Microvascular Research 99 (2015) 43–56

Contents lists available at ScienceDirect

Microvascular Research
journal homepage: www.elsevier.com/locate/ymvre

Numerical modeling of drug delivery in a dynamic solid

tumor microvasculature
M. Sefidgar a, M. Soltani a,b,⁎, K. Raahemifar c, M. Sadeghi d, H. Bazmara a, M. Bazargan a, M. Mousavi Naeenian a
a
Department of Mechanical Engineering, K.N.T. University of Technology, Tehran, Iran
b
Division of Nuclear Medicine, Department of Radiology and Radiological Science, Johns Hopkins University, School of Medicine, MD, USA
c
Electrical & Computer Department of Ryerson University, Toronto, Ontario, Canada
d
Digital Health Hub, Simon Fraser University, Surrey, BC, Canada

a r t i c l e i n f o a b s t r a c t

Article history: The complicated capillary network induced by angiogenesis is one of the main reasons of unsuccessful cancer
Accepted 11 February 2015 therapy. A multi-scale mathematical method which simulates drug transport to a solid tumor is used in this
Available online 24 February 2015 study to investigate how capillary network structure affects drug delivery. The mathematical method involves
processes such as blood flow through vessels, solute and fluid diffusion, convective transport in extracellular
Keywords:
matrix, and extravasation from blood vessels. The effect of heterogeneous dynamic network on interstitial
Solute transport
Convection and diffusion equation
fluid flow and drug delivery is investigated by this multi-scale method. The sprouting angiogenesis model is
Coupled interstitial and intravascular flow used for generating capillary network and then fluid flow governing equations are implemented to calculate
Remodeling capillary network blood flow through the tumor-induced capillary network and fluid flow in normal and tumor tissues. Finally,
Heterogeneous capillary network convection–diffusion equation is used to simulate drug delivery. Three approaches are used to simulate drug
transport based on the developed mathematical method: without a vascular network, using a static vascular
network, and a dynamic vascular network. The avascular approach predicts more uniform and higher drug
concentration than vascular approaches since the simplified assumptions are implemented in this method. The
dynamic network which uses more realistic assumptions predicts more irregular blood vessels, high interstitial
pressure, and more heterogeneity in drug distribution than other two approaches.
Crown Copyright © 2015 Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction
Based on findings from clinical applications, most drug treatments
fail to eliminate malignant tumors completely even though drug
delivery through systemic administration may inhibit their growth
(Goel et al., 2011; Jain et al., 2007; Soltani and Chen, 2011; Tamburrino
et al., 2013). The heterogeneity of microvascular network which is
generated by tumor angiogenesis is one of unsuccessful treatment
reasons. The microvascular network induced by a tumor disorganizes
vessel network with high microvascular permeability (Dreher et al.,
2006). The high leakiness of vessels in lack of lymphatic function leads
to create an elevated interstitial pressure (Sefidgar et al., 2014; Soltani
and Chen, 2013; Stylianopoulos et al., 2013). The high interstitial
pressure is known as the main barrier to drug delivery to the tumors
(Soltani and Chen, 2011). Therefore, better understanding of tumor for-
mation is crucial in developing more effective therapeutics (Soltani and
⁎ Corresponding author at: Division of Nuclear Medicine, Department of Radiology and
Radiological Science, Johns Hopkins University, School of Medicine, MD, USA.
E-mail addresses: sefidgar@dena.kntu.ac.ir (M. Sefidgar), msoltani@jhu.edu
(M. Soltani), kraahemi@ee.ryerson.ca (K. Raahemifar), maryam_sadeghi@sfu.ca
(M. Sadeghi), bazmara@dena.kntu.ac.ir (H. Bazmara), bazargan@kntu.ac.ir (M. Bazargan),
mousavi@kntu.ac.ir (M. Mousavi Naeenian).
Chen, 2011). For this purpose, nowadays, solid tumor modeling and
simulation results are used to predict how therapeutic drugs are
transported to tumor cells by blood flow through capillaries and tissues.
This facilitates identifying better methods for delivering targeted anti-
cancer therapies (Jain, 2001; Pozrikidis, 2010).
The modeling of drug delivery to solid tumor microvasculature has a
multi-scale nature. Multi-scale modeling involves: microvasculature gen-
eration induced by a tumor in the scale of nano-meter, blood flow con-
vection in capillaries by considering non-continuous behavior of blood
and adaption of capillary diameters in the scale of micro-meter, blood
flow distribution through a network in the scale of millimeter and the
fluid flow and drug transport in tumor and normal tissues in the scale
of centimeter. These scales considered in this study are shown in Fig. 1.
There are two approaches in the simulation of drug delivery to solid
tumors: macroscopic and microscopic. In the macroscopic approach,
only the distribution of variables, such as interstitial pressure and concen-
tration, over the length scale of the tumor radius is important. In the mi-
croscopic approach, characteristics such as structure of microvascular
network, blood flow through microvessel, interaction between microvas-
cular wall and flow in peripheral flow are involved directly in the model.
Baxter and Jain (1989, 1990, 1991) by considering macroscopic
view, used a continuum porous media model to simulate interstitial

http://dx.doi.org/10.1016/j.mvr.2015.02.007
0026-2862/Crown Copyright © 2015 Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
44 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56
Fig. 1. Schematic of different scales of drug delivery to solid tumor simulation.
flow and solute transport in solid tumors. Soltani and Chen (2011)
developed a mathematical model in the spherical tumors to introduce
two new parameters: the critical tumor radius and critical necrotic
radius. They also applied their model to different geometries of tumors
to study the effects of tumor shape and size in drug delivery (Soltani and
Chen, 2012).
The effect of heterogeneous tissue transport properties on interstitial
transport is studied by Zhao et al. (2007), Arifin et al. (2009) and Pishko
et al. (2011, 2012). Pishko et al. (2011) modeled tumor with spatially-
varying tissue transport properties. The effect of microvascular network
induced by tumor angiogenesis is not considered in abovementioned
studies.
Developed models of fluid flow in tumors have also focused on
incorporating spatial and temporal variations in blood flow using micro-
vascular network models. Pozrikidis (2010) presented a theoretical
framework for blood flow modeling through the microvascular network
coupled with the interstitial flow. The effect of interstitial pressure and
microvascular permeability on the plasma leakage is investigated in
Pozrikidis's study. Chauhan et al. (2012) developed a mathematical
model of solute concentration to investigate how changes in vascular
pore size distribution can improve the drug delivery. Stylianopoulos
et al. (2013) developed a mathematical framework to investigate
transvascular rate for different pore sizes of vessel wall in a microvascular
network induced by angiogenesis. The effect of extravasation, perme-
ability, and vascular heterogeneities on drug transport is investigated
by Welter and Rieger (2013). Their tumor growth model includes vessel
co-option, regression, and angiogenesis. Cai et al. (2011) developed a
mathematical model to couple intravascular, transvascular, and inter-
stitial fluid flow. This model was capable to consider the dynamic
behavior of tumor cell proliferation, death, and angiogenesis. They
could also investigate angiogenesis and tumor growth interactions in
their model. The dynamic behavior of capillaries is not considered in
above works and the capillary network is modeled such as a rigid and
static network in all of these studies.
Secomb et al. (2013) investigated theoretical simulations including
dynamic vascular network. Based on what they developed on vascular
network formation in normal and tumor tissues, one can predicts the
effects of ant-angiogenesis therapies. Chaplain et al. (2012) modeled
an adaptive network associated with tumor-induced angiogenesis.
Based on their simulation results, drug can bypass tumor vasculature
without reaching tumor cells. In the abovementioned studies, only
intravascular blood flow was considered and drug delivery in tissues
was not involved.
In the previous work of our group (Soltani and Chen, 2013), a math-
ematical model was developed that simultaneously couples interstitial
fluid flow with convective non-continuous blood flow through vessels
by considering a remodeling network based on hemodynamic and
metabolic stimuli. The effect of vasculature induced by tumor angiogen-
esis on interstitial pressure was also investigated statistically (Sefidgar
et al., 2014a, 2014b). In the present work, the mathematical model pre-
sented in our previous work is further developed to investigate drug de-
livery to the tumors with the present of heterogeneous capillary
network. In this study, only the drugs successfully delivered to the tis-
sues are considered and drug barriers in the tissue level are investigated.
The lymphatic system which has significant effect on interstitial flow
and drug delivery is also added to the current mathematical model.
Same as previous work, the model is simulated by considering three ap-
proaches: the first, fluid flow in normal and tumor tissues without a
capillary network; the second, fluid flow in normal and tumor tissues
with static capillary network; and the third, fluid flow in normal and
tumor tissues with a dynamic capillary network.
The blood flow is coupled by interstitial fluid flow (IFF) to calculate
intravascular blood pressure (IBP) for the second and third approaches
and interstitial fluid pressure and velocity (IFP and IFV) for all three
approaches. IFP and IFV are used to calculate the drug concentration
in interstitial flow (DCIF), via implementing solute transport equation
in porous media. Results show that the IFP, IFV, and DCIF for the first
approach are more uniform than that one for other approaches with
considering heterogeneous microvascular network. Dynamic simula-
tion predicts the highest IFP due to elevated IBP in tumor region. The
predicted DCIF for dynamic simulation (the third approach) has more
value and heterogeneity than that of static network since it has bigger
source term values of drug and IFV than the static network.
Material and methods
The mathematical model includes three steps: the modeling of vas-
culature formation by sprouting angiogenesis, fluid flow in interstitial
 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56 45

