- Email: [email protected]
Finite element modelling of primary hip stem stability: The effect of interference fit
ARTICLE IN PRESS
Journal of Biomechanics 41 (2008) 587–594 www.elsevier.com/locate/jbiomech www.JBiomech.com
Finite element modelling of primary hip stem stability: The effect of interference fit Mohammed Rafiq Abdul-Kadira,b, Ulrich Hansena,, Ralf Klabundec, Duncan Lucasa, Andrew Amisa,d a Department of Mechanical Engineering, Imperial College, London, UK Biomechanics and Tissue Engineering Group, Universiti Teknologi Malaysia, Malaysia c Zimmer GmbH, Winterthur, Switzerland d Division of Surgery, Imperial College London, UK
b
Accepted 13 October 2007
Abstract The most commonly reported complications related to cementless hip stems are loosening and thigh pain; both of these have been attributed to high levels of relative micromotion at the bone–implant interface due to insufficient primary fixation. Primary fixation is believed by many to rely on achieving a sufficient interference fit between the implant and the bone. However, attempting to achieve a high interference fit not infrequently leads to femoral canal fracture either intra-operatively or soon after. The appropriate range of diametrical interference fit that ensures primary stability without risking femoral fracture is not well understood. In this study, a finite element model was constructed to predict micromotion and, therefore, instability of femoral stems. The model was correlated with an in vitro micromotion experiment carried out on four cadaver femurs. It was confirmed that interference fit has a very significant effect on micromotion and ignoring this parameter in an analysis of primary stability is likely to underestimate the stability of the stem. Furthermore, it was predicted that the optimal level of interference fit is around 50 mm as this is sufficient to achieve good primary fixation while having a safety factor of 2 against femoral canal fracture. This result is of clinical relevance as it indicates a recommendation for the surgeon to err on the side of a low interference fit rather than risking femoral fracture. r 2007 Elsevier Ltd. All rights reserved. Keywords: Hip arthroplasty; Primary stability; Micromotion; Finite element; Loosening
1. Introduction Achieving good primary fixation is of crucial importance in cementless hip arthroplasty to ensure good short-term and long-term results. Lack of primary stability leads to thigh pain and eventual loosening of the prosthesis because of a continuous disruption of the bone formation process around the implant (Kim et al., 2003; Knight et al., 1998; Mont and Hungerford, 1997; Petersilge et al., 1997). The stability, or the lack of it, is commonly measured as the amount of relative motion at the interface between the bone and the stem under physiological load. Large interfacial relative movements reduce the chance of Corresponding author. Tel.: +44 207 5947061.
E-mail address: [email protected] (U. Hansen). 0021-9290/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2007.10.009
osseointegration, and cause the formation of a fibrous tissue layer at the bone–implant interface (Pilliar et al., 1986), which may eventually lead to loosening and failure of the arthroplasty. The threshold value of micromotion, above which a fibrous tissue layer forms, has been studied in both animals and humans. In a review of dental implants in animals, a threshold micromotion value between 50 and 150 mm was found (Szmukler-Moncler et al., 1998). A similar range of values was reported for orthopaedic implants in humans. In a retrieval study of cementless femoral components, Engh et al. (1992) found indications that micromotions less than 40 mm had resulted in osseointegration while micromotions of 150 mm had caused the interposition of a fibrous tissue layer at the stem–bone interface. It can be concluded from these reports that the value of micromotion, above
ARTICLE IN PRESS 588
M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
which osseointegration is disrupted, ranges from 50 to 150 mm, possibly skewed towards the lower end of this range. While many believe a sufficiently high interference fit is essential to achieve good primary stability, it is also clear that introducing a interference fit has caused a clinically significant increase in intra-operative femoral canal fractures (Cameron, 2004; Meek et al., 2004), an effect which has also been demonstrated during in vitro testing (Jasti et al., 1993; Monti et al., 2001). The appropriate range of interference fit that ensures primary stability without risking femoral fracture is not well understood. There has been very little work on the effect of interference fit and except for a few papers (Ramamurti et al., 1997; Viceconti et al., 2000; Shultz et al., 2006), previous FE studies do not include an interference fit. Therefore, the aim of this paper was to study the effect of interference fit in a realistic model of a hip stem and to establish the optimal range of interference fit that ensures primary stability without risking femoral fracture. Current surgical techniques specify under-reaming of 3% of the major cross-sectional dimension and values of interference fits of 0.3–0.5 mm are typically mentioned (Otani et al., 1995; Ramamurti et al., 1997). However, the achieved interference fit is probably much lower as removal of the reamer and insertion of a rough coated implant is likely to remove and crush further material resulting in a larger cavity than indicated by the size of the reamer (Shirazi-Adl et al., 1994). Furthermore, the surgeon is guided by visual and auditory clues when deciding if the implant is ‘‘firmly’’ seated and also visco-elastic effects in the bone will reduce the interference fit. As a consequence of all these effects, even a ball-park estimate of the effective interference fit is fundamentally unknown. There are in principle two parts to this study. In order to get a rough idea of the interference fit introduced using current surgical practise, in the first part of this study, finite element predictions were correlated with in vitro micromotion measurements. The aim of this was to enable back calculation of the real interference fit introduced by the surgeon during the in vitro experiment. In the second part of the study, the effect of a range of interference fits on micromotion predictions was investigated using finite element models of a more physiologically realistic loading scenario than was possible during the first part of the study.
