- Email: [email protected]
An experimental study of damage accumulation in cemented hip prostheses
Uinicul Biomechanics Vol. 11, No. 4, pp. 214-219. Copyright
0 1996 Elsevier
1996 Science Limited. All rights reserved Printed in Great Britain 0268-0033196 $15.00 + 0.00
ELSEVIER
B A 0 McCormack ‘Bioengineering College Dublin, Ireland
MSC~,
P J Prendergast
Research Centre, Department and 2Department of Mechanical
PhD*, D G Gallagher
MEngSc’
of Mechanical Engineering, University Engineering, Trinity College, Dublin 2,
Abstract Objective. To develop a methodology to characterize the pattern of crack initiation and damage accumulation in intramedullary fixated cemented prostheses. Design. An experimental physical model of intramedullary fixation was developed which both represents the implant structure and permits monitoring of fatigue crack growth. Background. Many joint replacement prostheses are fixed into the medullary cavity of bones using a poly(methylmethacrylate) ‘bone cement’, which forms a mantle around the prosthesis and locks it to the bone. The endurance of the replacement is, to a great extent, determined by the mechanical durability of the cement and the implant interfaces under cyclic stresses generated by dynamic loading. The cement mantle is subjected to complex multiaxial stresses which vary in particular distribution depending on the prosthesis design. Methods. Damage accumulation is reported in terms of the number of cracks, the location of cracks, and the rate of crack growth. Results. The results clearly show the nature of damage accumulation in the cement mantle, and that many of the cracks which propagate within the cement mantle are related to cement porosity. Conclusion. This study gives experimental evidence to support the hypothesis of a damage accumulation failure scenario in cemented hip reconstructions. Relevance Cementing is the most popular technique for the fixation of joint replacement prosthesis. However, the sequence of events leading to the failure of cemented fixation is not fully understood. In this paper it is shown that damage accumulation can be directly monitored in an experimental model of cemented intramedullary fixation. Copyright @ 1996 Elsevier Science Ltd. Key words: Damage accumulation, Clin. Biomech.
prosthesis fixation, hip replacement, failure scenario
Vol. 11, No. 4, 214-219,
1996
Introduction
Most joint replacement prostheses have an intramedullary component which is fixed into the bone using a poly (methylmethacrylate) ‘bone cement’. For example, the femoral side of hip replacement, the tibia1 component of knee replacement and both the humeral and radial components of elbow replacements can all be fixed in this way’. Breakdown in the mechanical integrity of the cement layer is one route to clinical failure of these joint reconstructions2. This breakdown .l---.--
Received: 13 April 1995; Accepted: 23 November
1995 Correspondence and reprint requests to: B A 0 McCormack MSC, Bioengineering Research Centre, Department of Mechanical Engineering, Ilniversity College Dublin, Dublin 4, Iretand
can be associated with failure of the cement mantle by the process of ‘damage accumulation’. Damage accumulation is defined as the propagation of small internal flaws and microcracks which eventually form critical defects’. Cracks have been reported in the cement mantles of autopsy-retrieved hip prostheses even though there were no signs of clinical failure at the time of death4. These cracks were seen both within the cement mantle and on the interfaces, and were present in all retrievals (including those which had been implanted for only 3 years). Other retrieval studies have shown that pores and voids are formed during the insertion of the cernen$(j, an observation that has been made many times in bone cement specimens prepared under laboratory conditions7s. Fatigue striations on the fracture surfaces of retrieved mantles prove that, in
McCormack
et al: Damage
certain cases at least, not only do cracks occur in viva but that they also propagate to cause failure’. The objective of this investigation was to develop an experimental methodology for the analysis of microcrack growth in intramedullary-fixated implant structures. There are many studies reporting the particular nature of the stress pattern in the cement mantle and on the interfaces of intramedullary prostheses”,“. These stresses induce a complex failure mechanism in the cement mantle’* that cannot be fully regenerated in a simple test specimen. Therefore an experimental model must be designed to reproduce the essential features of this stress distribution whilst at the same time being simple enough to facilitate crack observation. One of the common problems encountered when attempting to study cracks in real structures is simply one of crack observation and crack growth monitoring. In a hip arthroplasty, the cement mantle is encased in bone and, even if the mantle were exposed, it is usually difficult to see into the depth of the cement. Nondestructive crack detection techniques such as acoustic emission13 and ultrasound14 have been used, but the accurate location of new cracks, and accurate tracking of crack growth is not possible using these techniques. Destructive testing such as sectioning or splitting the test specimen have been developedi5. However, destructive testing means that several different specimens must be used to determine what happens over time and an error due to inter-specimen variation is introduced. In addition, destructive techniques do not allow the distinction between pre-test cracks and cracks initiated during testing, and accurate tracking of damage accumulation under fatigue loading is therefore difficult. An experimental model which represents the implant structure in respect of its material properties, interface conditions, and geometry was developed. The design is achieved subject to the constraint that damage accumulation must be continuously visible during fatigue loading. In this way we can successfully monitor crack initiation sites and the growth characteristics of cracks. The growth of many cracks is observed (as opposed to just the rapid emergence of one dominant crack) and this supports the hypothesis of a damage accumulation failure scenario for cemented intramedullary reconstructions, such as hip joint replacements.
accumulation
in cemented
hip prosthesis
215
Methods
The first part of this section describes the design of the experimental model and the second part describes the methods of crack observation and analysis. The experimental model
The design concept for the model is a structure comprising a tapered metal stem with a cement layer on either side and a layer of bone on the outside of the cement layers, held intact by a component which replicates the circumferential force transfer of the bone, see concept design of Figure 1. Hence the structure has five layers (bone/cement/prosthesis/cement/bone) and is open on one side to allow crack observation. To permit the transmission of light through the cement layers, three ‘windows’ were machined into the back wall of the model. To ensure internal visibility of the cracks and flaws, a clear poly (methylmethacrylate) cement was used. The experimental test model may be considered as a longitudinal slice taken from the centre of an idealized intramedullary fixation. Dimensions were calculated to ensure that the experimental model transferred the stress through the cement mantle in a similar way to the real intramedullary system. Beams-on-elastic-foundation theory is applied to calculate the loading in this system, as has been applied previously to the femoral component of the artificial hip joint l&i’ The theory assumes that bone and prosthesis are stiff beams separated by an elastic layer, which is modelled as a set of springs. Using this approach, Huiskes” resolved the joint loading into axial and transverse components and carried out separate analysis for each. He has shown that the fundamental load transfer pattern of intramedullary fixation involves proximal stress transfer to the bone, a central region where the structure behaves as a composite beam, and a distal region of load transfer to the prosthesis. Using the beams-on-elasticfoundation approach (see Appendix), we evaluated many alternative designs of experimental model. The dimensions given in Figure 2 were those of the final design. For the final design of the experimental model it was calculated that a peak load of 1 kN applied to the stem develops stresses in the range O-6 MPa in the cement layers. Specimen manufacture and testing method
BOW.2 Ph4MA Stem/Holder
Channel Section “holder”
Figure 1. The experimental model is a five-layered together with a channel section.
