ARTICLE IN PRESS

Journal of Biomechanics 37 (2004) 189–196

Experimental and theoretical investigation of directional permeability

of human vertebral cancellous bone for cement infiltration
G. Barouda,*, R. Falkb, M. Crookshankc, S. Sponageld, T. Steffena
a
Orthopaedic Research Laboratory, Division of Orthopaedic Surgery, McGill University, Royal Victoria Hospital(Rm L4.65) 687 Pine Ave West,
Montreal, QC, Canada H3A 1A1
b
Co-operative Education Program, Biological Sciences, Simon Fraser University, Vancouver, BC, Canada
c
Co-operative Education Service, Biological Engineering, University of Guelph, Guelph, ON, Canada
d
.
Department of Biomedical Engineering, University of Applied Science of Aachen (FH), Julich, Germany
Accepted 3 June 2003

Abstract

The use of acrylic polymers in infiltrating the porous bone structure is an emerging procedure for the augmentation of
osteoporotic vertebrae. Although this procedure is employed frequently, it is performed based on empirical knowledge, and
therefore, does not take into consideration the porosity-dependent permeability of human vertebral cancellous bone. The purpose of
this study was to: (a) experimentally and theoretically investigate interdependence of the vertebral cancellous bone permeability and
porosity, and (b) examine if the bone permeability of spinal cancellous bone can be predicted using bone mineral density
measurements. If these relations can be established, they can be useful in optimizing the injection conditions for predicable cement
infiltration. To determine the porosity-dependent and directional permeability, 34 bone cores—20 samples in the superior–inferior
(SI) direction and 14 in the anterior–posterior (AP) direction—were cut from 20 lumbar vertebrae and infiltrated with silicone oil
with a viscosity matching that of PMMA. The permeability of the cores was determined based on Darcy’s law. The mean
permeability of SI and AP cores was 4.4571.72
108 and 3.4471.26
108 m2, respectively. An interesting finding of this study
was that the permeability of the AP cores was approximately 78% of that of SI cores, though the porosity of the SI and AP cores
taken from the same vertebra was approximately equal. In addition, we provided a theoretical model for the porosity-dependent
permeability that accurately described non-linear dependency of the bone permeability and porosity in both directions. Although the
relation of the bone permeability and porosity was established, bone mineral density was a weak predictor of the bone permeability.
The experimental and theoretical results of this study can be used to understand polymer flow in cement infiltration procedures.
r 2003 Elsevier Ltd. All rights reserved.

Keywords: Human cancellous bone; Permeability; Porosity; Cement augmentation; Vertebroplasty

1. Introduction augmentation (vertebroplasty) is possibly the only

effective solution for the management or prevention of
The prevalence of osteoporosis-related vertebral osteoporosis-related vertebral fractures (Cotten et al.,
fractures is increasing rapidly as the population ages 1998; Heini et al., 2000; Mathis et al., 2001).
(Riggs and Melton, 1995). Adequate fracture treatment The typical bone cement used for augmentation is a
is often difficult because conventional internal fixation two-component system composed of a liquid and a
fails to anchor sufficiently (Hu, 1997). In addition, long- powder. After mixing the components, cement is
term drug therapy has been ineffective at reversing injected under fluoroscopic guidance into a vertebral
osteopenia. Therefore, the need for treating patients body through a transpedicular canula. The cement
with acute osteoporotic fragility fractures will remain displaces the bone marrow as it fills the cancellous bone
over the next few decades. Currently, bone cement cavities. A catalyst, usually present in the powder,
initiates curing of the bone cement within the cancellous
*Corresponding author. Tel.: +1-514-842-1231; fax: +1-514-843-
bone skeleton.
1699. Cement infiltration of cancellous bone is still being
E-mail address: [email protected] (G. Baroud). performed based solely on personal experience of the

