www.elsevier.com/locate/jmbbm

Available online at www.sciencedirect.com

Research Paper

Novel methodology for assessingbiomaterial–biofluid interaction in cancellous bone

Antony Bou-Francisa,b,n, Rene P. Widmer Soykac, Stephen J. Fergusonc,Richard M. Halla, Nikil Kapurb

aInstitute of Medical and Biological Engineering, School of Mechanical Engineering, University of Leeds, Leeds, UKbInstitute of Thermofluids, School of Mechanical Engineering, University of Leeds, Leeds, UKcInstitute for Biomechanics, ETH Zurich, Schafmattstrasse 30, HPP O 14, 8093 Zurich, Switzerland

a r t i c l e i n f o

Article history:Received 30 October 2014Accepted 24 February 2015Available online 6 March 2015

Keywords:Bone surrogatesBone cementCement injection behaviorCement leakageVertebral augmentationExperimental computational cross-validation

a b s t r a c t

Understanding the cement flow behaviour and accurately predicting the cement place-ment within the vertebral body is extremely challenging. Vertebral cancellous bonedisplays highly complex geometrical structures and architectural inhomogeneities over arange of length scales, thus making the scientific understanding of the cement injectionbehaviour difficult in clinical or cadaveric studies. Previous experimental studies oncement flow have used open-porous aluminum foam to represent osteoporotic bone.Although the porosity was well controlled, the geometrical structure of each of the foamswas inherently unique. This paper presents novel methodology using customized,reproducible and pathologically representative three-dimensional bone surrogates to helpstudy biomaterial—biofluid interaction. The aim was to provide a robust tool forcomprehensive assessment of biomaterial injection behaviour through controlling thebone surrogate morphology and the injection parameters (i.e. needle gauge, needleplacement, flow rate and injected volume), measuring the injection pressure, and allowingthe visualization and quantitative analysis of the spreading distribution. This methodologyprovides a clinically relevant representation of cement flow patterns and a tool forvalidating computational simulations.

& 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Osteoporosis and other skeletal pathologies such as spinal met-astasis and multiple myeloma, compromise the structural integ-rity of the vertebra, thus increasing its fragility and susceptibilityto fracture (Bouxsein, 2003; Currey, 2001; Myers and Wilson,

1997). During vertebral augmentation procedures, bone cementis injected through a cannula into the cancellous bone of afractured vertebra with the goal of relieving pain and restoringmechanical stability. Further, prophylactic surgical stabilizationis often performed to reinforce a structurally compromised ver-tebra adjacent to the index level and decrease its susceptibility to

http://dx.doi.org/10.1016/j.jmbbm.2015.02.0271751-6161/& 2015 Elsevier Ltd. All rights reserved.

nCorrespondence to: Room X101, School of Mechanical Engineering, University of Leeds, LS2 9JT Leeds, UK. Tel.: þ44 113 34 32154;fax: þ44 113 2424611.

E-mail address: a.boufrancis@leeds.ac.uk (A. Bou-Francis).

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7

fracture (Ortiz and Mathis, 2010; Mathis and Wong, 2003). Thebone cements used are chemically complex, multi-componentand non-Newtonian with their viscosity having differing degreesof time and shear rate dependency. These cements interact withthe porous structures though which they flow and with otherfluids present within the porous media. The most widely usedcement, poly(methyl methacrylate) (PMMA), is generally ass-umed insoluble in any biofluid (bone marrow) it comes intocontact with, thus the cement-marrow displacement is charac-terized as a two-phase immiscible flow in porous media (Pinderand Gray, 2008; Widmer and Ferguson, 2011). As vertebralcancellous bone displays highly complex geometrical structuresand architectural inhomogeneities over a range of length scales,the pore-scale cement viscosity varies due to its non-lineardependency on shear rates, which are affected by variations inthe local tissue morphology. Furthermore, the vertebral cancel-lous bone microarchitecture varies among the patients beingtreated, thus making the scientific understanding of the cementflow behaviour difficult in clinical or, indeed, cadaveric studies(Bohner et al., 2003; Widmer Soyka et al., 2013a,b). Previousexperimental studies on cement flow (Bohner et al., 2003; Baroudet al., 2006; Habib et al., 2010; Loeffel et al., 2008; Mohamed et al.,2010) have used open-porous aluminum foam to represent oste-oporotic bone. Although the porosity was well controlled, thegeometrical structure of each of the foams was inherentlyunique. This paper presents novel methodology using custo-mized, reproducible and pathologically representative three-dimensional bone surrogates to help study biomaterial–biofluidinteraction. The aim is to provide a clinical representation ofcement flow distribution and a tool for validating computationalsimulations.