structure
held
Six identical models were fabricated for the testing programme. The details of manufacture are as follows. An aluminium channel section component acted as a ‘holder’ to replicate the bone. Lengths of bovine rib bone were cleaned, sectioned longitudinally, rough ground to a thickness of approximately 2 mm, and finally cut to fit the inside face of the aluminium component. The cortical side of the bone was bonded to the inside walls of the aluminium (using Araldite epoxy resin); thus the cancellous bone remained exposed. Next, a mild steel tapered stem with a
216
Clin. Biomech.
20
A
Vol. 11, No. 4, 1996
t
shorter than 0.2 mm were not recorded. Figure 2b shows the location and reference numbers used for the viewing windows. The criterion for crack growth was that the final length must be greater than the initial length by 0.2 mm. The models were dynamically loaded at 20 degrees to the long axis, at room temperature in air. The aluminium holder was clamped 60 mm from its distal end. The fatigue loading followed a saw-tooth wave pattern, varying from zero to 1 kN at a frequency of 10 Hz, applied using a customized 4-station Instron hydraulic testing machine (Instron Limited, UK). After 500000 cycles the test was interrupted and the location and length of any new cracks recorded for each zone. The growth of existing cracks was also measured. This procedure was repeated every 500000 cycles up to five million cycles (i.e. a total of 11 times for each experimental model).
2s 0
i
.oi
!
Figura 2. (a) Front and cross-sectional views showing layered structure. (b) Rear view showing crack-viewing windows (‘cut-outs’) and the zone identification code for counting cracks. The code for the zones are as follows: P (proximal); M (middle); D (distal); suffix 1 (lateral side); suffix 2 (medial side). Design features are (i) tapered stem; (ii) cement mantle of Simplex Rapids cement; (iii) cement layers extend beyond the distal tip; (iv) bone layer with cancellous bone inner surface; (v) aluminium holder with windows to facilitate crack observation; (vi) loading applied through the stem. Dimensions are in mm. Note that the cement layers in the axperimental model are 5-10 mm thickto facilitate observation of cracks, whereas in clinical practice typical cement mantles are 2-5 mm thick.
Results The results clearly show that damage accumulation can occur in cemented intramedullary fixation. In the final count (i.e. after 5 million cycles), 389 cracks were observed altogether in the six specimens; 19.8% (n = 77) of these were pre-existing cracks. The growth of pre-existing cracks were studied as a subgroup. The majority of pre-existing cracks were at the bone/cement interface, see Figure 3. However, as Figure 3 also shows, the majority of pre-existing cracks that grew were from the group of cracks within the cement layers. Of the cracks that were initated during loading, almost all initiated within the cement layers, and were associated with pores in the cement. The distal tip of the cement in all cases broke away from the main body of the cement after several thousand cycles; after this,
50 -7
machined finish was located centrally between the two bony walls using an alignment device fabricated for that purpose. The remaining space between the stem and the cancellous bone was then packed with hand-mixed Simplex Rapid@cement mixed in a 2: 1 powder to liquid ratio at room temperature. This cement is transparent when polished. No pressurization or specialized cement insertion equipment were used. After 24 h the front face of the specimen was polished, using a rotating emery disc, to remove the cloudy surface on the cement layers. Next the surface was treated with a red dye penetrant (Johnson and Allen Ltd, UK) to highlight the cracks, see Boyer and Carnesi*. Thirty minutes were allowed for the dye to penetrate into the cracks before removal from the surface with a standard cleaner. Before loading, each model was examined for cracks in the cement layers using a 20 x magnification profile projector (lvlitutoyo PJ300, Japan). The location and size of all pre-existing cracks were recorded, to 0.1 mm accuracy, in each of the viewing windows. Cracks
45 -40 -35 --
Growing
~
n
Not Growing
j
;-2
30 -Number of Pre-test Cracks
0
25 -20 -15 -10 -5 -oCement/Bone Interface
In-Mantle
cemenVYrosmes1s Interface
Figure 3. The numbers of pre-existing cracks (n = 77 for all six specimens) which grew compared with the number which did not, for the bulk cement In = 27) and the bone-cement (n = 49) and cement-stem interfaces (n = 1).
McCormack
et al: Damage
a 350
T
250
hip prosthesis
217
Discussion
I
200-Ccmcnt/5one interface
150’-
CementIPmsthesis interface 0
1
2 3 4 NO.of Cycles (millions)
5
b 140 120
in cemented
crack type and zone. However, a much greater number of specimens will need to be tested to obtain statistical significance.
300
Number of cracks
accumulation
1
loo-Number go-of Cracks
Huiskes” addresses the issue of failure in load-bearing implants in terms of failure scenarios and he identifies damage accumulation as one failure scenario. Damage accumulation alone can lead to loosening, or it can generate poly (methylmethacrylate) wear particles due to the abrasion of crack surfaces and hence lead on to a wear particle reaction failure scenario*“. Lee21 emphasizes that the definition of failure initiation must be considered in relation to the implant design principles. However, no matter how failure initiation is defined, it would seem that damage accumulation within the bulk cement is a critical phase in the failure of cemented reconstructions. McCormack and Prendergast** proposed that, rather than attempting to prevent mechanical failure initiation, the design objective should be to maximize the .time taken to progress through the failure train. The experimental quantification of the progress of failure scenarios (such a
r \
0.9 0.8
In-mantle
0.7
I 0
1
2
3
4
5
No. of Cycles (millions) Figure 4. The continuous initiation of new cracks in the cement layers of the model throughout the testing for all six specimens added together. Crack accumulation is plotted for (a) crack type and (b) crack zone.