0021-9290/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0021-9290(03)00246-X
 ARTICLE IN PRESS
190 G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196

clinician. Consequently, minimizing complications de- The intrinsic permeability, k is generally a function of
pends on each clinician’s learning curve. Furthermore, the morphological properties of the cancellous bone.
avoiding intra-operative complications, such as cement This study characterizes the intrinsic bone permeability
leakage, requires the clinician to work within a narrow as a function of the bone porosity (see appendix):
range of cement viscosity. For example, depending on
a
the cement viscosity and/or bone porosity, cement may k¼A ; ð2Þ
spread quickly in an uncontrolled manner, violating the 1a
anatomical boundaries, penetrating into adjacent tissue,
where a (dimensionless) is the bone porosity and A (L2)
and resulting in adverse effects such as neurological
is a phenomenological parameter.
damage and pulmonary embolism (Cotten et al., 1998;
Heini et al., 2000; Mathis et al., 2001). However, if the
cement is injected while it is too viscous or if the bone 2.2. Experimental work
porosity is low, the pressure required for injection
becomes too high, resulting in poor filling because the Four lumbar spines from L1 to L5 (mean age 70; age
surgeon is unable to generate enough pressure to force range 49–80) were harvested from female donors during
the cement to infiltrate the trabecular skeleton. routine autopsies. The specimen were X-rayed in the
To optimize the injection parameters of the cement anterior–posterior (AP) and lateral directions to exclude
injection procedure, we need to understand the hydrau- spinal pathologies. The specimen were then sealed in
lic permeability of vertebral cancellous bone as a plastic bags and stored at –20 C.
function of its porosity and cement viscosity. The BMD measurements were obtained in lateral projec-
cement viscosity can be identified independently using tion using dual photo-absorptiometry (DPX-L, Lunar
standardized uniform shear tests (Krause et al., 2003; Radiation Corporation, Madison, Wisconsin). For
Baroud et al., 2003). However, very little is known about BMD scanning, the frozen specimen were placed in a
the permeability of vertebral cancellous bone, especially plastic container and covered with rice to simulate soft
the interdependence of permeability and porosity. tissue (Oxland et al., 1996). The BMD measurements
Previous studies have investigated the permeability of were repeated four times for each of the five lumbar
cancellous bone at several anatomical locations, includ- levels and for each of the four spines. The repeatability
ing the tibia (Beaudoin et al., 1991), calcaneus (Grimm of the BMD measurements was estimated using the
and Williams, 1997) and femur (Nauman et al., 1999); coefficient of variation and was between 2% and 4%.
however only one study has been published on the The four spines were then thawed and the excess soft
permeability of vertebral cancellous bone (Nauman tissue was removed. Subsequently, the spines were
et al., 1999). One of the limitations of the latter study dissected into 20 single vertebrae. Each vertebra was
is that it investigated only a few samples with a narrow sealed in a plastic bag and stored at 20 C until the day
range of porosity. the cores were cut. A diamond-coring tool with an inner
The main goal of our study is to investigate the diameter of 16.45 mm was used to cut the bone cores.
interdependence of permeability and porosity of verteb- Two cores were cut from each vertebra according to
ral cancellous bone. This study also examines whether proceedures described in Nauman et al. (1999). The first
vertebral bone mineral density (BMD) can be used to core, taken from the right hemivertebra, was cut
predict the permeability or porosity of cancellous bone. perpendicular to the superior endplate—superior–infer-
ior (SI) samples. The vertebral body was then rotated
90 , and a second core was taken in the anterior–
posterior direction from the left hemivertebra—AP
2. Method
samples.
Although we started with 20 SI and 20 AP samples,
2.1. Theoretical consideration for permeability
six AP samples were excluded from this study because of
their proximity to the lateral vertebral shell. The
Given that cement is a thick viscous fluid and its flow
remaining 34 (20 SI and 14 AP) samples were trimmed
in permeable bone is slow (i.e., small Reynolds number),
using two diamond disc saws, which were positioned in
the cement flow in cancellous bone can be approximated
parallel on a rotary shaft to ensure parallel cuts of the
using Darcy’s law (Collins, 1961):
bone cores. The space of the two disc saws corresponded
k qp to the desired core height (Table 1). Using this method,
v¼ ; ð1Þ all 20 SI samples were trimmed to a height of 16 mm.
m qx
Twelve AP samples were trimmed to 16 mm and the
where v is the infiltration rate (LT1), k is the intrinsic other two to a height of 12 mm because of the larger
permeability of the bone (L2), m is the viscosity (FTL2), cavities (e.g., sizable blood vessels) at the open ends of
and qp=qx is the pressure gradient (FL3). the cores.
 ARTICLE IN PRESS
G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196 191

Table 1
Summary of the results of superior–inferior (SI) and anterior–posterior (AP) cores

Age Vertebra BMD SI samples AP samples Ratio in %

(yr) (g/cm2)
Apparent density Porosity Permeability Apparent density Porosity Permeability Porosity Permeability
(kg/m2) (—) (10E8m2) (kg/m2) (—) (10E8m2) AP/SI AP/SI