2. Materials and methods

2.1. Bone surrogate development

Three-dimensional bone surrogates were developed to mimicthe human vertebral body (Fig. 1). The surrogates were designedin SolidWorks (Dassault Systèmes, Vélizy, France) then manu-factured using a rapid prototyping technique (Projet HD 3000, 3DSystems, Rock Hill, South Carolina, USA). The structure of thesurrogates was tailored to mimic three skeletal pathologies:osteoporosis (Osteo), spinal metastasis (Lesion) and multiplemyeloma (MM). Fig. 2 illustrates the developed bone surrogatesand Table 1 describes the elements incorporated into eachsurrogate. Once the surrogates were manufactured, microCT(mCT 100, Scanco Medical, Switzerland) was used to assess thevariability in their morphology. Eight of the Osteo surrogateswere scanned at a spatial resolution of 24.6 mm (isotropic voxelsize). Then, a cylindrical volume of interest 15mm in diameterand 15mm in length was consistently defined at the centre ofeach specimen. Within this volume of interest, a threshold of"120 HAmg/ccm (based on Ridler’s method (Ridler and Calvard,1978)) was applied and the three-dimensional morphometricindices were determined using proprietary software (ScancoMedical, Switzerland). Only the bone volume fraction, BV/TV(%), trabecular thickness, Tb.Th (mm), and trabecular separation,Tb.Sp (mm) were compared. The porosity of the specimens wasobtained from the micro-CT data and validated using Archi-medes’ suspension method of measuring volume (Hughes, 2005)which was performed using six cubes (2#2#2 cm3) with thesame structure as the Osteo surrogates. One of the six cubes wasalso used to measure the permeability of the Osteo structureusing Darcy’s law (Baroud et al., 2004; Widmer and Ferguson,2012). Furthermore, static contact angle analysis (FTA 4000, FirstTen Angstroms, Virginia, USA) was performed on the material tocompare the surface wettability to that of cortical bone from adry human femur and a fresh ovine vertebra.

2.2. Experimental protocol

The surrogates were filled with bone marrow substitute pre-pared using an aqueous solution of 2.5% w/w carboxymethylcellulose (Mw$250,000—sodium carboxymethyl cellulose,Sigma-Aldrich, Missouri, USA) with a nominal viscosity of0.4 Pa s which has been reported as the approximate valuefor red bovine marrow (Bryant et al., 1989). Superior andinferior plates were used to create a tight seal and confinethe marrow within the surrogates. Following this, the surro-gates were placed into the experimental set-up (Fig. 3) andSimplex P bone cement (Stryker Corporation, Michigan, USA)was injected into each bone surrogate under a constant flowrate of 3 mL/min. A modified formulation for Simplex P wasprepared using 10mL liquid monomer, 9 g PMMA powder and1 g BaSO4. The cement was mixed (Bou-Francis et al., 2014)then transferred into a 10mL syringe and injected at 4 and8 min from mixing (SP4 and SP8, respectively) to assess theeffect on the flow behaviour. The same syringe, needle andcement were used to perform the two injection time pointsinto separate surrogates. The flow behaviour was tested ineach structure, and all injections were repeated three times.

Fig. 1 – The boundary of the 3D bone surrogates showing: (a)two identical and symmetrical elliptical openings 2 mm inheight and 1 mm in width applied to mimic breaches due toanterior blood vessels, (b) one circular opening 3 mm indiameter applied to mimic posterior breaches due to thebasivertebral veins, and (c) the insertion channels that wereincorporated to allow consistent needle placement duringinjection. The superior and inferior surfaces of thesurrogates were kept open due to manufacturingrestrictions. All dimensions are in millimetres.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7 159

The injections were performed using a 10mL luer-loksyringe (Becton Dickinson, New Jersey, USA) with a 12-ga needle(Blunt SS 12-ga#4 in., W. W. Grainger Inc, Illinois, USA). All

injections were performed at room temperature (21.570.1 1C)using a unilateral approach. The needles were consistentlyplaced through the left insertion channel, while the entry pointon the outside of the right insertion channel remained closedoff to prevent any fluid from escaping through the cortex. Afluoroscope (BV-25, Philips, Eindhoven, Netherlands) was usedto radiographically monitor the injections, while a specimenholder was designed to keep the surrogates at the same heightwith respect to the fluoroscope. Further, the holder ensuredthat the top plane of the surrogates was parallel to the plane ofthe fluoroscope.

2.3. Data analysis

The displacement of the syringe plunger and the force appliedon the plunger were used to determine the injection flow rateand pressure, respectively. The measured force was divided bythe cross-sectional area of the syringe to determine the totalpressure in the system, which is the sum of two pressures: (1)ΔPn, the pressure required to overcome frictional forces and toinject the fluid through the needle and (2) ΔPs, the pressurerequired to inject the fluid into the structure of the bonesurrogates. For this reason, injections were also performedwithout the surrogates to determine ΔPn, which was then

Fig. 2 – The developed 3D bone surrogates. (Left) Section-view of the Osteo, Lesion and MM surrogates. (Right)Craniocaudal view showing the location of all elementsincorporated into each surrogate (refer to Table 1).

Table 1 – The location and size of all elements incorporated into the 3D bone surrogates. All coordinates are measured withrespect to the geometrical centre of each element.