2.1. In vitro experimental set-up The experiment was designed for direct comparison of micromotion values between experiment and FE analyses. Four cadaver femurs and Alloclassic (Zimmer GmbH, Winterthur, Switzerland) hip stems were used, and two points, one in the proximal part and another in the distal part of the stem (Fig. 1), were chosen for micromotion measurement. In order to avoid damaging the stem–bone interface during drilling action, the two points on the implant were drilled before implantation. A guide jig ensured that the bone, subsequent to stem insertion, was drilled in the position matching these same two points on the stem. Finally, steel pegs were glued into the holes in the stem and protruding through the bone (Fig. 1, right). A linear variable differential transducer (LVDT Model DFg5, DC Miniature series, Solartron Metrology, UK), was rigidly fixed to the outside of the femur (Fig. 1, right). The connecting rod of the LVDT core rested on the free-end of the steel peg. When the implant was loaded, the implant and hence the peg moved relative to the bone and the LVDT measured the axial movement of the peg relative to the transducer, thus providing an estimate of the relative axial movement between bone and stem. Implantation was carried out by an experienced orthopaedic surgeon (D.L.). The neck of the femur was first resected, and the femur was then reamed with firm impaction using a series of reamers to open the canal. A femoral stem was then implanted in the femur. The femur was sectioned 250 mm distal to the lesser trochanter and its distal end fixed inside a cylindrical metal container using polymethylmethacrylate (PMMA). These were then placed onto the table bed of a universal materials testing machine (Instron 5565, Instron Corp., Canton, MA). The specimen was adjusted so that the long axis of the stem was coaxial to the direction of loading. A cyclical axial compression load of 0–2 kN and triangular waveform was applied to the shoulder of the stem for 50 cycles at a rate of 1 kN/min using a 5 kN load cell. Micromotion readings via the LVDT were taken manually at maximum load of 2 kN and when fully unloaded at each cycle.
2.2. Finite element methodology for correlation study A 3D model of a hip stem (Alloclassic, Zimmer GmbH) was constructed from CAD files received from the manufacturer (Fig. 2). In the correlation part of the study, the finite element model needs to be as accurate a representation of the experimental set-up as possible. Hence, the FE simulations of this part of the study were based on CT scans of the specific bones used in the experiments. There were two sets of scans: one
proximal and distal peg
2. Methods In the first part of the study, the finite element models were based on CT scans from the specific bones used in the experiment. In the second part of the study, the CT scans from the visible human dataset were used. Also in the first part of the study, the purpose was simply to compare finite element predictions and experiments and to simplify the experiments, a simple load configuration was chosen. In the second part of the study, physiological loads including muscle loads were used.
Fig. 1. The jig used to position the holes in the bone and the pegs in the implant, respectively (left). The implant–bone specimen with LVDT attached to the femur loaded in compression in the mechanical testing machine (right).
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
SL=0.02
589
SL=0.1
SL=0.5
µm 200 150 100 50 0
Fig. 3. Contour plots of micromotion over the surface of the Alloclassic stem under stairclimbing loads and for different values of the SL parameter (SL describes the non-linear friction characteristics of the interface). Fig. 2. The hip stem used in the study indicating the FE mesh used (left) and the implant inserted in the femur (right).
