of the delivery system on the in vivo drug effect absorption. We further discuss the basic PD theory that links drug
would be evaluated and facilitate its development. As shown in concentration and therapeutic effect. A few case studies, including
the bottom panel of Figure 1, the mechanism-based PK-PD the classic drug delivery system and modified large molecules, are
models, developed based on the PK-PD data from preclinical presented to exemplify the application of this theoretical framework
studies, can be used to optimize the drug delivery system and in practice. A model-based simulation is conducted to explain the

FIGURE 1 | Schematic of PK-PD modeling in the drug delivery system development. In the development of the drug delivery system, PK-PD modeling could guide
the formulation design and dosing regimen selection based on the preclinical and clinical data. This technique connects the drug dose to the physiological response,
related to the properties of the drug delivery system and physiological system. A chain of events illustrates the flow from the administration, drug exposure (plasma
and target site), receptor binding and activation, transduction to effect, and the effect on physiological response.

Frontiers in Pharmacology | [Link] 2 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

assertion and misconception of a hypothetical minimal effective Figure 2 shows a one-compartment model with first-order absorption
concentration or threshold, which has been used to guide the and first-order elimination. The model equations are as follows:
development of many drugs. Furthermore, the limitations and
challenges of the mechanism-based PK-PD model are discussed. dA1
= −ka
A1 (1)
dt


dA2 CL
BASIC THEORY = k a
A1 −
A2 (2)
dt V
Pharmacokinetic Modeling
PK modeling is critical to understand the time courses of drug A2
Cp = (3)
concentration following administration of various formulations and V
quantify the dose-concentration relationship. The method of where A1 denotes the mass of drug at the administration site, ka
compartmental modeling is commonly used to characterize PK is the absorption rate, A2 denotes the mass of drug in the body,
(Jones and Rowland-Yeo, 2013). After the drug enters the central CL represents the clearance, V represents the volume of
compartment (blood) via intravenous (IV) administration, distribution, and Cp denotes the plasma drug concentration.
distribution, and elimination occur. A one-compartment model is The initial conditions for Eqs. 1 and 2 are:
often used to describe the PK showing a monoexponential decline.
It assumes the entire body (including blood, organs, and tissues) A10 = Dose
F (4)
acts like a single, uniform compartment (Shargel et al., 1999). A
two- or three-compartment model describes the PK curve that A20 = 0 (5)
shows multi-exponential decay. Blood and well-perfused organs are
usually lumped together and considered as a central compartment, where Dose denotes the amount of drug administered, and F
while tissues with relatively slow but similar distribution rate are represents the bioavailability. By solving the above differentiation
grouped together as one or more peripheral compartments equations, the drug concentration, Cp, can be expressed as:
(Ahmed, 2015). For drug administered via extravascular routes,
F
ka
Dose

(e− V
t − e−ka
t )
CL
its absorption to the central compartment is usually described by a Cp = (6)
first-order or zero-order process. V(ka − CL V )
New drug delivery systems often significantly influence the
PK by modifying the in vivo drug release and absorption process. Zero-Order Absorption
By tuning the drug release profile, the apparent half-life of the Zero-order processes have also been used to describe the
drug could be prolonged, and drug accumulation at the target absorption after extravascular drug administration, where the
site may be enhanced (Shargel et al., 1999). Complex absorption drug is absorbed at a constant rate. The equations (Eqs.7 and 8)
can involve multiple pathways that occur simultaneously or for the one-compartment model with zero-order absorption and
sequentially, which could be challenging to model. Thus, linear elimination are:
modelers often use a numerical deconvolution technique to


recover the complex drug absorption profile from the PK data dA2 CL
= K0 −
A2 (7)
so that an appropriate model structure can be used accordingly dt V
(Bonate and Steimer, 2011). More complex absorption models
where K0 represents zero-order input. The drug concentration in
may consist of sequential and/or parallel combinations of the
blood can be expressed as:
simple ones. Here, the simple model and basic techniques used in
the absorption modeling are discussed, including first-order and K0  

1 − e− V
t
CL
zero-order absorption kinetics, flip-flop kinetics, and Cp = (8)
CL
deconvolution. While other PK techniques, such as the
population PK modeling and the physiologically based PK
(PBPK) modeling, are also commonly used in the modeling of Flip-Flop Kinetics
drug delivery system, basic concepts and principles of population A popular formulation approach is to extend the release of drug
PK and PBPK modeling are beyond the scope of this report. For from the delivery system to reduce the dosing frequency and
readers who are interested with those two topics, several papers
in the literature provided recent advancement and excellent
review of these fields (Moss and Siccardi, 2014; Lestner et al.,
2016; Li et al., 2017; Chetty et al., 2018; Nikanjam et al., 2018;
Siccardi et al., 2018).

First-Order Absorption
FIGURE 2 | One compartment model with first-order absorption and first-
The most common method to describe drug absorption after
order elimination.
extravascular administration assumes a first-order absorption.

Frontiers in Pharmacology | [Link] 3 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

improve patient compliance (Stege et al., 1996; Jadhav et al., 2006). convolution of absorption and disposition functions and
When the absorption process is much slower than the elimination expressed as:
process, the apparent half-life significantly increases due to the
slow absorption step, resulting in flip-flop kinetics (Davis, 2018). F(t) = In(t) ∗ D(t) (9)
For instance, in a one-compartment model with first-order where F(t) represents the function describing the drug concentration-
absorption and elimination (Figure 2), when absorption rate ka time profile, and In(t) and D(t) denote the input/absorption function
is much smaller than the elimination rate kel (derived by CL/V), and output/disposition function, respectively. The symbol “*”
resulting in the flip-flop phenomenon. A schematic of flip-flop represents the convolution operation. The disposition function D(t)
kinetics in Figure 3 shows the simulated PK profile of a drug upon can be obtained from the PK profile after IV administration. As long
IV and extravascular administration. With a rapid absorption as PK profiles after IV and extravascular administrations are known,
(ka > kel), the terminal slope of the concentration-time profile is the profile for absorption function In(t) can be derived by numerical
similar to that after IV administration route, reflecting the kel. deconvolution algorithms in software. Currently, deconvolution
However, when drug absorption is slower than the elimination algorithms are readily available in commercial software such as
(ka < kel), the absorption process becomes the rate-limiting step. Phoenix WinNonlin 8.1 (Pharsight Corporation, Cary, North
The downward concentration-time curve is less steep and reflects Carolina), Kinetica (Thermo Scientific), and GastroPlus
the ka, while the upward curve reflects the elimination process, kel. (Simulations Plus, Lancaster, CA) (da Silva Honó rio et al., 2013;
Flip-flop kinetics is commonly observed in the sustained- and Balakrishnan, 2014).
controlled-release formulations (Idkaidek et al., 1998; Stepensky
et al., 2001). However, the unawareness of flip-flop kinetics can
lead to the incorrect characterization of the absorption process.
Pharmacodynamic Modeling
PD models quantify the relationship between drug concentration
In particular, the terminal phase of the PK profile might be
and therapeutic effect. In this section, the PD models that capture
controlled by drug absorption instead of the elimination, which
the main mechanisms of drug action are presented, including
cannot be distinguished with only PK data after extravascular
direct effect, biophase, and indirect response models. Similar to
administration. Therefore, the IV data is indispensable to
PK, PD is usually described by the compartmental models, and
recognize the flip-flop phenomenon and to accurately estimate
complex PD models are created by combining the simple model.
the PK parameters associated with the drug absorption (Reed
et al., 2017).
Direct Effect Models
Deconvolution At the beginning of the research field of pharmacodynamics, it is
Deconvolution has been widely used in PK modeling of drug recognized that the pharmacological effect is linearly related to drug
absorption. It generates an input profile that can be used to guide concentration or logarithm of drug concentration (Levy, 1964). It
the selection of the model structure for absorption (Deslandes et al., was supported by the clinical data of tubocurarine (Levy, 1966). The
1992). By deconvolution, one can estimate the rate and extent of plasma drug concentration decreased exponentially after
absorption of various formulations via extravascular routes, such as intramuscular injection, and the degree of muscle relaxation
subcutaneous, oral, intranasal, rectal, and transdermal (Larsen et al., decreased linearly with time. However, the relationship between
1991; Fiset et al., 1995; Björkman et al., 1997; Duquesnoy et al., 1998). the effect and tubocurarine concentration is linear only if the effect is
Deconvolution is achieved by the inverse operation of either less than 20% (linear) or within 20 to 80% of the maximum
convolution, which is an approach to create a new function by (log-linear) effect (Emax) (Mager et al., 2003). Due to the limitation,
combining two mathematical functions (Yanez et al., 2011). For the nonlinear Emax model was introduced (Wagner, 1968). The
example, the time courses of drug concentration in plasma after rationale for the Emax model is based on the classic receptor
extravascular administration (Eq. 9) can be considered as a occupancy theory, and it assumes the drug effect (E) is directly
proportional to the fraction of occupied receptors:

E = g
RC (10)
where RC represents the concentration of drug-receptors complex,
and g is the proportional factor. The drug-receptor complex
equilibrium function is described as below:

R0
Cp
RC = (11)
KD + Cp
where R0 is the total receptor concentration in the tissue, and KD is
the dissociation constant for the drug-receptor complex. By
combining Eq. 10 to 11, the Emax model equation could be derived:
FIGURE 3 | Schematic of extravascular administration with flip-flop kinetics.
Simulated PK profile of the same drug to illustrate the effect of differences in the Emax
Cp
absorption rate. The slow absorption process results in a prolonged half-life. E= (12)
EC50 + Cp