Surrogate Element Coordinate Description

x y z

Osteo, Lesion and MM 1 0.0 0.0 0.0 Reference point2 20.0 0.0 15.3 Outlet, circular Ø 3.0 mm3 13.5 "20.6 8.0 Inlet, circular Ø 2.1 mm4 8.2 "26.1 9.3 Outlet, elliptical width 1.0 and height 2.0 mm5 31.8 "26.1 9.3 Outlet, elliptical width 1.0 and height 2.0 mm

Lesion 6 14.3 0.0 10.0 Outlet, circular Ø 2.7 mm7 14.3 "8.9 10.0 Spherical void Ø 19.0 mm

MM 6 14.3 0.0 8.0 Outlet, circular Ø 2.6 mm7 14.3 "2.2 8.0 Spherical void Ø 6.0 mm8 14.3 "14.9 8.0 Spherical void Ø 6.0 mm9 14.3 "23.4 8.0 Spherical void Ø 6.0 mm

Fig. 3 – Photograph of the experimental set-up showing: (A)the specimen holder with (B) the 3D bone surrogate, (C) the12 gauge needle, (D) the 10 mL luer-lok syringe with (E) acustom built plastic plunger (Delrins, DuPont, Delaware,USA), (F) the syringe pump, (G) the load cell and (H)the LVDT.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7160

subtracted from the total pressure to determine ΔPs atthe inlet.

The video sequences showing the superior view of thefluoroscopy projection were analyzed in Matlab (R2009b, Math-Works, MA, USA). This allowed automated segmentation ofthe flow contours by subtracting the first frame of the videosequence, which shows the surrogates with no fluid injected,from the remaining frames. The error in the needle placementwithin the surrogates was quantified on the first video framethrough measuring the needle length and angle with respectto the posterior wall. The mean spreading distance (MSD) wasmeasured on the segmented flow contours. The shortest andlongest distances were determined along the geometric centerof each contour and MSD was calculated by taking the mean ofthe two distances. Furthermore, the time and type of leakage(i.e. anterior or posterior) were obtained directly from the videosequences.

The construct was scanned using microCT post-injectionat a spatial resolution of 73.6 mm (isotropic voxel size). TheDICOM stack of each scan was processed in imageJ (Schneideret al., 2012) to obtain the final shape of the cement bolus andmeasure the sphericity (Eq. 1), which is defined as the surfacearea of a sphere enclosing the same volume (V) as the 3Dobject, divided by the surface area (A) of the structure (Linand Miller, 2005).

Sphericity¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi36πV23

p

Að1Þ

All data were presented as mean7standard deviation. Theinfluence of the viscosity and the structure on the measuredparameters was evaluated using the Mann–Whitney U andthe Kruskal–Wallis one-way analysis of variance with asignificance level set at α¼0.05. All statistical analyses wereperformed in SPSS version 21 (SPSS Inc, IL, USA).

2.4. Experimental/computational cross-validation

The developed CAD model of each bone surrogate was con-verted into a continuum scale finite element model similar tothat previously utilized for assessing cement flow in bone(Widmer Soyka et al., 2013a,b). The computer simulation wasset up with the same boundary conditions as the experiments.Then, the flow behaviour of Simplex P bone cement wassimulated in each bone surrogate for an injected volume of5 mL at a constant flow rate of 3 mL/min matching thatundertaken in the experiments. The marrow substitute wasmodeled in the simulation using viscosity measurementsperformed using a rheometer (Malvern Instruments Ltd, Mal-vern, UK) in rotation mode with a cone on plate configuration.The cone had a 25mm radius and 11 angle. The temperaturewas set to 21 1C and the angular speed of the cone (Ω) wasramped to achieve varying shear rates from 0 to 100 s"1. Theramp time was 5 min and 10 samples per decade were rec-orded. The Simplex P formulation was modeled in the simula-tion for injections performed at 4 min after cement mixing(SP4). The viscosity was based on previously published rheo-logical tests (Widmer Soyka et al., 2013a,b).

3. Results

3.1. Viscosity measurements

The rheological tests revealed that the marrow substitute hada viscosity of approximately 0.39 Pa ( s for shear rates rangingfrom 1 to 100 s"1 and exhibited Newtonian fluid character-istics (viscosity independence to shear rate).

3.2. Bone surrogate characterization

The variability in the morphology of the Osteo surrogates wasvery low with an overall strut thickness (Tb.Th) of 0.2570.04mmand pore spacing (Tb.Sp) of 0.8970.03mm. The histogram of thethickness map revealed two distinct peaks at approximately 0.18and 0.27mm corresponding to the horizontal and vertical strutsof the 3D bone surrogates, which had a nominal thickness of 0.15and 0.25mm, respectively. The mean porosity, which wasobtained from the microCT data, was 82.671.1%. The measuredpermeability of the Osteo structure was 57.176.1#10"10 m2. Thesurface wettability was comparable between materials withcontact angles ranging from 49 to 771 for the substance used tomanufacture the bone surrogates, 60 to 751 for bone from a dryhuman femur and 58 to 691 for bone from a fresh ovine vertebra.

