Wear tests in a hip joint simulator of different material counterfaces on the Ultra High Molecular Weight Polyethylene (uhmwpe) the materials used as counterfaces were forged with hand polished and mass finished, CoCrMo coating applied on the forged CoCrMo alloy obtained by PVD. The results showed that the hand polished alloy caused the higher UHMWPE wear of the acetabular cups.
Wear tests in a hip joint simulator of different material counterfaces on the Ultra High Molecular Weight Polyethylene (uhmwpe) the materials used as counterfaces were forged with hand polished and mass finished, CoCrMo coating applied on the forged CoCrMo alloy obtained by PVD. The results showed that the hand polished alloy caused the higher UHMWPE wear of the acetabular cups.
Wear tests in a hip joint simulator of different material counterfaces on the Ultra High Molecular Weight Polyethylene (uhmwpe) the materials used as counterfaces were forged with hand polished and mass finished, CoCrMo coating applied on the forged CoCrMo alloy obtained by PVD. The results showed that the hand polished alloy caused the higher UHMWPE wear of the acetabular cups.
Materials Science and Engineering C 29 (2009) 153158
Contents lists available at ScienceDirect
Materials Science and Engineering C
j ournal home page: www. elsevier. com/ locate/ ms ec
Wear tests in a hip joint simulator of different CoCrMo counterfaces on UHMWPE V.A. Gonzlez-Mora a , M. Hoffmann a , R. Stroosnijder a , F.J. Gil b,
a Institute for Health and Consumer Protection, Joint Research Centre, European Commission, Ispra, Italy b CREB, Dept. Ciencia de Materiales e Ingeniera Metalrgica, ETSEIB, Universidad Politcnica de Catalua, Av. Diagonal 647, 08028-Barcelona, Spain
A R T I C L E I N F O
Article history: Received 13 January 2008 Received in revised form 27 May 2008 Accepted 4 June 2008 Available online 18 June 2008 Keywords: CoCrMo Metals Wear Hip joint simulator
1. Introduction A B S T R A C T
The objective in this work was to study the effect of different material counterfaces on the Ultra High Molecular Weight Polyethylene (UHMWPE) wear behavior. The materials used as counterfaces were based on CoCrMo: forged with hand polished and mass nished, CoCrMo coating applied on the forged CoCrMo alloy obtained by Physical Vapour Deposition (PVD). A hip joint simulator was designed and built for these studies. The worn surfaces were observed by optical and scanning electron microscopy. The results showed that the hand polished CoCrMo alloy caused the higher UHMWPE wear of the acetabular cups. The CoCrMo coating caused the least UHMWPE wear, while the mass nished CoCrMo alloy caused an intermediate UHMWPE wear. It is shown that the wear rates obtained in this work are closer to clinical studies than to similar hip joints simulator studies. 2008 Elsevier B.V. All rights reserved.
resembles exactly those of the hip joint replacements. The acetabular When the natural joint has to be replacing with articial materials, there is a change in the tribological situation due to the inability of the actual materials used to produce an articial permanent lubricating lm. Therefore, the materials used to articulating components in an articial joint are always subject to wear. Furthermore, there is no ideal bearing material currently fulls all the requirement of total joint replacement design. Nowadays, the penetration rates of the femoral heads into the acetabular cups are so low that there should not be any problem associated with the design or functionality of the implant. However, the wear particles, especially from UHMWPE debris, cause osteolysis due to a particle-initiated foreign body reaction (leading to aseptic loosening of the stem or acetabular cup) and nally make a revision operation necessary [1,2]. Joint simulators were developed for predicting the in vivo wear rates of total joint replacements by means of laboratory tests. In these tests the motion, load, lubrication, environment and geometries of the articulation are more similar to those found clinically than in screening wear test devices [3,4]. The objective of this work was to study the effect of different material counterfaces on the UHMWPE wear behavior by means of a hip joint simulator (HJS).