space, and blood flow through vasculature and solute transport in

0:645
−0:170R −0:120R
μ 45 ¼ 6e þ 3:2−2:44  e ð4Þ
interstitium.
The present angiogenesis model is motivated by the tumor-induced
angiogenesis model initially proposed by Anderson and Chaplain
(1998). The model is lattice-based which is derived from the H the hematocrit (volume fraction of red blood cells contained
discretization of a 2D continuous model and used the theory of in the blood),
reinforced random walk to track the motion of individual endothelial R the vessel radius in μm, and
cells (ECs) located at the capillary sprout tips and the subsequent forma- ζ the viscosity dependency on the hematocrit defined by:
tion of capillaries. The detail of rules for sprouting angiogenesis and
algorithm of this method is discussed in both our previous work  
1

1
−0:150R
(Soltani, 2012) and Anderson et al. (2012). ζ ¼ 0:8 þ e −1 þ −11
þ : ð5Þ
Fluid flow modeling and dynamic simulation are summarized in 1 þ 10 ð2RÞ
12
1 þ 10−11 ð2RÞ12
Blood flow through microvascular network section and Dynamic
structure adaptation method section, respectively. Further details of
Blood flow in microvascular network is considered a two-phase flow
the simulation are introduced in our previous work (Soltani and Chen,
as it includes cells and plasma. Since the component distribution is not
2013). Solute transport simulation which is one of the main goals of
proportional to the plasma distribution at the branches, it is possible
this paper is introduced in Solute transport section.
to have all hematocrit in one daughter vessel and nothing in the other
one (Fung, 1993). The empirical equation developed by Pries and
Blood flow through microvascular network
Secomb (2010) is used for phase separation at microvascular bifurca-
tions. The fractional flow of red blood cells into one daughter branch
Poiseuille's law can be applied for blood flow in capillaries since
H2/H1 is calculated from the respective fractional blood flow Q 2/Q1 as
Reynolds number is low:
defined in Pries and Secomb (2010)

π ΔP V R4
QV ¼ ð1Þ 8
8 Lμ b > H2 Q2
>
> ¼0 if ≤ X0
>
> H Q1
>
< 1 H 
 
where Q2 Q
log it 2 ¼ A þ B log it −X 0 =ð1−2X 0 Þ if X 0 b 2 b 1−X 0
>
> H1 Q1 Q1
>
>
QV the blood flow rate through vascular [m3/s] >
> H Q
: 2¼1 if1−X 0 ≤ 2
ΔPV the pressure difference in vascular [Pa] H1 Q1
R the vessel radius [m] ð6Þ
L the vessel length [m], and
μb the blood viscosity [Pa·s] where logit x = ln[x/(1 − x)] and the parameters A, B, and X0 defining
the phase separation characteristics of the bifurcation were obtained
Poiseuille's law is applicable for Newtonian flow. The suspension from linear fits to experimental data (Pries and Secomb, 2010):
characteristics of blood strongly affect the dynamics of blood flow in h . i
2 2 2 2
microvessels. Finite size of these suspended elements in microvessels A ¼ −13:29 D3 =D2 −1 D3 =D2 þ 1 ð1−H1 Þ=D1
results in a few important phenomena including non-continuum B ¼ 1 þ 6:98ð1−H1 Þ=D1 ð7Þ
behavior, variation of the apparent viscosity of blood with tube X 0 ¼ 0:964ð1−H 1 Þ=D1 :
diameter, and non-uniform distribution of hematocrit between
branches of diverging microvascular bifurcations (Herbert H. Here, D1 is the diameters of the main vessel, D2, and D3 are the
Lipowsky, 2007; Pries and Secomb, 2008). To cover these effects, diameters of the daughter branches, H1 is the hematocrit in the main
Poiseuille's law has been further modified to be applied in blood flow vessel and H2 and H3 are hematocrit in daughter branches, respectively
simulation. (Fig. 2).
Blood viscosity in capillaries depends on the vessel diameter and
hematocrit. Poiseuille's law is corrected by replacing apparent viscosity Dynamic structure adaptation method
(μ app) instead of viscosity term (μ b) in Eq. (1). Pries et al. (1992, 1996)
defined an empirical relation for apparent blood viscosity as a function As mentioned, capillaries are able to continuously adapt their radius
of vessel radius and hematocrit for a wide range of microvessels radius and remodel the network structure in response to the functional
(from 2 μm to 300 μm) as follows: requirements of the tissues that capillaries supply (Pries et al., 1998,
2001a). For each vessel in the network, the change of its radius (ΔD)
μ app ¼ μ plasma  μ rel ð2Þ

in which

μ plasma the plasma viscosity (Pa·s) and

μ rel the relative viscosity defined by:

"
2 #
2
ζ
ð1− H Þ 2R 2R
μ rel ¼ 1 þ ðμ 45 −1Þ ð3Þ
ð1− 0:45Þζ −1 2R−1:1 2R−1:1

where

μ 45 the relative apparent blood viscosity for a fixed hematocrit of

0.45 given by Fig. 2. Schematic of blood flow bifurcation.
46 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56

for a time step (Δt) is assumed to be proportional to the stimulus term, The ϕV term is governed by Starling's law (Fung, 1986; Salathe and
its initial radius, and the time step (Pries et al., 1998): An, 1976):

ΔR ¼ Stot  R  Δt: ð8Þ LP S

ϕV ¼ ðP V −P i −σ ðπV −πi ÞÞ ð13Þ
V
Stot [1/s] includes the influences of the wall shear stress (Swss), the
intravascular pressure (Sp), and a metabolic mechanism depending on where
the blood hematocrit (Sm) (Pries et al., 1998):
Pi the IFP [Pa],
Stot ¼ Sh þ Sm −ks ð9Þ PV the IBP [Pa],
πV the osmotic pressure of the plasma [Pa],
where πi the osmotic pressure of the interstitial fluid [Pa],
Lp the hydraulic conductivity of the vessel wall [m/Pa·s],
Sh the hemodynamic stimuli [1/s] and defined by: σ the average osmotic reflection coefficient for the plasma
  proteins, and
Sh ¼ log10 τw þ τ re f −kp log10 τe ð10Þ S
V the surface area per unit volume of tissue for transport in the
interstitium [1/m].