scan prior to inserting the implant in the femur and a subsequent scan after implantation. The first set of scans was used to derive bone geometry and material properties from the Houndsfield units of the scan, while the second set of scans was used to ensure that the implant position and orientation in the FE model precisely matched the implant position within the femur in the experiment. The reason for this two-step procedure is that it would be inappropriate to use the CT datasets from the implanted femur for bone property assignment due to artefacts in these datasets caused by the metal stem. The construction of 3D models of the hip was done using AMIRA software (Mercury Computer Systems, Inc., San Diego, CA). Segmentation was compiled automatically using the software’s marching cubes algorithm which generates a 3D triangular surface mesh. The completed model was then converted to solid linear tetrahedral elements using Marc.Mentat (MSC.Software, Santa Ana, CA) software. The mesh was inspected to ensure it was reasonably shaped throughout. The Marc finite element software package was used in this study. Material properties for the bone were assigned based on the grey-scale value of the CT images on an element-by-element basis. The grey-level of the CT images was related to the apparent density using a linear correlation (Cann and Genant, 1980; McBroom et al., 1985). This allowed for the transformation of the spatial radiological description into the description of bone density. The modulus of elasticity of individual elements was then calculated from the assigned apparent densities using the cubic relationship proposed by Carter and Hayes (1977). The material properties were assumed to be linear elastic and isotropic with Poisson’s ratio set to 0.35. The FE model was loaded at the centre of the shoulder of the stem with 2 kN, the stem being coaxial to the direction of loading, hence, matching the loading configuration in the experiment. Mesh convergence is a standard issue in any finite element analysis and in a contact analysis, there are many other numerical parameters that affect the predicted micromotions. The default contact strategy in Marc is a ‘‘direct constraint’’ algorithm (MSC.Marc-Manual, 2004) which most importantly requires the input of a ‘‘contact zone size’’ (CZS). Furthermore, Bernakiewicz and Viceconti (2002) described the importance of the convergence tolerance (CTol) in non-linear analyses. They also suggested that the appropriate parameter settings should be such that the
resultant change in predicted micromotion between models with different parameter settings should be small relative to 150 mm. A sequential sensitivity analysis involving mesh density, CZS and CTol was carried out and a model with 12,078 nodes, CZS ¼ 0.025 mm and CTol ¼ 1% was found to be sufficient for an accurate solution. We also chose a Coulomb friction model which in Marc requires the input of the friction coefficient (m) as well as a parameter (SL). The Marc software has introduced the parameter SL, which describes a smoothing of the step-function of the Coulomb model, only in order to deal with an otherwise numerically difficult to handle discontinuity. However, not only does this parameter dramatically affect the predicted micromotion (Fig. 3) it also has an important physical interpretation. Shirazi-Adl et al. (1993) showed that the bone–implant interface friction curve is highly non-linear, exhibiting micromotion on the order of 150 mm (that is in the order of the critical level for osseointegration) before the slip load predicted by the Coulomb model is reached. The implication of Shirazi-Adl et al.’s work is that adopting the ideal Coulomb model is inadequate. However, the SL parameter can be interpreted and used to represent this non-linear behaviour. To establish the appropriate setting of the SL parameter, we simulated Shirazi-Adl et al.’s relatively simple experiment consisting of a bone cube exposed to normal and tangential loads moving on a metal plate. In Fig. 4 is shown Shirazi-Adl’s experimental curve of tangential load versus tangential displacement. The tangential load that would initiate slip according to the Coulomb model is 30.6. The finite element predicted curves for various settings of SL is also shown and a setting of SL ¼ 0.1 predicts the experimental curve well. Hence, in the rest of this study, this setting was used. The effect of friction coefficient on micromotion is relatively minor for friction coefficients higher than 0.15 (Kuiper and Huiskes, 1996). Viceconti et al. (2000) found that a friction coefficient between 0.2 and 0.5 led to the best correlation with experiments. Rancourt et al. (1990) measured friction coefficients experimentally and found a coefficient of 0.4. Based on these previous studies, a friction coefficient of 0.4 was used in this study. The objective of this study was to estimate the effective interference fit. Hence, we varied the interference fit in the finite element models. The predictions were then compared to the experimentally measured values to estimate which level of interference best matched the experiment.
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
590
Fn=45N FT
bone Metal
30.6N: Coulomb prediction
Tangential force, FT (N)
35 30 25 20
SL=0.5 15 10 SL=0.1
5
SL=0.02
experimental
0 0
100 200 Micromotion (microns)
300
Fig. 4. Tangential load versus tangential displacement of bone cube sliding on metal plate. The finite element predicted non-linear friction behaviour for different levels of the parameter SL is shown as well as the experimental curve reported by Shirazi-Adl et al. (1993). The critical value at which sliding would initiate according to an ideal Coulomb friction model is also indicated.