Frontiers in Pharmacology | [Link] 4 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

where Emax is the maximum possible effect and equal to g∙RC, and mechanism of the delay, this phenomenon can often be explained
EC50 is a sensitivity parameter representing the drug concentration by either the biophase model or the indirect response model.
producing the half-maximal effect. The Emax model is frequently
used to describe the in vivo exposure-effect relationship of many Biophase Distribution Model
central system drugs and cardiovascular agents (Minematsu et al., The biophase model attributes the delay between the drug
2001; Friberg et al., 2005), where a rapid onset of drug effect is concentration and therapeutic effect to the time that it takes
induced. For instance, the Emax model has been used to describe the for the drug in the plasma to distribute to the target site (Sheiner
relationship between cocaine concentration and cardiovascular et al., 1979; Danhof et al., 2007). A biophase compartment was
effect, including systolic and diastolic blood pressures as well as proposed to represents the drug at the target site. In Figure 4, a
the heart rate (Laizure and Parker, 2009). biophase model has been utilized to explain the delayed effect in
In addition to the linear relationship between the effect and relation to the plasma drug concentration of d-tubocurarine
the drug-receptor complex, a nonlinear relationship has also (Sheiner et al., 1979), where the plasma concentration is linked to
been proposed as the operational model of agonism (Black and biophase concentration with the following differential equation:
Leff, 1983). The model is expressed as:
dCe
= keo
Cp − keo
Ce (17)
Emax
RC dt
E= (13)
KE + RC where Ce and Cp represent the concentrations in the biophase
where KE is the concentration of the drug-receptor complex that and plasma, respectively, and keo represents the first-order
triggers a half-maximum effect. By combining Eq.11 and Eq.13, distribution rate constant. As the delay in response due to the
the relationship between the drug effect and the concentration of distribution process, it could be affected by the physicochemical
agonist (A) can be derived as below: properties of the drug (e.g., molecule size), binding to plasma
protein, and transporter expression (Danhof et al., 2007).
Emax
t
A
E= (14) Indirect Response Models
KD + (t + 1)
A
Indirect response (IDR) models are widely used to describe the
where t represents the operational efficacy of agonist and is delayed response generated by the indirect mechanism.
defined by the total concentration of receptor (R0) divided by KE. Specifically, the drug could stimulate or inhibit either the
It should be noted that the concentration of drug achieving production or the dissipation of drug response, causing the
maximum effect is no longer EC50 as defined in Eq.12. As the delay. Jusko and his group formalized four basic indirect
concentration of drug (A) goes to infinity, the maximal effect is response models to describe diverse clinical pharmacodynamic
described by the asymptote parameter (a) of Eq.15: data (Dayneka et al., 1993; Jusko and Ko, 1994). As shown in
Emax
t Figure 5, Model I and Model II show the inhibitory effect on the
a= (15) input or loss of response, and Model III and Model IV show the
1+t
stimulatory effect. The models are expressed as:
The concentration of agonist producing the half-maximal
!
effect (A50) could be derived and shown in Eq. 16: dR Imax
Cp
= kin 1 − − kout
R Model I (18)
K dt IC50 + Cp
A50 = D (16)
1+t
t
In Eq.15, when t is large ( 1+t approaches to 1), a approaches to
Emax, and A50 is much smaller than KD, suggesting the drug is a full
t
agonist. However, when t is small ( 1+t approaches to 0), a is
smaller than Emax, and A50 approaches to KD, indicating a partial
agonism. The operational model of agonism suggested that the drug
triggering effect is a two-step process, where the first step is the
receptor binding process, and the second step is the signal-
transduction process. Therefore, it can simultaneously analyze
concentration-response data of compounds with different binding
affinities (KD) and efficacy (t), such as full and partial receptor
agonists (Danhof et al., 2007). It has been applied in m opioid (MOP)
receptor agonists with varying affinities of receptor and intrinsic
efficacies (Scott et al., 1991; Cox et al., 1999).
The models mentioned above (the Emax model and the
operational model of agonism) describe the direct effect of drugs,
when there is no time delay between plasma concentration and FIGURE 4 | Schematic of biophase model. Ce and Cp represent the
response. However, the lag between therapeutic effect and plasma concentration at the biophase and plasma, respectively. keo represents the
first-order distribution rate constant.
concentration is commonly observed. Depending on the

Frontiers in Pharmacology | [Link] 5 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

Application of PK-PD Modeling

In this section, we will discuss the application of PK-PD
modeling in the development of four different types of drug
delivery systems, including extended-release formulation,
modified protein, liposome, and antibody-drug conjugates.
Furthermore, a model-based simulation using a spectrum of
basic PD models was conducted to illustrate the hypothetical
FIGURE 5 | Indirect response model structure. kin and kout represent the minimum effective concentration (MEC).
first-order production constant and the first-order dissipation rate constant of
the response, respectively. The open and solid box represents the inhibitory
Complex PK-PD models, as shown in Figures 6–9, are often
and stimulatory effect, respectively. constructed by assembling basic PK and PD model components that
have been discussed in previous sections. Table 1 summarizes the
building blocks of PK and PD models that are used to characterize
!
Imax
Cp the activity of the compounds in the case studies. For example, the
dR
= kin − kout 1 −
R Model II (19) PK-PD model for rHuEPO in Figure 8 is a combination of the first-
dt IC50 + Cp
order absorption model, target-mediated drug disposition (TMDD)
model (Mager and Jusko, 2001a), operational model of agonism
!
dR Smax
Cp (Black and Leff, 1983), precursor-pool dependent IDR model
= kin 1 + − kout
R Model III (20) (Sharma et al., 1998), and transit compartment model for
dt SC50 + Cp
transduction process (Mager and Jusko, 2001b).
!
dR Smax
Cp Application to Paliperidone Palmitate
= kin − kout 1 +
R Model IV (21) The development of a 3-month extended-release (ER) formulation
dt SC50 + Cp
of paliperidone palmitate (PP3M), demonstrates how the PK-PD
where R represents the response; Imax and Smax are defined as the modeling can contribute to the development of new ER
maximal effect of inhabitation or stimulation, respectively; IC50 and formulation. Paliperidone (9-hydroxy-risperidone) is an
SC50 are the concentration trigger the half-maximal effect of antipsychotic agent for the acute and maintenance treatment of
inhibition or stimulation, respectively. The IDR model I has been schizophrenia (Samtani et al., 2009). PP3M is a long-acting
applied to describe the PK-PD relationship of Warfarin, a vitamin injectable formulation of paliperidone palmitate, which is the
K epoxide reductase inhibitor. Warfarin takes effect by inhibiting insoluble ester prodrug of paliperidone (Cleton et al., 2014; Gopal
the production of prothrombin and cause a delayed anticoagulant et al., 2015). After intramuscular injection, paliperidone palmitate
effect (Nagashima et al., 1969). A number of papers have reviewed particles slowly dissolve into the local fluids at the injection site. Due
the applications of IDR models (Sharma and Jusko, 1998; Mager to the poor aqueous solubility, the drug release from the
et al., 2003; Danhof et al., 2007). The IDR models can be extended formulation is a dissolution-driven process and becomes the rate-
to characterize more complex dynamics. A life-span based IDR
model has been developed for therapeutic growth factors that alter
the production of natural cells (Krzyzanski et al., 1999). A
precursor-dependent IDR model has been used to describe the
drug effect on the production of endogenous mediators from a
specific precursor (Sharma et al., 1998).
Furthermore, integration of IDR model I with the operational
model (Eq.22) has been applied in A1 adenosine receptor
agonists with different binding affinities, to describe the
relationship between their plasma concentrations and
antilipolytic effects (Van der Graaf et al., 1997; Van der Graaf
et al., 1999):


dR Emax
t
A
= kin
1 − − kout
R (22)
dt KA + (t + 1)
A

The operational model with IDR model I (Eq. 22) might be

preferred than the original IDR model I (Eq.18), because it FIGURE 6 | The PK-PD model of paliperidone palmitate. The absorption model
was divided into two components: f2 and (1- f2). f2 denotes the fraction that
assumes the nonlinear transduction between the drug-receptor
enters the central compartment via a zero-order process, and (1-f2) denotes the
complex and the effect of stimulation of kin. However, this model fraction that enters the central compartment via a first-order process. tlag is the lag
requires PK-PD data from both full and partial agonists to time, ka is the absorption rate, D2 is the duration of f2 for zero-process and equal
resolve the model parameters. Given such data are often not to tlag, CL represents clearance, RO represents receptor occupancy, ROmax
available, the original IDR model is more commonly applied to represents the maximal occupancy can be achieved, Cp represents the plasma
concentration, and KD is the equilibrium dissociation constant.
characterize the PK-PD relationship of a single compound.

Frontiers in Pharmacology | [Link] 6 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

FIGURE 7 | The PK-PD model of the free doxorubicin and liposomal doxorubicin. Free doxorubicin and liposomal doxorubicin are represented by white and grey
circles, respectively; the disposition of free doxorubicin and liposomal doxorubicin was described by a two-compartment and a one-compartment model,
respectively. The intra-tumor disposition is described by a physiological model linked with tumor blood flow rate (Q). Tumor tissue was divided into three
compartments: capillary (CAP), interstitial (INT), and tumor cell. The former two are considered in the extracellular compartment (ESC). kRES represents
reticuloendothelial system (RES) mediated elimination rate constant of liposomal doxorubicin; liposomal doxorubicin was unidirectionally transported from CAP into
INT (ktu); krel represents the first-order release rate constant of free doxorubicin from liposomes in blood, CAP and INT; distribution of free doxorubicin to tumor cells
was described using kte and ket; k21, k12, and k10 represent the micro-pharmacokinetic constants for free doxorubicin. For the pharmacodynamic model, a cell-kill
kinetic model was linked with the PK model. The mass balance equation describes the change rate of cell number, where Cs represents cell number, fb∙Cecs
represents the unbound drug concentration in ESC, ks is the cell proliferation rate constant, and k is the drug-induced irreversible cell-death rate constant.