The lymphatic drainage term distributed uniformly in normal tissues

Sm the metabolic stimuli [1/s] and defined by: is defined as Eq. (14):


Q re f LPL SL
Sm ¼ km log10 þ1 : ð11Þ ϕL ¼ ðP i −P L Þ ð14Þ
QV H V

In Eqs. (9) to (11), τw is the wall shear stress in a capillary vessel and where
4Q μ 
V app
calculated by πR3
,τref is a small constant included to avoid singular LPL SL
the lymphatic filtration coefficient [1/Pa·s], and
V
behavior at low wall shear stress, τe, the wall shear stress resulting from PL the hydrostatic pressure of the lymphatic [Pa].
the blood pressure, is τe = 10 − 8.6 ⋅ exp[−5000 ⋅ [log10(log10PV)]5.4],
Q ref is the largest value of Q V (blood flow through vessel) in the net- The momentum balance for fluid flow in porous medium is as
work, kp and km are non-negative constants, and ks is the shrinking Eq. (15) (Vafai, 2011):
stimuli which shows that the vessel has a tendency to reduce in size
    
∂vi   2μ μ
in the absence of positive growth stimuli. The detail of these stimulus ρ
T
þ ðvi :∇Þvi ¼ ∇  −P i þ μ ∇vi þ ð∇vi Þ − ð∇  vi Þ − v þF
and their effects are mentioned in our previous work (Soltani and ∂t 3 K i
Chen, 2013). ð15Þ
The parameters used for the adaptation model are listed in Table 1.
where
Interstitial flow
K the permeability of the porous medium [m2],
The normal and tumor tissues have characteristics the same as ρ the interstitial fluid density [kg/m3],
porous media. IFF in tissues is defined by coupling the fluid flow μ the interstitial fluid viscosity [Pa·s], and
governing equations: continuity and momentum balance. The mass F the volume forces [N].
balance, continuity, equation for steady state incompressible flow in
the porous media with source and sink of mass are (Soltani and Chen, Since interstitial fluid is Newtonian and has low velocity through
2011): tissues, Eq. (15) in steady state condition is changed to Darcy's law by
neglecting the friction within the fluid and exchange of momentum
∇:vi ¼ ϕV −ϕL ð12Þ between the fluid and solid phases:
μ 
in which
∇P i ¼ − v: ð16Þ
K i
vi the IFV vector [m/s],
ϕV the fluid source term [1/s], and K/μ is defined as the hydraulic conductivity of the interstitium
ϕL the lymphatic drainage term [1/s]. κ [m2/Pa·s]. Rearranging Eq. (16) results in

In biological tissues, the two last terms signify the rate of fluid flow vi ¼ −κ∇P i : ð17Þ
per unit volume from blood vessels into the interstitial space and from
the interstitial space into lymph vessels, respectively. Combination of Darcy's law (Eq. (17)) and the continuity equation
(Eq. (12)) results in

Table 1 −∇  ðκ∇P i Þ ¼ ϕV −ϕL ð18Þ

Parameter values used in the adaptation process.

Parameter Value Reference when κ is constant, and the interstitial pressure can be expressed by
τref [Pa] 0.103 Pries et al. (2001b) 2
Qref [mm3/s] 4.87 ×10−3 Calculated based on network situation −κ ∇ P i ¼ ϕV −ϕL : ð19Þ
kp [1/s] 0.1 Stephanou et al. (2006)
km [1/s] 0.07 Stephanou et al. (2006) The material properties for tumor and normal tissues used for inter-
ks [1/s] 0.35 Stephanou et al. (2006)
stitial flow calculation are listed in Table 2.
 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56 47

Table 2 fluid across capillary membranes. The Starling's law introduced before
Material properties of normal and tumor tissues used in numerical simulations. is used to calculate transvascular flow as:
Parameter Baseline value Reference
k
LP [m/Pa s] Normal 2.7 × 10−12 Soltani and Chen (2011) Q T ¼ 2πRL Lp ðP V −P i −σ ðπV −πi ÞÞ: ð21Þ
Tumor 21.1 × 10−12
κ [m2/Pa s] Normal 6.41 × 10−15 Soltani and Chen (2011) PV and Pi are the intravascular pressure through each capillary
Tumor 30.0 × 10−15
S/V [m −1
] Normal 7000 Soltani and Chen (2011)
and surrounding interstitial pressure, respectively. The detail of Q kC
Tumor 20,000 calculation is mentioned in our previous work (Soltani and Chen,
PV [Pa] for approach Normal 2080 Soltani and Chen (2011) 2013). To calculate these values, staggered grid as shown in Fig. 4 is
without a network Tumor 2080 considered. This grid which is the same as grid traditionally used in
πV [Pa] Normal 2666 Soltani and Chen (2011)
computational fluid dynamic, includes a main grid (shown in Fig. 4
Tumor 2666
πi [Pa] Normal 1333 Soltani and Chen (2011) by blue line) used for intravascular pressure and a virtual network
Tumor 15 (shown in Fig. 4 by dashed line) used for interstitial pressure. In fact,
σ Normal 0.91 Soltani and Chen (2011) in this grid the interstitial pressure is found at the center of each cell
Tumor 0.82 and intravascular pressure is calculated in the node of the grid. This
PL [Pa] Normal 0 Pishko et al. (2011)
LPL SL/V [1/mm Hg s] Normal 1 × 10−7 Pishko et al. (2011)
method gives physical acknowledgment of simultaneous solution of
intravascular and interstitial flow; however, the results of this method
do not differ significantly from the results of the method with collocated
grid for intravascular and interstitial pressure.
The interstitial pressure for peripheral tissue of a vascular network
Interstitial flow and intravascular flow coupling
is calculated by solving Eq. (19). ϕV is only considered to the right-
hand side of Eq. (19) wherever there is a capillary (Soltani and
Blood flow in microvascular network is solved by finding pressure at
Chen, 2013). The uniform distribution of ϕL is considered only in
interconnecting points in a network (Soltani and Chen, 2013). Intersti-
normal tissues (Baxter and Jain, 1989). The mathematical form can be
tial flow is simulated by solving Darcy's equation in the considered
written as:
domain and finding interstitial pressure in each node of the numerical
grid. 8(
>
>
2
−∇ P i ¼ M−N Normal tissue
IBP is calculated by applying volumetric flow rate conservation law >
> for existence of blood source
< 2
−∇ P i ¼M Tumor tissue
at interconnecting points (for example “C” (Fig. 3)) in the network (
> −∇2 P
> ¼ −N Normal tissue
>
> i otherwise
N 
X  : 2
k −∇ P i ¼0 Tumor tissue
QC ¼0 ð20Þ
k¼1
ð22Þ

in which where

LP S
k the index which refers to adjacent nodes and N is the number M : kV
ðP V −P i −σ ðπ V −πi ÞÞ, and
of adjacent nodes,
Q kC the net flow rate for each capillary which includes the flow LPL SL
N : ðP i −P L Þ:
through the capillary and transvascular flow from each V
capillary.