3.3. Injection parameters

The measured displacement of the syringe plunger confirmedthe injection flow rate to be constant at 2.9970.01 mL/min.Multiplying the total displacement of the syringe plunger bythe cross-sectional area of the syringe confirmed that theinjected volume was 4.9570.01 mL. The mean needle lengthfrom the posterior wall was 2371 mm and the mean needleangle with respect to the posterior wall was 70721. As thenominal CAD distance is 25 mm and the nominal CAD angleis 69.81, the mean error was 8% for the needle length (d) and0.1% for the needle angle.

3.4. Flow distribution

In the Osteo surrogate, the experiments showed that cementinjected at 4 min (SP4) had a tendency to flow laterally into theright needle insertion channel (Fig. 4). This behaviour was notobserved in the MM and Lesion surrogates, independent of thecement injection time, mainly due to the flow being guided bythe lesions. The flow distribution obtained numerically via thecomputer simulation was similar to that observed experimen-tally in the Osteo and MM surrogates, although the simulationpredicted lateral flow in the MM surrogate which was notobserved experimentally. In the Lesion surrogate, the simulatedflow was towards the anterior boundary and along the rightneedle insertion channel, while the experimental flow waschannelled by the void towards the posterior wall.

Qualitative analysis of the experimental data also showedthat in the Osteo and MM surrogates there was a high tendencyfor anterior leakage only, because the cements never reached theposterior wall. In the Lesion surrogate, there was a high ten-dency for posterior leakage only, although the cement injected at

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7 161

8min (SP8) reached the anterior wall. The computer simulationpredicted anterior leakage in all the bone surrogates, thusmatching the experiments in the Osteo and MM surrogates only.

3.5. Sphericity

Increasing the injection time from 4 to 8min after cementmixing significantly (po0.05) increased the sphericity by 12.6%in the Lesion surrogate only (Table 2 and Fig. 5). Furthermore,introducing structural voids significantly (po0.05) decreased thesphericity of the SP4 cement only. Relative to the Osteosurrogate, the presence of the large void in the Lesion surrogatecaused a more pronounced decrease (11.0%) in sphericitycompared to the presence of three small voids in the MMsurrogate (7.7%). The sphericity obtained numerically for theSP4 cement was similar to that measured experimentally in theOsteo surrogate (5.9% difference). In the MM surrogate, thesphericity obtained numerically was 14.9% higher than that

measured experimentally, although the simulated flow distri-bution was qualitatively similar. In the Lesion surrogate, thedifference in the sphericity obtained numerically and thatmeasured experimentally was less pronounced (10%), which isinteresting especially when the simulated flow was preferen-tially horizontal towards the anterior boundary while theexperimental flow was along the posterior void.

3.6. Mean spreading distance (MSD)

Increasing the cement injection time did not have a signifi-cant (p¼0.321) effect on the MSD, independent of the struc-ture (Table 2). Similarly, introducing structural voids did nothave a significant (p¼0.104) effect on the MSD, independentof the cement injection time. The MSD obtained numericallyfor the SP4 cement was overestimated by 22.7%, 16.3% and34.2% in the Osteo, MM and Lesion surrogates, respectively.

Fig. 4 – Representative images of typical flow patterns after 5 mL of Simplex P cement was separately injected at 4 and 8 minafter mixing (SP4 and SP8) into each 3D bone surrogate. The solid line surrounding the contours is the outline of the imagesegmentation performed in Matlab. The simulated flow distribution is included for comparison.

Table 2 – Summary of the experimental results showing the sphericity, the mean spreading distance (MSD), the leakagetime and the peak injection pressure after 5 mL of Simplex P cement was separately injected at 4 and 8 min after cementmixing (SP4 and SP8, respectively) into each surrogate at a constant flow rate of 3 mL/min. The experimental data ispresented as mean7SD. The sphericity and MSD obtained numerically for SP4 are included for comparison.

Parameter Osteo MM Lesion

SP4Comp

SP4 Exp SP8 Exp SP4Comp

SP4 Exp SP8 Exp SP4Comp

SP4 Exp SP8 Exp

Sphericty 0.81 0.77(70.01)

0.79(70.03)

0.81 0.71(70.02)

0.75(70.02)

0.75 0.68(70.02)

0.77(70.04)

MSD (mm) 28.86 23.53(70.27)

23.13(70.57)

29.28 25.18(70.18)

24.07(70.76)

33.56 24.99(72.62)

23.72(70.46)

Leakage Time(s)

– 28.37(71.52)

39.45(74.08)

– 29.94(73.06)

44.73(72.03)

– 55.40(710.7)

54.45(713.7)

Peak Pressure(MPa)

– 0.14(70.01)

0.40(70.01)

– 0.15(70.02)

0.24(70.08)

– 0.10(70.02)

0.24(70.03)

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7162

3.7. Leakage time

The experimental results showed that increasing the cementinjection time significantly (po0.05) increased the leakagetime in the Osteo surrogate by 39.1% and in the MM surrogateby 49.4% (Table 2). The leakage time in the Lesion surrogateremained the same independent of the cement injectiontime. For SP4 cement only, the structural void in the Lesionsurrogate caused a significant (po0.05) increase (95.3%) in theleakage time, relative to the Osteo surrogate. Fig. 6 presentstypical leakage observed in the Osteo and Lesion surrogates.