FIGURE 8 | Mechanism-based PK-PD model of ESAs. The upper panel of the flow schematic is based on the TMDD model to describe the PK process. D and AD
represent the duration of the zero-order input and absorption compartment for the SC route, respectively. F is the bioavailability, and ka is the first-order absorption
rate constant. CA, CB, and RC represent the concentration of free ESAs, endogenous EPO (eEPO), and the EPO-receptor complex, respectively. ATA and ATB are
tissue compartments, ktpA, ktpB, kptA, and kptB are tissue distribution rate constants; konA, konB, koffA, and koffB represents second-order rate constants for forming RC
and first-order dissociation rate constants of ESAs and eEPO; Rtot represents total receptor concentration; CLA and CLB are the first-order elimination rate constants;
kint is the first-order internalization and degradation rate constant; kEPO represents the synthesis rate of eEPO. The PD model contains delay parameters, indicating
the life span of various erythropoietic cells. P1, P2, and P3 represent three different maturation-level erythroid precursor cell compartments; RET is reticulocytes, and
RBCM is mature red blood cell; HGB represents hemoglobin. KIN0 represents production process for P1 cells, Smax represents the maximal effect of RC on the
proliferation of precursor cells, and SRC50 is the concentration of drug-receptor complex to induce half of Smax; TP, TR, TB, mean residence time for precursor cell,
reticulocytes, and mature red blood cell. PD effect of the expansion of precursor cell on Rtot mediates the PK process in return, and x is the factor of proportionality.

Frontiers in Pharmacology | [Link] 7 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

FIGURE 9 | The multiscale PK-PD model of brentuximab-vedotin. The PK model is an integration of a modified two-compartment model in tumor to simultaneously
characterize the plasma and intracellular PK of ADC and the payload. After the IV administration of ADC, partial ADC dissociates and releases payloads. ADC and
the free drug in the central compartment can be eliminated, distribute to the peripheral compartment, or distribute to the tumor compartment. In the tumor tissue,
the free drug could enter the cell by passive diffusion and bind to the target or go back into the extracellular environment. ADC in the tumor environment interacts
with the antigen on the tumor cell membrane, and is internalized into the tumor cell. Subsequently, ADC is degraded in lysosomes to release the free drug
intracellularly. The intracellular concentration is considered as the concentration in the biophase compartment. X1ADC, X2ADC, X1PL, and X2PL, the amount of ADC or
payload in the central or peripheral compartment; V1ADC, V2ADC, V1PL, and V2PL, volume of distribution of ADC or payload in the central or peripheral compartment;
CLADC, plasma clearance of ADC; CLDADC, distribution clearance of ADC; ADCFree and ADCBound, free and bound ADC concentration in tumor tissue; Ag: total
antigen; PLFree and PLBound, free and bound payload concentration in cancer cell; DAR, average drug antibody ratio; kdis, dissociation rate of payload from antibody-
drug conjugate; kintAg, internalization rate of antigen inside the cell; konADC and koffADC, association and dissociation rate constants between antibody-drug conjugate
and antigen; konPL and koffPL, association and dissociation rate constants between drug and intracellular drug target; kinPL, drug nonspecific uptake rate in cancer cell;
koutPL, efflux rate of payload from the cancer cell; V1, V2, V3, and V4, tumor volume in the growth.

TABLE 1 | PK-PD model components and related mechanisms of action of drugs.

Drug PK model components PD model components Contribution of PK-PD modeling

Paliperidone One-compartment model with Dopamine D2 receptor occupancy Optimized formulation and dose regimen
palmitate First-order elimination regulation (Kapur et al., 2000) Accelerated the clinical trial
First-order absorption
Zero-order absorption
Flip-flop kinetics
Liposomal Two-compartment model for doxorubicin Biophase model (Sheiner et al., 1979) Evaluated the influence of drug-, carrier-, and
doxorubicin One-compartment model for liposome Cell-killing kinetic model (Jusko, 1973) system-specific parameter on anti-tumor efficacy
Flip-flop kinetics
Erythropoiesis Two-compartment model The operational model of agonism Quantified minimal-effect concentration
stimulating First-order absorption (Black and Leff, 1983) Explained the relationship between in vivo binding
agents Flip-flop kinetics Precursor-Pool dependent indirect affinity and effect
TMDD model (Mager and Jusko, 2001a) response model (Sharma et al., 1998)
Transit compartment model (Mager and
Jusko, 2001b)
Brentuximab- Two-compartment model with Emax model (Wagner, 1968) Predicted the clinical response by preclinical data
vedotin first-order elimination Biophase model (Sheiner et al., 1979)
Transit compartment model (Mager and
Jusko, 2001b)

limiting step of absorption. Ultimately, the slow absorption resulted affecting the release rate (Samtani et al., 2009). The PK-PD model of
in flip-flop kinetics and prolonged apparent half-life. paliperidone palmitate is shown in Figure 6. For the PK modeling in
The formulation of PP3M was developed from a 1-month PP1M, the deconvolution technique was utilized to identify the
formulation (PP1M) based on the knowledge gained from PK-PD appropriate absorption model. A fraction of the dose was initially
modeling, which identified the injection volume as a significant factor absorbed via a zero-order process, and the remaining fraction entered

Frontiers in Pharmacology | [Link] 8 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

the central compartment via a first-order process after a delay. Since medium release rate suggested by the model was a balance
paliperidone is a dopamine receptor D2 antagonist, the clinical between the drug elimination from blood and the drug
response is linked with the D2 receptor occupancy (Nordström accumulation in the tumor site.
et al., 1993; Kapur et al., 2000; Arakawa et al., 2008). The previous In this mechanism-based PK-PD model, the separation of the
study has indicated that over 70% receptor occupancy is needed for delivery-specific properties from the drug- and system-specific
the therapeutic effect (Arakawa et al., 2008). A PK model-based properties allows predicting the in vivo anti-tumor effect under
covariate analysis suggested the absorption rate (ka) of PP1M is different experimental conditions. One can predict the in vivo
negatively associated with injection volume. Thus, the PP3M outcome in various tumor models by changing the system-
formulation was developed with an increase in the injection specific parameter (e.g., the sensitivity of the tumor to the anti-
volume. Together with the benefit of an increased drug cancer drug). Similarly, by simulating PD associated with various
concentration in the suspension, the half-life of the new carrier-specific parameters (e.g., clearance of the carrier), one can
formulation is long enough to maintain an effective concentration compare the performance of different carriers. Hence, PK-PD
over three months (Samtani et al., 2009; Samtani et al., 2016). modeling and simulation can predict the drug effect in different
PK-PD modeling also accelerated the clinical development of disease models for carriers with varying PK properties.
PP3M. Based on the Phase I study of PP3M, a population PK model
was developed. Model-based simulations were conducted to find Application to Erythropoiesis-Stimulating Agents
the dose with a PK profile matching that of the PP1M formulation in Mechanism-based PK-PD modeling was applied to quantify the
the Phase III study (Samtani et al., 2016). The PK matching strategy MEC of erythropoiesis-stimulating agents (ESAs), and explain the
hypothesized that the drug effect was dependent on the drug paradox that ESAs with lower binding affinity has a higher in vivo
concentration in plasma above a targeting concentration activity (Yan and Krzyzanski, 2013). By binding to the erythropoietin
supported by the PP1M PK/PD studies. Eventually, based on the receptor (EPOR) on the membrane of erythroid precursor cells, ESAs
PK-PD study of PP1M and limited single-dose Phase I data of stimulate the proliferation and differentiation of erythroid precursor
PP3M, a prospective dose in Phase III was selected without cells (Elliott et al., 2008). Current ESAs include epoetin alfa,
conducting any Phase II dose-finding study. The Phase III study darbepoetin alfa, and continuous erythropoietin receptor activator
in the end achieved predicted PK and efficacy. The development of (CERA) (Locatelli and Del Vecchio, 2011). Epoetin, the first
the PP3M formulation is a successful case, demonstrating that PK- recombinant human erythropoietin (rHuEPO), has a half-life
PD modeling can significantly accelerate the clinical development between 5 and 12 hours and requires thrice-weekly dosing.
of a drug delivery system and reduce the cost. Darbepoetin, a hyperglycosylated analog of rHuEPO, has 3- to 4-
fold longer half-life than epoetin. However, the receptor binding
Application to Long-Circulating Liposomal affinity of darbepoetin is 4.3-times lower than that of epoetin, yet it
Doxorubicin has higher in vivo efficacy (Egrie and Browne, 2001). CERA was
The PK-PD model in Figure 7 described the disposition of developed by incorporating a 30 kDa methoxy polyethylene glycol
liposomal doxorubicin and free doxorubicin, and quantitatively polymer chain to rHuEPO (Macdougall, 2005). It has a longer half-
evaluated the relationship between drug exposure and anti-tumor life, but its receptor binding affinity is much lower (50–100 times)
effect by separating the carrier specific-, drug-specific- and system- than that of epoetin.
specific parameters (Harashima et al., 1999). The model was It was hypothesized that the effect of ESAs is not dependent on
developed based on reported and experimentally obtained the peak concentration but on the duration of drug concentration
preclinical data. The PK profile of liposomal doxorubicin and free above a ‘critical concentration’, also known as the MEC (Kiss et al.,
doxorubicin in the blood were described by one- and two- 2010). ESAs with a lower receptor-binding affinity are considered to
compartment models, respectively. Extracellular (ESC) have a higher MEC level to ensure sufficient receptor binding. Given
compartment at the tumor site is considered as the biophase MEC varies among various ESAs with different receptor binding
compartment, where the concentration has been linked with a affinity, the dosing of darbepoetin and CERA cannot be directly
cell-killing function. A sensitivity analysis suggested that the rates of derived by PK matching with their predecessor, epoetin. Without
liposome clearance (kRES) and drug release (krel) play critical roles in the quantitative definition of MEC, a large number of clinical trials
drug delivery. Lower kRES could mitigate the loss of liposomal had been conducted to optimize the dosing regimen of darbepoetin
doxorubicin in blood to maintain a longer blood circulation period (Glaspy et al., 2001; Glaspy et al., 2002; Vansteenkiste et al., 2002).
of liposomes. Subsequently, the accumulation of liposomes in the It was believed that establishing a relationship between the
tumor site (ESC) would increase, resulting in a better tumor-killing MEC and receptor binding affinity may help to find the optimal
effect. Furthermore, the simulation results suggested that a medium regimen for various ESAs. A mechanism-based PK-PD model
release rate (krel at 0.06 h−1) was optimal to achieve higher efficiency was, therefore, developed to quantify the MEC of various ESAs
in rodents. Compared with the rapid-release formulations, a slower based on the clinical PK-PD data of rHuEPO (Yan et al., 2012;
release rate could increase the free drug accumulation in the ESC Yan and Krzyzanski, 2013). The model structure is provided in
compartment, instead of being cleared in blood. However, if the Figure 8. This model incorporated the operational model of
release rate was too low (krel at 0.006 h−1), the formulation failed to agonism into the PK-PD modeling, which helped to dissect the
achieve the critical drug concentration to inhibit tumor influence of receptor binding affinity on the drug effect.
proliferation (Harashima et al., 1999). Therefore, the optimal Furthermore, the binding between ESAs and EPOR results in