2. Materials and methods
The HJS used in this study is a three-station machine that operates with a type of motion termed biaxial rocking motion (BRM). As a simulator device, the geometric conguration of the components
Corresponding author. E-mail address: francesc.xavier.gil@upc.edu (F.J. Gil). 0928-4931/$ see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.msec.2008.06.006 components (cups) and the femoral components (heads) are mounted in the so-called anatomical position [13] that is, with the cups sliding over the heads, as is the case of the natural joint. Fig. 1 shows a schematic representation of the HJS motion/loading conguration. In the HJS, a biaxial rocking motion is applied to the femoral heads via a rotating inclined block (angle of inclination of 23) mounted below the femoral head. The BRM provides a practical-engineering imple- mentation of the walking cycle in vitro. The BRM includes exion- extension and abductionadduction motion, both with a sinusoidal excursion of 23 but with a phase difference of /2. The excursion of the abductionadduction motion is four times higher than the true abductionadduction motion of the human hip joint when walking. Thus, BRM provides exaggerated walking kinematics. However, the HJS has been designed to provide a practical imitation of the motions and loads seen by the hip joint during a typical walking cycle [46]. In the HJS, the motion can be synchronized to physiological loading proles [710] describing a double-peak load waveform, that results in a closer reproduction of the true dynamic load of the hip joint when walking, as performed by Mejia [1], Wang [5] and Clarke [3]. However, Saikko [4] has used constant load, hypothesizing that since the BRM is not a close reproduction of the true situation the load need not be, either. The constant load applied (1000 N) corresponds to the average load of the Paul load cycle [2]. The tests in the following described were also performed under a constant load, based on the results by Saikko [4], which addressed the possibility that static load in conjunction with BRM can be sufcient to produce realistic wear simulation for UHMWPE in hip joint replacements [1114]. Additionally, static or constant load simplies the wear test device as well as the method. All the before mentioned studies involving BRM joint simulators have been proven as a realistic and practical wear test method. 154 V.A. Gonzlez-Mora et al. / Materials Science and Engineering C 29 (2009) 153158
Table 2 Roughness and hardness of each material tested Material Roughness Ra (m) Hardness (HVN) Hand polished 0.030.01 67321 Mass nished 0.050.01 84062 CoCrMo coating 0.100.01 88428
antibacterial agent. The serum was purchased at Sigma-Aldrich SrI (Calf serum, bovine donor; product No. C9676). The soak adsorption of the UHMWPE cups was determined using an additional control cup. The cleaning and drying of the UHMWPE cups were performed according to the ASTM 1715 standard. Weighing was carried out with a Mettler Toledo AT261DeltaRange microbalance with an accuracy of 10 g. All specimens were observed in an optical microscope Olympus and a JEOL 6400 scanning electron microscope at an electron beam voltage of 20 kV. This equipment has a microanalysis EDS. Acetabular cups made of UHMWPE GUR1020 and previously sterilized with 25 KGy (2.5 Mrad) gamma radiation were used. The cups were supplied by SAMO S.p.A. The density was 0.9737 g/cm 3 . The cups are designed to be used with a metal back component and in conjunction with 28 mm femoral heads. The articulating counterfaces were therefore 28 mm femoral heads. Three different femoral head materials were studied; listed below the standard material in this study was a hot-forged CoCrMo alloy. Table 1 shows the chemical composition of this material. The microhardness was made on the 18 samples obtaining 10 measurements in each sample. A Leitz Miniload 2 instrument was used applying a 490.3 mN load on a pyramidal diamond indenter. The surface roughness measurements were made using a prolometer (Mitutoyo Surftest SV-512; Mitutoyo, IL, USA) with a 5 nm resolution and assisted with appropriate software (Surfpack, v 3.0, Mitutoyo, Japan). Before evaluation, a Gaussian lter was used to remove errors of form and waviness. For each specimen ve different lengths (sampling length 0.8 mm, and transversing length 2.5 mm) were analyzed following the ISO/JIS B0601. Ra was used to give a numerical characterization of the surface roughness. Ra is the arithmetical mean deviation of the prole and is calculated as the arithmetical mean of Fig. 1. Motion/loading conguration of the HJS wear test machine.
