Available online at www.sciencedirect.com

ng C 28 (2008) 366–373www.elsevier.com/locate/msec

Materials Science and Engineeri

Direct metal fabrication of titanium implants withtailored materials and mechanical propertiesusing electron beam melting technology

Ola L.A. Harrysson a,⁎, Omer Cansizoglu a, Denis J. Marcellin-Little b,Denis R. Cormier a, Harvey A. West II a

a Edward P. Fitts Department of Industrial and Systems Engineering, USAb Department of Clinical Sciences, North Carolina State University, Raleigh, North Carolina, 27695, USA

Available online 14 April 2007

Abstract

The design of custom or tailored implant components has been the subject of research and development for decades. However, the economicfeasibility of fabricating such components has proven to be a challenge. New direct metal fabrication technologies such as Electron Beam Melting(EBM) have opened up new possibilities. This paper discusses the design and fabrication of titanium implant components having tailoredmechanical properties that mimic the stiffness of bone to reduce stress shielding and bone remodeling. Finite Element Analysis was used to designthe tailored structures, and results were verified using mechanical testing.© 2007 Elsevier B.V. All rights reserved.

Keywords: Orthopedic implants; Electron beam melting; Theoretical modeling; Materials; Experimentation

1. Introduction

Hip implants have been successfully used since the 18thcentury [1]. Rigid fixation methods became popular withCharnley's technique [2] and have had a high rate of successamongst older patients [3]. However, lower success rates havebeen reported for younger patients. The average useful implantlife is typically 10 to 15 years and depends on the patient,implant type, fixation method, and material used for theimplant. Major issues related to hip implants include implantwear, bone resorption, osteolysis, and pain. Revision surgeriesare costly, painful and sometimes not possible due to thepotential loss of bone mass and mineral content around theimplant. Cementless implants have been developed as analternative to cemented implants to provide long-term stability[4]. In cementless implants, the bone-interface surface is coatedwith metal beads, mesh, or fine particles to provide a porousbone ingrowth surface [5]. Bone systems adapt to mechanical

⁎ Corresponding author.E-mail address: harrysson@ncsu.edu (O.L.A. Harrysson).

0928-4931/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.msec.2007.04.022

conditions by adding bone mass in dynamically loaded areasand reducing bone mass in unloaded or statically loaded areas.This presents a challenge for implant designers, as the mismatchbetween the stiffness of the prosthetic stem and the patient'sbone results in stress shielding. Stress shielding causes boneresorption and may result in premature loosening of the implant.Bone loss may result in challenging revision surgery with lowersuccess rate than primary surgeries [6]. Insertion of a relativelystiff titanium or Co–Cr stem generally changes the loadingpattern around the bone. Engh et al. [7] reported that thick stemsresulted in a five fold increase in incidence of bone resorptioncompared to thin stems. Huiskes et al. [8] reported that flexiblematerials reduce stress shielding and bone resorption, howeverthey increase interface stresses between the implant and bone.The reduction in the loss of cortical bone mass with reducedstem stiffness has been shown in several animal studies [9–11].Different designs and manufacturing techniques have been usedto address the issue of stress shielding by considering flexibilityand interface properties. Early composite implants weredesigned with greater flexibility, but often failed due toinsufficient interface strength between bone and implant orfailure of the structure itself [12–15].

Fig. 1. a) Relative dimensions of one face of the rhombic dodecahedron unit cell b) front view of the unit cell c) rectangular flexural test bar with rhombicdodecahedron unit cells.

367O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

The strength of a cementless implant fixation has been shown tobe largely a function of the pore size. Pore sizes from 50 to 800 μmhave been investigated for different implants [16]. Bobyn et al.reported that pore sizes from 50 to 400 μm provide maximumfixation strength [17]. However, Clemow et al. reported decrease inbone strength and bone ingrowth for increasing pore sizes in the175–325 μm range [18]. The literature shows several contradictingfindings regarding the pore size for optimal tissue ingrowth, but therange is most often within a 50 to 800 μm range.