Frontiers in Pharmacology | [Link] 9 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

receptor-mediated drug elimination (Yan and Krzyzanski, 2013). observed in clinical trials (Bartlett et al., 2009; Younes et al., 2010;
A model-based simulation was conducted to simulate the PK-PD Fanale et al., 2012). This example shows that the PK-PD model could
profile of epoetin and darbepoetin under a thricely weekly IV integrate the in vitro and in vivo preclinical information to predict the
bolus regimen. The PK profiles were overplayed with the C50 (Eq. ultimate clinical response.
23) and showed that C50 could be considered as the MEC: Furthermore, this ADC disposition model offers a conceptual
framework for the design of ADC and facilitates the preclinical-to-
KD clinical translation. The PK model described above has been applied
C50 = (23)
1+t in auristatin-based anti-5T4 antibody-drug conjugates (Shah et al.,
where C50 is the ESA concentration that triggers the half-maximal 2014). It was discovered that the stability of the payload on the ADC
effect, t is the efficacy parameter in the operational model of agonism, and tumor size are the two most influential factors to the payload
and KD is the dissociation equilibrium constant of ESAs. Consistent exposure in plasma. The linker controls the stability of payload and
with the MEC hypothesis, darbepoetin has with a lower binding its dissociation from the antibody. The occurrence of extracellular
affinity (higher KD), but a higher C50 value compared to epoetin. dissociation increases the systemic exposure of toxic payload and
However, the simulation also demonstrated that lower binding results in a severe adverse effect. Therefore, it offered a rationale to
affinity of darbepoetin leads to lower receptor-mediated drug modify the linker to increase the stability of the payload on ADC.
elimination and hence a slower clearance of the drug compared to Also, the sensitivity analysis suggested an increase in tumor size may
epoetin. Therefore, the duration of darbepoetin concentration lead to a rise in payload exposure in plasma. When tumors are small
maintained above C50 is longer than that of epoetin, leading to a and avascular, ADC could only diffuse from the tumor periphery to
higher in vivo activity (Yan and Krzyzanski, 2013). In other words, the the tumor. However, as a tumor grows and becomes vascularized,
higher concentration and prolonged duration above MEC will ADC can quickly enter the tumor by diffusion from the vasculature.
eventually compensate for the counteracting effects of lower Subsequently, the payload released from ADC inside of the tumor
receptor binding affinity, thereby increasing their in vivo activities. would also increase and diffuse to the systemic circulation. The
Model-based simulations further successfully predicted that CERA author suggested that the difference in tumor size between animals
had the highest in vivo potency among the three ESAs when and patients should be considered during the preclinical-to-
administered with the same molar dose at any of the approved clinical translation.
dosing regimens (i.e., thrice weekly, once weekly, and once every 2
weeks). The model also predicted that if the receptor binding affinity Model-Based Simulation to Illustrate the
was too low, the benefit (i.e., longer half-life) of lower receptor binding Hypothetical MEC
affinity dissipated and eventually led to a lower in vivo activity. During the development of delivery system, maintaining drug
concentration at a threshold concentration (e.g., IC50 and EC50) in
Application to Brentuximab-Vedotin blood or target site has often been used as a simple method to
PK-PD modeling is applied to improve the understanding of the evaluate the performance of diverse formulation (Mishra et al.,
complex PK-PD relationship of ADC. ADC consists of a 2011). For example, drug release from nanocarriers is often
monoclonal antibody (mAb), a cytotoxic payload, and a linker deliberately controlled to maintain the plasma drug level within
(Tsuchikama and An, 2018). After binding to the antigen expressed the therapeutic window between the MEC and the minimum toxic
on the surface of a tumor cell, ADC is internalized and transported concentration (Siegel and Rathbone, 2012). The presented case
to lysosomes. Once inside of lysosomes, the linker is cleaved to studies also implied the presence of MEC and the importance of
release the cytotoxic payload. It can enhance the selective delivery of maintaining drug concentration above a MEC.
the payload to the tumor cell and reduce the systemic toxicity. In this section, we use PK-PD modeling and simulation to
Brentuximab-vedotin is an anti-CD30 based ADC (Shah et al., illustrate the assertion and misconception of MEC. Simple direct
2012). A multiscale PK-PD model (Figure 9) was developed to response model (Eq. 12), indirect response model III (Eq. 20), and
simultaneously captures the disposition of brentuximab-vedotin indirect model III with the operational model of agonism (Eqs. 15
and its payload, at the cellular and physiological (plasma and tumor and 20) are used in PD simulations to accommodate major
tissue) level. The detailed description of the model is provided in the mechanisms of drug action. A one-compartment model with
legend of Figure 9. A cellular PK model was developed based on in first-order absorption and first-order elimination is used to
vitro experiments, to mimic the intracellular and extracellular simulate PK. The duration of drug concentration above a
(tumor) PK activities of the payload and the ADC. Then, based threshold was manipulated by changing the value of the
on the preclinical and clinical in vivo data, the PK model of ADC absorption rate (ka), mimicking the different release rates of
and payload in plasma and tumor was developed. In terms of the PD various drug delivery systems. EC50 and A50 are used as the
model, the Emax model was used to describe the nonlinear concentration threshold in the simulation. The relationship
relationship between the intracellular concentration and the between duration of drug concentration above a threshold, and
tumor-killing rate. The PD parameters were estimated based on the the response was investigated. All the simulations were performed
preclinical tumor growth inhibition data in mice. By integration of the using mrgsolve, an open-source R package (Elmokadem et al.,
multiscale PK-PD model and PK data of brentuximab-vedotin in 2019). Parameter values are provided in the legend of Figure 10.
patients, clinical responses were predicted. A retrospective Figure 10 shows the simulated PK, PD, and the area under the
comparison suggested that model-predicted progression-free effect versus time curve (AUEC) for each model with different
survival and complete response rates were comparable to those absorption rates (ka), which controls the duration of concentration

Frontiers in Pharmacology | [Link] 10 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

FIGURE 10 | Model-based simulations with different absorption rates. Drug concentration, response (effect), and area under response (AUEC) versus time profiles
have been simulated with different PD models: (A) direct response model, (B) indirect model, and (C) indirect response model with operational model of agonism.
The dose is 1000 mg. The absorption rate varies from 0.001 to 1 h−1. The red dashed line shows the EC50 values or the A50 values, which are equal to 1 mg/L.
Clearance (CL) is 27 L/h, the volume of distribution (V) is 90 L (kel = 0.3 h−1), concentration of receptor complex that triggers the half-maximum effect (KE) is 5 mg/L,
Emax is 1, response production rate constant (kin) is 0.1 h−1, response elimination rate constant (kout) is 0.02 h−1, equilibrium dissociate rate (KD) is 2, target
production (ksyn) are 1 h−1, and degradation (kdeg) rate are 0.2 h−1.