The tests on the HJS wear machine were performed as follows. The acetabular cups were mounted on a home-made metal back support to transmit the load to each of the three stations. The cups were then pressed against the femoral heads. While the cups remain static, the heads slide according to the biaxial rocking motion. A constant load of 1000 N (102 kg) per station was applied during the test. The frequency of the motion was 1.23 Hz (810 ms/cycle) [1820]. In the HJS a cycle is considered as completion of one rotation of the head. The wear of the UHMWPE cups was determined by weight loss measurements every 333,333 cycles up to a total test length of 3 million cycles. The test lubricant was replaced with fresh solution after every weighing stop and distilled water was added during the test for compensating water evaporation. Each station has an environmental test chamber made of a transparent polycarbonate wall. Each chamber was lled with 350 ml of lubricant. As test lubricant, a solution consisting of bovine serum and distilled water was used with a total protein concentration of 30 mg/ml. The lubricant includes sodium azide (0.1 mg/l) as
Table 1 Chemical composition of the forged CoCrMo alloy (%) Element Cr Mo Mn Ni Si Fe C N Balance 2630 57 max 1 max 1 max 1 max 0.7 max 0.35 max 0.25 the absolute values of the prole deviations from the mean line. Table 2 the hardness and roughness of each material. CoCrMo forged (hand polished) CoCrMo forged (mass nished)
Fig. 2. Average wear losses of the UHMWPE cups in the HJS wear tests. V.A. Gonzlez-Mora et al. / Materials Science and Engineering C 29 (2009) 153158 155
Hand polished 90 0.998 Mass nished 68 0.993 CoCrMo coating 44 0.998
CoCrMo forged with a CoCrMo coating obtained by means of Physical Vapour Deposition (PVD). The coating had the same chemical composition as the substrate. Vapour deposition process required the creation of material vapour via sputtering and their subsequent condensation onto a substrate to form the lm. In the magnetron sputter deposition process was applied a magnetic eld parallel to the cathode surfaces forms electron traps and restricts the primary electron motion to the vicinity of the of the cathode. The magnetic eld strength was around 200 G. Therefore, it can inuence the plasma electrons but not the ions. The pressure was 130 mPa. For each articulating counterface material three heads were tested. A total of 18 wear tests were performed.
3. Results and discussion
The wear of the UHMWPE specimens (cups) in the HJS wear test method is shown in Fig. 2, where the volumetric wear rates (mm 3 ) of the UHMWPE cups are represented as a function of test duration in cycles for each the head material. The volumetric wear results are calculated from the average weight loss of three specimens. The results after 3 million cycles show a high linearity. However, it is more interesting to compare the wear rates of the sliding couples than to compare the volumetric wear after the one million cycles of the wear test. The wear rate values are calculated from the slope of the linear regression tting to the volumetric wear data and are presented in Table 3. The results presented show that the hand polished CoCrMo alloy caused the higher UHMWPE wear of the acetabular cups, when tested with the HJS. The CoCrMo coating caused the least UHMWPE wear, while the mass nished CoCrMo alloy caused an intermediate UHMWPE wear. It is generally considered that the simulator wear tests produce lower wear rates than prostheses in vivo. The reasons for that have been widely discussed in different studies [1518]. In Table 4, the results of this work are compared to those of other hip joint simulator studies employing similar femoral heads (28 mm CoCrMo alloy), UHMWPE and serum-based lubricant. The comparison shows that a higher UHMWPE cup wear is obtained in this work than those found for other laboratory tests. Furthermore, in Table 5, it is shown that the
Table 4 Hip joint simulator wear tests for 28 mm CoCrMo femoral heads Study Wear rates (mm 3 /10 6 cycles) Smith and Unsworth [17] 48.2 51.4 Barbour et al. [19] 41.6 Greer et al. [20] 30 Hamilton et al. [21] 35 McKellop et al. [22] 30 Sanford et al. [23] 22.5 Saikko and Ahlroos [4] 24.5 Smith and Unsworth [24] 51.7 49.2 McKellop et al. [25] 34.3 Park and Mckellop [26] 38.1 AverageSD 3810 Results of this work 44/68/90 Table 5 Clinical wear rates for 28 mm CoCrMo femoral heads Study Wear rates (mm 3 /10 6 cycles) Wooson and Murphy [27] 86 Kabo et al. [28] 144 Bankston et al. [29] 31 Callaghan et al. [30] 86 74 68 43 Bankston et al. [31] 31 68 Hop et al. [32] 84 59 48 65 Shaver et al. [33] 47 27 Paul et al. [34] 64 AverageSD 6429 Results of this work 44/68/90
wear rates obtained in this work are closer to clinical studies than to similar hip joint simulator studies. This suggests that the testing conditions selected for this work reproduce quite realistically the UHMWPE wear found in vivo. Specially, the selection of the adequate lubricant, the protein concentration, additives, are of the maximum importance and may lead to high variation in the wear results, even if the tested materials are the same. Observation of the UHMWPE cups with an optical microscope was performed after the wear tests. In Fig. 3, the differences between an UHMWPE cup before and after the wear test can be seen. A cup after three million cycles is compared to the cup used as control that represents the surface of the cups before the wear tests. The cups after began wear tests present a highly polished surface when compared to the control cup. Note the evident difference in light reection between them in Fig. 3. The highly polished surface was also present in the other cups. During the observation of the cups, three main wear zones were identied, as shown one form the other. This three wear zones are shown in Fig. 3. The wear zone corresponding to the polar region of the cup is denominated as polar zone (A). The next regions are denominated as medial zone (B) and border zone (C). Additionally, there exist a region where there was no contact between cup and head which is denominated as non-wear zone (D). Fig. 4, shows the differences between the wear zone (contact zone) and non-wear zone (D) on the acetabular cup. Near the rim of the cup the non-wear zone can be identied by its less light-reectivity, while the wear zone highly reects the light, proving thus to be a highly polished surface [3538].