0.6 Average 0.5 Changein Length(mm) 0.4
Cement/Bone interface
0.3
Cement/Prosthesis
0.2
the rate of stem subsidence increased to a maximum subsidence of approximately 5 mm. Cracks initiated continuously during fatigue loading. There is a steady increase in the number of new cracks initiating within the mantle from zero to five million cycles, as shown in Figure 4a. The number of new interfacial cracks was small in comparison. In this experimental model, most new cracks were generated in the middle zones and the least number of cracks occurred in the proximal zones, see Figure 4b. Of importance for the longevity of cemented reconstructions is the rate of damage accumulation. The rate of crack growth (which is a measure of the rate of damage accumulation) is calculated as the average change in crack length per cycle. Cracks from pores within the mantle grew steadily during testing, whereas cracks from the interfaces showed much smaller amounts of growth, see Figure 5a. The most rapid crack growth occurs in the medial M2 zone, and the distal zones, Dl and D2, whereas the cracks in the proximal zones, Pl and P2, and the medial zone, Ml, grow very slowly, see Figure 5b. Table 1 presents means and standard deviations of all the cracks in the six specimens tested, according to
1
0
2 3 4 No. of Cycles (millions)
5
b 0.7
T
0.6 --
Average Change in f--en& (mm)
Ml
0
1
2
No. of Cycles (millions) Figure 5. The rate of damage accumulation in terms of the continuous increase in crack length throughout testing plotted for (a) cracktype and (b) crack zone. The data is plotted as the average over six specimens of the 77 pre-existing cracks.
218
C/in. Biomech.
Vol. 11, No. 4, 1996
Table 1. Means and standard deviations of numbers in all six specimens after 5 million cycles of loading Zme
Pl P2 Ml M2 D? lx? rotai
nt bone-cement interface (Mean (SD)/ _~~.
---
Within the cement mantle (Mean (SD)) I.0
22 2.3 2.8 1.5
12.2) (2.6) (3.8) (1.9)
23
(2.1)
ll,i
1.2 75.8 16.7 6.0 12.2 52.9
of cracks occurring At metal-cement interface (Mean (SD))
(1.1)
(1.21 (14.2) 112.5) (7.5) (11.0)
0.8 (1.2) 0.8
The table shows the distribution of cracks in both cement layers wth respect pfox~mal, middle and distal regions. For each zone the data has been subgrouped rrack type P ‘--’ tndicated no crack observation
to by
as damage accumulation presented in this paper) could facilitate this design objective. Because of the technical approach to design of the experimental model. the difficulty that others have reported regarding detection of mechanical failure was avoided”. The frustration of not being able to see the failure process has been overcome and direct evidence of damage accumulation has been provided. However, the trade-off is that the complex stress distributions of r~eal intramedullary fixation has not been completely replicated in the model. Specifically, the threedimensional nature of the real cement mantle is represented hv an aluminium channel component which provides circumferential force transfer. Many finite element analyses have shown this to be a satisfactory approach for the calculation of the bending stresses so long as the bone and the model have the same second moment of area. Nevertheless, circumferential stresses are not generated in the cement layers of this experimental model and it is not possible to determine the effect of this simplification except by testing a fully three-dimensional model. In addition, the interfacial conditions occurring clinically are not replicated precisely in the model because the cement/ bone interface in the experimental model is for bovine rib cancellous bone. which obviously differs somewhat in roughness from human femoral cancellous bone. When pail; (methylmethacrylate) samples are tested in simple tension tests. fracture occurs typically from a dominant crack23. Because the experimental model presented in this paper is held together by a ‘holder’ component. the cement layers are contained in a physiological way. and such failure is not dominated b!, a single crack; instead distributed cracking is observed. Indeed. in this experimental model one single crack could span the cement layer without causing faihne of the mechanical integrity of the cemented structure. In this respect, the present investigation highlights the value of testing the cement under the conditions it experiences in the total biomechanical system. Initially. the majority of cracks were located at the bone-cement interface, whereas after 5 million cycles of testing the majority of cracks were located within the cement mantle. These latter cracks were associated with cement porosity. Figure 5a shows that, in this experimental model at least, the cement layer cracks
are the ones that grow most rapidly. This observation suggests that distributed cracking occurs continuously under fatigue loading, and that the rapid emergence of a critical crack is not the failure mechanism of the cement mantle. From Figure 5a, it can be read that the most rapid growth is about 0.9 mm in 5 million cycles. At this rate of growth, a 3-mm mantle would develop one or more cracks through the thickness in lo-15 years of normal use. This timescale is not dissimilar to revision times reported in the clinical literature*“. This result would support the clinical practice which goes to extreme effort to reduce air bubbles and cement defects by, for example, vacuum*‘j or centrifugal mixing” of the cement and by pressurization*s. Although it is surely impossible to eliminate every pore. a reduction should mean fewer crack initiation sites, and this will reduce the rate of damage accumulation and should improve the endurance of the fixation. In conclusion, an experimental model has been developed that allows crack growth/damage accumulation to be directly observed. The model has many of the features of a intramedullary fixation. Additional experiments, and a thorough stress analysis of the experimental model in the form of a finite element analysis will allow a more complete interpretation of the results. This experimental model may prove useful to investigate the effect of specific design factors (e.g. stem texture and roughness, stem taper, cement reinforcement) on damage accumulation. Acknowledgements We are grateful to Luke Curley for his advice regarding the manufacturing and testing of the models. Financial support was provided by the UCD President’s Research Award to Brendan McCormack and by a Forbairt Applied Research Award to Donnachadha Gallagher.