Conceptual designs of hollow stems, grooved stems, andmodular stem systems that show lower stress shielding havebeen investigated using Finite Element Analysis (FEA) [19–23]. Despite the early failures of composite stems, the Epochstem by Zimmer (Warsaw, Indiana) has been reported to showacceptable results [24]. Porous tantalum has been developedand used for bone ingrowth applications to improve themechanical bond between bone and implant [25]. Currently,cellular metal structures are commercially available in differentdensities for almost any metal. Casting with foaming agents ordirect gases, investment casting, chemical vapor deposition, andpowder metallurgy (PM) has been used to make stochasticcellular metals [26,27]. Properties of the stochastic foamsdepend on the relative density, processing methods, andmaterial. Porous Ti–6Al–4V has been manufactured usingPM techniques, extrusion, and mold replication [28,29].

Arcam's Electron Beam Melting (EBM) process is a directmetal layered fabrication technique that has been used to makecomplex 3D parts such as tools with conformal coolingchannels and medical implants [30,31]. It has a thermionicemission gun that uses a tungsten filament to produce anelectron beam with a maximum power of 4.8 kW. The EBMprocess selectively melts metal powder in 0.07 to 0.25 mm-thick layers. Each layer is first preheated by scanning the beam

Fig. 2. Principal fabrication of thin struts using layered fabrication technology. Theresulting in a thin disc. If the strut angle to the build plane is too small the overlap

at low power and high velocity to lightly sinter the particles.Sintered powder surrounding the part helps support downwardfacing surfaces during the build process. However, the lightlysintered powder will break up during the post build siftingprocess allowing for most of the unmelted powder to berecovered and reused. The elevated temperature also helpslessen residual stresses between the cooling melt pool andpreviously solidified layers. The entire process takes placeunder a vacuum of 10−4 mbar in the chamber and 10−6 mbar inthe gun. The EBM process is currently in use for low volumeproduction of medical components in Europe and the U.S. Thispaper describes a new method for making tailored hip stemswith bone ingrowth surfaces from biocompatible Ti–6Al–4Vusing the EBM process. This work addresses the issue of stressshielding and reports on early test results of hip stems withtailored high strength mesh structures.

2. Materials and methods

The purpose of this research is to design and fabricatetitanium hip stems with tailored mechanical properties that willreduce stress shielding. The ultimate goal is to reduce theimplant's bending stiffness while maintaining its strength. Non-stochastic mesh structures have been designed, fabricated,tested, and evaluated using FEA.

2.1. Design

2.1.1. Lattice structuresIn order to reduce the implant stiffness, the solid stems were

replaced with stems having a repeating 3-dimensional latticestructure designed to produce the desired stiffness. Possible 3Dlattice structures include cubes, truncated octahedrons, truncated

energy beam melts a small dot of metal powder with a predetermined thicknessbetween the layers will be minimal resulting in very weak structures.

Fig. 3. Different configurations of hip stems a) Mesh block, Solid hip stem, and resulting mesh hip stem b) Complete mesh hip stem after Boolean operations. c) Solidhip stem with holes.

Fig. 4. Structures fabricated via Electron BeamMelting: a) Cubes with 40% relative density (60% porous) b) Cubes with relative densities of 8.0%, 5.0%, and 3.8% c)Bending specimens with 8 mm and 6 mm cell sizes (7.3% and 11.9% relative density) d) hip stems with mesh configuration, hole configuration, and solid.

368 O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

Fig. 5. a) Loading of mesh hip stem for flexure testing b) Three point flexure testing of rectangular mesh beam.