Frontiers in Pharmacology | [Link] 11 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

above EC50 or A50. After the last dose treatment on day 21, the order systems, such as the erythropoietic system, the method of scaling
of the AUEC under varying ka values are ka = 0.06 > ka = 0.2 > ka= from animal to human has already been established (Mager et al.,
0.006 > ka = 0.6 > ka=1 > ka = 0.001, which is consistent with the 2009). Thus, given the available prior information, it is relatively more
order of duration of drug concentration maintained above the acceptable to scale from animal to human. Fourth, PK-PD modeling
threshold (red dashed line) in the PK profiles. The medium value of based on data collected from tumor xenograft mice model has been
absorption ka = 0.06 triggers the highest effect in all three models. extensively used in oncology. This is supported by that the predicted
Compared with ka = 0.06 group, a larger ka value (ka = 0.2, 0.6, and 1) threshold concentration derived from xenograft experiments
results in faster absorption and subsequent elimination, and a correlates with the active dose in humans for several marketed
shorter duration of concentration above the threshold. On the chemotherapy drug (Rocchetti et al., 2007). Such a correlation can
other hand, when ka is smaller (ka = 0.006 and 0.001), the absorption also be observed in targeted therapies, including both small and large
is too slow to achieve the threshold concentration during the dosing molecules (Yamazaki, 2013; Lindauer et al., 2017). Therefore, the
period. These simulations demonstrate that the duration of drug effective concentration in preclinical species could serve as a
concentration above a threshold is the determining factor for the minimum target concentration that needs to be achieved in humans.
therapeutic effect. Therefore, many formulations have focused on It should be pointed out that modeling assumptions are usually
optimizing the drug release rate for maintaining drug concentration based on prior experience and knowledge and are not equal to
above MEC. random guessing (even they are often not be provided by modelers).
Although EC50 and A50 are used as the threshold in the present They contribute to the power of modeling. Having assumptions helps
simulation, it should be noted that the threshold is a hypothetical us to fill the missing pieces among existing information and simplify
concept and arbitrarily selected. The concentration-effect the understanding of a complex system. However, modelers should
relationship (as expressed as Emax model or the operational model) be mindful of and acknowledge explicitly the assumptions and
is a continuous function without such a threshold. It means limitations of the model, and develop models that are fit for their
maintaining concentration above a specific threshold may not be purpose (Zhang et al., 2008). Methods like sensitivity analysis and
appropriate for formulation selection, unless the concentration and external validation can check model’s dependency on assumptions,
response relationship is known, and the response associated with this and allows to identify alternative assumption and the information
threshold concentration is desired. Therefore, simulation based on that is needed to in future experiments. An optimal study design can
the PK-PD relationship is preferred than a single threshold to select be achieved by simulation and re-estimation, to ensure that the
the formulation and dosing regimens. information collected in future experiments is informative on the new
assumption (Suryawanshi et al., 2010). Once the new data become
available to validate the predicted outcome, the assumption can be
CHALLENGE AND STRATEGY further refined and updated. Such a process can be repeated until the
final goal of modeling is achieved.
Although PK-PD modeling has facilitated drug development,
several challenges are associated with its application. First,
assumptions that are used to simplify the drug delivery process in SUMMARY
PK-PD modeling might not always be appropriate. A reliable
In this review, we have presented the basic theory and techniques
assumption requires a detailed understanding of both the
of PK-PD modeling and highlighted its application in the
physiological system and a drug delivery system. Due to the lack
development of different drug delivery systems and modified
of relevant information, it’s challenging to validate these
large molecules. PK-PD modeling and simulation are used to
assumptions (Suryawanshi et al., 2010). For instance, the PK-PD
illustrate the misconception of the concentration threshold. The
model of ADC we discussed previously assumed the drug
development of new technology can improve the understanding
concentration in tumor cells was related to the cell-killing effect
of the physicochemical properties of the delivery system and
(Shah et al., 2012). However, without actual measurement in the
their interactions with the physiological system. Thus, the
distribution cascade, the validity of this assumption is uncertain.
predictive ability of PK-PD modeling can be enhanced to guide
Furthermore, extrapolation from animals to humans based on the
the development of new drug delivery systems.
PK-PD modeling is also challenging. For instance, in the case study of
liposomal doxorubicin, the author suggested that the optimal rate of
drug release in rodents was different from that in humans AUTHOR CONTRIBUTIONS
(Harashima et al., 1999). Although there is no universal solution,
some methods could be considered to minimize the error of scaling. HZ performed the model simulation and wrote the manuscript.
First, we should always select the appropriate animal species for XY supervised and contributed to the simulation and editing of
scalling to humans (Tibbitts et al., 2010). For instance, it is known that the review. SL and PB contributed to the revision of the review.
dog is the most relevant species to human for preclinical cardiacsafety
assessment (Gralinski, 2003; Dubois et al., 2017). Second, allometric
scaling using multiple species has been applied to minimize the error FUNDING
and increase the accuracy of prediction (Huh et al., 2011). This
method is commonly used to predict human PK (Amantana et al., This work was supported by the start‐up grant from School of
2013; Di Martino et al., 2019). Third, for some pharmacological Pharmacy, The Chinese University of Hong Kong.

Frontiers in Pharmacology | [Link] 12 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

REFERENCES classification of biomarkers. Pharm. Res. 22 (9), 1432–1437. doi: 10.1007/