Fig. 3. Photograph of a wear-tested UHMWPE cup (left side) and a control cup (right side). The wear-tested cup is divided in four zones denominated: A, polar zone; B, medial zone; C, border zone; and D, non-wear zone. 156 V.A. Gonzlez-Mora et al. / Materials Science and Engineering C 29 (2009) 153158
CoCrMo coated sliding pair, as it has been indicated by weight loss measurements. The medial zone (B) is in contrast a region with fewer polished regions and represents a transition zone between the border and polar zones. It reects less the light due to bands of the UHMWPE material which were not removed during the wear process. These bands are present in almost the whole cup diameter, forming a ring-like band on the cup. For the UHMWPE/hand polished sliding pair, the bands were smaller than those found for the UHMWPE/CoCrMo coated sliding pair, indicating a bigger removal of UHMWPE material caused by the hand polished heads. The last coincides again with the weight loss results. Finally the border zone (C), showed scratches directed to the centre of the cup and a very highly polished appearance, more than in the polar zone. In this zone, more scratches were found for the cups sliding against hand polished forged CoCrMo heads. UHMWPE cups sliding against hand polished forged CoCrMo heads presents a highly polished surface on the polar zone with multi- Fig. 4. Detail of the wear (C) and non-wear zones (D) in an UHMWPE cup when tested in the HJS. On the left image a photograph of the cup rim showing the limit of the wear zone.
The cups were sectioned after the wear tests for observation. The results are presented in the Fig. 5 showing the best and the least sliding couples: UHMWPE against hand polished and CoCrMo coating. Fig. 5 has shown interesting features of the UHMWPE cups after three million cycles: The majority of the polar zone (A) showed a very polished appearance. This is the zone in the cup where the effects of wear were more pronounced. The main feature of the polar zone is the multidirectional scratches. Differences can be found for the UHMWPE/ hand polished and UHMWPE/CoCrMo coated sliding pairs. While for the rst the polar zone appears to be totally free of UHMWPE remaining material, for the UHMWPE sliding against the CoCrMo coated heads, isolated islands of UHMWPE material remained after the three million cycle wear test. The presence of UHMWPE was determined by EDS microanalysis. This indicates higher wear damage for the UHMWPE/hand polished sliding pair than for the UHMWPE/ directional scratches, although these cups were somehow prone to unidirectional scratches. The standard deviation noticed from the weight loss measurements where noted during the optical observa- tion, where differences from cup to cup could be appreciated. UHMWPE cups sliding against CoCrMo coated heads, presented quantitatively fewer scratches than the cups sliding against hand polished forged CoCrMo heads. This superior number of scratches on the worn surfaces indicates a higher roughness of the hand polished heads compared to the CoCrMo heads, this may be responsible for the higher UHMWPE wear found for the cups sliding against hand polished heads. In the polar zone, many multidirectional scratches were produced and isolated islands of UHMWPE material. The wear zones were imaged with a SEM to observe the very ne microstructure remaining after the wear tests. The SEM observation was focused in the formation of UHMWPE particles that detach from the worn surface producing wear debris [3941]. Images were obtained with secondary electrons. The border zones in all cups have a similar ripple-like microstructure; there is almost no UHMWPE particle formation in this wear zone. For all the cups observed, the
Fig. 5. Composition of optical micrographs of a UHMWPE cup after three millions cycles sliding against a hand polished (forged) CoCrMo and CoCrMo coated head. (A) Border zone, (B) medial zone and (C) polar zone. V.A. Gonzlez-Mora et al. / Materials Science and Engineering C 29 (2009) 153158 157
Fig. 6. SEM micrographs of the medial zone in a cup sliding against a hand polished (forged) and coated CoCrMo head at different magnications.