369O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

cubes, truncated cuboctahedrons, triangular prisms, rectangularprisms, hexagonal prisms, octagonal prisms, and rhombicdodecahedras [32]. Several mesh configurations were tested inpreliminary stages of this research, and the 12-faced rhombicstructures were selected for space filling properties and geometricproperties suitable for EBM mesh fabrication. Dodecahedronstructures were also used to model orthotropic properties of thebone [33]. Each strut of the rhombic dodecahedron unit cell wasbuilt in the same orientation with respect to the build plane at an

angle of Φ fixed at 35.26° based on tan/ ¼ 1ffiffiffi2

p as shown in

Fig. 1. The strut length was equal for all edges of the unit cell.Structures having the same angular beam orientation and equalbeam length are preferred to reduce the variability in the SolidFreeform Fabrication (SFF) processes caused by layeredmanufacturing of mesh structures [34]. Fig. 2 shows the principleof building thin struts using the layered approach. In our case, theelectron beam melts a dot in each layer. If the strut angle to thebuild plane is too small, there will be no or little overlap betweenthe successive melted layers resulting in a very weak structure.

Unit cells were designed in the SolidWorks CAD packageusing beam segments with rectangular cross sections instead of

Fig. 6. a) Finite Element Analysis tetrahedral meshes of the femur–hip i

circular cross sections. When beams with circular cross sectionsare used, the size and complexity of the resulting CAD modeloverwhelms the capabilities of most workstations when thenumber of unit cells exceeds a certain number. The use of rect-angular cross sections drastically reduces the file size in SolidWorksthus allowing larger mesh structures to be modeled. Further, thisdramatically reduces the STL-file size and subsequent computa-tional requirements. When the STL-file is sliced, the square crosssection of each strut is effectively a dot. When the electron beammelts the metal powder to form the struts, the minimum beamdiameter is used and a small circular disc is produced as shown inFig. 2. Compression test and bend test specimens weremodeled fordifferent sizes of unit cells in SolidWorks, andwere then transferredto Materialise's Magics software as STL files for slicing prior tofabrication on the EBM machine. Relative densities of thestructures were calculated based on the weight of the structure/volume/density of solid Ti–6Al–4V (4.42 g/cm3).

2.1.2. Hip stemsHip stems were designed using CAD files from BioMedtrix,

Boonton, NJ a veterinary orthopedic implant company. Thismodel was first modified by creating through holes in the lower

mplant assembly and hip stem b) ANSYS simulated mesh structure.

Table 1Compression testing of rhombic structures with cell sizes from 3 mm to 12 mm

Structures(cell size–orientation-sample#)

Peak load(N)

Relativedensity

Compressivestrength (MPa)

3 mm-XY-#1 18,777.23 0.41 85.72995 ⁎⁎

3 mm-XY-#2 19,530.29 0.41 91.34936 ⁎⁎

3 mm-XY-#3 19,324.38 0.41 89.9208 ⁎⁎

3 mm-Z-#1 ⁎ 18,717.37 0.41 91.7393 ⁎⁎

3 mm-Z-#2 ⁎ 18,879.29 0.41 94.04639 ⁎⁎

3 mm-Z-#3 ⁎ 19,132.2 0.41 94.93092 ⁎⁎

8 mm-XY-#1 1682.04 0.08 2.9775898 mm-XY-#2 1799.76 0.08 3.185988 mm-XY-#3 1661.27 0.08 2.9408218 mm-Z-#1 1990.24 0.08 2.8901028 mm-Z-#2 2131.88 0.08 3.09578310 mm-XY-#1 1197.87 0.05 1.40576910 mm-XY-#2 1160.47 0.05 1.36187810 mm-XY-#3 1066.3 0.05 1.25136410 mm-Z-#1 713.47 0.05 0.8471112 mm-XY-#1 976.17 0.04 0.81896212 mm-XY-#2 1010.66 0.04 0.847898

⁎ Same sample tested also for Z orientation.⁎⁎ Not peak strength (loading limitation).

370 O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

portion to reduce bending stiffness as shown in Fig. 3c). Then,as shown in Fig. 3 a) and b), Boolean operations were used inMagics to create a hip implant with a lattice structure below theneck of the stem. Solid necks on the implants were retained topreserve compatibility with standard prosthetic heads and tohave a stable head-to-stem interface. Solid material at the neckis also needed due to the impact the implant absorbs duringinsertion into the patient. Square pegs were added to the distalend of the stem to facilitate fixturing during subsequent 4-axisfinish machining.