3.8. Injection pressure

The experimental results also showed that increasing thecement injection time significantly (po0.05) increased the peakpressure by 183.0% in the Osteo surrogate, 54.2% in the MMsurrogate and 133.7% in the Lesion surrogate (Table 2 andFig. 7). Furthermore, the presence of structural voids generallydecreased the peak pressure independent of the cement injec-tion time. For the SP4 cement, the peak pressure recorded in

the MM surrogate was similar to that recorded in the Osteosurrogate (7.6% difference). However, the structural void in theLesion surrogate decreased the peak injection pressure by29.5% relative to the Osteo surrogate. For the SP8 cement, thestructural voids in the MM and Lesion surrogates decreasedthe peak injection pressure by 41.4% relative to the Osteosurrogate.

4. Discussion

The developed bone surrogates can be assumed constant interms of their geometrical structure as the variability inmorphology was very low. This is crucial to reduce the incon-sistency, render the experiments reproducible and shift thefocus onto understanding the influence of viscosity and struc-ture on the flow behaviour. The pore spacing of the Osteosurrogates (0.8970.03 mm) was comparable to that reported inthe literature for human osteoporotic vertebral cancellous bone(Chen et al., 2008; Gong et al., 2006; Hildebrand et al., 1999;Hulme et al., 2007; Lochmuller et al., 2008), suggesting that thesurrogates were pathologically representative. The boundary ofthe surrogates simulated the vertebral shell which confines theflow and controls the intravertebral pressure, significantlyaffecting the filling pattern (Baroud et al., 2004). This is con-sistent with previous studies in which Loeffel et al., (2008)sealed their surrogates with a reusable acrylic enclosure, whileBaroud et al., (2006) and Mohamed et al., (2010) coated theirsurrogates with a 1mm layer of acrylic cement to mimic thevertebral shell. The boundary including the inlet and the flowexit points were kept constant for all the surrogates withinthis study. The openings in the boundary simulate breachesthrough the cortex due to a lesion and/or a vessel exchangingblood in and out of the vertebral body. This is important as suchbreaches create paths of least resistance providing means forleakage into the surrounding structures. Furthermore, theproposed surrogates simulate the rheological environmentwithin the vertebral body. The measured permeability of theOsteo surrogates (57.176.1#10"10 m2) was comparable to thatreported by Nauman et al. (1999) for human vertebral cancellousbone. Also, the surface wettability of the surrogates matches thatof bone and the presence of the marrow substitute simulates thetwo-phase flow that occurs within the vertebral body. In pre-vious studies, Bohner et al. (2003) and Loeffel et al. (2008) bothused melted cow butter, whereas Baroud et al. (2006) andMohamed et al. (2010) both used a water/gelatin solution. A truerepresentation of the rheological properties of human red bonemarrow present within the cancellous bone is extremely impor-tant as such properties significantly affect the cement flowbehaviour (Bohner et al., 2003)).

In all the experiments, the flow rate was kept constantat 3 mL/min and maintained for the duration of the injection.The chosen flow rate was similar to that reported in previouscadaveric studies (Ahn et al., 2007; Reidy et al., 2003) and fallswithin the range reported during clinical percutaneous ver-tebroplasty (PV) which is 1.2 to 12.0 mL/min (Krebs et al.,2005). However, in clinical PV intermittent injections aretypically performed with pauses due to changing of syringesand/or the needle position, which is often retracted back-wards due to excessive pressurization (Krebs et al., 2005).

Fig. 5 – (a) The sphericity after 5 mL of Simplex P cement wasseparately injected at 4 and 8 min after cement mixing (SP4and SP8) into each 3D bone surrogate at a constant flow rateof 3 mL/min. The experimental data is presented asmean7SD. The sphericity obtained numerically for SP4 isincluded for comparison. (b) Representative images of theSP4 cement bolus obtained from the experiments after 5 mLwas injected into each 3D bone surrogate. The flat surface ofthe cement bolus in the Osteo and MM surrogates indicatesthat the cement had reached the superior plate of thespecimen holder.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7 163

The continuous injection was necessary to simplify theinjection parameters.

Loeffel et al. (2008) were the only group to study the infl-uence of viscosity on the cement spreading pattern in humancadaveric vertebrae as well as foam material with similarmorphological properties. Their results showed that the

measured circularity, which was used to quantitatively describethe final shape of the cement bolus, was similar in the cadavericsamples as well as the foam material, suggesting that bonesurrogates can be used to achieve a clinical representation ofcement flow distribution. Loeffel et al. (2008) also found thatincreasing the cement viscosity from 50 to 100 Pa s, resulted in

Fig. 6 – Photographs of the Osteo and Lesion surrogates showing the typical leakage observed after 5 mL of Simplex P cementwas injected.

Fig. 7 – The inlet pressure as a function of time after Simplex P cement was separately injected at 4 and 8 min after mixing(SP4 and SP8) into each 3D bone surrogate at a constant flow rate of 3 mL/min. The pressure obtained numerically for SP4 isincluded for comparison. The experimental data is presented as mean7SD.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7164

significantly denser and more circular cement patterns (highercircularity). In the current study however, all the viscositiestested were below 100 Pa s (Widmer Soyka et al., 2013a,b) whichmay explain why increasing the cement starting viscosity didnot have a significant (po0.05) effect on the MSD, independent ofstructure.