References Walker PS. Human Joints and their Artificial Replacement. Charles C Thomas: Springfield, IL, 1977 Huiskes R. Mechanical failure in total hip arthroplasty with cement. Curr Orthop 1993; 7: 239-47 Lemaitre J, Chaboche J-L. MPchanique des Mate’riaux solides. Dunod: Paris, 1985. Tr. Mechanics of Solid Materials. Cambridge: Cambridge University Press, 1994 Jasty M, Maloney WJ, Bragdon CR et al. The initiation of failure in cemented femoral components of hip arthroplasties. J Bone Joint Surg 1991; 73B: 551-8 Helmke HW, Lednicky CL, Tullos HS. Porosity of the cement/metal interface following cemented hip replacement. Transactions of the 38th nieeting of the Orthopaedic Research Society 1992; 364
James SP, Schmalzried TP, McGarry FJ, Harris WH. Extensive porosity at the cement-femoral prosthesis interface: a preliminary study. J Biomed Mater Res 1993; 27: 71-8 Lee AJC, Ling RSM, Vangala SS. Some clinically relevant variables affecting the mechanical behaviour of bone-cement. Arch Orthop Trauma Surg 1978; 92: l- 18
McCormack
et al: Damage
8 Linden U. Mechanical properties of bone cement importance of the mixing technique. Clin Orthop Rel Res 1991; 272: 274-S 9 Culleton TP, Prendergast PJ, Taylor D. Fatigue failure in the cement mantle of an artificial hip joint. Clin Mats 1993; 12: 95-102 10 Prendereast PJ. Monaahan J. Tavlor D. Materials , selectiogin the artificial hip joint using finite element stress analysis. Clin Mats 1989; 4: 361-76 11 Huiskes R. The various stress patterns of press-fit, ingrown, and cemented femoral stems. Clin Orthop Rel Res 1990; 261: 27-38 12 Verdonschot N, Huiskes R. A combination of continuum damage mechanics and the finite element method to analyze acrylic bone cement cracking around implants. In: Middleton J, Pande G, Jones M, ed. Second International Conference on Computer Methods in Biomechanics and Biomedical Engineering. Gordon and
Breach, Amsterdam, 1966; 25-33 13 Sugiyama H, Whiteside LA, Kaiser AD. Examination of rotational fixation of the femoral component in total hip arthroplasty. A mechanical study of micromovement and acoustic emission. Clin Orthop Rel Res 1989; 249: 122-8 14 Davies J, Tse M-K, Harris WH. Prospective demonstration of debonding of the cement metal interface of the femoral THR using acoustic emission and ultrasound in situ. Transactions of the 41st Meeting of the Orthopaedic Research Society. 1995; 712 15 Taylor D, McCormack BAO, Clarke F, Sheehan J. Reinforcement of bone cement using metal meshes. Proc I Mech E (Lond) 1989; 203H: 49-53 16 Gola MM, Gugliotta A. An analytical estimate of stresses in bones and prosthesis stems. J Strain Anal 1979; 14:
accumulation
in cemented
hip prosthesis
219
23 Chao EYS, Chin HC, Stauffer RN. Roentgenographic and mechanical performance of centrifuged cement in a simulated total hip arthroplasty model. Clin Orthop Rel Res 1992; 285: 91-101 24 Carter DR, Gates EI, Harris WH. Strain-controlled fatigue of acrylic bone cement. J Biomed Mater Res 1982; 16: 647-57 25 Malchau H, Herberts P, Ahnfelt L. Prognosis of total hip replacement in Sweden. Follow up of 92,675 operations performed 1978- 1990.Acta Orthop Stand 1993; 64: 497-506 26 Wixson RL. Do we need COvacuum mix or centrifuge cement? Clin Orthop Rel Res 1992; 285: 84-90
27 Schreurs BW, Spierings PTJ, Huiskes R, Slooft TJJH. Effects of preparation techniques on the porosity of acryclic cements. Acta Orthop Stand 1988; 59: 403-9 28 Fowler JC, Gie GA, Lee AJC, Ling RSM. Experience with the Exeter total hip replacement since 1970. Clin Orthop North Am 1988; 19: 477-89 Appendix The load (denoted P) arises at each point in the cement layer along the length of the stem (i.e. in the z-direction). It is given by
P(z) = G(z) {us(z) - %(Z)>
(Al)
where u, = deflection of the stem neutral axis, u,, = deflection of the bone neutral axis and C, = stiffness in the cement layer against transverse loading. From the beams-on-elastic foundations theory, the following coupled equations arise
+ P(z) = 0
(AZ)
29-33
17 Huiskes R. Some fundamental aspectsof human joint replacement. Actu Orthop Stand 180; Supplementum 185 18 Boyer HE, Carnes WJ. (Eds.) Metals handbook; Vol. 11; Non-Destructive Inspection and Quality Control; Liquid Penetrant Inspection. American Society for Metals,
Eighth edn, 1976 19 Huiskes R. Stresspatterns, failure modes and bone remodelling. In: Fitzgerald R. ed. Noncemented Total Hip Arthroplusty. Raven Press, New York, 1988; 283-302 20 Horowitz SM. Dotv SB. Lane JM. Burstein AH. Studies of the mechanism by which the mechanical failure of polymethylmethacrylate leads to bone resorption. J Bone Joint Surg 1993; 75A: 802- 13 21 Lee AJC. Rough or polished surface on femoral anchorage stems? In: Buchhorn GH, Willert H-G ed. Technical Principles, Design and Safety of Joint Implants.
Hogrefe and Huber, Gottingen, 1994; 209- 11 22 McCormack BA, Prendergast PJ. Interface failure in implants cemented with different bone cements: a fracture mechanics analysis. In: Middleton J, Pande G, Jones M. ed. Second Znternational Symposium on Computer Methods in Biomechanics and Biomedical Engineering.
Gordon and Breach, Amsterdam, 1996; 35-45
and
-
d2
( Fbd2ub)dz* \ dz* 1
P(z) = 0
where F, and F, denote the flexural stiffnesses of the stem and bone respectively. These equations were solved to produce equations for u\(z), ~~(2). P(z) along the length L. The distributed axial force, Q(z), is
Q(z) = G(z)x
(A41
where C,(z) is the shear stiffness of the cement layer. The final load distribution is then combined by adding P(z) [equation (Al)] and Q(z) (equation (A4)) for the top and bottom cement layers. Dividing by the depth of the cement layer gives the final stress distribution in the cement layer. We note that this analysis is only approximate for two main reasons. Firstly, the experimental model is not symmetric in the plane of bending (and therefore the shear centre and the centroid do not coincide) and this gives rise to some torsional stresses which the beams-on-elastic foundation mode1does not calculate. Secondly, soon after loading, the metal debonds from the cement and this causesa redistribution of stress in the cement mantle. The beams-on-elastic foundations stress analysis was used as a guide in the design of the physical model. During the design process, Perspex (Plexiglas) and aluminium were considered as possible materials for the circumferential component, the latter being the final choice as it produced the closest fit with the stress distributions reported for the cement layer of a hip prosthesis.