2.1.3. EBM manufacturingRectangular Ti–6Al–4V compression test and bend test

specimens with different densities were fabricated using theEBM process as shown in Fig. 4. Solid hip stems, mesh hipstems and hip stems with holes were also fabricated as shown inFig. 4 d). The mesh stems were fabricated with reduced pre-heating to lessen the amount of powder sintering and to make iteasier to remove unmelted powder from the meshed regionfollowing processing. All mesh specimens were fabricated using

Fig. 7. a) Average compression strength versus relative density for tested and estimatrelative density in XY orientation and Z orientation.

the same beam segment thickness and beam segment orientationwith respect to the build plane.

2.1.4. TestingMaterial testing was conducted using an ATS 1605C universal

tester. Compression tests on the cube specimens in Fig. 3 wereconducted at a crosshead speed of 5.1 mm/min. One set ofcompression samples was tested parallel to the build directionwhile a second set was tested perpendicular to the build direction.The overall area of the parts was used to calculate the compressionstrength and modulus. A 3-point flexure test was used for thebending specimens as shown in Fig. 5 b). Hip stems were testedby fixing the square peg at the distal end in a vise as shown inFig. 5a. The loadwas applied at the head at a rate of 1.27mm/min.

2.1.5. Analysis and FEAABAQUS CAE 6.4 was used to model the effect of stem

stiffness on bone and implant stresses. A Computed Tomogra-phy (CT)-scan of a broached femur was converted into anABAQUSmodel usingMaterialise's Mimics 9.11 and assembledwith different hip stem models (see Fig. 6 a) [35]. Due to thelimitations of Mimics, only a tetrahedral mesh could be used forthe femur. Materials and properties were assigned as follows: Ti–6Al–4V stem, E=110 GPa; bone, E=15 GPa; stem-shell,E=20 GPa; Stem–CoCr, E=200 GPa; Stem–Mesh, E=20 GPa.A Poisson's ratio of 0.3 was used for all cases. The bone ingrowtharea was bounded to the stem as if it were one unit. Frictionlesscontact was assigned at the interface between the distal stem andthe bone. The distal end of the bone was fixed, and the head of thestem was loaded normal to the surface. Input files were generatedfor each material assignment and submitted to the ABAQUSsolver at the High Performance Computing Center (NCSU).

To compare the unit cell test results for the physical structures,scaling rules were modeled in ANSYS using 3D beam elements.Rhombic structures were created in ANSYS with 4 unit cells in Xand Yand 3 unit cells in Z as shown in Fig. 6b. Bottom nodes werefixed in the Z-direction and top nodes were moved 0.1 mmdownward to simulate compression testing in the plane parallel tothe build orientation. Nodes at the left face were fixed in the Y-direction and nodes at the right face weremoved 0.1mm to the leftto simulate compression testing in the plane perpendicular to thebuild orientation. Both compression test directionswere simulated

ed mesh structures b) ANSYS simulated results for compression stiffness versus

Table 2Comparison of tested structures and simulated structures with 0.49 mm2 squareprofile beams

Sample Tested EMPa (XYorientation)

Tested EMPa (Zorientation)

ANSYS(XY)

Ratio-XYANSYS/test

ANSYS(Z)

Ratio-ZANSYS/test

6 mm – – 747.43 – 1592.438 mm 60 78.81 241.30 4.02 532.30 6.7510 mm 25 23.63 99.79 3.99 223.88 9.4712 mm 12 – 48.38 4.03 109.57 –

371O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

for 6 mm, 8mm, 10mm, and 12mm cell sizes to show the densityeffect for different orientations. The cross sectional area of eachbeam was 0.49 mm2 with a square profile. A modulus ofE=110 GPa and Poisson's ratio of 0.3 were assumed. Relativedensities of the simulated structures were calculated based on theircross section and overall dimensions in ANSYS.

