Professional Documents
Culture Documents
& Design
Materials and Design 26 (2005) 391–401
www.elsevier.com/locate/matdes
Abstract
This paper is a retrospective presentation of the authorsÕ articles concerning the design and prototype manufacturing of a com-
posite femoral prosthesis. A composite hip femoral prosthesis has been designed and development process is described. The novel
hip prosthesis is a multi-material structure composed of a cobalt–chrome core surrounded by a flexible controlled stiffness composite
structure and was first assessed by means of finite-element analysis and its predicted performance compared with conventional
bonded and press-fit Co–Cr and titanium prostheses. A conceptual prototype was then manufactured and the concept validated
through an idealised numerical-experimental procedure that allowed the determination of its graded stiffness.
2004 Elsevier Ltd. All rights reserved.
0261-3069/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2004.07.024
392 J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401
shielding, but lower interface stresses. The design chal- [34]. Currently, it is not clear what cost savings, if any,
lenge is to know how to minimise stress shielding main- are possible by using composites as compared to the tra-
taining interface stresses and interface micromotions ditional technology. In spite of this, many of the tradi-
within acceptable levels [3,5]. tional metal devices are being replaced by composite
Conventional proximal femoral prostheses are man- ones and are, therefore, outperforming metal alloys [4].
ufactured with single constant modulus material (metal Most of the developed composite material prostheses
alloy). Such prostheses can only be designed optimally have been manufactured from laminated composites
to minimise stress shielding or interface stresses, and [11,13,16,20,29–32,34]. Christel et al. [12] developed
never both simultaneously. Therefore, adequate stiff- hip prostheses made from carbon fibres infiltrated either
ness is necessary to address simultaneously, within cer- with dense pyrolytic carbon or silicon carbide through
tain restrictions, stress shielding and micromotion [2,4]. chemical vapour infiltration. The implantation of these
Aiming this objective, numerical schemes have been carbon–carbon hip prostheses has been stopped due to
developed and tested to generate optimised designs. significant mechanical failures. Davidson et al. [14] man-
Kuiper [6] and Kuiper and Huiskes [5,7] developed a ufactured Corin C-Fit stems from carbon fibre rein-
numerical scheme, together with a mathematical formu- forced plastic laminates. Claes et al. [13] obtained
lation of the design objectives, and obtained material carbon fibre reinforced carbon prostheses from lami-
properties for a two-dimensional and three-dimensional nates where the carbon fibres were arranged in several
bone-prosthesis configuration. The design objectives layers in angles of 0, ±45 and 90. Reinhardt et al.
within this study have been formulated considering a [25] studied composite hip prostheses manufactured
limit to the amount of bone loss while generating a using resin transfer moulding technology. In the re-
favourable interface stress distribution. The results show search program performed by Li et al. [20], a numerical
that the implanted femur with variable modulus pros- method was developed to predict the progressive failure
thesis produces the desired load transfer distribution, of a thick laminated composite femoral component for
minimizing both calcar stress shielding and interface total hip arthroplasty. Srinivasan et al. [29] examined
micromotion. in situ laminated composite hip prosthesis structural re-
Some investigations have addressed the thematic of sponse and relative strengths as a function of walking
designing optimised hip prosthesis, although the objec- and stair climbing for better appreciation of its struc-
tive for using composite materials is not always very tural performance. In terms of relative strength, the
clear. In fact, there has been substantial active research authors found that the neck of the composite prosthesis
with composites for biomedical applications, namely for failed for all activities with the exception of the mid-
hip prostheses [2,4,8–32]. The possibilities of mainly stance phase of level walking. However, the prosthesis
stress shielding reduction, and therefore bone resorption appears to have sufficient relative strength for function
minimization, and thigh pain alleviation, have stimu- at positions distal to the neck of the prosthesis. De San-
lated the development of compliant and tailored pros- tis et al. [15] designed a composite hip prosthesis made
theses designs made from these materials. from poly(ether-imide) reinforced with carbon and glass
Composite materials allow the designer to customise fibres. The YoungÕs modulus and the strength in tension
the desired optimized stiffness and strength and are for the prosthesis was uniform along the stem axis (40
potentially excellent materials for hip prostheses appli- GPa and 600 MPa, respectively) while the elastic modu-
cations. However, it is necessary to prove the benefits lus in bending varied from 10 to 60 GPa in the tip-head
of these materials clearly, namely those properties re- direction. Yildiz et al. [31,32] studied the numerical
lated to biocompatibility and biofunctionality perform- mechanical performance of fibre reinforced composite
ance [4]. The manufacturing costs of composite hip prostheses in a femur. The main goal was to study
materials for the design of prostheses or other medical the effect of fibre orientation in the prosthesis on the re-
equipment are an important design consideration. Ten sponse of the surrounding femoral bone. Akay and
years ago, Havelin et al. [33] have estimated that nearly Aslan [8] performed a comparative stress analysis with