0 1996 Elsevier
1996 Science Limited. All rights reserved Printed in Great Britain 0268-0033196 $15.00 + 0.00
ELSEVIER
B A 0 McCormack ‘Bioengineering College Dublin, Ireland
MSC~,
P J Prendergast
Research Centre, Department and 2Department of Mechanical
PhD*, D G Gallagher
MEngSc’
of Mechanical Engineering, University Engineering, Trinity College, Dublin 2,
Abstract Objective. To develop a methodology to characterize the pattern of crack initiation and damage accumulation in intramedullary fixated cemented prostheses. Design. An experimental physical model of intramedullary fixation was developed which both represents the implant structure and permits monitoring of fatigue crack growth. Background. Many joint replacement prostheses are fixed into the medullary cavity of bones using a poly(methylmethacrylate) ‘bone cement’, which forms a mantle around the prosthesis and locks it to the bone. The endurance of the replacement is, to a great extent, determined by the mechanical durability of the cement and the implant interfaces under cyclic stresses generated by dynamic loading. The cement mantle is subjected to complex multiaxial stresses which vary in particular distribution depending on the prosthesis design. Methods. Damage accumulation is reported in terms of the number of cracks, the location of cracks, and the rate of crack growth. Results. The results clearly show the nature of damage accumulation in the cement mantle, and that many of the cracks which propagate within the cement mantle are related to cement porosity. Conclusion. This study gives experimental evidence to support the hypothesis of a damage accumulation failure scenario in cemented hip reconstructions. Relevance Cementing is the most popular technique for the fixation of joint replacement prosthesis. However, the sequence of events leading to the failure of cemented fixation is not fully understood. In this paper it is shown that damage accumulation can be directly monitored in an experimental model of cemented intramedullary fixation. Copyright @ 1996 Elsevier Science Ltd. Key words: Damage accumulation, Clin. Biomech.
prosthesis fixation, hip replacement, failure scenario
Vol. 11, No. 4, 214-219,
1996
Introduction
Most joint replacement prostheses have an intramedullary component which is fixed into the bone using a poly (methylmethacrylate) ‘bone cement’. For example, the femoral side of hip replacement, the tibia1 component of knee replacement and both the humeral and radial components of elbow replacements can all be fixed in this way’. Breakdown in the mechanical integrity of the cement layer is one route to clinical failure of these joint reconstructions2. This breakdown .l---.--
Received: 13 April 1995; Accepted: 23 November
1995 Correspondence and reprint requests to: B A 0 McCormack MSC, Bioengineering Research Centre, Department of Mechanical Engineering, Ilniversity College Dublin, Dublin 4, Iretand
can be associated with failure of the cement mantle by the process of ‘damage accumulation’. Damage accumulation is defined as the propagation of small internal flaws and microcracks which eventually form critical defects’. Cracks have been reported in the cement mantles of autopsy-retrieved hip prostheses even though there were no signs of clinical failure at the time of death4. These cracks were seen both within the cement mantle and on the interfaces, and were present in all retrievals (including those which had been implanted for only 3 years). Other retrieval studies have shown that pores and voids are formed during the insertion of the cernen$(j, an observation that has been made many times in bone cement specimens prepared under laboratory conditions7s. Fatigue striations on the fracture surfaces of retrieved mantles prove that, in
McCormack
et al: Damage
certain cases at least, not only do cracks occur in viva but that they also propagate to cause failure’. The objective of this investigation was to develop an experimental methodology for the analysis of microcrack growth in intramedullary-fixated implant structures. There are many studies reporting the particular nature of the stress pattern in the cement mantle and on the interfaces of intramedullary prostheses”,“. These stresses induce a complex failure mechanism in the cement mantle’* that cannot be fully regenerated in a simple test specimen. Therefore an experimental model must be designed to reproduce the essential features of this stress distribution whilst at the same time being simple enough to facilitate crack observation. One of the common problems encountered when attempting to study cracks in real structures is simply one of crack observation and crack growth monitoring. In a hip arthroplasty, the cement mantle is encased in bone and, even if the mantle were exposed, it is usually difficult to see into the depth of the cement. Nondestructive crack detection techniques such as acoustic emission13 and ultrasound14 have been used, but the accurate location of new cracks, and accurate tracking of crack growth is not possible using these techniques. Destructive testing such as sectioning or splitting the test specimen have been developedi5. However, destructive testing means that several different specimens must be used to determine what happens over time and an error due to inter-specimen variation is introduced. In addition, destructive techniques do not allow the distinction between pre-test cracks and cracks initiated during testing, and accurate tracking of damage accumulation under fatigue loading is therefore difficult. An experimental model which represents the implant structure in respect of its material properties, interface conditions, and geometry was developed. The design is achieved subject to the constraint that damage accumulation must be continuously visible during fatigue loading. In this way we can successfully monitor crack initiation sites and the growth characteristics of cracks. The growth of many cracks is observed (as opposed to just the rapid emergence of one dominant crack) and this supports the hypothesis of a damage accumulation failure scenario for cemented intramedullary reconstructions, such as hip joint replacements.
accumulation
in cemented
hip prosthesis
215
Methods
The first part of this section describes the design of the experimental model and the second part describes the methods of crack observation and analysis. The experimental model
The design concept for the model is a structure comprising a tapered metal stem with a cement layer on either side and a layer of bone on the outside of the cement layers, held intact by a component which replicates the circumferential force transfer of the bone, see concept design of Figure 1. Hence the structure has five layers (bone/cement/prosthesis/cement/bone) and is open on one side to allow crack observation. To permit the transmission of light through the cement layers, three ‘windows’ were machined into the back wall of the model. To ensure internal visibility of the cracks and flaws, a clear poly (methylmethacrylate) cement was used. The experimental test model may be considered as a longitudinal slice taken from the centre of an idealized intramedullary fixation. Dimensions were calculated to ensure that the experimental model transferred the stress through the cement mantle in a similar way to the real intramedullary system. Beams-on-elastic-foundation theory is applied to calculate the loading in this system, as has been applied previously to the femoral component of the artificial hip joint l&i’ The theory assumes that bone and prosthesis are stiff beams separated by an elastic layer, which is modelled as a set of springs. Using this approach, Huiskes” resolved the joint loading into axial and transverse components and carried out separate analysis for each. He has shown that the fundamental load transfer pattern of intramedullary fixation involves proximal stress transfer to the bone, a central region where the structure behaves as a composite beam, and a distal region of load transfer to the prosthesis. Using the beams-on-elasticfoundation approach (see Appendix), we evaluated many alternative designs of experimental model. The dimensions given in Figure 2 were those of the final design. For the final design of the experimental model it was calculated that a peak load of 1 kN applied to the stem develops stresses in the range O-6 MPa in the cement layers. Specimen manufacture and testing method
BOW.2 Ph4MA Stem/Holder
Channel Section “holder”
Figure 1. The experimental model is a five-layered together with a channel section.