49 L1 0.79 290.55 0.85 2.24 — — — — —

49 L2 0.81 285.36 0.85 2.22 290.38 0.84 1.59 98.98 71.60
49 L3 0.84 270.36 0.86 2.32 — — — — —
49 L4 0.95 278.55 0.86 2.35 311.34 0.83 1.83 97.47 77.74
49 L5 1.38 307.62 0.84 2.15 492.27 0.72 1.39 85.85 64.76
74 L1 0.58 147.48 0.92 7.43 151.29 0.92 4.35 99.53 58.61
74 L2 0.88 267.62 0.86 3.26 161.51 0.91 3.97 106.68 121.79
74 L3 0.84 157.32 0.92 5.70 165.10 0.91 4.09 99.30 71.73
74 L4 0.65 191.23 0.90 5.04 145.74 0.91 3.63 100.76 72.06
74 L5 0.52 150.35 0.92 4.83 — 0.92 3.75 100.16 77.53
78 L1 0.53 109.01 0.94 6.58 137.03 — — — —
78 L2 0.52 132.21 0.93 6.79 119.71 0.93 4.14 99.79 60.91
78 L3 0.52 134.75 0.93 6.67 135.02 0.93 5.39 100.51 80.87
78 L4 0.58 137.61 0.93 5.69 182.74 0.93 4.95 100.23 86.86
78 L5 0.67 158.19 0.92 4.61 — 0.90 3.21 97.78 69.56
80 L1 0.31 84.48 0.96 6.57 — — — — —
80 L2 0.35 103.69 0.94 5.75 — — — — —
80 L3 0.37 124.18 0.93 4.99 135.11 — — — —
80 L4 0.40 170.37 0.91 3.15 148.96 0.93 3.69 102.13 117.23
80 L5 0.53 176.15 0.91 3.40 148.96 0.91 2.32 100.72 68.10

The last two columns contain the AP/SI ratio (in percentage) for bone porosity and permeability. The AP permeability was approximately 78% of
the SI permeability, though the porosity of the AP and SI samples from the same vertebra was equal.

The volume of the core cylinders (L3) was calculated: Archimedes’ principle (Nauman et al., 1999; Kohles
pd 2 et al., 2001). For this purpose, the bone samples were
Vcore ¼ h; ð3Þ degassed in distilled water under constant vacuum. Each
4
sample was then removed from the vacuum chamber
where h is the core height (L) and d is the core diameter and weighed to obtain the submerged weight. The
(L). The height and diameter of each core were volume of cancellous bone core can be determined as
determined with a caliper (10 mm resolution) based on follows:
the mean of five random measurements. The variation
of these measurements, using the coefficient of variation, mdefatted  msubmerged
Vb ¼ ; ð5Þ
was in the range of 5%. 0:997
The cores were then defatted by placing them in an where 0.997 is the density (g/cm3) of the distilled water
ultrasonic bath filled with a 1% solution enzymatic at 25 C.
detergent (Terg-A-Zymes, Alconox, New York, NY) The porosity of the sample was
for 3 h. The cores were rinsed with a low-power water jet
to wash out the digested marrow. The procedure was Vb
a¼1  : ð6Þ
repeated until no bone marrow could be detected upon Vcore
visual inspection (Nauman et al., 1999). The repeatability of the bone volume fraction
An air jet was used to dry each sample. The water-free measurements was estimated, using the coefficient of
weight was recorded using a micro-scale (PB 3002 variation, to be less than 2%. The defatted cores were
DeltaRanges, Mettler Toledo, Columbus, OH), then placed in a 0.9% saline solution and stored at 4 C
throughout the drying process until the core weight until the testing day.
stopped dropping. The mass (M) of defatted and water- Thirty-four uniform infiltration experiments were
free cores, mdefatted ; was recorded and the apparent conducted on isolated cores of osteoporotic bone using
density (ML3), rapparent ; of the bone was calculated a custom-built infiltration device (Fig. 1). The bone
using the equation. cores were placed at the bottom of a custom-made
mdefatted
rapparent ¼ : ð4Þ metallic tube with an inner diameter of 16.52 mm (Micro
Vcore Group, Medway, MA). Each bone was circumferentially
To determine the sample porosity, the bone volume wrapped with Teflon thread seal tape to ensure that it fit
fraction of the core (dimensionless) was measured using tightly in the metallic tube.
 ARTICLE IN PRESS
192 G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196