Table 1 Applied joint and muscle loads corresponding to peak loads during stairclimbing loads (Heller et al., 2005) Force (Percentage of body weight) Hip contact force Abductor Ilio-tibial tract, proximal part Ilio-tibial tract, distal part Tensor fascia lata, proximal part Tensor fascia lata, distal part Vastus lateralis Vastus medialis
251 114 17 17 6 7 137 270
2.3. Finite element methodology for parametric study on the effect of interference fit In this part of the study, geometry and material characterisation was based on the CT scans available from the visible human project (VHP) dataset. The hip stem model was aligned inside the femur according to the recommendations of the manufacturer (Alloclassic surgical technique, Zimmer Ltd., Warsaw, IN) and an experienced surgeon (D.L.) inspected the resulting configurations and considered them appropriate. The models were restrained distally and loaded with physiological stair-climbing loads including all relevant muscle forces. Stair-climbing loads were applied as this loading scenario has been shown to be more critical than other activities (Kassi et al., 2005). Similarly, Kassi et al. (2005) showed that it is essential to include muscle loads although this issue has been debated (Cristofolini and Viceconti, 2006). This load configuration was based on the extensive work by Bergmann (2001) and Heller et al. (2005) in which load directions and muscle attachments are described. The magnitudes of the loads in percentage body weight are shown in Table 1 and a body weight of 82 kg was used. All other aspects of the model were as described in Section 2.2.
3. Results 3.1. In vitro micromotion measurements and correlation with model predictions During the experiments, the stem initially subsided into the bone but after a sufficient number of cycles the stem settled. Even in this relatively stable state, there continued to be low levels of reversible motion at the stem–bone interface as a result of the continued load–unload cycle. It is high levels of this continued disruption of the interface that is thought to prevent osteogenic cells from bonding to the surface of the stem (Pilliar et al., 1986). Hence, in terms of evaluating the ingrowth potential of an arthroplasty, it is the reversible micromotion rather than subsidence which is the relevant constituent of the overall relative motion between bone and implant. The reversible micromotion during a load cycle was estimated as the difference between the total micromotion measured at maximum load and the total micromotion when the specimen was unloaded. For the remainder of this study, we will refer to this quantity as reversible micromotion or just micromotion. In Fig. 5 is shown the experimentally determined reversible micromotion for each of the four specimens plotted against load cycles for the distal and proximal parts, top and bottom, respectively. During the initial cycles, this micromotion was high but then stabilised at a lower value. Relatively high levels of micromotion during the first few times a patient exposes a joint to loading, are probably not critical. It is the long term or stabilised value of reversible micromotion that will continue to disrupt the implant–bone interface and prevent osseointegration. Hence, it is the converged values of Fig. 5 which are relevant. Based on the data of Fig. 5, the converged average value in the distal and proximal regions were 1872 and 1975 mm, respectively. The results of the FE analyses using different levels of interference fit and simulating the experiment are shown in Fig. 6. The figure shows that with just 1 mm of interference, the level of micromotion is predicted to be in the range of 20–30 mm. With 2 mm of interference, this drops to 10–20 mm. Comparing this to the experimental values of 18 and 19 mm also shown in the figure, this implies that the interference fit introduced by the surgeon is only 1 or 2 mm. This seems perhaps unrealistically low. Shultz et al. (2006) considered an interference fit of 100 mm to cause bone interface damage and reported this level of interference as a threshold value. Therefore, we included an interference fit of 100 mm in one of the finite element models and inspected the resulting tensile hoop stresses (Fig. 7). This model was not exposed to any other loads. As can be seen from the figure, interference induced hoop stresses are on the order of 50 MPa on the surface of the bone (internally the stresses are somewhat higher). Comparing this stress level with the transverse tensile strength of cortical bone of approximately 50 MPa (Reilly and Burstein, 1975), it would seem that 100 mm represents
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
591
Distal micromotion 60 Spec 1
Micromotion (micrometer)
Spec 2
50
Spec 3 Spec 4
40 MPa 30
50
20
40 30
10
20 0 0
10
20
30
40
50
Cycles
10 0
Proximal micromotion 45 Spec 1
Micromotion (micrometer)
40
Spec 2 Spec 3
35
Spec 4
Fig. 7. Hoop stresses in femoral bone caused by an interference fit of 100 mm. No external loads are applied in this model.
30 25 20 15 10 5 0 0
10
20
30
40
50
Cycles Fig. 5. Distal micromotion (top) and proximal micromotion (bottom) results from the experiment.
no interference
1 micron
2 microns
5 microns
the critical level of effective interference fit above which the femoral canal will fracture. The location of high hoop stresses towards the distal end of the implant seen in Fig. 7 also matches the location of 77% of intra-operative fractures (Meek et al., 2004). Considering that femoral canal fractures are not infrequently occurring intra-operatively (Cameron, 2004; Meek et al., 2004), it would seem that surgeons are introducing close to the critical level of interference fit of 100 mm. Assuming that surgeons are able to control the insertion process within a factor of 2, perhaps a realistic range of interference fit can be argued to be in the range of 50–100 mm. In summary, this first part of the study indicates that the range of realistic interference fits may be within a range of very low levels (just a few microns) and up to 100 mm. 3.2. The effect of interference fit on micromotion
µm
19 ± 5µm experiment
40 30 20 10 0
experiment 18 ± 2µm
Fig. 6. Contour plots of micromotion over the surface of the stem under an axial load of 2 kN, using interference fits of (from left to right) 0, 1, 2 and 5 mm, respectively. The experimentally determined proximal and distal micromotion is also indicated.