structure
held
Six identical models were fabricated for the testing programme. The details of manufacture are as follows. An aluminium channel section component acted as a ‘holder’ to replicate the bone. Lengths of bovine rib bone were cleaned, sectioned longitudinally, rough ground to a thickness of approximately 2 mm, and finally cut to fit the inside face of the aluminium component. The cortical side of the bone was bonded to the inside walls of the aluminium (using Araldite epoxy resin); thus the cancellous bone remained exposed. Next, a mild steel tapered stem with a
216
Clin. Biomech.
20
A
Vol. 11, No. 4, 1996
t
shorter than 0.2 mm were not recorded. Figure 2b shows the location and reference numbers used for the viewing windows. The criterion for crack growth was that the final length must be greater than the initial length by 0.2 mm. The models were dynamically loaded at 20 degrees to the long axis, at room temperature in air. The aluminium holder was clamped 60 mm from its distal end. The fatigue loading followed a saw-tooth wave pattern, varying from zero to 1 kN at a frequency of 10 Hz, applied using a customized 4-station Instron hydraulic testing machine (Instron Limited, UK). After 500000 cycles the test was interrupted and the location and length of any new cracks recorded for each zone. The growth of existing cracks was also measured. This procedure was repeated every 500000 cycles up to five million cycles (i.e. a total of 11 times for each experimental model).
2s 0
i
.oi
!
Figura 2. (a) Front and cross-sectional views showing layered structure. (b) Rear view showing crack-viewing windows (‘cut-outs’) and the zone identification code for counting cracks. The code for the zones are as follows: P (proximal); M (middle); D (distal); suffix 1 (lateral side); suffix 2 (medial side). Design features are (i) tapered stem; (ii) cement mantle of Simplex Rapids cement; (iii) cement layers extend beyond the distal tip; (iv) bone layer with cancellous bone inner surface; (v) aluminium holder with windows to facilitate crack observation; (vi) loading applied through the stem. Dimensions are in mm. Note that the cement layers in the axperimental model are 5-10 mm thickto facilitate observation of cracks, whereas in clinical practice typical cement mantles are 2-5 mm thick.
Results The results clearly show that damage accumulation can occur in cemented intramedullary fixation. In the final count (i.e. after 5 million cycles), 389 cracks were observed altogether in the six specimens; 19.8% (n = 77) of these were pre-existing cracks. The growth of pre-existing cracks were studied as a subgroup. The majority of pre-existing cracks were at the bone/cement interface, see Figure 3. However, as Figure 3 also shows, the majority of pre-existing cracks that grew were from the group of cracks within the cement layers. Of the cracks that were initated during loading, almost all initiated within the cement layers, and were associated with pores in the cement. The distal tip of the cement in all cases broke away from the main body of the cement after several thousand cycles; after this,
50 -7
machined finish was located centrally between the two bony walls using an alignment device fabricated for that purpose. The remaining space between the stem and the cancellous bone was then packed with hand-mixed Simplex Rapid@cement mixed in a 2: 1 powder to liquid ratio at room temperature. This cement is transparent when polished. No pressurization or specialized cement insertion equipment were used. After 24 h the front face of the specimen was polished, using a rotating emery disc, to remove the cloudy surface on the cement layers. Next the surface was treated with a red dye penetrant (Johnson and Allen Ltd, UK) to highlight the cracks, see Boyer and Carnesi*. Thirty minutes were allowed for the dye to penetrate into the cracks before removal from the surface with a standard cleaner. Before loading, each model was examined for cracks in the cement layers using a 20 x magnification profile projector (lvlitutoyo PJ300, Japan). The location and size of all pre-existing cracks were recorded, to 0.1 mm accuracy, in each of the viewing windows. Cracks
45 -40 -35 --
Growing
~
n
Not Growing
j
;-2
30 -Number of Pre-test Cracks
0
25 -20 -15 -10 -5 -oCement/Bone Interface
In-Mantle
cemenVYrosmes1s Interface
Figure 3. The numbers of pre-existing cracks (n = 77 for all six specimens) which grew compared with the number which did not, for the bulk cement In = 27) and the bone-cement (n = 49) and cement-stem interfaces (n = 1).
McCormack
et al: Damage
a 350
T
250
hip prosthesis
217
Discussion
I
200-Ccmcnt/5one interface
150’-
CementIPmsthesis interface 0
1
2 3 4 NO.of Cycles (millions)
5
b 140 120
in cemented
crack type and zone. However, a much greater number of specimens will need to be tested to obtain statistical significance.
300
Number of cracks
accumulation
1
loo-Number go-of Cracks
Huiskes” addresses the issue of failure in load-bearing implants in terms of failure scenarios and he identifies damage accumulation as one failure scenario. Damage accumulation alone can lead to loosening, or it can generate poly (methylmethacrylate) wear particles due to the abrasion of crack surfaces and hence lead on to a wear particle reaction failure scenario*“. Lee21 emphasizes that the definition of failure initiation must be considered in relation to the implant design principles. However, no matter how failure initiation is defined, it would seem that damage accumulation within the bulk cement is a critical phase in the failure of cemented reconstructions. McCormack and Prendergast** proposed that, rather than attempting to prevent mechanical failure initiation, the design objective should be to maximize the .time taken to progress through the failure train. The experimental quantification of the progress of failure scenarios (such a
r \
0.9 0.8
In-mantle
0.7
I 0
1
2
3
4
5
No. of Cycles (millions) Figure 4. The continuous initiation of new cracks in the cement layers of the model throughout the testing for all six specimens added together. Crack accumulation is plotted for (a) crack type and (b) crack zone.