The leakage patterns observed in the Osteo surrogate weresimilar to those reported by Lador et al. (2010) who studied thepoints and pattern of cement extravasation in 23 humanvertebrae and showed that the most common type of leakage,classified as severe, was through small breaches in the cortexdue to anterior blood vessels. This type of leakage occurredin 83% of the samples, thus highlighting the importanceof monitoring all vertebral walls for cement extravasationthrough breaches in the cortex to avoid complications andminimize possible life-threatening risks to the patients. Thisis mainly because leakage into the surrounding vasculaturecan reach remote areas of the body, such as the lungs, andcause pulmonary embolisms (Gosev et al., 2013; Habib et al.,2012; Tourtier and Cottez, 2012). The posterior leakage typi-cally observed in the Lesion surrogate was similar to thatreported by Reidy et al. (2003) who studied the cement fillingpattern in seven osteoporotic human vertebrae with simu-lated lytic lesion. Their results showed that the cementleakage through the posterior venous foramen of the vertebrainto the spinal canal occurred in 86% of the samples. Thishighlights the importance of monitoring cement extravasa-tion through the posterior wall in patients with metastaticinvolvement to avoid complications such as nerve root orspinal cord compression (Sidhu et al., 2013). It is interesting tonote that the structural void in the current Lesion surrogategenerally increased the leakage time under the chosen injec-tion parameters. This is not surprising especially whenposterior leakage occurred and the target site for cementdeposition was at the anterior third. Also, the filled volumeis large due to the presence of the structural void. Further-more, the leakage time in the Osteo and MM surrogatesincreased with the cement injection time. This is consistentwith the study by Baroud et al. (2006) who also reportedimmediate leakage at the early injection time point of 5 minafter cement mixing.

As expected, the injection pressure measured in this studysignificantly (po0.05) increased with an increase in the cementinjection time. This is consistent with the study by Krebs et al.(2005) who also reported significantly (po0.001) higher in vivocement injection pressure during the later phase of thecement polymerization (7 vs. 11min from mixing). Further-more, the presence of structural voids generally decreased thepeak injection pressure independent of the cement injectiontime. This is not consistent with the study by Reidy et al. (2003)who reported no significant difference in the force required toinject bone cement into osteoporotic human vertebrae withand without simulated lytic lesion. In their study however, thelesion was filled with soft tumour tissue. It is important tonote that the overall flow resistance of the vertebra is mainly afunction of the cortical shell, the porosity and the rheologicalproperties of the “fluid” phases (tumour and/or marrow)present within the trabecular network (Reidy et al., 2003).From a geometrical perspective only, the presence of lyticlesion increases the porosity and the hydraulic permeability

thus decreases the overall flow resistance of the vertebra,which decreases the peak injection pressure as evident in theresults of this study. On the other hand, soft tumour tissue(with higher viscosity relative to bone marrow) present withinthe lesion increases the overall flow resistance of the vertebra.This may explain why there was no significant difference inthe force reported by Reidy et al. (2003) required to inject bonecement into osteoporotic human vertebrae with and withoutsimulated lytic lesion. Further experiments are required toelucidate the effects of soft tumour tissue on the injectionbiomechanics.

This study demonstrates that the computational modeldeveloped to simulate the flow of two immiscible fluids throughporous media (Widmer Soyka et al., 2013a,b) generally agreeswith the experimental data in the Osteo and MM surrogates.The difference between the experimental data and the simula-tions was more pronounced in the Lesion surrogate. Thecomputational flow model is based on the assumption thatthe flow process of two immiscible fluids is mainly governed bythe structural properties of the porous medium and therheological properties of the fluids present within the porousmedium. Thus, the accuracy of the simulation depends inher-ently on how the permeability coefficients are determined andhow the viscosities of the two fluids are measured. In thecurrent simulation (Widmer Soyka et al., 2013a,b), the perme-ability was estimated using the Kozeny equation which is basedon hydraulic radius theories. The assumptions of such theorieshold well in the Osteo surrogate and may explain the strongagreement between the simulations and the experimental data.In the Lesion and MM surrogates, the voids were modelled ashigh-permeable regions. Moreover, the equations hold onaverage in these surrogates and may explain the weaker agre-ement. A further source of error for the discrepancy in theLesion surrogate may be due to the difference between theexperimental needle tip position and the nominal position ofthe injection inlet in the CAD model, which was adopted in thesimulation. The needle length from the posterior wall measuredexperimentally on the superior view of the fluoroscopy projec-tion was on average 2mm shorter (i.e. closer to the posteriorwall) compared to nominal position of the inlet in the simula-tions. This could have a significant effect on the flow behaviourand may explain the difference between the simulated and theexperimental flow patterns. Additional experiments are neededto clarify the effect of needle tip position, especially in thepresence of a large structural void.