microstructure of the polar zones appeared likely to be ripple-like but less accentuate than in the border zone. On the other hand, the UHMWPE particle formation could not be seen. The medial zones presented the most interesting surface features. For the cups sliding against hand polished heads, the low magnica- tion micrograph (Fig. 6A) shows the bands, which present a different morphology as the surface surrounding. When increasing the magnications (Fig. 6B) it can be seen that the bands in the medial zone consist of UHMWPE material, where the production of UHMWPE brils is evident as well as a ripple-like microstructure. The brils are several microns in length and extend over several ripples (Fig. 6C). For the cups sliding against CoCrMo coated heads, the microstructure resembles what was found for the UHMWPE cups surface sliding against the hand polished heads. An example of the bands founded as isolated island is presented in Fig. 6D. At higher magnications it can be seen a ripple-like microstructure and UHMWPE brils (Fig. 6E and F). However, brils are smaller and fewer when compared with those formed in the UHMWPE/hand polished sliding pair. It seems therefore that wear is in a less advanced stage to what observed for the UHMWPE/hand polished sliding pair [4244]. This lower bril production demonstrates the lesser damage causes on the worn cup surface for the UHMWPE/CoCrMo coated sliding pair.
4. Conclusions
The results of the HJS wear tests have shown that the wear resistance of UHMWPE sliding against the CoCrMo coating was higher than when sliding against mass nished and hand polished CoCrMo alloys. Besides, the mass nishing treatment on CoCrMo alloys reduced the UHMWPE wear when compared to hand polished CoCrMo. The HJS parameters used in this work reproduce adequately the in vivo conditions, according to the agreement of the results obtained with the clinical results.
References
[1] L.C. Mejia, T.J. Brierley, Biomed. Mater. Eng. 4 (4) (1994) 259. [2] J. Paul, Proc. Inst. Mech. Eng. Part 3J 181 (1996) 8. 158 V.A. Gonzlez-Mora et al. / Materials Science and Engineering C 29 (2009) 153158
[3] I. Clarke, V. Good, L. Anissian, A. Gustafson, Proc. Inst. Mech. Eng. Part H 211 (No H1) (1997) 25. [4] V. Saikko, T. Ahlroos, J. Eng. Med. W51 (2004) 123. [5] A. Wang, C. Stark, J. Dumbleton, Proc. Inst. Mech. Eng. Part H 210 (No H3) (1996) 141. [6] C. Bragdon, D. O'Connor, J. Lowestein, M. Jasty, W. Syniuta, Proc. Inst. Mech. Eng. Part H 210 (No H3) (2006) 157. [7] ASTM F732, Standard practice for reciprocating pin-on-at evaluation of friction and wear properties of polymeric materials for use in total joint prostheses, in 1998 Annual book of ASTM standards, Medical devices, vol. 13.01, emergency medical services, 1998. [8] J. Davidson, R. Poggie, A. Mishra, Bio-Med. Mater. Eng. 4 (1994) 213. [9] N. Suh, M. Mosleh, J. Arinez, Wear 214 (1998) 231. [10] M. Fernndez-Fairn, P. Sala, FJ. Gil, Bio-Med. Mater. Eng. 16 (2006) 415. [11] J. Pea, J.M. Guilemany, F.J. Gil, Metall. Mater.Trans., A 37A (2006) 