The plunger was lubricated to reduce the friction

Cross head
between the plunger and the barrel. Repeated measure-
ments, using the testing apparatus described (Fig. 1),
have shown the error related to frictional force between
plunger and barrel was in the range of 4% of the
infiltration force. To compensate for the friction-related
error, the frictional forces were subtracted from the
Test tube measured infiltration forces.
Plunger
Before testing, the linear relation between the pressure
gradient and the infiltration speed (i.e., the suitability of
Darcy’s law for this investigation) was verified as
described in Nauman et al. (1999) and recommended
Silicone oil Bone core
in Arramon and Nauman (2001). Using a dedicated
bone core, independent infiltration tests were conducted
at speeds of 1, 5, 10, and 20 mm/s. Injections at each
speed were repeated four times. Based on the coefficient
Test stand of variation, permeability measurements were repeata-
ble within 4%. The pressure gradient increased linearly
Load platform with respect to the infiltration speed and when a
regression line of the pressure gradient with respect to
the infiltration speed was forced to pass through the
Collecting cup origin, the correlation coefficient ðR2 Þ of 0.99 was
obtained. The reliability of the testing procedure was
Fig. 1. Experimental set-up for the determination of the permeability
of cancellous bone. A cancellous bone core was placed at the bottom of
also examined by forcing the silicone oil to flow through
a tube. Silicone oil was poured into the tube, leveled to the upper rim. a die (i.e., filter) instead of a bone sample. Since the
The initial height of the silicone oil column was approximately 60 mm. geometry of the die was known (diameter 2 mm, length
Using a material testing machine, the plunger was pressed down at a 40 mm), the viscosity of the silicone oil was calculated to
constant injection speed, forcing the silicone oil to flow through the be 93.6 Pa  s, using Hagen–Poiseuille’s law. The error
interstitial space of the cancellous bone. The pressure, p ¼ F =A; in
response to the linear injection was recorded over the course of the
was within 2%.
experiment.

3. Results
Permeability is a material property (i.e., independent
of fluid that flows though porous media). Bone cement is Table 1 summarizes all measurements obtained from
highly non-linear with respect to shear-rate and time, the SI and AP core samples. Bone permeability was
and thus it is not recommended for permeability testing dependent on flow direction and bone porosity. The
(Arramon and Nauman, 2001). This study was therefore mean (71 standard deviation) permeability of the SI
conducted using silicone oil of a constant Newtonian cores was 4.4571.72
108 m2. The mean permeability
viscosity (Viscosity standard 100 000, Brookfield En- of the AP cores was 3.4471.26
108 m2.
gineering, Middleboro, MA). The silicone oil viscosity On average, the permeability of the AP cores was
(95 Pa  s) was similar to bone cement viscosity and it approximately 78% (po0:001) of the permeability of the
was not sensitive to temperature changes in a laboratory SI cores from the same vertebra (Table 1), even though
(73 C). As a result, determining intrinsic permeability there was no significant difference in the porosity of the
of the samples was straightforward. SI and AP core samples (p > 0:5).
After placing a prepared bone core at the bottom of The bone permeability as a function of bone porosity
the tube, silicone oil was poured into the tube. A rigid was found to be strongly non-linear for both flow
plunger with an o-ring was then inserted into the top of directions. The porosity-dependence of the bone perme-
the tube. Using a material testing machine, the plunger ability can be approximated by Eq. (2). Because of the
was pressed down at a constant speed, forcing the directional properties of permeability, Eq. (2) was fit
silicone oil to flow through the bone core. Approxi- independently to the permeability values of the SI and
mately 20 cm3 of silicone oil were forced through each AP samples (Figs. 2 and 3).
sample. Pressure versus time curves were recorded Significant regressions at the po0:0001 level were
during the experiment. The pressure gradient was found for SI permeability with respect to bone porosity
calculated assuming the outlet pressure was atmo- (A ¼ 0:394
108 m2, R2 ¼ 0:66) and for the AP
spheric. The permeability of the cores was calculated permeability with respect to bone porosity
at speeds of 5 and 10 mm/s according to Eq. (1). (A ¼ 0:347
108 m2, R2 ¼ 0:82).
 ARTICLE IN PRESS
G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196 193