The flow rate through capillary is calculated by Poiseuille's law. The
transvascular flow rate is calculated by the Starling's law, which repre-
sents the role of hydrostatic and oncotic pressures in the movement of
Solute transport
Solute transport is due to diffusion and convection mechanisms in
porous media. The diffusion mechanism is obtained by first Fick's law
and the convection mechanism is calculated by product of velocity
and concentration (Truskey et al., 2004). Then, by applying the second

Fig. 3. Schematic of intravascular and transvascular flow in an interconnecting point. Fig. 4. Schematic of staggered grid used in numerical simulation.
48 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56

Fick's law, the mass conservation is obtained (Swabb et al., 1974; Wu Since lack of lymphatic system in tumor tissue (Jain and Fenton,
et al., 2013; Zhao and Toksoz, 1994) (Eq. (23)): 2002; Jain et al., 2007; Leu et al., 2000), the solute transport rate across
the lymphatic vessels can be considered as (Pishko et al., 2011):
∂C
¼ −∇  J ð23Þ 
∂t ϕL C Normal Tissue
ΦL ¼ : ð29Þ
0 Tumor Tissue
where
Finally, the general form of solute transport in tissue is:
C the solute concentration [kg/m3], and
J the solute mass flux defined by Eq. [kg/m2·s] (24) (Swabb
et al., 1974) ∂C 2
¼ Deff ∇ C − ∇  ðv i C Þ
∂t |fflfflfflfflffl{zfflfflfflfflffl} |fflfflfflfflffl{zfflfflfflfflffl}
J ¼ −Deff ∇C þ vi C: ð24Þ Diffusion transport in interstitium Convection transport in interstitium

0 1
BLP S   C
Eq. (24) includes two mechanisms of transport, the first term is þB
PS Pe C
@ V ðP V −P i −σ ðπb −πi ÞÞ 1−σ f C P þ V ðC P −C Þ ePe −1 A− ΦL
|{z}
:
diffusion flux caused by concentration gradient (first Fick's law) and |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Lymphatic term
Convection rate from or to vessel
the second term is convection flux caused by fluid flow velocity. The Diffusion rate from or to vessel

Deff [m2/s] is the effective diffusion coefficient. By substituting Eq. (24) ð30Þ
into Eq. (23), Eq. (25) is obtained:

∂C h i In the modeling of solute distribution in tissue with capillary

¼ ∇  Deff ∇C −∇  ½vi C : ð25Þ network, the third and fourth terms of Eq. (30) are implemented
∂t
wherever the capillary exists and for other places these terms are
zero. To show this more systematically, the mathematical form of our
By assuming that the solute transport does not affect the density
elaborate solute transport model in tissue is given by:
of interstitial fluid and consequently on the interstitial fluid flow, the
velocity field, vi, is calculated independent of solute transport. 88
> > ∂C
This equation is acceptable for porous media when there is no fluid >>
> 2
> < ¼ Deff ∇ C−∇  ðvi C Þ þ ΦV −ΦL Normal tissue
>
> ∂t
>
> for existence of blood source
source or sink in the medium, but biological tissues have sources and > > ∂C
<>
> 2
: ¼ Deff ∇ C−∇  ðvi C Þ þ ΦV Tumor tissue
sinks for solute; therefore, the solute transport equation in this case 8 ∂t
>
>>> ∂C
can be written as (Sefidgar et al., 2014): >
> < 2
¼ Deff ∇ C−∇  ðvi C Þ−ΦL Normal tissue
>
> ∂t
>
> otherwise
> ∂C
:>
> >
: 2
¼ Deff ∇ C−∇  ðvi C Þ Tumor tissue:
∂C h i ∂t
¼ ∇  Deff ∇C −∇½vi C  þ ΦV −ΦL ð26Þ ð31Þ
∂t

where
ΦV the rate of solute transport per unit volume from blood
vessels into the interstitial space [kg/m3s], and
ΦL the rate of solute transport per unit volume from the
interstitial space into lymph vessels [kg/m3s].
Same as solute transport in tissue, the transvascular solute from
vessel wall to interstitium occurs in the form of both diffusion and
convection. The pressure gradient which causes fluid movement leads
to convective transport of solute and diffusion transport is caused by
concentration gradient between plasma and interstitium. Patlak et al.'s
(1963) equation which is the most complete description of solute trans-
port from vessel to tissue is used in this study:
   
S Pe
ΦV ¼ P ðC P −C Þ Pe þ ϕV 1−σ f C P
V e −1
The parameters of solute transport model taken from Baxter and Jain
(1989) and are listed in Table 3. Although, the numerical model is
applicable for any type of macromolecular therapeutic agents, in the
present study, the properties of Fragment antigen-binding (F(ab′)2) as
a sample of macromolecule is used.
Geometry and boundary conditions
The 2D domain (shown in Fig. 5) is considered for the numerical
model.
Boundary conditions
Intravascular and interstitial flow. IBP is 3325 Pa (25 mm Hg) for inlet
and 1330 Pa (10 mm Hg) for outlet. The boundary condition used
for interstitial flow is shown in Fig. 5. A symmetry boundary condition
ð27Þ
is considered for the right, upper, and lower edges of the domain, i.e.,

where: ∇P i ¼ 0: ð32Þ

P the diffusive permeability [m/s],

CP the plasma concentrations of the solute,
σf the solvent-drag reflection coefficient, and
Table 3
Pe referred to as the Peclet number, and is given by:
Parameters of solute transport model used in numerical simulations.
 
ϕV 1−σ f Parameter Baseline value Reference
Pe ¼ : ð28Þ σf Normal 0.9 Baxter and Jain (1989)
PS=V
Tumor 0.9
Deff [m2/s] Normal 0.16 × 10−12 Baxter and Jain (1989)
The Peclet number indicates the ratio of convection to diffusion Tumor 2.0 × 10−12
across the microvessel wall to interstitium. In Eq. (27), the first term P [m/s] Normal 2.2 × 10−10 Baxter and Jain (1989)
in the right hand side is the diffusion part and the second term is the Tumor 17.3 × 10−10
τ [h] Plasma 6.1 Baxter and Jain (1989)
convection part.
 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56 49

Fig. 5. Schematic of domain and boundary conditions used in numerical simulation. A circular tumor surrounded by normal tissues is considered in this study.

For the left edge of the domain where the interstitial pressure is section. The initial radius of each capillary segment is 6 μm. During the
constant; the relative interstitial pressure is considered to be zero remodeling process, allowable radius changes are in the range of 3 μm
(Soltani and Chen, 2011): to 12 μm (Ciofalo et al., 1999).
The Cp in this study is considered based on two types of injection:
P i ¼ 0: ð33Þ bolus and continuous injections. In bolus injection, the plasma concen-
tration decreases with time exponentially Cp = C 0pe− ln(0.5)t/τ, in which τ
The continuity of interstitial pressure and interstitial velocity (pres- is the drug half-life in plasma. The continuous injection leads to a
sure gradient) are considered for the boundary between the tumor and constant plasma concentration (Cp = constant).
normal tissues:
Algorithm of numerical simulation
−κ t ∇P i jΩ− ¼ −κ n ∇P i jΩþ
ð34Þ
P i jΩ − ¼ P i jΩ þ
Three approaches are used to investigate DCIF distribution. The
− +
where Ω and Ω indicate the tumor and normal tissues at the outer general algorithm of numerical simulation is shown in Fig. 6. In the
edge of the solid tumor; the t and n indices show the tumor and normal first approach, the governing equations of interstitial flow and solute
tissues, respectively. transport are solved in normal and tumor tissues without a capillary
network. The uniform and constant IBP is considered, in the first
Solute transport. The boundary condition required for solute transport approach. IBP in the second approach is calculated in the static network,
equation is shown in Fig. 5. The right, upper, and lower edges of the iteratively. The capillaries diameters are considered constant during
domain are in symmetry (Soltani and Chen, 2011): calculation in the static network. For the third approach, IBP is calculated
in a dynamic network in which the capillaries adapt their diameters
Deff ∇C þ vi C ¼ 0: ð35Þ based on signals received from metabolic and hemodynamic stimuli. In
the second and third approaches, the interstitial flow is simultaneously
Same as interstitial flow, the continuity of concentration and its solved with intravascular flow and then the drug delivery is obtained
flux are considered for the boundary between tumor and normal in the domain.
tissues: Since the third approach is a comprehensive model in this study,
 