In the current simulation, the viscosities of the fluid phaseswere governed using a power law, which adjusts the viscosity ofeach fluid phase depending on the time and the shearing rate(Widmer Soyka et al., 2013a,b). Although the viscosities werecharacterized with respect to time and shear rate, the accuracyof the simulation is inherently dependent on how the shear rateis determined. In this simulation, the shear rate was computedfrom the Darcy flux and empirically related to the microscopiclength scale (Widmer Soyka et al., 2013a,b). Furthermore, thesimulation neglects other factors that may affect the rheology ofthe fluids such as the fluid temperature or the yield stresses.More importantly, the simulation expresses Darcy’s law in theform of one single constitutive equation in terms of the mixtureof both fluid phases and does not take into account the potentialeffect of the fluid–fluid interface induced by the surface tension.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7 165

However, this study shows that computer simulation can beused to predict the cement placement in cancellous bone, whichhas been identified as a critical parameter in the biomechanicalbehaviour of the construct post-augmentation (Liebschner et al.,2001; Tschirhart et al., 2005).

The advantage of the experimental methodology presentedhere is that it provides a clinically relevant representation ofcement flow patterns and a tool for validating computationalsimulations. The design of the experimental set-up facilitates themodeling aspect of the simulation. The injection parameters (i.e.needle gauge, needle placement, flow rate and injected volume)were well controlled and reproducible for all the injections. Theboundary including the openings and the needle insertionchannels were kept constant for all surrogates. Furthermore,the structure of the surrogates was uniform to reduce the varia-bility in the morphology (i.e. thickness and spacing) and thestructural geometry, thus achieving a near constant intrinsicpermeability coefficient, particularly in the Osteo surrogate. It isimportant to control the structural properties as these dictate thepermeability tensor and in combination with the applied pres-sure gradient the flow in the porous medium.

5. Conclusion

This paper presents novel methodology using reproducible andpathologically representative three-dimensional bone surro-gates to help study biomaterial-biofluid interaction providinga clinical representation of cement flow distribution and a toolfor validating computational simulations. The proposed bonesurrogates overcome the limitations of previous materials astheir geometrical structure is well controlled and can be tailoredto mimic the morphology of specific bone conditions at differ-ent skeletal sites. This allows for the pathological representa-tion to remain fixed between investigations and the effects ofsubtle differences in the injection behaviour to be assessed in areproducible manner.

Funding

This work was funded by the European Union under the FP7Marie Curie Action, grant agreement no. PITN-GA-2009-238690-SPINEFX.

r e f e r e n c e s

Ahn, H., et al., 2007. The effect of pre-vertebroplasty tumorablation using laser-induced thermotherapy onbiomechanical stability and cement fill in the metastaticspine. Eur. Spine J. 16 (8), 1171–1178.

Baroud, G., et al., 2004. Experimental and theoretical investigationof directional permeability of human vertebral cancellousbone for cement infiltration. J. Biomech. 37 (2), 189–196.

Baroud, G., et al., 2004. Effect of vertebral shell on injectionpressure and intravertebral pressure in vertebroplasty. Spine(Phila Pa 1976) 30 (1), 68–74.

Baroud, G., Crookshank, M., Bohner, M., 2006. High-viscositycement significantly enhances uniformity of cement filling invertebroplasty: an experimental model and study on cementleakage. Spine (Phila Pa 1976) 31 (22), 2562–2568.

Bohner, M., et al., 2003. Theoretical and experimental model todescribe the injection of a polymethylmethacrylate cementinto a porous structure. Biomaterials 24 (16), 2721–2730.

Bou-Francis, A., et al., 2014. Assessing cement injectionbehaviour in cancellous bone: an in vitro study using flowmodels. J. Biomater. Appl. 29 (4), 582–594.

Bouxsein, M.L., 2003. Bone quality: where do we go from here?#.Osteoporos. Int. Suppl 5, S118–S12714 Suppl 5, S118–S127.

Bryant, J.D., et al., 1989. Rheology of bovine bone marrow. Proc.Inst. Mech. Eng. H 203 (2), 71–75.

Chen, H., et al., 2008. Regional variations of vertebral trabecularbone microstructure with age and gender. Osteoporos. Int. 19(10), 1473–1483.

Currey, J.D., 2001. Bone strength: what are we trying to measure?.Calcif. Tissue Int. 68 (4), 205–210.

Gong, H., et al., 2006. Regional variations in microstructuralproperties of vertebral trabeculae with structural groups.Spine (Phila Pa 1976) 31 (1), 24–32.

Gosev, I., et al., 2013. Right ventricular perforation and pulmonaryembolism with polymethylmethacrylate cement afterpercutaneous kyphoplasty. Circulation 127 (11), 1251–1253.

Habib, M., et al., 2010. Cement leakage and filling pattern study oflow viscous vertebroplastic versus high viscous confidencecement. Int. Soc. Adv. Spine Surg. 4, 26–33.

Habib, N., et al., 2012. Cement pulmonary embolism afterpercutaneous vertebroplasty and kyphoplasty: an overview.Heart Lung 41 (5), 509–511.