0.6 Average 0.5 Changein Length(mm) 0.4
Cement/Bone interface
0.3
Cement/Prosthesis
0.2
the rate of stem subsidence increased to a maximum subsidence of approximately 5 mm. Cracks initiated continuously during fatigue loading. There is a steady increase in the number of new cracks initiating within the mantle from zero to five million cycles, as shown in Figure 4a. The number of new interfacial cracks was small in comparison. In this experimental model, most new cracks were generated in the middle zones and the least number of cracks occurred in the proximal zones, see Figure 4b. Of importance for the longevity of cemented reconstructions is the rate of damage accumulation. The rate of crack growth (which is a measure of the rate of damage accumulation) is calculated as the average change in crack length per cycle. Cracks from pores within the mantle grew steadily during testing, whereas cracks from the interfaces showed much smaller amounts of growth, see Figure 5a. The most rapid crack growth occurs in the medial M2 zone, and the distal zones, Dl and D2, whereas the cracks in the proximal zones, Pl and P2, and the medial zone, Ml, grow very slowly, see Figure 5b. Table 1 presents means and standard deviations of all the cracks in the six specimens tested, according to
1
0
2 3 4 No. of Cycles (millions)
5
b 0.7
T
0.6 --
Average Change in f--en& (mm)
Ml
0
1
2
No. of Cycles (millions) Figure 5. The rate of damage accumulation in terms of the continuous increase in crack length throughout testing plotted for (a) cracktype and (b) crack zone. The data is plotted as the average over six specimens of the 77 pre-existing cracks.
218
C/in. Biomech.
Vol. 11, No. 4, 1996
Table 1. Means and standard deviations of numbers in all six specimens after 5 million cycles of loading Zme
Pl P2 Ml M2 D? lx? rotai
nt bone-cement interface (Mean (SD)/ _~~.
---
Within the cement mantle (Mean (SD)) I.0
22 2.3 2.8 1.5
12.2) (2.6) (3.8) (1.9)
23
(2.1)
ll,i
1.2 75.8 16.7 6.0 12.2 52.9
of cracks occurring At metal-cement interface (Mean (SD))
(1.1)
(1.21 (14.2) 112.5) (7.5) (11.0)
0.8 (1.2) 0.8
The table shows the distribution of cracks in both cement layers wth respect pfox~mal, middle and distal regions. For each zone the data has been subgrouped rrack type P ‘--’ tndicated no crack observation
to by
as damage accumulation presented in this paper) could facilitate this design objective. Because of the technical approach to design of the experimental model. the difficulty that others have reported regarding detection of mechanical failure was avoided”. The frustration of not being able to see the failure process has been overcome and direct evidence of damage accumulation has been provided. However, the trade-off is that the complex stress distributions of r~eal intramedullary fixation has not been completely replicated in the model. Specifically, the threedimensional nature of the real cement mantle is represented hv an aluminium channel component which provides circumferential force transfer. Many finite element analyses have shown this to be a satisfactory approach for the calculation of the bending stresses so long as the bone and the model have the same second moment of area. Nevertheless, circumferential stresses are not generated in the cement layers of this experimental model and it is not possible to determine the effect of this simplification except by testing a fully three-dimensional model. In addition, the interfacial conditions occurring clinically are not replicated precisely in the model because the cement/ bone interface in the experimental model is for bovine rib cancellous bone. which obviously differs somewhat in roughness from human femoral cancellous bone. When pail; (methylmethacrylate) samples are tested in simple tension tests. fracture occurs typically from a dominant crack23. Because the experimental model presented in this paper is held together by a ‘holder’ component. the cement layers are contained in a physiological way. and such failure is not dominated b!, a single crack; instead distributed cracking is observed. Indeed. in this experimental model one single crack could span the cement layer without causing faihne of the mechanical integrity of the cemented structure. In this respect, the present investigation highlights the value of testing the cement under the conditions it experiences in the total biomechanical system. Initially. the majority of cracks were located at the bone-cement interface, whereas after 5 million cycles of testing the majority of cracks were located within the cement mantle. These latter cracks were associated with cement porosity. Figure 5a shows that, in this experimental model at least, the cement layer cracks
are the ones that grow most rapidly. This observation suggests that distributed cracking occurs continuously under fatigue loading, and that the rapid emergence of a critical crack is not the failure mechanism of the cement mantle. From Figure 5a, it can be read that the most rapid growth is about 0.9 mm in 5 million cycles. At this rate of growth, a 3-mm mantle would develop one or more cracks through the thickness in lo-15 years of normal use. This timescale is not dissimilar to revision times reported in the clinical literature*“. This result would support the clinical practice which goes to extreme effort to reduce air bubbles and cement defects by, for example, vacuum*‘j or centrifugal mixing” of the cement and by pressurization*s. Although it is surely impossible to eliminate every pore. a reduction should mean fewer crack initiation sites, and this will reduce the rate of damage accumulation and should improve the endurance of the fixation. In conclusion, an experimental model has been developed that allows crack growth/damage accumulation to be directly observed. The model has many of the features of a intramedullary fixation. Additional experiments, and a thorough stress analysis of the experimental model in the form of a finite element analysis will allow a more complete interpretation of the results. This experimental model may prove useful to investigate the effect of specific design factors (e.g. stem texture and roughness, stem taper, cement reinforcement) on damage accumulation. Acknowledgements We are grateful to Luke Curley for his advice regarding the manufacturing and testing of the models. Financial support was provided by the UCD President’s Research Award to Brendan McCormack and by a Forbairt Applied Research Award to Donnachadha Gallagher.