 
the calculation procedure of this approach is described in more detail
t
n
here:
Deff ∇C þ vi C
− ¼ Deff ∇C þ vi C
−
Ω Ω ð36Þ By considering initial guesses for P0V, P0i and R0 (mentioned in Initial
CjΩ− ¼ CjΩþ :
condition section), the IBP (PV) at each node of microvascular network is
calculated based on a set of non-linear equations by iterative SOR
The open boundary condition is used for the left edge of domain. The
algorithm. The calculated PV is used to find IFP (Pi). The Pi is calculated
open boundary is used to set up mass transport across the boundaries
by solution of Eq. (22) with numerical method described below.
where both convective inflow and outflow can occur and this open
The calculated IBP and IFP are used to update the vessels radius,
boundary condition is defined by Eq. (37):
apparent viscosity, and hematocrit in vessels based on Eqs. (2), (6),
−n  ∇C ¼ 0 ð37Þ and (8). The relative error for PV, Pi and R is calculated by max(XN −
X0)/X0, where X can be each of PV, Pi and R. If the maximum relative
where error is lower than the threshold (10− 6), the solution procedure of
IBP and IFV is stopped and these values are used to calculate DCIF.
n the normal vector. Otherwise, the procedure is repeated until finding new values of IBP
and IFP.
Initial condition The DCIF equation (Eq. (31)) and also interstitial pressure equation
P0i and C0 are initialized as zero, and P0V is considered as 1330 Pa, (Eq. (22)) are solved by an element based finite volume method
based on boundary conditions explained in Boundary conditions (EB-FVM). The EB-FVM has the capability of the finite element method
50 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56

Fig. 6. Algorithm of three approaches of DCIF calculation used in this study.

(FEM) in handling complex geometries and also the sound physical-
based properties of the finite volume method (FVM) (Soltani and
Chen, 2009). The discretized form of the governing equations, in their
general form, is then linearized and solved implicitly. Finally, the
converged form of the solution is calculated using an iterative method.
The criterion for the convergence is to reduce the residual by 4 orders
of magnitudes. In order to check the grid independency of the code,
the results for three different grids are compared, indicating the
conservative property of the numerical method. Final choice of the
grid includes 26,794 elements.
and blood flow simulation through the capillary network (shown in
Fig. 7) for the second and third approaches are investigated in this
section. The contour of IBP in capillary network for the second and
third approaches is shown in Fig. 8. IBP is non-dimensionalized by the
maximum pressure in the network which is the inlet pressure in the
parent vessel (25 mm Hg). Vessels which do not receive enough flow
rate (less than 0.01 maximum flow rate of the network) are pruned

Results

Network growth simulation

Capillary network obtained by discrete angiogenesis model is

presented in Fig. 7. The growth of capillary network is simulated by 9
sprouts as initial condition. This network is used in the second and
third approaches for IBP calculation. Since the mathematical model
used for network generation is not a physical model, all vessels do not
make a loop in the network. Those vessels that are not in such a loop
situation are removed by a mathematical pruning method before
simulation of blood flow through the network is started.

Fluid flow

Results of fluid flow simulation in normal and tumor tissues for three
approaches mentioned in Algorithm of numerical simulation section Fig. 7. Results of discrete sprouting angiogenesis.
 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56 51

Fig. 8. Contour of IBP for static and dynamic networks.

Fig. 9. Radius distribution in static and dynamic networks.

from the network. The numbers of pruned vessels from the static
network are much more than that of the dynamic network because
the dynamic network adapts its vessels diameter to reach enough
blood flow through each vessel. IBP of the dynamic network, the third
approach, is higher than the IBP of the static network especially near
the tumor because the flow resistance in capillaries is reduced by
adaption method in the dynamic network.
The effect of dynamic modeling on the network structure is shown
in Fig. 9. In Fig. 9, the distribution of vessel radius is compared between
the static and dynamic network. The static approach is simulated by
considering initial radius of dynamic network (6 μm) for all vessels in
the network. As mentioned, the radius of dynamic network is changed
between 3 and 12 μm. The low radius density (10.5 μm) shows that
vessels at this radius are unstable and they attend to higher diameter
to response to the hemodynamic stimuli (Shafer et al., 2011). Also,
the same phenomena are reported in the work of Stephanou et al.
(2006).
IFP distribution for three approaches is shown in Fig. 10. Calculated
IFP is elevated in the tumor because of high leaky vessel and lack of
lymphatic system. In the first approach, IFP is uniform within tumor
but in the second and third approaches, the predicted IFP is heteroge-
neous in the tumor. The heterogeneous fluid source is the main reason
of non-uniform IFP distribution for simulations with network (the
second and third approaches). IFP of the dynamic network has
the highest value among all approaches because of the elevated IBP in
the tumor which increases the transvascular flow rate. IFP value is
decreased rapidly at the tumor boundary as the transvascular flow is
absorbed by lymph vessels in the normal area. IFP profiles along the
lines shown in Fig. 11 are illustrated in Fig. 12. The tumor boundary is
shown by a black dashed line in Fig. 12.

Fig. 10. The contour of IFP in the domain for both normal and tumor tissues in three approaches.
52 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56

tumor for simulations with capillary network. The highest IFV value
predicted in the tumor region is 0.3 μm/s, which is almost match to
the value reported by Pishko et al. (2012) and near the values measured
by Butler et al. (1975).

Drug concentration

DCIF distribution for three approaches is shown in Fig. 15 at 1, 8, and

Fig. 11. Lines in which results are shown.
IFV contours in the domain and its profile along line 1 are shown in
Fig. 13 and Fig. 14, respectively. IFV has its highest value at or close to
the tumor boundary and outwardly directed from the tumor. IFV for
simulations with network is heterogeneous and non-zero in the tumor
region in comparison to the model without a network. As required by
Darcy's law (Eq. (17)), non-uniform pressure profiles and existing
pressure gradient in tumor region lead to non-zero velocities in the
20 h post bolus injection. DCIF are non-dimensionalized by Cp for con-
tinuous injection and C0p for bolus injection, respectively. In comparison
of uniform distribution of DCIF in results of the first approach, DCIF is
heterogeneous in the second and third approaches. Since in early time
drug delivery depends on vessel source terms and transport by diffusion
and convection is narrow, DCIF in the approaches with network (with
spatially varying source terms instead of having uniform source
terms) is low. The results of blood flow distribution for static network
leads to an unrealistic distribution of DCIF; therefore, the static
approach is not considered in the rest of this paper. The low concentra-
tion area near the tumor core is predicted in the approaches with
network because of the lack of drug source and limited interstitial
transport (diffusion and convection) in the tissue. The DCIF in normal
tissue has low and smooth distribution. This is due to the lower vascular
permeability and presence of lymphatic vessels.
Fig. 16 compares average DCIF in a tumor during calculation for
approaches without a network and with dynamic network. The drug
concentration initially increases with time due to increasing plasma
concentration and source terms of drug extravasation. The maximum
value of average DCIF occurs at 8 h post injection; however, this

Fig. 12. IFP profile along line 1.