Fig. 8 shows contour plots of predicted micromotion over the stem surface under stairclimbing loads and for four different levels of interference fit. Fig. 9 shows the change in micromotion with levels of interference fit for the two points labelled P (proximal) and D (distal) shown on the left model of Fig. 8. Also in Fig. 9 is indicated, by the grey-coloured region, the threshold range of micromotion above which soft tissue formation will be predicted and below which osseointegration would be expected. From these two figures, it is clear that the interference fit had a very large effect on micromotion predictions. In the case of no interference fit, the entire surface area of the implant was in or above the grey area indicating that the
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
592
No interference
5 microns
25 microns
50 microns
P
µm 200 150 100 50 0
D
Fig. 8. Contour plots of micromotion over the surface of the stem under stairclimbing loads and with interference fits of (from left to right) 0, 5, 25 and 50 mm, respectively.
micromotion (microns)
200
150 micromotion at 'P' 100
50 micromotion at 'D' 0 0
25
50 75 interference (microns)
100
Fig. 9. Micromotion at points P (proximal) and D (distal) as a function of the level of interference fit. Locations of point P and D are shown in Fig. 8 (left). The grey area indicates the range of the critical micromotion threshold. Above this level, fibrous tissue formation would be expected; below, osseointegration is anticipated.
primary stability of the implant is at risk. In contrast, with 50 mm of interference, all but the most proximal part of the implant was predicted to osseointegrate. Interestingly, increasing the level of interference beyond 50 mm had negligible effect. Also, it is clear that the effect of the interference fit was most dramatic at low levels of interference. Including just 5 mm of interference causes almost a 50% reduction in micromotion and including more interference only has a relatively small effect. 4. Discussion This study has shown that modelling the interference fit characteristic of hip stems is crucial for quantitative predictions of micromotion. Ignoring the interference fit will probably lead to an under-estimation of the stability of
the stem. In contrast, ignoring the non-linear friction behaviour reported by Shirazi-Adl et al. (1993) and reproduced in Fig. 4, will probably to lead to too optimistic predictions of stem stability. The magnitude of interference fit is fundamentally unknown and may be the reason most previous works have omitted this parameter from their finite element analyses. Indeed, during this study it became clear just how difficult it is to estimate this parameter. Nevertheless, this study demonstrates the importance of the interference fit as including only a small level of interference changed the evaluation of the investigated stem from that of an unstable stem to that of a stable stem. Our predictions showed high levels of micromotion distally and proximally while micromotion at the stem midsection was lower (Fig. 8, left). This is qualitatively consistent with the finite element predictions by Keaveny and Bartel (1993). Keaveny and Bartel did not include an interference fit and predicted very high absolute values of micromotion (0–550 mm). Keaveny and Bartel simulated a cylindrical stem which is likely to be less resistant to torsional loads and that may explain the higher levels of micromotion as compared to our results. Viceconti et al. (2000) did simulate a press-fit although it is not possible to quantify this press-fit in a manner that allows a direct correlation with our results. Viceconti et al. predicted micromotions ranging from 17 to 49 mm across the surface of the implant which is reasonably consistent with our results simulating an interference fit of 25 mm (Fig. 8). The results of Fig. 6 indicate that surgeons introduce very low interference fits, on the order of 1–2 mm. Apart from any aspects of the model that may cause inaccurate predictions, it is of course also possible that the experimental results are inaccurate. Notably, our experiment, like the vast majority of other experimental micromotion studies, does not measure the actual interface micromotion but instead measures the motion between the LVDT fixation point on the bone and the point of the peg insertion on the implant. The motion measured, therefore, includes other flexibilities such as bone deformation and will tend to overestimate micromotion (Bu¨hler et al., 1997). If these flexibilities are