s11095-005-5882-3
Agoram, B. M., Martin, S. W., and van der Graaf, P. H. (2007). The role of Danhof, M., de Jongh, J., De Lange, E. C., Della Pasqua, O., Ploeger, B. A., and Voskuyl,
mechanism-based pharmacokinetic-pharmacodynamic (PK-PD) modelling in R. A. (2007). Mechanism-based pharmacokinetic-pharmacodynamic modeling:
translational research of biologics. Drug Discovery Today 12 (23-24), 1018– biophase distribution, receptor theory, and dynamical systems analysis. Annu. Rev.
1024. doi: 10.1016/[Link].2007.10.002 Pharmacol. Toxicol. 47, 357–400. doi: 10.1146/[Link].47.
Ahmed, T. (2015). “Pharmacokinetics of drugs following IV Bolus, IV infusion, and 120505.105154
oral administration,” in Basic Pharmacokinetic Concepts and Some Clinical Davis, J. L. (2018). “Pharmacologic Principles,” in Equine Internal Medicine
Applications. Ed. T. A. Ahmed (Rijeka: IntechOpen), 53–98. doi: 10.5772/61573 (Philadelphia: W.B. Saunders), vol. 79–137.
Ait-Oudhia, S., Mager, D. E., and Straubinger, R. M. (2014). Application of Dayneka, N. L., Garg, V., and Jusko, W. J. (1993). Comparison of four basic models
pharmacokinetic and pharmacodynamic analysis to the development of of indirect pharmacodynamic responses. J. Pharmacokinet. Pharmacodyn. 21
liposomal formulations for oncology. Pharmaceutics 6 (1), 137–174. (4), 457–478. doi: 10.1007/bf01061691
doi: 10.3390/pharmaceutics6010137 Deslandes, A., Westphal, J. F., Trouvin, J. H., and Farinotti, R. (1992). Adaptive
Ait-Oudhia, S., Zhang, W., and Mager, D. E. (2017). A mechanism-based PK/PD computer program for determination of absorption profiles by numerical
model for hematological toxicities induced by antibody-drug conjugates. AAPS deconvolution: application to amoxicillin absorption. J. Pharm. Sci. 81 (8),
J. 19 (5), 1436–1448. doi: 10.1208/s12248-017-0113-5 802–807. doi: 10.1002/jps.2600810816
Amantana, A., Chen, Y., Tyavanagimatt, S. R., Jones, K. F., Jordan, R., Di Martino, M. T., Arbitrio, M., Fonsi, M., Erratico, C. A., Scionti, F., Caracciolo, D.,
Chinsangaram, J., et al. (2013). Pharmacokinetics and interspecies allometric et al. (2019). Allometric Scaling Approaches for Predicting Human
scaling of ST-246, an oral antiviral therapeutic for treatment of orthopoxvirus Pharmacokinetic of a Locked Nucleic Acid Oligonucleotide Targeting Cancer-
infection. PloS One 8 (4), e61514. doi: 10.1371/[Link].0061514 Associated miR-221. Cancers (Basel) 12 (1), 27. doi: 10.3390/cancers12010027
Anselmo, A. C., and Mitragotri, S. (2014). An overview of clinical and commercial Diao, L., and Meibohm, B. (2015). Tools for predicting the PK/PD of therapeutic
impact of drug delivery systems. J. Controlled Release 190, 15–28. doi: 10.1016/ proteins. Expert Opin. Drug Metab. Toxicol. 11 (7), 1115–1125. doi: 10.1517/
[Link].2014.03.053 17425255.2015.1041917
Arakawa, R., Ito, H., Takano, A., Takahashi, H., Morimoto, T., Sassa, T., et al. Dubois, V. F., Smania, G., Yu, H., Graf, R., Chain, A. S., Danhof, M., et al. (2017).
(2008). Dose-finding study of paliperidone ER based on striatal and Translating QT interval prolongation from conscious dogs to humans. Br.
extrastriatal dopamine D 2 receptor occupancy in patients with J. Clin. Pharmacol. 83 (2), 349–362. doi: 10.1111/bcp.13123
schizophrenia. Psychopharmacology 197 (2), 229–235. doi: 10.1007/s00213- Duquesnoy, C., Mamet, J., Sumner, D., and Fuseau, E. (1998). Comparative
007-1029-z clinical pharmacokinetics of single doses of sumatriptan following
Asiri, A. M., and Mohammad, A. (2018). Applications of nanocomposite materials subcutaneous, oral, rectal and intranasal administration. Eur. J. Pharm. Sci. 6
in drug delivery (Cambridge: Woodhead Publishing). (2), 99–104. doi: 10.1016/S0928-0987(97)00073-0
Balakrishnan, N. (2014). Methods and applications of statistics in clinical trials, Egrie, J. C., and Browne, J. K. (2001). Development and characterization of novel
volume 2: Planning, analysis, and inferential methods (New Jersey: John Wiley erythropoiesis stimulating protein (NESP). Br. J. Cancer 84 (1), 3–10. doi:
& Sons). 10.1054/bjoc.2001.1746
Bartlett, N., Forero-Torres, A., Rosenblatt, J., Fanale, M., Horning, S., and Elliott, S., Pham, E., and Macdougall, I. C. (2008). Erythropoietins: a common
Thompson, S. (2009). Complete remissions with SGN-35 weekly dosing: a mechanism of action. Exp. Hematol. 36 (12), 1573–1584. doi: 10.1016/
phase I dose-escalation study in relapsed/refractory Hodgkin lymphoma or [Link].2008.08.003
systemic ALCL patients. J. Clin. Oncol. 17, 8500. doi: 10.1200/jco.2009.27. Elmokadem, A., Riggs, M. M., and Baron, K. T. (2019). Quantitative Systems
15_suppl.8500 Pharmacology and Physiologically-Based Pharmacokinetic Modeling With
Benchimol, M. J., Bourne, D., Moghimi, S. M., and Simberg, D. (2019). mrgsolve: A Hands-On Tutorial. CPT: Pharmacometrics Syst. Pharmacol. 8
Pharmacokinetic analysis reveals limitations and opportunities for (12), 883–893. doi: 10.1002/psp4.12467
nanomedicine targeting of endothelial and extravascular compartments of Fanale, M. A., Forero-Torres, A., Rosenblatt, J. D., Advani, R. H., Franklin, A. R.,
tumours. J. Drug Target 27 (5-6), 690–698. doi: 10.1080/1061186X.2019.1566339 Kennedy, D. A., et al. (2012). A phase I weekly dosing study of brentuximab vedotin
Björkman, S., Rigemar, G., and Idvall, J. (1997). Pharmacokinetics of midazolam in patients with relapsed/refractory CD30-positive hematologic malignancies.
given as an intranasal spray to adult surgical patients. Br. J. Anaesth. 79 (5), Clin. Cancer Res. 18 (1), 248–255. doi: 10.1158/[Link]-11-1425
575–580. doi: 10.1093/bja/79.5.575 Fiset, P., Cohane, C., Browne, S., Brand, S. C., and Shafer, S. L. (1995).
Black, J. W., and Leff, P. (1983). Operational models of pharmacological agonism. Proc. Biopharmaceutics of a new transdermal fentanyl device. Anesthesiology 83 (3),
R. Soc. London Ser. B: Biol. Sci. 220 (1219), 141–162. doi: 10.1098/rspb.1983.0093 459–469. doi: 10.1097/00000542-199509000-00004
Bonate, P. L., and Steimer, J.-L. (2011). Pharmacokinetic-pharmacodynamic Friberg, L. E., Isbister, G. K., Hackett, L. P., and Duffull, S. B. (2005). The population
modeling and simulation (New York: Springer). pharmacokinetics of citalopram after deliberate self-poisoning: A Bayesian
Chetty, M., Johnson, T. N., Polak, S., Salem, F., Doki, K., and Rostami-Hodjegan, approach. J. Pharmacokinet. Phar. 32 (3-4), 571–605. doi: 10.1007/s10928-005-
A. (2018). Physiologically based pharmacokinetic modelling to guide drug 0022-6
delivery in older people. Adv. Drug Delivery Rev. 135, 85–96. doi: 10.1016/ Glaspy, J., Jadeja, J. S., Justice, G., Kessler, J., Richards, D., Schwartzberg, L., et al.
[Link].2018.08.013 (2001). A dose-finding and safety study of novel erythropoiesis stimulating
Cleton, A., Rossenu, S., Crauwels, H., Berwaerts, J., Hough, D., Gopal, S., et al. (2014). A protein (NESP) for the treatment of anaemia in patients receiving
single-dose, open-label, parallel, randomized, dose-proportionality study of multicycle chemotherapy. Br. J. Cancer 84 (1), 17–23. doi: 10.1054/
paliperidone after intramuscular injections of paliperidone palmitate in the bjoc.2001.1748
deltoid or gluteal muscle in patients with schizophrenia. J. Clin. Pharmacol. 54 Glaspy, J., Jadeja, J., Justice, G., Kessler, J., Richards, D., Schwartzberg, L., et al. (2002).
(9), 1048–1057. doi: 10.1002/jcph.295 Darbepoetin alfa given every 1 or 2 weeks alleviates anaemia associated with cancer
Cox, E. H., Langemeijer, M. W., Gubbens-Stibbe, J. M., Muir, K. T., and Danhof, M. chemotherapy. Br. J. Cancer 87 (3), 268–276. doi: 10.1038/[Link].6600465
(1999). The comparative pharmacodynamics of remifentanil and its metabolite, Glassman, P. M., and Muzykantov, V. R. (2019). Pharmacokinetic and
GR90291, in a rat electroencephalographic model. Anesthesiol.: J. Am. Soc. Pharmacodynamic Properties of Drug Delivery Systems. J. Pharmacol. Exp.
Anesthesiol. 90 (2), 535–544. doi: 10.1097/00000542-199902000-00030 Ther. 370 (3), 570–580. doi: 10.1124/jpet.119.257113
da Silva Honó rio, T., Pinto, E. C., Rocha, H. V. A., Esteves, V. S. A. D., dos Santos, Gopal, S., Vermeulen, A., Nandy, P., Ravenstijn, P., Nuamah, I., Buron Vidal, J. A.,
T. C., Castro, H. C. R., et al. (2013). In vitro–in vivo correlation of efavirenz et al. (2015). Practical guidance for dosing and switching from paliperidone
tablets using GastroPlus®. AAPS Pharm. Sci. Tech. 14 (3), 1244–1254. doi: palmitate 1 monthly to 3 monthly formulation in schizophrenia. Curr. Med.
10.1208/s12249-013-0016-4 Res. Opin. 31 (11), 2043–2054. doi: 10.1185/03007995.2015.1085849
Danhof, M., Alvan, G., Dahl, S. G., Kuhlmann, J., and Paintaud, G. (2005). Gralinski, M. R. (2003). The dog’s role in the preclinical assessment of QT interval
Mechanism-based pharmacokinetic-pharmacodynamic modeling-a new prolongation. Toxicol. Pathol. 31 (1_suppl), 11–16. doi: 10.1080/01926230390174887

Frontiers in Pharmacology | [Link] 13 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