Fig. 13. The contour of IFV in the domain for both normal and tumor tissues in three approaches.
 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56 53

Fig. 14. IFV profile along line 1.

Fig. 16. The average DCIF in tumor region in three approaches.

Fig. 15. The contour of DCIF in the domain at different times for both normal and tumor tissues with bolus injection in three approaches.
54 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56

time occurs later in simulations with network. The DCIF value in the
approach without a network is higher than DCIF value in dynamic
network approach. The continuous and uniform distribution of the
source terms in the without a network model leads to a higher and
more uniform distribution of DCIF in the tumor region. Result shows
that the heterogeneous capillary network decreases the drug delivery
to the solid tumor.
The distribution of DCIF for a dynamic network is also calculated
for a continuous injection. Fig. 17 shows DCIF distribution at 1, 8, 20,
and 72 h post continuous injection. In spite of well and continuous
perfusion, DCIF has a heterogonous distribution even after 72 h. Results
show that drug takes a long time to reach in tumor region where there is
no vessel.
Fig. 18 and Fig. 19 show DCIF along the lines shown in Fig. 11
for bolus and continuous injections, respectively. The tumor region is
specified with a dashed line in Fig. 18 and Fig. 19. The “r” in these figures
shows the distance from center of the tumor (Fig. 11) and is non-
dimensionalized by the domain length. The heterogeneity of DCIF
distribution is clearly illustrated for both types of injection. At the
earlier time (1 h post injection), the drug concentration only depends
Fig. 18. The DCIF distribution along specified lines shown in Fig. 11 for bolus injection.
on transvascular rate of drug to tissue. The highest value of DCIF
is taken place close to the tumor boundary and DCIF in the center
of the tumor has the lowest value most of the time. DCIF along line Discussion
2 has its lowest value because of the lowest blood source in this
direction. DCIF along line 1 is higher and more uniform than DCIF Current work presents the first solute transport study in a dynamic
along other lines since the vessel density along this line is more than capillary network induced by a tumor, within normal and tumor tissues.
vessel density in other lines. DCIF in the normal tissue is affected by This model simultaneously predicts intravascular blood flow, interstitial
the lymphatic vessels as a sink term; therefore, DCIF is the same along fluid flow, and solute transport within tissues. The mathematical
all lines. model used in previous studies for generating the capillary network