1175. [12] J. Charnley, Low friction and arthroplasty of the hip, theory and practice, Springer, New York, 1979. [13] J. Fisher, D. Dowson, H. Hamdzah, H. Lee, Wear 175 (1994) 219. [14] S. Young, M. Lolito, T. Keller, Wear 222 (1998) 29. [15] W.L. Sauer, M.E. Anthony, in: Jacobs, Craig (Eds.), ASTM STP, vol. 1346, 1998, p. 1. [16] V. Saikko, T. Alroos, Proc. Inst. Mech. Eng. Part H 213 (4) (1999) 301. [17] S.L. Smith, A. Unsworth, Proc. Inst. Mech. Eng. Part H 213 (6) (1999) 475. [18] A. Wang, D. Sun, C. Stark, J. Dumbleton, Wear 181183 (1995) 241. [19] P.S. Barbour, M.H. Stone, J. Fisher, Proc. Inst. Mech. Eng. Part H. 211 (H3) (1997) 265. [20] K.W. Greer, M.B. Scmidt, I.B. Tariah, 21st Annual Meeting of the Society for Biomaterials, 1995, p. 227. [21] J.V. Hamilton, M.B. Schmidt, C. Shah, 43rd Annual Meeting of the Orthopaedic Research Society, 1997, p. 782. [22] H.A. McKellop, F.W. Shen, Y.J. Yu, B. Lu, R. Salavey, 43rd Annual Meeting of the Orthopaedic Research Society, 1997, p. 94. [23] W.M. Sanford, W.C. Moore, D. McNulty, C. Frisinger, T.P. Schmalzired, 43rd Annual Meeting of the Orthopaedic Research Society, 1997, p. 95. [24] S.L. Smith, A. Unsworth, Proc. Inst. Mech. Eng. Part H 214 (3) (2000) 223. [25] H.A. McKellop, F. Shen, W. DiMaio, J.G. Lancaster, Clin. Orthop. 369 (1999) 73. [26] S.H. Park, H.A. Mckellop, Spine 33 (6) (2008) 597. [27] S.T. Wooson, M.G. Murphy, J. Bone Joint Surg. 75B (2003) 1311. [28] J.M. Kabo, J.S. Beghard, G. Loren, H.C. Amstutz, J. Bone Joint Surg. 75B (1993) 254. [29] A.B. Bankston, H. Cates, M.A. Ritter, E.M. Keating, P.M. Paris, Clin. Orthop. 317 (2005) 227. [30] J.J. Callaghan, D.R. Pedersen, J.P. Olejniczak, D.D. Goetz, R.C. Johnston, Clin. Orthop. 317 (2005) 14. [31] A.B. Bankston, E.M. Keating, C. Ranawat, P.M. Paris, M.A. Ritter, Clin. Orthop. 317 (2005) 37. [32] J.D. Hop, J.J. Callaghan, J.P. OLejniczak, D.R. Pedersen, T.D. Brown, R.C. Johston, Clin. Orthop. 344 (1997) 20. [33] S.M. Shaver, T.D. Brown, S.L. Hillis, J.J. Callaghan, J. Bone Joint. Surg 79 A (1997) 690. [34] A. Paul, Y. Yamamoto, T. Masako, Tribol. Trans. 12 (2007) 103. [35] D. Dowson, S. Taheri, N. Wallbridge, Wear 119 (1987) 277. [36] J. Pea, E. Solano, A. Mendoza, J. Casals, J.A. Planell, F.J. Gil, BioMed. Mater. Eng. 15 (2005) 289. [37] J.M. Manero, F.J. Gil, M.P. Ginebra, J.A. Planell, A. Artola, I. Goi, M. Gurruchaga, J. Biomater. Appl. 18 (04) (2004) 305. [38] R. Hall, A. Unsworth, P. Siney, B. Wroblewski, Proc. Inst. Mech. Eng. Part H 210 (1996) 197. [39] F.J. Gil, R. Canedo, A. Padrs, M.V. Baeres, J.M. Arano, Bio-Med. Mater. Eng. 01 (2002) 1. [40] A. Wang, V. Polineni, C. Stark, J. Dumbleton, 44th Annual Meeting, Orthopaedic Research Society, New Orleans, USA, March 1619, 1998. [41] M. Northeld, T. Scmalzried, G. Belcher, H. Amstutz, J. Bone Joint Surg. 80-A (1998) 54. [42] J. Pea, F.J. Gil, J.M. Guilemany, Acta Mater 50 (2002) 31153124. [43] F.J. Gil, J.A. Planell, Proc. Inst. Mech. Eng. Part H 212 (No H6) (1998) 473. [44] F.J. Gil, J.A. Planell, J.Biomed. Mater. Res. 48 (5) (1999) 682.