substantial compared to the true interface micromotion, it would cause our methodology to predict very small levels of interference which is of course what seems to be the case. In connection with Fig. 7, we proposed that surgeons are in fact more likely to introduce interference fits of 50–100 mm. Shultz et al. (2006) predicted that with an interference fit of 100 mm, the hoop stresses in the bone would visco-elastically relax by approximately 50%. In other words, if a surgeon introduces an interference fit of 100 mm, this would relax and represent an effective interference of 50 mm. Shultz reported that interference fits lower than 100 mm would relax less than 50%. Therefore, even if a surgeon only achieves the lower range of the 50–100 mm interference, we have estimated, there should be
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
at least 25 mm of effective interference left after relaxation, well above the 1–2 mm estimated from the experiment. We have no evidence to explain the small levels of interference fit predicted from the experiments but we are inclined to believe that the experiment overestimated the micromotion, for the reasons noted above. We have assumed a uniform interference fit over the entire surface of the implant. Accordingly, the press-fit (pressure) varied considerably from the proximal cancellous femur to the cortical distal femur as modelled through the variation in the local Young’s modulus of the bone adjacent to the implant. This variation in press-fit between the proximal and distal region is undoubtedly qualitatively correct. However, our study was not set up to investigate variation in interference fit. This was not included due to the practical difficulty in quantifying the variation and generalising such variation that is likely to vary between implants. It is also probable, given the very small interferences calculated, that surgeons cannot create implant cavities with uniform interference across the interface area, so that clinical cases would include variations from the micromotions predicted. The effects of a more realistic scenario are not yet known. The results of this study support the suggestion made earlier (Shirazi-Adl et al., 1994) that the cavity that is created in the femur is larger than is indicated by the nominal interference of 0.3–0.5 mm (Otani et al., 1995; Ramamurti et al., 1997); such a large interference would cause the femur to fracture, according to our results. Perhaps the most important result of the study and the result with direct clinical relevance relates to Figs. 7 and 9. Fig. 7 predicts that surgery is safe against femoral canal fracture at interference fits lower than 100 mm. Fig. 9 predicts that the stem would osseointegrate at interference levels of 50 mm. Therefore, the recommendation is for the surgeon to err on the side of a low interference fit during surgery as only 50 mm is enough to achieve stability and provides a safety factor of 2 against femoral canal fracture. If considering a stem likely to be successful as long as just the distal part of the stem (embedded in the strong cortical bone) osseointegrates, Fig. 9 indicates that just 10 mm of interference fit is necessary for stability and provides a safety factor of 10 against femoral canal fracture. Of course, our computational predictions should be further investigated before being applied in clinical practise. It is likely, that stems with different geometry or material will behave differently. The Alloclassic stem in this study, for example, has a rectangular cross-section, which might be advantageous in resisting torsional loading during the stairclimbing simulated. Nevertheless, the predictions clearly indicate a recommendation to modify surgical practise thereby reducing or even eliminating the 7% intra-operative femoral canal fractures during primary hip surgery reported by Cameron, (2004) and the 6–50% fracture rates reported by Meek et al. (2004) in connection with revision hip surgery.
593