10 Bone Porosity Versus Bone Mineral Density

1.1
SI test data from the current study
−1
κ=A∗α(1−α) , A=0.394*10-8 m2, R2=0.66
8
Nauman et al., 1999
Permeability [10-8 m-2]

1.0

Bone Porosity [-]

0.9

0.8

0.7 Test data (SI)

Regression line
0 Test data (AP)
0.80 0.85 0.90 0.95 1.00
0.6
Bone Porosity [-] 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
2
Fig. 2. Comparison of the theoretical model Eq. (2), experimental data Bone Mineral Density [g/cm ]
of this study (superior–inferior direction) and data published by
Nauman et al. (1999). The superior–inferior permeability of vertebral Fig. 4. A linear regression of the bone porosity of the superior–inferior
cancellous bone increases in a non-linear fashion with respect to bone (SI) and anterior–posterior (AP) cores as a function of bone mineral
porosity. The theoretical model presented an accurate description of density ðR2 ¼ 0:81Þ:
bone permeability as a function of bone porosity. The model was only
fit to the data from the present study.

Bone Permeability Versus Bone Mineral Density

10 12

SI cores
AP test data from the current study AP cores
κ=A*α(1-, α)-1, A=0.347, R2=0.82 10 2
Regression line (-5.03*BMD+7.45, R =0.45) for SI cores
8 Nauman et al., 1999
Bone Permeability [10-8 m2]

2
Regression line (-3.57*BMD+5.64, R =0.47) for AP cores
Permeability [10-8m-2]

8
6

4
4

2
2

0 0
0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Bone Porosity [-] Bone Mineral Density [g/cm2]

Fig. 3. Comparison of the theoretical model Eq. (2), experimental data Fig. 5. A linear regression of the bone permeability of the superior–
of this study (anterior–posterior direction) and data published by inferior (SI) and anterior–posterior (AP) cores as a function of bone
Nauman et al. (1999). The anterior–posterior permeability of the mineral density.
cancellous bone increases in a non-linear fashion with respect to bone
porosity. The theoretical model presented an accurate description of
the bone permeability as a function of bone porosity. The model was
only fit to the data from the present study. However, BMD was a weak predictor of bone
permeability. Consequently, BMD, which was again
calculated from the central portion of the vertebral
body, was of limited use in predicting the permeability
The apparent bone density ðrapparent Þ was the best of both SI (po0:0001; R2 ¼ 0:45) and AP samples
predictor of the bone porosity. A significant linear (po0:0001; R2 ¼ 0:47) (Fig. 5).
regression at the po0:0001 level was obtained for the
pooled data of SI and AP samples
(a ¼ 0:99  0:0005rapparent ; R2 ¼ 0:99). 4. Discussion
Similarly, significant linear correlations (po0:0001;
R2 ¼ 0:81) were obtained for the SI and AP pooled data Infiltrating osteoporotic bone with acrylic cement has
of bone porosity as a function of the bone mineral been recently introduced as an effective treatment for
density calculated from the central portion of the stabilizing osteoporotic vertebrae. Knowledge of
vertebral body (Fig. 4). the effects of bone porosity and rheological cement
 ARTICLE IN PRESS
194 G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196