800,000 hip replacements were being carried out world- a short carbon–fibre reinforced PEEK (CF/PEEK) pros-
wide. This number must have increased significantly, thesis manufactured by injection moulding. Canine
being probably around 1,000,000. When carrying out re- composite stems have also been developed and clinically
search, we tend to concentrate efforts on increasing the evaluated [16,21,22]. Most of these studies had as objec-
service of life of implants rather than on lowering tech- tive the design of prostheses with identical stiffness char-
nology costs. Implants with longer life service will com- acteristics of the host bone, which does not seem to be
pensate the possible high composite technology costs, the correct design approach. These designs will probably
since one of the biggest obstacles in switching from met- allow the reduction of stress shielding, and probably
al-based to composite-based technology has to do with bone resorption, but will provoke higher interface stres-
the high investment made over the years on the technol- ses [2]. It is more or less the consensus that these types of
ogy for fabricating implants using traditional materials designs produce larger relative micromotions at the
J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401 393
bone-prosthesis interface, which are known to cause the layer (for low distal stiffness). The proposed design is
formation of an unstable fibrous tissue layer [35,36]. It based on the concept discussed by Kuiper [6], with the
was mainly due to a very low proximal stiffness of the prosthesis having a relatively high proximal-medial
first version of the RM isoelastic prosthesis that a high modulus and a receding distal modulus matching the
incidence of loosening was observed [37]. A second ver- cortical bone modulus at the prosthesis tip. The novel
sion of the same prosthesis was developed, but with a prosthesis consists of an inner cobalt–chrome (E = 200
larger proximal metallic core to higher proximal stiffness GPa) core surrounded by a flexible composite outer
and so avoid high micromotion. layer (E = 10 GPa, modulus obtained from specimen
The novel prosthesis hereby described is based on a testing).
cobalt–chrome core with a flexible composite outer As shown in Fig. 1, the geometry has been based on a
layer. By varying the composite layer thickness, and modified version of the Freeman hip and included an in-
intrinsically the geometry of the core, it was possible creased thickness (2 mm) composite layer on the proxi-
to control the prosthesis stiffness. The design concept mal medial aspect of the prosthesis for a distance of 5
was critically assessed using the finite element method. mm below the neck resection level [38]. The prosthesis
Prototypes were manufactured embedding the design was separated in two parts, the straight-backed cylindri-
concept. cal stem and the proximal medial wedge. The modulus
of the straight-backed stem was varied from 80 GPa
proximally to 20 GPa distally. The modulus of the prox-
imal medial wedge was varied from 150 GPa proximally
2. Materials and methods to 100 GPa at the intersection with the straight-backed
stem. The tip and proximal-lateral modulus values were
2.1. The design concept derived from Kuiper [6]. A different proximal-medial
modulus was used, in this case 150 GPa to avoid high
Although composite materials offer a very high de- peak interface stresses observed with a 100 GPa
gree of design freedom, it can however be extremely dif- modulus.
ficult to manufacture prosthesis with modulus materials To obtain the composite layer thickness, a mathemat-
ranging from 20 to 200 GPa. To overcome this design ical model was implemented base on the hypothesis that
obstacle, the present design concept discussed is based the stiffness of the optimised prosthesis is the same as the
on a combination of a metal core (to give high proximal stiffness of the controlled stiffness prosthesis (core and
stiffness) bonded to a reinforced polymeric composite composite layer) [38]
Composite
Co-Cr
Titanium
Cemented
Fig. 2. Comparison of the medial and lateral periosteal strain energy distribution produced by bonded prostheses.
J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401 395
4
Lateral Bonded Press-fit
Medial
3
Fig. 3. Mean and peak medial and lateral strain energies for bonded and press-fit prostheses.
energy. Similar differences can also be observed on the design, since it generates the lowest peak and mean min-
lateral face of the femur. The composite prosthesis pro- imum principal stresses. The cemented and the compos-
duces the best strain energy distribution, either as a ite designs produce very similar endosteal cancellous
bonded or press-fit implant. bone stresses. It is interesting to note that the composite
The endosteal cancellous bone minimum principal design produces lower distal peak cancellous bone stres-
stresses were also examined. All prostheses generated ses than the cemented prosthesis. The titanium stem pro-
similar cancellous bone stress distribution patterns, with duces a mean cancellous bone stress very similar to the
the peak cancellous bone stresses occurring just below cobalt–chrome stem but has a very high peak value of
the neck resection level. For the cemented stem, the 7.2 MPa. The press-fit prostheses generate significantly
stress reduces gradually to 2 MPa at almost the distal higher cancellous bone stresses (Fig. 5). The composite
portion of cancellous bone, before rising to 4 MPa at stem incorporates, as said previously, a proximal soft
the cortical-cancellous bone junction. For the cementless layer of 2 mm thick and 5 mm in length, as shown in
protheses, the cancellous bone stresses gradually de- Fig. 1, which causes a drop in the peak cancellous bone
crease distally to a value of 1 MPa. stress at this level.