Hildebrand, T., et al., 1999. Direct three-dimensionalmorphometric analysis of human cancellous bone:microstructural data from spine, femur, iliac crest, andcalcaneus. J. Bone Min. Res. 14 (7), 1167–1174.

Hughes, S.W., 2005. Archimedes revisited: a faster, better, cheapermethod of accurately measuring the volume of small objects.Phys. Educ. 40, 5.

Hulme, P.A., Boyd, S.K., Ferguson, S.J., 2007. Regional variation invertebral bone morphology and its contribution to vertebralfracture strength. Bone 41 (6), 946–957.

Krebs, J., et al., 2005. Clinical measurements of cement injectionpressure during vertebroplasty. Spine (Phila Pa 1976) 30 (5),E118–E122.

Lador, R., et al., 2010. A pictorial classification atlas of cementextravasation with vertebral augmentation. Spine J. 10 (12),1118–1127.

Liebschner, M.A., Rosenberg, W.S., Keaveny, T.M., 2001. Effects ofbone cement volume and distribution on vertebral stiffnessafter vertebroplasty. Spine (Phila Pa 1976) 26 (14), 1547–1554.

Lin, C.L., Miller, J.D., 2005. 3D characterization and analysis ofparticle shape using X-ray microtomography (XMT). PowderTechnol. 154 (1), 61–69.

Lochmuller, E.M., et al., 2008. Does thoracic or lumbar spine bonearchitecture predict vertebral failure strength more accuratelythan density?#. Osteoporos. Int. 19 (4), 537–545.

Loeffel, M., et al., 2008. Vertebroplasty: experimentalcharacterization of polymethylmethacrylate bone cementspreading as a function of viscosity, bone porosity, and flowrate. Spine (Phila Pa 1976) 33 (12), 1352–1359.

Mathis, J.M., Wong, W., 2003. Percutaneous vertebroplasty:technical considerations. J. Vasc. Interv. Radiol. 14 (8), 953–960.

Mohamed, R., et al., 2010. Cement filling control and bonemarrow removal in vertebral body augmentation byunipedicular aspiration technique: an experimental studyusing leakage model. Spine (Phila Pa 1976) 35 (3), 353–360.

Myers, E.R., Wilson, S.E., 1997. Biomechanics of osteoporosis andvertebral fracture. Spine (Phila Pa 1976) 22 (24 Suppl), 25S–31S.

Nauman, E.A., Fong, K.E., Keaveny, T.M., 1999. Dependence ofintertrabecular permeability on flow direction and anatomicsite. Ann. Biomed. Eng. 27 (4), 517–524.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7166

Ortiz, O., Mathis, J.M., 2010. Vertebral body reconstruction:techniques and tools. Neuroimaging. Clin. N. Am. 20 (2),145–158.

Pinder, G.F., Gray, W.G., 2008. Essentials of Multiphase Flow andTransport in Porous Media, .. John Wiley & Sons, Inc..

Reidy, D., et al., 2003. A biomechanical analysis of intravertebralpressures during vertebroplasty of cadaveric spines with andwithout simulated metastases. Spine (Phila Pa 1976 28 (14),1534–1539.

Ridler, T.W., Calvard, S., 1978. Picture thresholding using aniterative selection method. IEEE Trans. Syst. Man Cybern. 8 (8),630–632.

Schneider, C.A., Rasband, W.S., Eliceiri, K.W., 2012. NIH Image toImageJ: 25 years of image analysi. Nat. Meth., 9; 671–675.

Sidhu, G.S., et al., 2013. Neurological deficit due to cementextravasation following a vertebral augmentation procedure. J.Neurosurg. Spine 19 (1), 61–70.

Tourtier, J.P., Cottez, S., 2012. Pulmonary cement embolism aftervertebroplasty. N. Eng. J. Med. 366 (3) 258-258.

Tschirhart, C.E., Roth, S.E., Whyne, C.M., 2005. Biomechanicalassessment of stability in the metastatic spine followingpercutaneous vertebroplasty: effects of cement distributionpatterns and volume. J. Biomech. 38 (8), 1582–1590.

Widmer, R.P., Ferguson, S.J., 2011. A mixed boundaryrepresentation to simulate the displacement of a biofluid by abiomaterial in porous media. J. Biomech. Eng. 133 (5), 051007.

Widmer, R.P., Ferguson, S.J., 2012. On the interrelationship ofpermeability and structural parameters of vertebral trabecularbone: a parametric computational study. Comput. MethodsBiomech. Biomed. Eng. 16 (8), 908–922.

Widmer Soyka, R.P., et al., 2013a. Numerical description andexperimental validation of a rheology model for non-Newtonian fluid flow in cancellous bone. J. Mech. Behav.Biomed. Mater. 6161 (13) 00214-2.

Widmer Soyka, R.P., et al., 2013b. Numerical description andexperimental validation of a rheology model for non-Newtonian fluid flow in cancellous bone. J. Mech. Behav.Biomed. Mater. 27, 43–53.

j o u r n a l o f t h e m e c h a n i c a l b e h a v i o r o f b i o m e d i c a l m a t e r i a l s 4 6 ( 2 0 1 5 ) 1 5 8 – 1 6 7 167