properties on bone permeability will improve our tational prediction. Arramon and Cowin (1997) and
understanding of cement spread through the porous Arramon and Nauman (2001) employed Kozeny–Car-
bone structure. This knowledge has the potential to man and drag-flow models in an attempt to explain the
improve the infiltration procedure by making it repro- bone permeability as a function of the porosity and
ducible and predictable. The main goal of this study was specific surface1 of the porous media (Collins, 1961). In
to investigate the interdependence of permeability and this study, we did not consider the drag-flow theories
porosity of vertebral cancellous bone. This study also because they define the permeability as function of flow
examined whether vertebral bone mineral density rate. This dependence is inconsistent with the definition
(BMD) can be used to predict the permeability of of permeability being a material property (Arramon and
cancellous bone. Nauman, 2001). Although the Kozeny–Carman model
In this study, we examined a wide range of bone and similar theories (e.g., Hydraulic radius-based
porosities (0.72–0.96). All samples that showed any theories) were developed for low porosity media and
damage or irregularities were excluded from the study or were based on the assumption that the pores are
trimmed further to ensure the cylindrical geometry of geometrically simple and uniformly distributed, they
the samples. In addition, all samples were wrapped in provided a fairly good fit—using a total of seven
Teflon tape to ensure that the cores fit tightly within the parameters—to historical animal data (Arramon and
tubes. Silicone oil with a viscosity independent of Nauman, 2001). However, in the Kozeny–Carman
temperature and shear rate was selected for this study. model the permeability is defined as a function of
As a result of careful selection and treatment of the porosity, a; and specific surface, SL1 :
samples, measurements were repeatable within 4%.
Bone permeability (Figs. 2 and 3) increased in an a3
k¼C ; ð7Þ
over-proportional fashion with respect to porosity. This S2
type of relation is not surprising because permeability is
typically dependent on geometry in a higher polynomial where C is a dimensionless constant (Collins, 1961).
fashion (e.g., in Hagen–Poiseuille’s flow through a tube, Previous studies showed that the specific surface and
the polynomial order with respect to the tube diameter is porosity of bone are interrelated (Fazzalari et al., 1989;
four). All of the permeability values of this study were Martin, 1984). Thus, it is possible to express k as a
within one order of magnitude, which is excellent function of bone porosity alone. In this paper, we
considering the large variation of permeability values proposed such a permeability function using only a
previously reported (Arramon and Cowin, 1997; Arra- single parameter and performed the parameter identifi-
mon and Nauman, 2001; Grimm and Williams, 1997). cation. The theoretical model provided a good fit of the
There are a few reports on the permeability of bone permeability as a function of bone porosity (Figs. 2
trabecular bone (Beaudoin et al., 1991; Nauman et al., and 3).
1999; Kohles et al., 2001). However, there is only one Interestingly, the permeability of the AP samples was
report on the permeability of human vertebral cancel- approximately 78% of the permeability of the SI
lous bone (Nauman et al., 1999), which examined a samples, though there was no significant difference in
narrow range of porosity. In Figs. 2 and 3, the the porosity of the SI and AP core samples. This finding
permeability values of our study confirm and expand suggested that, in addition to porosity, other morpho-
upon the values reported in Nauman et al. (1999). The logical parameters (e.g., trabecular separation) contri-
dependency of bone permeability on flow direction was bute to bone permeability. Therefore, the parameter A
also reported by Nauman et al. (1999) and more recently in Eq. (2) is considered to express the directional
by Kohles et al. (2001). However, the latter study dependence of bone morphology, and therefore the
measured the permeability of bovine cancellous bone, degree of anisotropy of bone permeability. In addition
which has different morphology and permeability than to the good fit of the theoretical model in both flow
that of human cancellous bone (Nauman et al., 1999). In directions, it properly predicts the limits of bone
addition, our study provided permeability data over a permeability as a function of porosity. The model
broader range of porosity. predicts that k approaches zero as porosity approaches
Only a few reports exist on the theoretical modeling of zero, and k approaches infinity as porosity approaches
bone permeability with respect to its bone porosity. For 1. More importantly, this theoretical model accurately
example, Beaudoin et al. (1991) used a finite-element represents the function relation of bone porosity and
model of fluid flow through a series of tori (Torus–plural permeability for the examined range of porosity.
Tori–is a donut-shaped, three-dimensional geometrical 1
The specific surface, S; of a porous material is defined as the
solid formed by rotating a closed curve, especially a
interstitial surface area of the pores per unit of bulk volume of porous
circle, about an axis) and predicted an over-proportional material. It is obvious that finely structured materials exhibit a greater
increase of bone permeability with respect to bone specific surface than coarse materials. Since specific area is the ratio of
porosity. Our experimental data confirmed his compu- area to volume, its dimension is L1.
 ARTICLE IN PRESS
G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196 195