The minimum principal cancellous bone stresses for All the press-fit prostheses produce mean cancellous
the bonded prostheses are shown in Fig. 4. The co- bone stresses higher than the cemented prosthesis (Fig.
balt–chrome prosthesis performs better than any other 6). The cobalt–chrome, titanium, composite stems
6
8,1
Distal 7,4
Co-Cr
Titanium
2
Composite
Proximal
0
120 140 160 180 200
Prosthesis axial position ( mm )
Fig. 4. Comparison of the minimum principal cancellous bone stresses for bonded prostheses.
396 J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401
6
7,2
Co-Cr
Titanium
Cemented
4
Composite
distal proximal
0
120 140 160 180 200
Prosthesis axial position ( mm )
Fig. 5. Comparison of the minimum principal cancellous bone stresses for press-fit prostheses.
10
bonded
press-fit
8
8,0
7,2 7,4
6
5,5 5,6
5,1
4 4,6
Fig. 6. Mean and peak cancellous bone stresses for bonded and press-fit prostheses.
produce 17%, 28% and 40%, respectively, higher mean to be anisotropic and heterogeneous in the proximal fe-
endosteal minimum principal stresses than the cemented mur. The bonded prostheses were assumed to have
stem. achieved 100% osseointegration with the surrounding
bone, whereas in reality this is unlikely to occur. Thus,
2.2.2. Discussion these results should be treated as qualitative, rather than
Two-dimensional finite element models have shown quantitative, in nature.
to give a reasonable approximation of the stress distri- The long-term success of any implant is dependent
bution in the mid-frontal plane [41]. In order to simplify on many parameters. However, bone resorption due
the models a number of assumptions were made. The to stress shielding and excessive implant migration
models were simplistically loaded with a vertically ap- [42–44] has been identified as major contributing fac-
plied joint reaction force. However, Taylor et al. tors to the aseptic loosening of proximal femoral pros-
[39,40] have shown that the cancellous bone stress distri- theses. Within the numerical study, both have been
bution is dominated by the vertical component of the evaluated by examining the periosteal strain energy dis-
joint reaction force. The cancellous bone was assumed tribution as a measure of the potential stress shielding
to be isotropic and homogeneous, whereas it is known [41], and hence bone resorption, and the initial cancel-
J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401 397
All press-fit prostheses produce cancellous bone stress of 2.14 (Tokyo Sokki Kenkyujo Co., Ltd.) were bonded
distributions that are above the 4 MPa fatigue failure to the specimens and the four elastic constants (Ez, Eh,
threshold. The cobalt–chrome prosthesis produces the mhz and mzh) obtained (Table 1).
lowest cancellous bone stresses of all the press-fit im-
plants. It appears to be desirable to have a very stiff stem 3.2. Femoral component manufacturing
in order to minimise cancellous bone stresses. Essen-
tially, this study suggest that, regardless of the implant Prototypes of the novel design were manufactured
material, a press-fit prosthesis will always generate high using compression moulding composite technology.
migration rates and a corresponding high incidence of The same type of composite structure manufactured
late aseptic loosening due to the fatigue failure of the for the tubular specimens was applied in the manufac-
surrounding cancellous bone. The composite stem stud- turing of prototypes. Pre-forms were wrapped around
ied transfers the majority of the load in the most proxi- a metal core, impregnated with resin and placed in a
mal region, causing significantly higher cancellous bone mould, which was closed using a hot press stamp and
stresses than the cobalt–chrome stem, which distributes left curing at room temperature for 24 h. The prototype
the applied joint reaction force evenly over the available was left post curing for 24 h at 80 C. Fig. 7 illustrates
load-bearing area. schematically the manufacturing of the prototype stems.
For the bonded prostheses, the composite design Fig. 8 illustrates one of the manufactured prototypes [2].
analysed combines the most favourable features of
the load-transfer mechanisms of a stiff and a flexible
stem. It should be noted that it still provokes stress 4. Concept validation
shielding in the calcar region. Another important
observation is that the composite prosthesis hereby To determine the variation of the prototype prosthe-
studied did not improve the predicted performance of sis stiffness, the method idealised consisted on the com-
press-fit designs. parison of the prosthesis lateral displacement field
measured using non-contact experimental technique,
namely electronic speckle pattern interferometry (ESPI)