References Walker PS. Human Joints and their Artificial Replacement. Charles C Thomas: Springfield, IL, 1977 Huiskes R. Mechanical failure in total hip arthroplasty with cement. Curr Orthop 1993; 7: 239-47 Lemaitre J, Chaboche J-L. MPchanique des Mate’riaux solides. Dunod: Paris, 1985. Tr. Mechanics of Solid Materials. Cambridge: Cambridge University Press, 1994 Jasty M, Maloney WJ, Bragdon CR et al. The initiation of failure in cemented femoral components of hip arthroplasties. J Bone Joint Surg 1991; 73B: 551-8 Helmke HW, Lednicky CL, Tullos HS. Porosity of the cement/metal interface following cemented hip replacement. Transactions of the 38th nieeting of the Orthopaedic Research Society 1992; 364
James SP, Schmalzried TP, McGarry FJ, Harris WH. Extensive porosity at the cement-femoral prosthesis interface: a preliminary study. J Biomed Mater Res 1993; 27: 71-8 Lee AJC, Ling RSM, Vangala SS. Some clinically relevant variables affecting the mechanical behaviour of bone-cement. Arch Orthop Trauma Surg 1978; 92: l- 18
McCormack
et al: Damage
8 Linden U. Mechanical properties of bone cement importance of the mixing technique. Clin Orthop Rel Res 1991; 272: 274-S 9 Culleton TP, Prendergast PJ, Taylor D. Fatigue failure in the cement mantle of an artificial hip joint. Clin Mats 1993; 12: 95-102 10 Prendereast PJ. Monaahan J. Tavlor D. Materials , selectiogin the artificial hip joint using finite element stress analysis. Clin Mats 1989; 4: 361-76 11 Huiskes R. The various stress patterns of press-fit, ingrown, and cemented femoral stems. Clin Orthop Rel Res 1990; 261: 27-38 12 Verdonschot N, Huiskes R. A combination of continuum damage mechanics and the finite element method to analyze acrylic bone cement cracking around implants. In: Middleton J, Pande G, Jones M, ed. Second International Conference on Computer Methods in Biomechanics and Biomedical Engineering. Gordon and
Breach, Amsterdam, 1966; 25-33 13 Sugiyama H, Whiteside LA, Kaiser AD. Examination of rotational fixation of the femoral component in total hip arthroplasty. A mechanical study of micromovement and acoustic emission. Clin Orthop Rel Res 1989; 249: 122-8 14 Davies J, Tse M-K, Harris WH. Prospective demonstration of debonding of the cement metal interface of the femoral THR using acoustic emission and ultrasound in situ. Transactions of the 41st Meeting of the Orthopaedic Research Society. 1995; 712 15 Taylor D, McCormack BAO, Clarke F, Sheehan J. Reinforcement of bone cement using metal meshes. Proc I Mech E (Lond) 1989; 203H: 49-53 16 Gola MM, Gugliotta A. An analytical estimate of stresses in bones and prosthesis stems. J Strain Anal 1979; 14:
accumulation
in cemented
hip prosthesis
219
23 Chao EYS, Chin HC, Stauffer RN. Roentgenographic and mechanical performance of centrifuged cement in a simulated total hip arthroplasty model. Clin Orthop Rel Res 1992; 285: 91-101 24 Carter DR, Gates EI, Harris WH. Strain-controlled fatigue of acrylic bone cement. J Biomed Mater Res 1982; 16: 647-57 25 Malchau H, Herberts P, Ahnfelt L. Prognosis of total hip replacement in Sweden. Follow up of 92,675 operations performed 1978- 1990.Acta Orthop Stand 1993; 64: 497-506 26 Wixson RL. Do we need COvacuum mix or centrifuge cement? Clin Orthop Rel Res 1992; 285: 84-90
27 Schreurs BW, Spierings PTJ, Huiskes R, Slooft TJJH. Effects of preparation techniques on the porosity of acryclic cements. Acta Orthop Stand 1988; 59: 403-9 28 Fowler JC, Gie GA, Lee AJC, Ling RSM. Experience with the Exeter total hip replacement since 1970. Clin Orthop North Am 1988; 19: 477-89 Appendix The load (denoted P) arises at each point in the cement layer along the length of the stem (i.e. in the z-direction). It is given by
P(z) = G(z) {us(z) - %(Z)>
(Al)
where u, = deflection of the stem neutral axis, u,, = deflection of the bone neutral axis and C, = stiffness in the cement layer against transverse loading. From the beams-on-elastic foundations theory, the following coupled equations arise
+ P(z) = 0
(AZ)
29-33
17 Huiskes R. Some fundamental aspectsof human joint replacement. Actu Orthop Stand 180; Supplementum 185 18 Boyer HE, Carnes WJ. (Eds.) Metals handbook; Vol. 11; Non-Destructive Inspection and Quality Control; Liquid Penetrant Inspection. American Society for Metals,
Eighth edn, 1976 19 Huiskes R. Stresspatterns, failure modes and bone remodelling. In: Fitzgerald R. ed. Noncemented Total Hip Arthroplusty. Raven Press, New York, 1988; 283-302 20 Horowitz SM. Dotv SB. Lane JM. Burstein AH. Studies of the mechanism by which the mechanical failure of polymethylmethacrylate leads to bone resorption. J Bone Joint Surg 1993; 75A: 802- 13 21 Lee AJC. Rough or polished surface on femoral anchorage stems? In: Buchhorn GH, Willert H-G ed. Technical Principles, Design and Safety of Joint Implants.
Hogrefe and Huber, Gottingen, 1994; 209- 11 22 McCormack BA, Prendergast PJ. Interface failure in implants cemented with different bone cements: a fracture mechanics analysis. In: Middleton J, Pande G, Jones M. ed. Second Znternational Symposium on Computer Methods in Biomechanics and Biomedical Engineering.
Gordon and Breach, Amsterdam, 1996; 35-45
and
-
d2
( Fbd2ub)dz* \ dz* 1
P(z) = 0
where F, and F, denote the flexural stiffnesses of the stem and bone respectively. These equations were solved to produce equations for u\(z), ~~(2). P(z) along the length L. The distributed axial force, Q(z), is
Q(z) = G(z)x
(A41
where C,(z) is the shear stiffness of the cement layer. The final load distribution is then combined by adding P(z) [equation (Al)] and Q(z) (equation (A4)) for the top and bottom cement layers. Dividing by the depth of the cement layer gives the final stress distribution in the cement layer. We note that this analysis is only approximate for two main reasons. Firstly, the experimental model is not symmetric in the plane of bending (and therefore the shear centre and the centroid do not coincide) and this gives rise to some torsional stresses which the beams-on-elastic foundation mode1does not calculate. Secondly, soon after loading, the metal debonds from the cement and this causesa redistribution of stress in the cement mantle. The beams-on-elastic foundations stress analysis was used as a guide in the design of the physical model. During the design process, Perspex (Plexiglas) and aluminium were considered as possible materials for the circumferential component, the latter being the final choice as it produced the closest fit with the stress distributions reported for the cement layer of a hip prosthesis.