In a clinical setting, BMD is fairly easy to measure, Appendix A. A poroelastic theory of fluid flow for the
and will allow us to estimate bone porosity with fairly infiltration of cancellous bone
high reliability (Fig. 4, R2 ¼ 0:81). However, the degree
of correlation between measured BMD and permeability Cancellous bone is a porous structure, which is made
drops considerably (AP samples: R2 ¼ 0:45; SI samples: up of pore space and a solid matrix (cancellous
R2 ¼ 0:47) because of the sum of individual errors. skeleton). The mechanical behavior of fluid and solid
With typical porosity levels of osteoporotic bone bone matrix during the infiltration process can be
(i.e., above 90%) permeability can vary by a factor of described using a poroelastic mixture theory (Mow
three (see Figs. 2 and 3). Although being able to et al., 1980; Spilker et al., 1992). The effect of fluid
explain nearly 50% of this variation in permeability viscosity (i.e. friction between solid and cement) can be
with BMD measurements is very helpful, it would be incorporated through drag forces.
highly desirable to improve the level of permeability The assumptions made for this cement infiltration
prediction in clinical setting by means of non-invasive model are: (1) the mixture (bone and fluid) are
methods. Better prediction methods would allow us chemically inert and immiscible; (2) the fluid is
to better standardize or individually optimize incompressible; (3) the solid is linear elastic since the
the injection pressure for a controlled infiltration of infiltration forces are small; and (4) the inertial forces of
cement. Non-invasive methods such as quantitative the fluid are negligible since the infiltration process is
computed tomography (qCT) or ultrasonography, used slow (i.e. small Reynolds number).
alone or in combination with BMD measurements The governing equations for the fluid and solid
in a multivariate model, could likely improve this interaction are:
prediction. Mass balance
The permeability measurements in the two directions
r  ðas vs þ af vf Þ ¼ 0: ðA:1Þ
of AP and SI were considered sufficient for the purposes
of this study; however, permeability measurements can Momentum balance
be further extended to a third anatomical direction
r  T s þ Rs ¼ 0; ðA:2aÞ
(Kohles et al., 2001). To determine the porosity of the
bone cores, they were freed from bone marrow.
r  T f þ Rf ¼ 0: ðA:2bÞ
Consequently, our permeability measurements were
performed using marrow-free cores. Since the viscosity Constitutive equations
of silicone oil is much higher than that of bone marrow
T ¼ as pI þ leI þ 2mes ¼ 0;
s
ðA:3aÞ
(Bryant et al., 1989), defatting the bone cores was not
expected to significantly affect the permeability mea-
surements. Previous permeability measurements were T f ¼ af pI: ðA:3bÞ
typically conducted with water. Our permeability In the above equations, boldfaces denote a tensor of
measurements were conducted with silicone oil of a first or second order, subscripts s and f denote solid and
similar viscosity to bone cement (Baroud et al., 2003) fluid, respectively. The symbol r denotes the gradient
since it has the advantage of maintaining large and operator; a denotes the constituents volume fractions
easily measurable pressure. (note that af þ as ¼ 1 for a fully saturated mixture); v
The experimental and theoretical results of this study velocity; T stress; p pore pressure; I identity tensor; l
on porosity-dependent permeability and the viscosity and m the Lame constants of the solid; e volumetric
data of acrylic cements (Krause et al., 1982; Baroud strains; and e solid deviatoric strains.
et al., 2003) can be implemented into a theoretical frame The drag forces between a fluid and solid are equal
using Darcy’s law (Eq. (1)), which can then be used in because of the momentum balance in the mixture; and
conjunction with computational mechanics to investi- according to Spilker et al. (1992), they can be written as
gate adequate injection conditions for a predictable and

reproducible cement infiltration. Rf ¼ Rs ¼ k  vf  vs  praf ; ðA:4Þ
where k is a drag coefficient, which in the case of fluid
flow through porous media can be assumed to depend
on fluid viscosity, m; and porosity of the mixture, af : The
Acknowledgements drag coefficient depends on the porosity of solid. For a
non-porous solid (af ¼ 0) and for a fluid (af ¼ 1), the
This work has been supported by the Canadian drag forces (A.4) are null and thus the drag coefficient
Institute of Health Research (CIHR) Grant Number approaches zero. The following formulation is therefore
MOP 57835. No benefits were, or will be, received, from assumed for the drag coefficient, k:
a commercial party related directly or indirectly to this


study. k ¼ l af ; m af 1  af ; ðA:5Þ
 ARTICLE IN PRESS
196 G. Baroud et al. / Journal of Biomechanics 37 (2004) 189–196

where lðaf ; mÞX0 is a material function that describes Cotten, A., Boutry, N., Cortet, B., Assaker, R., Demondion, X.,
the dependence of drag forces on viscosity. Because of Leblond, D., Chastanet, P., Duquesnoy, B., Deramond, H., 1998.
the shear-rate dependence of bone cement (Baroud et al., Percutaneous vertebroplasty: state of the art. Radiographics 18,
311–320.
2003), l also depends on porosity, af : Fazzalari, N.L., Crisp, D.J., Vernon-Roberts, B., 1989. Mathematical
Taking Eq. (A.5) and combining Eqs. (A.4) and modelling of trabecular bone structure: the evaluation of analytical
(A.3b), the final equation for the fluid momentum (flow) and quantified surface to volume relationships in the femoral head
becomes and iliac crest. Journal of Biomechanics 22, 901–910.