Fig. 17. The contour of DCIF in the domain at different times for both normal and tumor tissues with continuous injection in the dynamic network.
 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56
Fig. 19. The DCIF distribution along specified lines shown in Fig. 11 for continuous
injection.
(Stylianopoulos et al., 2013; Chaplain et al., 2012; Cai et al., 2011) is used
here to generate the heterogeneous vasculatures.
The IBP in the vessel near the tumor for a dynamic network simula-
tion is 1.3 times greater than IBP of a static network (Fig. 8). This is
due to low resistance of blood flow from parent vessel to vessels near
the tumor because of better distribution of vessels radius in network.
The ratio of lowest IBP to highest IBP is 0.4 in the network which is
close to the prediction of Cai et al. (2011) which is reported around
0.3. Also, the IBP distribution pattern for static network is similar
to the results of Stylianopoulos and Jain (Supplementary Fig. 4s)
(Stylianopoulos and Jain, 2013). The larger vessel radius in the dynamic
network near the tumor region (shown in Fig. 9) leads to low blood flow
resistance and increases blood pressure around the tumor. The metabolic
and hemodynamic stimuli lead to increase vessel radius for conducting
blood flow to the tumor area.
The higher value of IBP has both desirable and undesirable effects on
drug delivery. In one hand, the IBP increases convection rate of drug
from vessels based on the third term of Eq. (30). On the other hand, it
also increases the transvascular flow based on Eq. (13) which leads to
increase IFP in tumor domain which is one of the main barriers of
drug delivery to the tumors. This effect will be discussed later in more
detail.
The interstitial fluid flow results based on the first approach
(without a network) shows that the IFP is at its maximum value at the
center of the tumor and is uniform and symmetry in tumor tissue.
Boucher et al. (1990) measured the IFP in different depth of a tumor.
The IFP profiles predicted in this study is consistent with their
experimental observations (Fig. 3 in Boucher et al., 1990). Huber et al.
(2005) measured the IFP for different tumors between 1.1 kPa and
1.8 kPa. The maximum IFP for all approaches in current study is
1.53 kPa which is very close to what they reported. The predicted IFP
for the second and third approaches is proportional to the vascular den-
sity and shows heterogeneity in the tumor region. The heterogeneous
microvessels as source terms in interstitial flow equation (Eq. (22))
cause non-uniform IFP distribution in the tumor region. This effect is
also seen by Stylianopoulos and Jain (Fig. 2 of Stylianopoulos et al.,
2013), and Wu et al. (Fig. 6 of Wu et al., 2013). Since the dynamic
approach predicts the higher IBP in tumor region than the static
approach does and vessel in tumor region is so leaky and is not capable
to withstand the intravascular and oncotic pressure gradients across the
microvessel wall, which leads to increase the drainage rate of vessel to
tissue, the IFP for the third approach is around 1.3 times greater than
IFP of the first and second approaches. The higher IFP for dynamic
55
network shows the significant dependency of intestinal fluid flow to
vasculature distribution and intravascular fluid flow. This effect is also
shown by Netti et al. (1996).
IFV within the tumor is predicted to be on the order of 0.2–0.3 μm/s
for three approaches which is supported by experimental results of
Hompland et al. (2013) and simulation results of Jain et al. (2007).
The maximum velocity happens at the tumor boundary. The IFV distri-
bution is non-uniform in approaches with static and dynamic networks.
Based on Darcy's law (Eq. (17)), IFV depends on IFP gradient in tissues.
Since IFP has heterogeneous distribution for approaches with network,
IFP gradient and consequently IFV have non-uniform distribution in
tumor region. This effect is presented in studies of heterogeneous
tumor in works of Zhao et al. (2007) and Pishko et al. (2012).
DCIF distribution for the first approach, without a network (Fig. 15),
is uniform in tumor region which this pattern is seen in work of Jain and
Baxter (1988), Baxter and Jain (1989) and Arifin et al. (2007). The
concentration decreasing in the interstitium is the result of plasma
concentration reduction below the interstitial concentration due to
plasma clearance. The fourth term of Eq. (30) shows the mathematical
reason of this effect. Drug delivery, especially at the tumor region in
the first approach, is more dependent on diffusion rate from microvessel
and diffusion transport in tissue than convection rate from microvessel
and convection transport in tissue. The uniform (smooth) IFP leads to
zero pressure gradient and zero velocity (Eq. (17) and Fig. 13), there-
fore, the convection transport which depends on the velocity has
negligible effect in drug delivery, also elevated IFP in tumor region limits
the convective drug transport from vessel to tissue (the third term
of Eq. (30)). The unrealistic assumption of uniform distribution of
vasculature in the considered domain for the first approach causes
that DCIF for the first approach is more than DCIF for the approaches
with a network.
The DCIF contour for dynamic network illustrated in Fig. 15 and
Fig. 17 shows that the irregular vascular network causes heterogeneous
drug distribution. In well injection system (continuous injection), the
drug delivery needs about 72 h to distribute in all regions of the
tumor. Since the diffusion transport of drug in tissue is so slow (because
of low diffusion coefficient) and convection transport is negligible
(because of low velocity in tumor region), the drug delivery depends
on the leakage rate from vessels. Because of these reasons, in tumor
region without vessel such as necrosis or hypoxia in real physiology of
a tumor, the drug delivery needs continuous injection and spending
time for reaching drug to these regions and for rapid drug injection
(bolus injection) (Carlson and Sikic, 1983) drug delivery to these
regions is not successful (Fig. 15 and Fig. 17).
The highest value of DCIF is taken place near the tumor boundaries
(Fig. 18 and Fig. 19). In the tumor boundaries, IFV has its highest value
and IFP is decreased; therefore, the convection transport in tissue and
convection rate from microvessel are increased.
As mentioned before, the elevated IBP in the tumor region for
approach with dynamic capillary network can increase the convection
rate from vessel wall (the third term in the right hand side of
Eq.(30)). However, the results (Figs. 15, 17 to 19) show that the
undesirable effect of elevated IBP (elevated IFP in tumor) has more
effect on drug delivery than positive effect of elevated IBP (increase
convection rate). Consequently, this undesirable effect decreases drug
transport from vessel to tissue and also convection transport in tumor
tissue.
As shown in Fig. 18 and Fig. 19, in the along line 2, the microvessels
have the lowest density in all considered directions and DCIF has its
lowest value in this direction. The dependency of DCIF to microvessel
distribution indicates one of drug delivery barriers to solid tumors.
Because of the complexity of network in our model, the drug delivery
to solid tumor is not done properly as introduced in previous study
(Baish et al., 2010; Chaplain et al., 2012).
The multi-scale numerical method used in this study indicates that
drug delivery to solid tumors depends on the tumor induced capillary
56 M. Sefidgar et al. / Microvascular Research 99 (2015) 43–56
network structure. The dynamic approach which uses more realistic as-
sumptions than other two approaches, changes the capillary network
structure based on signals sent by hemodynamic and metabolic stimuli.
This approach generates the more irregular capillary network around
the tumor and predicts a higher IFP in the tumor region. This elevated
IFP with irregular capillary network leads to a heterogeneous distribu-
tion of DCIF in the tumor region similar to in vivo observations.
References
Anderson, A.R., Chaplain, M.A., 1998. Continuous and discrete mathematical models of
tumor-induced angiogenesis. Bull. Math. Biol. 60, 857–899.
Anderson, A.R., Chaplain, M.A., Mcdougall, S., 2012. A hybrid discrete-continuum model of
tumour induced angiogenesis. In: Jackson, T.L. (Ed.), Modeling Tumor Vasculature.
Springer, New York, NY, pp. 105–134.
Arifin, D.Y., Wang, C., Smith, K.A., 2007. Patient-specific chemotherapeutic drug delivery
to brain tumors brain tissue tumor ventricle. Mimics Innov. Awardpp. 1–9.
Arifin, D.Y., Lee, K.Y.T., Wang, C.-H., 2009. Chemotherapeutic drug transport to brain
tumor. J. Control. Release 137, 203–210.
Baish, J.W., Stylianopoulos, T., Lanning, R.M., Kamoun, W.S., Fukumura, D., Munn, L.L., Jain,
R.K., 2010. Scaling rules for diffusive drug delivery in tumor and normal tissues. PNAS
108, 1–5.
Baxter, L.T., Jain, R.K., 1989. Transport of fluid and macromolecules in tumors. (I) Role of
interstitial pressure and convection. Microvasc. Res. 37, 77–104.
Baxter, L.T., Jain, R.K., 1990. Transport of fluid and macromolecules in tumors. (II) Role of
heterogeneous perfusion and lymphatics. Microvasc. Res. 40, 246–263.
Baxter, L.T., Jain, R.K., 1991. Transport of fluid and macromolecules in tumors III. Role of