Harashima, H., Tsuchihashi, M., Iida, S., Doi, H., and Kiwada, H. (1999). Lindauer, A., Valiathan, C. R., Mehta, K., Sriram, V., de Greef, R., Elassaiss-Schaap, J.,
Pharmacokinetic/pharmacodynamic modeling of antitumor agents encapsulated et al. (2017). Translational Pharmacokinetic/Pharmacodynamic Modeling of
into liposomes. Adv. Drug Deliv. Rev. 40 (1-2), 39–61. doi: 10.1016/S0169-409x(99) Tumor Growth Inhibition Supports Dose-Range Selection of the Anti-PD-1
00039-3 Antibody Pembrolizumab. CPT Pharmacometrics Syst. Pharmacol. 6 (1), 11–20.
He, H., Yuan, D., Wu, Y., and Cao, Y. (2019). Pharmacokinetics and Pharmacodynamics doi: 10.1002/psp4.12130
Modeling and Simulation Systems to Support the Development and Regulation of Locatelli, F., and Del Vecchio, L. (2011). Erythropoiesis-stimulating agents in renal
Liposomal Drugs. Pharmaceutics 11 (3), 110. doi: 10.3390/pharmaceutics11030110 medicine. Oncologist 16, 19–24. doi: 10.1634/theoncologist.2011-S3-19
Hedrich, W. D., Fandy, T. E., Ashour, H. M., Wang, H., and Hassan, H. E. (2018). Macdougall, I. (2005). CERA (Continuous Erythropoietin Receptor Activator): a
Antibody-Drug Conjugates: Pharmacokinetic/Pharmacodynamic Modeling, new erythropoiesis-stimulating agent for the treatment of anemia. Curr.
Preclinical Characterization, Clinical Studies, and Lessons Learned. Clin. Hematol. Rep. 4 (6), 436–440.
Pharmacokinet. 57 (6), 687–703. doi: 10.1007/s40262-017-0619-0 Mager, D. E., and Jusko, W. J. (2001a). General pharmacokinetic model for drugs
Huh, Y., Smith, D. E., and Feng, M. R. (2011). Interspecies scaling and prediction exhibiting target-mediated drug disposition. J. Pharmacokinet. Pharmacodyn.
of human clearance: comparison of small- and macro-molecule drugs. 28 (6), 507–532. doi: 10.1023/a:1014414520282
Xenobiotica 41 (11), 972–987. doi: 10.3109/00498254.2011.598582 Mager, D. E., and Jusko, W. J. (2001b). Pharmacodynamic modeling of time-
Idkaidek, N. M., Amidon, G. L., Smith, D. E., Najib, N. M., and Hassan, M. M. (1998). dependent transduction systems. Clin. Pharmacol. Ther. 70 (3), 210–216.
Determination of the population pharmacokinetic parameters of sustained-release doi: 10.1067/mcp.2001.118244
and enteric-coated oral formulations, and the suppository formulation of Mager, D. E., Wyska, E., and Jusko, W. J. (2003). Diversity of mechanism-based
diclofenac sodium by simultaneous data fitting using NONMEM. Biopharm. pharmacodynamic models. Drug Metab. Dispos. 31 (5), 510–518. doi: 10.1124/
Drug Dispos. 19 (3), 169–174. doi: 10.1002/(SICI)1099-081X(199804)19:3<169:: dmd.31.5.510
AID-BDD83>[Link];2-C Mager, D. E., Neuteboom, B., and Jusko, W. J. (2005). Pharmacokinetics and
Jadhav, P. R., Agersø, H., Tornøe, C. W., and Gobburu, J. V. (2006). Semi- pharmacodynamics of PEGylated IFN-b 1a following subcutaneous administration
mechanistic pharmacodynamic modeling for degarelix, a novel gonadotropin in monkeys. Pharm. Res. 22 (1), 58–61. doi: 10.1007/s11095-004-9009-z
releasing hormone (GnRH) blocker. J. Pharmacokinetics Pharmacodyn. 33 (5), Mager, D. E., Woo, S., and Jusko, W. J. (2009). Scaling pharmacodynamics from in
609–634. doi: 10.1007/s10928-006-9025-1 vitro and preclinical animal studies to humans. Drug Metab. Pharmacokinet.
Jones, H., and Rowland-Yeo, K. (2013). Basic concepts in physiologically based 24 (1), 16–24. doi: 10.2133/dmpk.24.16
pharmacokinetic modeling in drug discovery and development. CPT McSweeney, M., Wessler, T., Price, L., Ciociola, E., Herity, L., Piscitelli, J., et al.
Pharmacometrics Syst. Pharmacol. 2, e63. doi: 10.1038/psp.2013.41 (2018). A minimal physiologically based pharmacokinetic model that predicts
Jusko, W. J., and Ko, H. C. (1994). Physiologic indirect response models anti-PEG IgG-mediated clearance of PEGylated drugs in human and mouse.
characterize diverse types of pharmacodynamic effects. Clin. Pharmacol. J. Controlled Release 284, 171–178. doi: 10.1016/[Link].2018.06.002
Ther. 56 (4), 406–419. doi: 10.1038/clpt.1994.155 Minematsu, T., Ohtani, H., Yamada, Y., Sawada, Y., Sato, H., and Iga, T. (2001).
Jusko, W. J. (1973). A pharmacodynamic model for cell-cycle-specific chemotherapeutic Quantitative relationship between myocardial concentration of tacrolimus and
agents. J. Pharmacokinetics Biopharm. 1 (3), 175–200. doi: 10.1007/BF01062346 QT prolongation in guinea pigs: Pharmacokinetic/pharmacodynamic model
Kadakia, E., Bottino, D., and Amiji, M. (2019). Mathematical Modeling and incorporating a site of adverse effect. J. Pharmacokinet. Phar. 28 (6), 533–554.
Simulation to Investigate the CNS Transport Characteristics of Nanoemulsion- doi: 10.1023/A:1014460404352
Based Drug Delivery Following Intranasal Administration. Pharm. Res. 36 (5), Mishra, G. P., Bagui, M., Tamboli, V., and Mitra, A. K. (2011). Recent applications
75. doi: 10.1007/s11095-019-2610-y of liposomes in ophthalmic drug delivery. J. Drug Deliv. 2011, 1–14. doi:
Kapur, S., Zipursky, R., Jones, C., Remington, G., and Houle, S. (2000). 10.1155/2011/863734
Relationship between dopamine D2 occupancy, clinical response, and side Moss, D. M., and Siccardi, M. (2014). Optimizing nanomedicine pharmacokinetics
effects: a double-blind PET study of first-episode schizophrenia. Am. using physiologically based pharmacokinetics modelling. Br. J. Pharmacol. 171
J. Psychiatry 157 (4), 514–520. doi: 10.1176/[Link].157.4.514 (17), 3963–3979. doi: 10.1111/bph.12604
Kiss, Z., Elliott, S., Jedynasty, K., Tesar, V., and Szegedi, J. (2010). Discovery and Nagashima, R., O’reilly, R., and Levy, G. (1969). Kinetics of pharmacologic effects
basic pharmacology of erythropoiesis-stimulating agents (ESAs), including the in man: the anticoagulant action of warfarin. Clin. Pharmacol. Ther. 10 (1), 22–
hyperglycosylated ESA, darbepoetin alfa: an update of the rationale and clinical 35. doi: 10.1002/cpt196910122
impact. Eur. J. Clin. Pharmacol. 66 (4), 331–340. doi: 10.1007/s00228-009- Nikanjam, M., Capparelli, E. V., Lancet, J. E., Louie, A., and Schiller, G. (2018).
0780-y Persistent cytarabine and daunorubicin exposure after administration of novel
Krzyzanski, W., Ramakrishnan, R., and Jusko, W. J. (1999). Basic pharmacodynamic liposomal formulation CPX-351: population pharmacokinetic assessment. Cancer
models for agents that alter production of natural cells. J. Pharmacokinetics Chemother. Pharmacol. 81 (1), 171–178. doi: 10.1007/s00280-017-3484-5
Biopharm. 27 (5), 467–489. doi: 10.1023/A:1023249813106 Nordström, A.-L., Farde, L., Wiesel, F.-A., Forslund, K., Pauli, S., Halldin, C., et al.
Krzyzanski, W., Sutjandra, L., Perez-Ruixo, J. J., Sloey, B., Chow, A. T., and Wang, Y.-M. (1993). Central D2-dopamine receptor occupancy in relation to antipsychotic
(2013). Pharmacokinetic and pharmacodynamic modeling of romiplostim in drug effects: a double-blind PET study of schizophrenic patients. Biol.
animals. Pharm. Res. 30 (3), 655–669. doi: 10.1007/s11095-012-0894-2 Psychiatry 33 (4), 227–235. doi: 10.1016/0006-3223(93)90288-O
Laizure, S. C., and Parker, R. B. (2009). Pharmacodynamic Evaluation of the Park, K. (2019). PK/PD model for liposomal chemophototherapy. J. Control
Cardiovascular Effects after the Coadministration of Cocaine and Ethanol. Release 297, 102. doi: 10.1016/[Link].2019.02.022
Drug Metab. Dispos. 37 (2), 310–314. doi: 10.1124/dmd.108.023531 Peck, C. C., Barr, W. H., Benet, L. Z., Collins, J., Desjardins, R. E., Furst, D. E., et al.
Larsen, F., Clausen, T., Wolff, J., Hansen, P., and Rasmussen, S. (1991). Absorption (1992). Opportunities for integration of pharmacokinetics, pharmacodynamics,
profiles of rectally administered midazolam estimated by deconvolution. Eur. and toxicokinetics in rational drug development. Clin. Pharmacol. Ther. 51 (4),
J. Drug Metab. Pharmacokinetics 291–296. 465–473. doi: 10.1038/clpt.1992.47
Lestner, J. M., Groll, A. H., Aljayyoussi, G., Seibel, N. L., Shad, A., Gonzalez, C., Reed, S. M., Bayly, W. M., and Sellon, D. C. (2017). Equine Internal Medicine-E-
et al. (2016). Population pharmacokinetics of liposomal amphotericin B in Book (New York: Elsevier Health Sciences).
immunocompromised children. Antimicrob. Agents Chemother. 60 (12), 7340– Rocchetti, M., Simeoni, M., Pesenti, E., De Nicolao, G., and Poggesi, I. (2007).
7346. doi: 10.1128/AAC.01427-16 Predicting the active doses in humans from animal studies: a novel approach in
Levy, G. (1964). Relationship between elimination rate of drugs and rate of decline of their oncology. Eur. J. Cancer 43 (12), 1862–1868. doi: 10.1016/[Link].2007.05.011
pharmacologic effects. J. Pharm. Sci. 53 (3), 342–343. doi: 10.1002/jps.2600530325 Rodallec, A., Fanciullino, R., Lacarelle, B., and Ciccolini, J. (2018). Seek and destroy:
Levy, G. (1966). Kinetics of pharmacologic effects. Clin. Pharmacol. Ther. 7 (3), improving PK/PD profiles of anticancer agents with nanoparticles. Expert Rev.
362–372. doi: 10.1002/cpt196673362 Clin. Pharmacol. 11 (6), 599–610. doi: 10.1080/17512433.2018.1477586
Li, M., Zou, P., Tyner, K., and Lee, S. (2017). Physiologically Based Pharmacokinetic Sager, J. E., Yu, J., Ragueneau-Majlessi, I., and Isoherranen, N. (2015).
(PBPK) Modeling of Pharmaceutical Nanoparticles. AAPS J. 19 (1), 26–42. Physiologically based pharmacokinetic (PBPK) modeling and simulation
doi: 10.1208/s12248-016-0010-3 approaches: a systematic review of published models, applications, and