Conflict of interest statement Each author has been involved in the design and interpretation of the study and writing of the manuscript. All authors have read and concur with the content of the manuscript. The under-signed warrant that the article is original, does not infringe upon any copyright or other proprietary right of any third party, is not under consideration by another journal and has not been published previously. Acknowledgements This work has been funded by the Government of Malaysia and partly funded by Zimmer Inc., Warsaw, IN, USA. References Bergmann, G. (Ed.), 2001. HIP98—Loading of the Hip Joint. Compact Disc. Free University of Berlin. Bernakiewicz, M., Viceconti, M., 2002. The role of parameter identification in finite element contact analyses with reference to orthopaedic biomechanics applications. Journal of Biomechanics 35, 61–67. Bu¨hler, D.W., Oxland, T.R., Nolte, L.P., 1997. Design evaluation of a device for measuring three-dimensional micromotions of press-fit femoral stem prostheses. Medical Engineering & Physics 19 (2), 187–199. Cameron, H.U., 2004. Intraoperative hip fractures: ruining your day. Journal of Arthroplasty 19 (4 (Suppl. 1)), 99–103. Cann, C., Genant, H., 1980. Precise measurement of vertebral mineral content using computed tomography. Journal of Computer Assisted Tomography 4, 493–500. Carter, D., Hayes, W., 1977. The compressive behaviour of bone as a twophase porous structure. Journal of Bone and Joint Surgery—American Volume 59, 954–962. Cristofolini, L., Viceconti, M., 2006. Comments on ‘‘Stair climbing is more critical than walking in pre-clinical assessment of primary stability in cementless THA in vitro’’ by Jean-Pierre Kassi, Markus O. Heller, Ulrich Stoeckle, Carsten Perka, Georg N. Duda, published in Journal of Biomechanics 2005; 38, 1143–1154. Journal of Biomechanics 2006; 39 (16), 3085–3087. Engh, C.A., Oconnor, D., Jasty, M., Mcgovern, T.F., Bobyn, J.D., Harris, W.H., 1992. Quantification of implant micromotion, strain shielding, and bone-resorption with porous-coated anatomic medullary locking femoral prostheses. Clinical Orthopaedics and Related Research, 13–29. Heller, M.O., Bergmann, G., Kassi, J.-P., Claes, L., Haas, N.P., Duda, G.N., 2005. Determination of muscle loading at the hip joint for use in pre-clinical testing. Journal of Biomechanics 38, 1155–1163. Jasti, M., Henshaw, R.M., O’Connor, D.O., Harris, W.H., 1993. High assembly strains and femoral fractures produced during insertion of uncemented femoral components. A cadaver study. Journal of Arthroplasty 8 (5), 479–487. Kassi, J.-P., Heller, M.O., Stoeckle, U., Perka, C., Guda, G.N., 2005. Stair climbing is more critical than walking in pre-clinical assessment of primary stability in cementless THA in vitro. Journal of Biomechanics 38, 1143–1154. Keaveny, T.M., Bartel, D.L., 1993. Effects of porous coating, with and without collar support, on early relative motion for a cementless hipprosthesis. Journal of Biomechanics 26, 1355–1368. Kim, Y.H., Oh, S.H., Kim, J.S., 2003. Primary total hip arthroplasty with a second-generation cementless total hip prosthesis in patients younger than fifty years of age. Journal of Bone and Joint Surgery—American Volume 85A, 109–114.
ARTICLE IN PRESS 594
M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
Knight, J.L., Atwater, R.D., Guo, J., 1998. Clinical results of the midstem porous-coated anatomic uncemented femoral stem in primary total hip arthroplasty: a five- to nine-year prospective study. The Journal of Arthroplasty 13, 535–545. Kuiper, J.H., Huiskes, R., 1996. Friction and stem stiffness affect dynamic interface in total hip replacement. Journal of Orthopaedic Research 14, 36–43. Mcbroom, R.J., Hayes, W.C., Edwards, W.T., Goldberg, R.P., White, A.A., 1985. Prediction of vertebral body compressive fracture using quantitative computed-tomography. Journal of Bone and Joint Surgery—American Volume 67A, 1206–1214. Meek, R.M., Garbuz, D.S., Masri, B.A., Gredanus, N.V., Duncan, C.P., 2004. Intraoperative fracture of the femur in revision total hip arthroplasty with a diaphyseal fitting stem. Journal of Bone and Joint Surgery 86-A (3), 480–485. Mont, M.A., Hungerford, D.S., 1997. Proximally coated ingrowth prostheses—a review. Clinical Orthopaedics and Related Research, 139–149. Monti, L., Cristofolini, L., Toni, A., Ceroni, R.G., 2001. In-vitro testing of the primary stability of the VerSys enhanced taper stem: a comparative study in intact and intraoperative cracked femora. Proceedings of the Institution of Mechanical Engineers, Part H—Journal of Engineering in Medicine 215 (1), 75–83. MSC.Marc-Manual, 2004. Vol. A Theory and User Information. MSC.Software Corporation. Otani, T., Whiteside, L.A., White, S.E., McCarthy, D.S., 1995. Reaming technique of the femoral diaphysis in cementless total hip arthroplasty. Clinical Orthopaedics and Related Research 311, 210–221. Petersilge, W.J., Dlima, D.D., Walker, R.H., Colwell, C.W., 1997. Prospective study of 100 consecutive Harris-Galante porous total hip arthroplasties-4- to 8-year follow-up study. Journal of Arthroplasty 12, 185–193.