binding and metabolism. Microvasc. Res. 41, 5–23.
Boucher, Y., Baxter, L.T., Jain, R.K., 1990. Interstitial pressure gradients in tissue-isolated
and subcutaneous tumors: implications for therapy. Cancer Res. 50, 4478–4484.
Butler, T.P., Grantham, F.H., Gullino, P.M., 1975. Bulk transfer of fluid in the interstitial
compartment of mammary tumors. Cancer Res. 35, 3084–3088.
Cai, Y., Xu, S., Wu, J., Long, Q., 2011. Coupled modelling of tumour angiogenesis, tumour
growth and blood perfusion. J. Theor. Biol. 279, 90–101.
Carlson, R.W., Sikic, B.I., 1983. Continuous infusion or bolus injection in cancer chemo-
therapy. Ann. Intern. Med. 99, 823–833.
Chaplain, M.A.J., McDougall, S.R., Anderson, A.R., 2012. Blood flow and tumour-induced
angiogenesis: dynamically adapting vascular networks. In: Jackson, T.L. (Ed.), Modeling
Tumor Vasculature. Springer, New York, NY, pp. 167–212.
Chauhan, V.P., Stylianopoulos, T., Martin, J.D., Chen, O., Kamoun, W.S., Bawendi, M.G.,
Fukumura, D., Jain, R.K., 2012. Normalization of tumour blood vessels improves the
delivery of nanomedicines in a size-dependent manner. Nat. Nanotechnol. 7,
383–388.
Ciofalo, M., Collins, M.W., Hennessy, T.R., 1999. Microhydrodynamics phenomena in the
circulation. In: Mala, N. (Ed.), Nanoscale Fluid Dynamics in Physiological Processes,
A Review Study. WIT Press, pp. 219–236.
Dreher, M., Liu, W., Michelich, C., Dewhirst, M., Yuan, F., Chilkoti, A., 2006. Tumor vascular
permeability, accumulation, and penetration of macromolecular drug carriers. J. Natl.
Cancer Inst. 98, 335–344.
Fung, Y.C., 1986. Blood rheology in microvessels. In: He, X.X. (Ed.), Biomechanics—
Mechanical Properties of Living Tissues. Science Technique Publisher, China.
Fung, Y.C., 1993. Biomechanics Mechanical Properties of Living Tissues. second. ed.
Springer.
Goel, S., Duda, D.G., Xu, L., Munn, L.L., Boucher, Y., Fukumura, D., Jain, R.K., 2011. Normal-
ization of the vasculature for treatment of cancer and other diseases. Physiol. Rev. 91,
1071–1121.
Lipowsky, Herbert H., 2007. Blood rheology aspects of the microcirculation. In: Baskurt,
O.K. (Ed.), Handbook of Hemorheology and Hemodynamics. IOS Press, pp. 307–321.
Hompland, T., Gulliksrud, K., Ellingsen, C., Rofstad, E.K., 2013. Assessment of the Interstitial
Fluid Pressure of Tumors by Dynamic Contrast-enhanced Magnetic Resonance Imaging
with Contrast Agents of Different Molecular Weights 627–635.
Huber, P.E., Bischof, M., Heiland, S., Peschke, P., Saffrich, R., Gro, H., Lipson, K.E., Abdollahi,
A., 2005. Trimodal cancer treatment: beneficial effects of combined antiangiogenesis,
radiation, and chemotherapy. Cancer Res. 65, 3643–3655.
Jain, R.K., 2001. Delivery of molecular and cellular medicine to solid tumors. Adv. Drug
Deliv. Rev. 46, 149–168.
Jain, R.K., Baxter, L.T., 1988. Mechanisms of heterogeneous distribution of monoclonal
antibodies and other macromolecules in tumors: significance of elevated interstitial
pressure. Cancer Res. 48, 7022–7032.
Jain, R.K., Fenton, B.T., 2002. Intratumoral lymphatic vessels: a case of mistaken identity or
malfunction? J. Natl. Cancer Inst. 94, 417–421.
Jain, R.K., Tong, R.T., Munn, L.L., 2007. Effect of vascular normalization by antiangiogenic
therapy on interstitial hypertension, peritumor edema, and lymphatic metastasis:
insights from a mathematical model. Cancer Res. 67, 2729–2735.
Leu, A.J., Berk, D.A., Lymboussaki, A., Alitalo, K., Jain, R.K., 2000. Absence of functional
lymphatics within a murine sarcoma: a molecular and functional evaluation. Cancer
Res. 60, 4324–4327.
Netti, P.A., Roberge, S., Boucher, Y., Baxter, L.T., Jain, R.K., 1996. Effect of transvascular fluid
exchange on pressure–flow relationship in tumors: a proposed mechanism for tumor
blood flow heterogeneity. Microvasc. Res. 52, 27–46.
Patlak, C.S., Goldstein, D.A., Hoffman, J.F., 1963. The flow of solute and solvent across a
two-membrane system. J. Theor. Biol. 5, 426–442.
Pishko, G.L., Astary, G.W., Mareci, T.H., Sarntinoranont, M., 2011. Sensitivity analysis of an
image-based solid tumor computational model with heterogeneous vasculature and
porosity. Ann. Biomed. Eng. 39, 2360–2373.
Pishko, G.L., Astary, G.W., Zhang, J., Mareci, T.H., Sarntinoranont, M., 2012. Role of convec-
tion and diffusion on DCE-MRI parameters in low leakiness KHT sarcomas. Microvasc.
Res. 84, 306–313.
Pozrikidis, C., 2010. Numerical simulation of blood and interstitial flow through a solid
tumor. J. Math. Biol. 60, 75–94.
Pries, A.R., Secomb, T.W., 2008. Blood flow in microvascular networks. Direct 3–36.
Pries, A.R., Secomb, T.W., 2010. Blood flow in microvascular networks. In: Tuma, R.F.,
Duran, W.N., Ley, K. (Eds.), Microcirculation. Academic Press.
Pries, A.R., Neuhaus, D., Gaehtgens, P., 1992. Blood viscosity in tube flow: dependence on
diameter and hematocrit. Am. J. Philol. 263, 1770–1778.
Pries, A.R., Secomb, T.W., Gaehtgens, P., 1996. Biophysical aspects of blood flow in the
microvasculature. Cardiovasc. Res. 32, 654–667.
Pries, A.R., Secomb, T.W., Gaehtgens, P., 1998. Structural adaptation and stability of micro-
vascular networks: theory and simulations. Am. J. Physiol. Heart Circ. Physiol. 275,
349–360.
Pries, A.R., Reglin, B., Secomb, T.W., 2001a. Structural adaptation of vascular networks:
role of the pressure response. J. Am. Hear. Assoc. 38, 1476–1479.
Pries, A.R., Reglin, B., Secomb, T.W., 2001b. Structural adaptation of microvascular
networks: functional roles of adaptive responses. Am. J. Physiol. Heart Circ. Physiol.
281, 1015–1025.
Salathe, E., An, K., 1976. mathematical analysis of fluid movement across capillary walls.
Microvasc. Res. 11, 1–23.
Secomb, T.W., Alberding, J.P., Hsu, R., Dewhirst, M.W., Pries, A.R., 2013. Angiogenesis: an
adaptive dynamic biological patterning problem. PLoS Comput. Biol. 9, 1–12.
Sefidgar, M., Raahemifar, K., Bazmara, H., Bazargan, M., Mousavi, S.M., Soltani, M., 2014a.
Effect of remodeled tumor-induced capillary network on interstitial flow in cancerous
tissue. 2nd Middle East Conference on Biomedical Engineering. IEEE, pp. 212–215.
Sefidgar, M., Soltani, M., Bazmara, H., Mousavi, M., Bazargan, M., Elkamel, A., 2014b. Inter-
stitial flow in cancerous tissue: effect of considering remodeled capillary network.
J. Tissue Sci. Eng. 4, 1–8.
Sefidgar, M., Soltani, M., Raahemifar, K., Bazmara, H., Nayinian, S., Bazargan, M., 2014c.
Effect of tumor shape, size, and tissue transport properties on drug delivery to solid
tumors. J. Biol. Eng. 8, 12.
Shafer, I., Nancollas, R., Boes, M., Sieminski, A.L., Geddes, J.B., 2011. Stability of a
microvessel subject to structural adaptation of diameter and wall thickness. Math.
Med. Biol. 28, 271–286.
Soltani, M., Chen, P., 2009. Shape design of internal flow with minimum pressure loss.
Adv. Sci. Lett. 2, 347–355.
Soltani, M., Chen, P., 2011. Numerical modeling of fluid flow in solid tumors. PLoS ONE 6,
1–15.
Soltani, M., Chen, P., 2012. Effect of tumor shape and size on drug delivery to solid tumors.
J. Biol. Eng. 6, 4.
Soltani, M., Chen, P., 2013. Numerical modeling of interstitial fluid flow coupled with
blood flow through a remodeled solid tumor microvascular network. PLoS ONE 8,
e67025.
Soltani, M., 2012. Numerical Modeling of Drug Delivery to Solid Tumor Microvasculature.
PhD thesis, Chem. Eng. (Nanotechnology), Waterloo, Ontario, Canada.
Stephanou, A.S., Mcdougall, S.R.R., Anderson, A.R.A., Chaplain, M.A.J., 2006. Mathematical
modelling of the influence of blood rheological properties upon adaptative tumour-
induced angiogenesis. Math. Comput. Model. 44, 96–123.
Stylianopoulos, T., Jain, R.K., 2013. Combining two strategies to improve perfusion and
drug delivery in solid tumors. PANS 110, 18632–18637.
Stylianopoulos, T., Soteriou, K., Fukumura, D., Jain, R.K., 2013. Cationic nanoparticles have
superior transvascular flux into solid tumors: insights from a mathematical model.
Ann. Biomed. Eng. 41, 68–77.
Swabb, E.A., Wei, J., Gullino, P.M., 1974. Diffusion and convection in normal and neoplastic
tissues. Cancer Res. 34, 2814–2822.
Tamburrino, A., Piro, G., Carbone, C., Tortora, G., Melisi, D., 2013. Mechanisms of resistance
to chemotherapeutic and anti-angiogenic drugs as novel targets for pancreatic cancer
therapy. Front. Pharmacol. 4, 56.
Truskey, G.A., Yuan, F., Katz, D.F., 2004. Transport Phenomena in Biological Systems.
Pearson Education.
Vafai, K., 2011. Porous Media, Application in Biological Systems and Biotechnology. Taylor
and Francis.
Welter, M., Rieger, H., 2013. Interstitial fluid flow and drug delivery in vascularized
tumors: a computational model. PLoS ONE 8, e70395.
Wu, L., Gao, B., Tian, Y., Muñoz-Carpena, R., 2013a. Analytical and experimental analysis of
solute transport in heterogeneous porous media. J. Environ. Sci. Health A 49,
338–343.
Wu, M., Frieboes, H.B., McDougall, S.R., Chaplain, M. a J., Cristini, V., Lowengrub, J., 2013b.
The effect of interstitial pressure on tumor growth: coupling with the blood and
lymphatic vascular systems. J. Theor. Biol. 320, 131–151.
Zhao, X., Toksoz, M.N., 1994. Solute transport in heterogeneous porous media. Mass. Inst.
Technol. Earth Resour. Lab.
Zhao, J., Salmon, H., Sarntinoranont, M., 2007. Effect of heterogeneous vasculature on
interstitial transport within a solid tumor. Microvasc. Res. 73, 224–236.