Frontiers in Pharmacology | [Link] 14 July 2020 | Volume 11 | Article 997

Zou et al. PK-PD Modeling in Drug Delivery

model verification. Drug Metab. Dispos. 43 (11), 1823–1837. doi: 10.1124/ clinical trials with novel biopharmaceuticals. Regul. Toxicol. Pharmacol. 58 (2),
dmd.115.065920 243–251. doi: 10.1016/[Link].2010.06.007
Samtani, M. N., Vermeulen, A., and Stuyckens, K. (2009). Population pharmacokinetics Tsuchikama, K., and An, Z. (2018). Antibody-drug conjugates: recent advances in
of intramuscular paliperidone palmitate in patients with schizophrenia: a novel conjugation and linker chemistries. Protein Cell 9 (1), 33–46. doi: 10.1007/
once-monthly, long-acting formulation of an atypical antipsychotic. Clin. s13238-016-0323-0
Pharmacokinet. 48 (9), 585–600. doi: 10.2165/11316870-000000000-00000 Van der Graaf, P., Van Schaick, E., Mathot, R., Ijzerman, A., and Danhof, M.
Samtani, M. N., Nandy, P., Ravenstijn, P., Remmerie, B., Vermeulen, A., Russu, A., (1997). Mechanism-based pharmacokinetic-pharmacodynamic modeling of
et al. (2016). Prospective dose selection and acceleration of paliperidone palmitate the effects of N6-cyclopentyladenosine analogs on heart rate in rat:
3-month formulation development using a pharmacometric bridging strategy. Br. estimation of in vivo operational affinity and efficacy at adenosine A1
J. Clin. Pharmacol. 82 (5), 1364–1370. doi: 10.1111/bcp.13050 receptors. J. Pharmacol. Exp. Ther. 283 (2), 809–816.
Scott, J. C., Cooke, J. E., and Stanski, D. R. (1991). Electroencephalographic Van der Graaf, P. H., Van Schaick, E. A., Visser, S. A. G., De Greef, H. J. M. M.,
quantitation of opioid effect: comparative pharmacodynamics of fentanyl and Ijzerman, A. P., and Danhof, M. (1999). Mechanism-based pharmacokinetic-
sufentanil. Anesthesiology 74 (1), 34–42. doi: 10.1097/00000542-199101000-00007 pharmacodynamic modeling of antilipolytic effects of adenosine A(1) receptor
Shah, D. K., Haddish-Berhane, N., and Betts, A. (2012). Bench to bedside agonists in rats: Prediction of tissue-dependent efficacy in vivo. J. Pharmacol.
translation of antibody drug conjugates using a multiscale mechanistic PK/ Exp. Ther. 290 (2), 702–709.
PD model: a case study with brentuximab-vedotin. J. Pharmacokinet. Vansteenkiste, J., Pirker, R., Massuti, B., Barata, F., Font, A., Fiegl, M., et al. (2002).
Pharmacodyn. 39 (6), 643–659. doi: 10.1007/s10928-012-9276-y Double-blind, placebo-controlled, randomized phase III trial of darbepoetin
Shah, D. K., King, L. E., Han, X. G., Wentland, J. A., Zhang, Y. H., Lucas, J., et al. alfa in lung cancer patients receiving chemotherapy. J. Natl. Cancer Inst. 94
(2014). A Priori Prediction of Tumor Payload Concentrations: Preclinical Case (16), 1211–1220. doi: 10.1093/jnci/94.16.1211
Study with an Auristatin-Based Anti-5T4 Antibody-Drug Conjugate. AAPS J. Wagner, J. (1968). Kinetics of pharmacologic response I. Proposed relationships
16 (3), 452–463. doi: 10.1208/s12248-014-9576-9 between response and drug concentration in the intact animal and man.
Shargel, L., Andrew, B., and Wu-Pong, S. (1999). Applied biopharmaceutics and J. Theor. Biol. 20 (2), 173–201. doi: 10.1016/0022-5193(68)90188-4
pharmacokinetics (New York: Appleton & Lange Stamford). Yamashita, F., and Hashida, M. (2013). Pharmacokinetic considerations for
Sharma, A., and Jusko, W. J. (1998). Characteristics of indirect pharmacodynamic targeted drug delivery. Adv. Drug Delivery Rev. 65 (1), 139–147. doi: 10.1016/
models and applications to clinical drug responses. Br. J. Clin. Pharmacol. 45 [Link].2012.11.006
(3), 229–239. doi: 10.1046/j.1365-2125.1998.00676.x Yamazaki, S. (2013). Translational pharmacokinetic-pharmacodynamic modeling
Sharma, A., Ebling, W. F., and Jusko, W. J. (1998). Precursor-dependent indirect from nonclinical to clinical development: a case study of anticancer drug,
pharmacodynamic response model for tolerance and rebound phenomena. crizotinib. AAPS J. 15 (2), 354–366. doi: 10.1208/s12248-012-9436-4
J. Pharm. Sci. 87 (12), 1577–1584. doi: 10.1021/js980171q Yan, X., and Krzyzanski, W. (2013). Quantitative assessment of minimal effective
Sheiner, L. B., Stanski, D. R., Vozeh, S., Miller, R. D., and Ham, J. (1979). Simultaneous concentration of erythropoiesis-stimulating agents. Cpt-Pharmacometrics Syst.
modeling of pharmacokinetics and pharmacodynamics: application to d- Pharmacol. 2 (8), 1–10. doi: 10.1038/psp.2013.39
tubocurarine. Clin. Pharmacol. Ther. 25 (3), 358–371. doi: 10.1002/cpt1979253358 Yan, X., Lowe, P. J., Fink, M., Berghout, A., Balser, S., and Krzyzanski, W. (2012).
Siccardi, M., Rannard, S., and Owen, A. (2018). The emerging role of physiologically Population pharmacokinetic and pharmacodynamic model-based
based pharmacokinetic modelling in solid drug nanoparticle translation. Adv. comparability assessment of a recombinant human Epoetin Alfa and the
Drug Deliv. Rev. 131, 116–121. doi: 10.1016/[Link].2018.06.016 Biosimilar HX575. J. Clin. Pharmacol. 52 (11), 1624–1644. doi: 10.1177/
Siegel, R. A., and Rathbone, M. J. (2012). “Overview of controlled release 0091270011421911
mechanisms,” in Fundamentals and applications of controlled release drug Yanez, J. A., Remsberg, C. M., Sayre, C. L., Forrest, M. L., and Davies, N. M. (2011).
delivery (New York: Springer), 19–43. Flip-flop pharmacokinetics–delivering a reversal of disposition: challenges and
Singh, A. P., and Shah, D. K. (2018). Utility of PK-PD Modeling and Simulation to opportunities during drug development. Ther. Delivery 2 (5), 643–672.
Improve Decision Making for Antibody-Drug Conjugate Development. Innovations doi: 10.4155/tde.11.19
Next-Gener. Antibody-Drug Conjugates 73–97. doi: 10.1007/978-3-319-78154-9 Younes, A., Bartlett, N. L., Leonard, J. P., Kennedy, D. A., Lynch, C. M., Sievers, E. L.,
Singh, A. P., Shin, Y. G., and Shah, D. K. (2015). Application of Pharmacokinetic- et al. (2010). Brentuximab vedotin (SGN-35) for relapsed CD30-positive
Pharmacodynamic Modeling and Simulation for Antibody-Drug Conjugate lymphomas. New Engl. J. Med. 363 (19), 1812–1821. doi: 10.1056/NEJMoa1002965
Development. Pharm. Res. 32 (11), 3508–3525. doi: 10.1007/s11095-015-1626-1 Zhang, L., Pfister, M., and Meibohm, B. (2008). Concepts and Challenges in
Soininen, S. K., Vellonen, K. S., Heikkinen, A. T., Auriola, S., Ranta, V. P., Urtti, A., Quantitative Pharmacology and Model-Based Drug Development. AAPS J. 10
et al. (2016). Intracellular PK/PD Relationships of Free and Liposomal (4), 552–559. doi: 10.1208/s12248-008-9062-3
Doxorubicin: Quantitative Analyses and PK/PD Modeling. Mol. Pharm. 13 Zheng, Y., Scheerens, H., Davis, J. [Link]., Deng, R., Fischer, S. K., Woods, C., et al.
(4), 1358–1365. doi: 10.1021/[Link].6b00008 (2011). Translational pharmacokinetics and pharmacodynamics of an FcRn-
Stege, R., Gunnarsson, P. O., Johansson, C. J., Olsson, P., Pousette, Å., and Carlström, K. variant anti-CD4 monoclonal antibody from preclinical model to phase I
(1996). Pharmacokinetics and testosterone suppression of a single dose of study. Clin. Pharmacol. Ther. 89 (2), 283–290. doi: 10.1038/clpt.2010.311
polyestradiol phosphate (Estradurin®) in prostatic cancer patients. Prostate 28 (5),
307–310. doi: 10.1002/(SICI)1097-0045(199605)28:5<307::AID-PROS6>[Link];2-8 Conflict of Interest: The authors declare that the research was conducted in the
Stepensky, D., Friedman, M., Srour, W., Raz, I., and Hoffman, A. (2001). Preclinical absence of any commercial or financial relationships that could be construed as a
evaluation of pharmacokinetic–pharmacodynamic rationale for oral CR potential conflict of interest.
metformin formulation. J. Controlled Rel. 71 (1), 107–115. doi: 10.1016/S0168-
3659(00)00374-6 Copyright © 2020 Zou, Banerjee, Leung and Yan. This is an open-access article
Suryawanshi, S., Zhang, L., Pfister, M., and Meibohm, B. (2010). The current role distributed under the terms of the Creative Commons Attribution License (CC BY).
of model-based drug development. Expert Opin. Drug Discovery 5 (4), 311–321. The use, distribution or reproduction in other forums is permitted, provided the
doi: 10.1517/17460441003713470 original author(s) and the copyright owner(s) are credited and that the original
Tibbitts, J., Cavagnaro, J. A., Haller, C. A., Marafino, B., Andrews, P. A., and publication in this journal is cited, in accordance with accepted academic practice. No
Sullivan, J. T. (2010). Practical approaches to dose selection for first-in-human use, distribution or reproduction is permitted which does not comply with these terms.

Frontiers in Pharmacology | [Link] 15 July 2020 | Volume 11 | Article 997