2 Grimm, M.J., Williams, J.L., 1997. Measurements of permeability in
kvf ¼ af rp: ðA:6Þ human calcaneal trabecular bone. Journal of Biomechanics 30,
743–745.
Given Eq. (A.5) for drag forces, the fluid flow Heini, P.F., Walchli, B., Berlemann, U., 2000. Percutaneous transpe-
equation can be written in a form similar to Darcy’s law: dicular vertebroplasty with PMMA: operative technique and early
results. A prospective study for the treatment of osteoporotic
vf ¼ k  rp; ðA:7Þ compression fractures. European Spine Journal 9, 445–450.
where the permeability Hu, S.S., 1997. Internal fixation in the osteoporotic spine. Spine 22,
43S–48S.
af 1 Kohles, S.S., Roberts, J.B., Upton M, L., Wilson C, G., Bonassar,
k¼ : ðA:8Þ L.J., Schlichting, A.L., 2001. Direct perfusion measurements of
ð 1  af Þ lð af ; ZÞ
cancellous bone anisotropic permeability. Journal of Biomechanics
In this paper, we identified the permeability, k; as 34, 1197–1202.
function of bone porosity, af ; according to (A.8). The Krause, W.R., Miller, J., Ng, P., 1982. The viscosity of acrylic bone
superscripts have been omitted in the manuscript for cements. Journal of Biomedical Materials Research 16, 219–243.
Martin, R.B., 1984. Porosity and specific surface of bone. Critical
easier readability.
Reviews in Biomedical Engineering 10, 179–222.
Mathis, J.M., Barr, J.D., Barr, M.S., Jensen, M.E., Deramond, H.,
2001. Percutaneous vertebroplasty: a developing standard of care
References for vertebral compression fractures. American Journal of Neuror-
adiology 22, 373–381.
Arramon, Y.P., Cowin S, C., 1997. Hydraulic stiffening of cancellous Mow, V.C., Kuei, S.C., Lai, W.M., Armstrong, C.G., 1980. Biphasic
bone. Forma 12, 209–221. creep and stress relaxation of articular cartilage in compression?
Arramon, Y.P., Nauman, E.A., 2001. The intrinsic permeability of theory and experiments. Journal Biomechanical Engineering 102
cancellous bone. In: Cowin, S.C. (Ed.), Bone Mechanics Hand- (1), 73–84.
book. CRC Press, New York, pp. 25.1–25.17. Nauman, E.A., Fong, K.E., Keaveny, T.M., 1999. Dependence of
Baroud, G., Wu, J.Z., Bohner, M., Sponagel, S., Steffen, T., 2003. intertrabecular permeability on flow direction and anatomic site.
How to determine the permeability for cement infiltration of Annals Biomedical Engineering 27 (4), 517–524.
osteoporotic cancellous bone. Journal of Medical Engineering and Oxland, T.R., Lund, T., Jost, B., Cripton, P., Lippuner, K.,
Physics 6, 133–139. Jaeger, P., Nolte, L.P., 1996. The relative importance of vertebral
Beaudoin, A.J., Mihalko, W.M., Krause, W.R., 1991. Finite element bone density and disc degeneration in spinal flexibility and
modeling of polymethylmethacrylate flow through cancellous bone. interbody implant performance. An in vitro study. Spine 21,
Journal of Biomechanics 24, 127–136. 2558–2569.
Bryant, J.D., David, T., Gaskell, P.H., King, S., Lond, G., 1989. Riggs, B.L., Melton, L.J., 1995. The worldwide problem of osteo-
Rheology of bovine bone marrow. Proceedings of the Institution of porosis: insights afforded by epidemiology. Bone 17, 505S–511S.
Mechanical Engineers Part H. Journal of Engineering in Medicine Spilker, R.L., Suh, J.K., Mow, V.C., 1992. A finite element analysis of
203, 71–75. the indentation stress-relaxation response of linear biphasic
Collins, R.E., 1961. Flow of Fluid Through Porous Materials. articular cartilage. Journal Biomechanical Engineering 114 (2),
Reinhold Publishing Corporation, New York, pp. 1–10. 191–201.