3. Prototype manufacturing with the one obtained with an analytical model. The
detailing of the numerical–experimental procedure can
3.1. Material stiffness characteristics be seen in [45].
The experimental displacements were obtained with a
For the numerical simulation and determination of prototype to determine the variation of its stiffness and
stiffness properties of the composite material, three compared with identical results derived from an analyt-
tubular specimens of hybrid carbon–glass fibre rein-
forcement were manufactured. Commercial braided
pre-forms (Cobrasoxe – A&T Technology, Kentucky,
USA) with diameter of 38.1 mm and a braiding angle
of 45 were used. For a volume fraction of 50%, the
mean thickness of these reinforcement structures is 0.3
mm. Three pre-forms were impregnated with an epoxy
resin system (Beresin L 84A resin with 20% Beresin L
84B catalyst), wrapped around a 40 mm diameter poly-
ethylene tubular bag (Walsroder Perfect) and placed in-
side an aluminium mould. A pressure of 5 bar was
introduced into the polyethylene bag, pressing the
impregnated laminate towards the inner mould surface.
The laminate was then left to cure for 24 h at room tem-
perature. All the specimens were post-cured at 80 C for
12 h. The dimensions and volume fibre of the specimens
are given in Table 1. The volume fibre content was deter-
mined according to the Portuguese Standard NP 2216/
1988 (determination of mass loss by calcinations of glass
fibre reinforced plastics) [4].
The axial and circumferential strains developed at the
outer surface of the tubular specimens were measured
performing internal pressurisation tests. To determine
these, biaxial rosettes GFCA-3-350-50 with gauge factor Fig. 8. Controlled stiffness composite femoral prosthesis.
J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401 399
ical model. The prosthesis was fixed and loaded as as- with the ESPI technique was 4.3 lm; for the analytical
sumed in the analytical model, as a cantilever beam. A model, the maximum displacement was 4.1 lm, a rela-
stiff support was used to hold the prosthesis 30 mm from tive difference of 5%.
its distal end. The magnitude of the displacement field
measured was of the order of the micron, therefore a
very low load of P = 0.2 N was used to obtain a reason- 5. Concluding remarks
able amount of fringes and to calculate the displacement
values. The controlled stiffness composite concept prosthesis
The displacement field obtained resulted from the showed its adequacy for the design of a hip prosthesis,
phase map calculation by using an image-processing considering a compromise between strain energy distri-
algorithm based on phase stepping techniques. The 2p bution (a parameter used to determine the stress shield-
phase difference corresponds to a k/2 (k = 632 nm) dis- ing effect and therefore the degree of bone resorption)
placement; therefore, the displacement point on the and cancellous bone stresses (parameter to determine
prosthesis surface was obtained by y(z) = fringe or- the prosthesis migration). By combining a stiff material
der · 316 nm. with a more flexible one, it is possible to obtain a design
Fig. 9 shows the graphic comparison of the ESPI and with desirable controlled stiffness to meet the require-
numerical integration displacement fields. A relatively ments of the compromise between the two parameters
good correlation between these was obtained using the studied: stress shielding and migration.
method described, and so we can state with some cer-
tainty that the prototype has the optimised stiffness de-
fined in Section 2. However, at the distal part of the Acknowledgements
prosthesis, where it was rigidly fixed, the highest dis-
placement difference was observed, which is a natural The authors express their acknowledgments to all col-
phenomena due to a typical experimental perturbing re- leagues that somehow contributed to this work. Spe-
gion. The displacement difference near the prosthesis fix- cially thanks to Prof. Mario Vaz and Jaime Monteiro
ation device was 400% and gradually diminished up to a (LOME-INEGI) for the help given with the ESPI work;
difference of 2.5% at the last observed fringe, which was Dr. Mark Taylor from the University of Southampton
localised at a distance of 15 mm from the applied load. for the help given with the numerical simulations; Carlos
The higher differences observed at the fixation region of Relvas from the Department of Mechanical Engineering
the prosthesis may have resulted from a less efficient fix- of the University of Aveiro and José Carlos from Nor-
ation, which in the analytical–numerical model is con- cam-Engenharia e Design Industrial for the machining
sidered ideally. The maximum displacement obtained of the moulds.
400 J.A. Simões, A.T. Marques / Materials and Design 26 (2005) 391–401
[42] Freeman MAR, Plante-Bordeneuve P. Early migration and late [44] Walker PS, Mai SF, Cobb AG, Bentley G, Hua J. Prediction of
aseptic failure of proximal femoral prostheses. J Bone Joint Surg clinical outcome of THR from migration measurements on
B 1994;76:432–438. standard radiographs. J Bone Joint Surg B 1995;77:705–714.
[43] Kärrholm J, Borssen B, Löwenhielm G, Snorrason F. Does early [45] Simões JA, Monteiro J, Vaz MA. Numerical–experimental
micromotion of the femoral stem prostheses matter. J Bone Joint method for the validation of a controlled stiffness femoral
Surg B 1994;76:912–917. prosthesis. J Biomech Eng 2001;123:234–238.