2.2. Results

2.2.1. Compression testingPhysical compression test results are given in Table 1 for 4

lattice densities. Specimens loaded parallel to the build directionare labeled with an “XY” orientation. Specimens loaded per-pendicularly to the build direction are labeled with a “Z”orientation. The 3 mm unit cell structures could not be testeduntil failure due to the load limit of the testing machine(5000 lbs). The cell height is taller in the XYorientation than theZ orientation due to the geometry of the structure as shown inFig. 1b. As can be seen in Table 1, compressive strength in boththe XY and Z direction decreases with increasing cell sizes. Thismakes intuitive sense, as the structure's relative density de-creases when the cell size increases.

A graph of compressive strengths for tests in the XY orien-tation at different relative densities is shown in Fig. 7a. Thecompressive scaling equation is fitted for test data according tothe formula

rc ¼ C1⁎rc;spps

� �1:5

where C1=0.341 and σc,s=897 MPa.

Fig. 8. Bending test results for specimens with unit cell size of 3 mm and 40%relative density.

A comparison of the modulus in the XY and Z direction wasmade using ANSYS as shown in Fig. 7b. In general, the struc-tures are stiffer in the Z-direction than the XY-direction, and thedifference increases with decreasing cell size. Results ofaverage tested modulus (MPa) for 8 mm, 10 mm, and 12 mmstructures are compared to the predicted results from ANSYS inTable 2. The actual and predicted modulus values differed bynearly a factor of 4. In the FEA model, struts are assumed to besmooth with cross sectional areas of 0.49 mm2. The actual strutshave a rough texture with thicknesses that range from 0.3 to0.7 mm2. These results can be used to modify the models of thelattice structures so that the physical behavior of the as-fabricated structures is predictable.

2.2.2. Flexure testingThe results from flexure tests of 3 mm unit cell specimens are

reported in Fig. 8. The average bending strength was approx-imately 150 MPa which is higher than porous tantalum at110MPa [25]. The 3mm unit cell samples showed a linear load–displacement relationship as shown in Fig. 8 with an averagemodulus of 12 GPa. The average modulus for 6 mm and 8 mmstructures in the XY orientation were 349.5 MPa and 47 MParespectively Table 3.

2.2.3. Hip stem testingThree EBM fabricated hip stems were tested in the same

fixture and setup. The weight of the stems after cleaning thepowder was 46.72 g for the solid stem, 33.83 g for the stem withholes, and 26.19 g for the mesh stem. The relative load/dis-placement ratio with respect to the solid stem was 1.16 for thestem with holes and 2.38 for the mesh stem. Linear load–displacement relationships are shown in Fig. 9a.

2.2.4. FEA stress modeling resultsThe results of the hip stem analyses for different material

configurations are shown in Fig. 10. Threematerial configurationswere used for the hip stems— solid Co–Cr, solid Ti–6Al–4V, andmesh Ti–6Al–4V. Not surprisingly, the stresses on the distalportion of the stems were higher for stiffer materials. The contactregions were defined as the same at the bone ingrowth region for allthree cases. The stress distribution on the femurs was different forthe three cases. The proximal portion of the femur exhibited higherstresses with decreasing stem stiffness as shown in Fig. 10b. The

Fig. 9. Results of bend testing of solid stem, stem with holes, and stem withmesh structure are shown in the graph.

Fig. 10. Left)Von Mises stresses at the same scale show results for solid Co–Cr stem, solid Ti–6Al–4V stem, and mesh Ti–6Al–4V stem from left to right. Right) VonMises stresses on the femur for a) Co–Cr stem b) Ti–6Al–4V stem c) mesh Ti–6Al–4V stem at the same stress color coding.

372 O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

proximal portion of the femur with a Co–Cr stem shows a morepronounced stress shielding than the other two cases, and the meshTi6Al4V stem shows the lowest level of stress shielding.

3. Discussion and conclusions

Direct metal fabrication technologies such as Electron BeamMelting show considerable promise for the fabrication of cus-tom orthopedic implants with tailored material properties.However, the processes have limitations that must be consid-ered during the design of such implants. Building non-stochas-tic lattice structures is possible, however the orientation of thelattice struts during the fabrication is important. The beamelements must be oriented such that the inclination is within acertain range. While FEA models generally assume struts withsmooth surfaces and constant cross sections, the fabricated strutshave textured surfaces with slightly varying cross sections. Thesedifferences must be reflected during the design of implantsthrough use of safety factors or other methods. By changing theunit cell size, a predictable compressive and bend strength can beachieved with a predictable modulus. By using such non-stochastic mesh structures, components with tailored mechanicalproperties can be designed and fabricated.

In this paper, the non-stochastic mesh structures have beenused to design a hip stem with a lower bend modulus aimed atlowering stress shielding and uneven bone remodeling. The FEAclearly shows that a stem with a lower bend modulus results in a

Table 3Bending test results for 3 mm, 6 mm, and 8 mm structures

Specimen(cell size-orientation-sample #)

Peak loadN

EMpa

My/IMpa

Relativedensity %

3 mm XY #1 528 11,990 151 39.83 mm XY #2 537 12,749 154 39.53 mm XY #3 486 11,291 139 40.06 mm XY #1 471 261 10 11.96 mm XY #2 688 438 15 11.96 mm Z #1 293 133 6 11.98 mm XY #1 183 47 2 7.38 mm Z #1 434 55 4 7.38 mm Z #2 376 47 3 7.3

much more even stress distribution in the proximal portion of thefemur. This more even stress distribution would likely lead to areduction in bone remodeling.

An important future research topic is fatigue testing of thesenon-stochastic mesh structures to ensure that the non-stochasticmesh structures will perform equally well over time. The per-formance of the non-stochastic mesh structures can be furtherimproved using design optimization algorithms based on aspecific loading scenario. Distribution of the solid regions of theimplant can be calculated based on topology optimization thatminimizes compliance for a given material volume [36].Manual optimization studies conducted on such complex 3dimensional structures is a computational and time consumingtask that would not necessarily be feasible on a case to casebasis. Further research and development on semi-automateddesign optimization procedures is therefore needed. A func-tionally gradient design approach is being considered as wellwhere the stem is fully dense at the core and is graduallytransitioned into a mesh structure. Such a structure can moreeasily be designed and verified using FEA to ensure appropriatebend stiffness as opposed to using optimization. Further, theactual shape and size of the stem can be custom designed tobetter fit a specific patient based upon a model of the femurderived from a CT-scan. Such a design would be more timeconsuming and more expensive than a standard implant design,yet the potential for increasing the longevity of the implantcould more than make up for the additional cost. Anotheradvantage of fabricating a hip stem using EBM technology isthat the porous bone ingrowth surfaces can be built in at thesame time, thus saving several fabrication steps. If boneingrowth is not desired along the entire stem, a thin skin oftitanium can be added to those areas and later polished duringthe finishing step to prevent bone ingrowth.

References

[1] J.T. Scales, Proc. Inst. Mech. Eng. 181 (1966) 64.[2] J. Charnley, J. Bone Jt. Surg. 42B (1960) 28.[3] S.M. Madey, J.J. Callaghan, J.P. Oleiniczak, D.D. Goetz, R.C. Johnston,

J. Bone Jt. Surg. 79A (1997) 53.[4] R.H. Rothman, J.C. Cohn, Clin. Orthop. Relat. Res. 254 (1990) 153.[5] R.M. Pilliar, Orthop. Clin. North Am. 36 (1) (2005) 113.

373O.L.A. Harrysson et al. / Materials Science and Engineering C 28 (2008) 366–373

[6] T.A. Gruen, G.M. McNeic, H.C. Amstutz, Clin. Orthop. Relat. Res. 141(1979) 17.

[7] C.A. Engh, J.D. Bobyn, Clin. Orthop. Relat. Res. 231 (1988) 7 Jun.[8] R. Huiskes, H. Weinans, B. van Rietbergen, Clin. Orthop. Relat. Res. 274

(1992) 124 Jan.[9] J.D. Bobyn, E.S. Mortimer, A.H. Glassman, C.A. Engh, J.E. Miller, C.E.

Brooks, Clin. Orthop. Relat. Res 274 (1992) 79.[10] G.L. Maistrelli, V. Fornasier, A. Binnington, K. McKenzie, V. Sessa,

I. Harrington, J. Bone Jt. Surg. 73B (1991) 43.[11] T.M. Turner, D.R. Sumner, R.M. Urban, J.O. Galante, J. Bone Jt. Surg.

79A (1997) 1381.[12] F. Adam, D.S. Hammer, S. Pfautsch, K. Westermann, J. Arthroplast. 17

(2002) 217.[13] I. Jakim, C. Barlin, M.B. Sweet, J. Arthroplast. 3 (1988) 191.[14] T. Niinimaki, J. Puranen, P. Jalovaara, J. Bone Jt. Surg. 76B (1994) 413.[15] M. Au, J. Formos Med. Assoc. 93 (1994) 497.[16] H. Kienapfel, C. Sprey, A. Wilke, R. Griss, J. Arthroplast. 14 (3) (1999)

355.[17] J.D. Bobyn, R.M. Pilliar, H.U. Cameron, G.C. Weatherly, Clin. Orthop.

Relat. Res. 150 (1980) 263.[18] A.J. Clemow, A.M. Weinstein, J.J. Klawitter, J. Koeneman, J.J. Anderson,

Biomed. Mater. Res. 15 (1) (1981) 73 Jan.[19] J.A. Mandell, D.R. Carter, S.B. Goodman, D.J. Schurman, G.S. Beaupré,

Clin. Biomech. 19 (2004) 695.[20] S. Gross, E.W. Abel, J. Biomech. 34 (2001) 995.[21] R. Poss, P. Walker, M. Spector, D.T. Reilly, D.D. Robertson, C.B. Sledge,

Clin. Orthop. Relat. Res. 235 (1988) 181.

[22] C. Tai, C.H. Shih, W. Chen, S. Lee, Y. Liu, P. Hsieh, W. Chen, Clin.Biomech. 18 (2003) 53.

[23] H.U. Cameron, J. Arthroplast. 8 (1993) 239.[24] J. Kärrholm, C. Anderber, F. Snorrason, J. Thanner, N. Langeland,

H. Malchau, P. Herberts, J. Bone Jt. Surg. 84A (2002) 1651.[25] L.D. Zardiackas, D.E. Parsell, L.D. Dillon, D.W. Mitchell, L.A. Nunnery,

R. Poggie, J. Biomed. Mater. Res. 58 (2) (2001) 180.[26] J. Banhart, JOM 52 (2000) 22.[27] M.F. Ashby, A.G. Evans, N.A. Fleck, L.J. Gibson, J.W. Hutchinson,

H.N.G. Wadley, Metal foams: a design guide, Butterworth-heinemann,UK, 2000.

[28] D.C. Dunand, Adv. Eng. Mater. 6 (6) (2004) 369.[29] J.P. Li, J.R. Wijn, C.A.V. Blitterswijk, K. Groot, Biomaterials 27 (2006)

1223.[30] O. Harrysson, D. Cormier, D. Marcellin-Little, K. Jajal, Rapid Prototyping

J. 9 (1) (2003) 37.[31] D.R. Cormier, H. West, O.L.A. Harrysson, K. Knowlson, Solid Freeform

Fabrication Symposium, Austin, Texas, USA, August 2–4, 2004.[32] C.K. Chua, K.F. Leong, C.M. Cheah, S.W. Chua, Int. J. Adv. Manuf.

Technol. 21 (2003) 291.[33] P. Kowalczyk, Comput. Methods Biomech. Biomed. Eng. 9 (3) (13) (2006)

135.[34] O. Cansizoglu, O. Harrysson, D. Cormier, H. West, T. Mahale, in press.[35] D. Brooks, Personal Communication, 2006.[36] O. Sigmund, Mech. Mater. 20 (4) (1995) 351.