Pilliar, R.M., Lee, J.M., Maniatopoulos, C., 1986. Observations on the effect of movement on bone ingrowth into porous-surfaced implants. Clinical Orthopaedics and Related Research, 108–113. Ramamurti, B.S., Orr, T.E., Bragdon, C.R., Lowenstein, J.D., Jasty, M., Harris, W.H., 1997. Factors influencing stability at the interface between a porous surface and cancellous bone: a finite element analysis of a canine in vivo micromotion experiment. Journal of Biomedical Materials Research 36, 274–280. Rancourt, D., Shirazi-Adl, A., Drouin, G., Paiement, G., 1990. Friction properties of the interface between porous-surfaced metals and tibial cancellous bone. Journal of Biomedical Materials Research 24, 1503–1519. Reilly, D.T., Burstein, A.H., 1975. The elastic and ultimate properties of compact bone tissue. Journal of Biomechanics 8, 393–405. Shirazi-Adl, A., Dammak, M., Paiement, G., 1993. Experimental determination of friction characteristics at the trabecular bone/ porous-coated metal interface in cementless implants. Journal of Biomedical Materials Research 27, 167–175. Shirazi-Adl, A., Dammak, M., Zukor, D.J., 1994. Fixation pull-out response measurement of bone screws and porous-surfaced posts. Journal of Biomechanics 27 (10), 1249–1258. Shultz, T.R., Blaha, J.D., Gruen, T.A., Norman, T.L., 2006. Cortical bone viscoelasticity and fixation strength of press-fit femoral stems: a finite element model. ASME Journal of Biomechanical Engineering 128 (1), 7–12. Szmukler-Moncler, S., Salama, H., Reingewirtz, Y., Dubruille, J.H., 1998. Timing of loading and effect of micromotion on bone-dental implant interface: review of experimental literature. Journal of Biomedical Materials Research 43, 192–203. Viceconti, M., Muccini, R., Bernakiewicz, M., Baleani, M., Cristofolini, L., 2000. Large-sliding contact elements accurately predict levels of bone–implant micromotion relevant to osseointegration. Journal of Biomechanics 33, 1611–1618.
Journal of Biomechanics 41 (2008) 587–594 www.elsevier.com/locate/jbiomech www.JBiomech.com
Finite element modelling of primary hip stem stability: The effect of interference fit Mohammed Rafiq Abdul-Kadira,b, Ulrich Hansena,, Ralf Klabundec, Duncan Lucasa, Andrew Amisa,d a Department of Mechanical Engineering, Imperial College, London, UK Biomechanics and Tissue Engineering Group, Universiti Teknologi Malaysia, Malaysia c Zimmer GmbH, Winterthur, Switzerland d Division of Surgery, Imperial College London, UK
b
Accepted 13 October 2007
Abstract The most commonly reported complications related to cementless hip stems are loosening and thigh pain; both of these have been attributed to high levels of relative micromotion at the bone–implant interface due to insufficient primary fixation. Primary fixation is believed by many to rely on achieving a sufficient interference fit between the implant and the bone. However, attempting to achieve a high interference fit not infrequently leads to femoral canal fracture either intra-operatively or soon after. The appropriate range of diametrical interference fit that ensures primary stability without risking femoral fracture is not well understood. In this study, a finite element model was constructed to predict micromotion and, therefore, instability of femoral stems. The model was correlated with an in vitro micromotion experiment carried out on four cadaver femurs. It was confirmed that interference fit has a very significant effect on micromotion and ignoring this parameter in an analysis of primary stability is likely to underestimate the stability of the stem. Furthermore, it was predicted that the optimal level of interference fit is around 50 mm as this is sufficient to achieve good primary fixation while having a safety factor of 2 against femoral canal fracture. This result is of clinical relevance as it indicates a recommendation for the surgeon to err on the side of a low interference fit rather than risking femoral fracture. r 2007 Elsevier Ltd. All rights reserved. Keywords: Hip arthroplasty; Primary stability; Micromotion; Finite element; Loosening
1. Introduction Achieving good primary fixation is of crucial importance in cementless hip arthroplasty to ensure good short-term and long-term results. Lack of primary stability leads to thigh pain and eventual loosening of the prosthesis because of a continuous disruption of the bone formation process around the implant (Kim et al., 2003; Knight et al., 1998; Mont and Hungerford, 1997; Petersilge et al., 1997). The stability, or the lack of it, is commonly measured as the amount of relative motion at the interface between the bone and the stem under physiological load. Large interfacial relative movements reduce the chance of Corresponding author. Tel.: +44 207 5947061.
E-mail address: [email protected] (U. Hansen). 0021-9290/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2007.10.009
osseointegration, and cause the formation of a fibrous tissue layer at the bone–implant interface (Pilliar et al., 1986), which may eventually lead to loosening and failure of the arthroplasty. The threshold value of micromotion, above which a fibrous tissue layer forms, has been studied in both animals and humans. In a review of dental implants in animals, a threshold micromotion value between 50 and 150 mm was found (Szmukler-Moncler et al., 1998). A similar range of values was reported for orthopaedic implants in humans. In a retrieval study of cementless femoral components, Engh et al. (1992) found indications that micromotions less than 40 mm had resulted in osseointegration while micromotions of 150 mm had caused the interposition of a fibrous tissue layer at the stem–bone interface. It can be concluded from these reports that the value of micromotion, above
ARTICLE IN PRESS 588
M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
which osseointegration is disrupted, ranges from 50 to 150 mm, possibly skewed towards the lower end of this range. While many believe a sufficiently high interference fit is essential to achieve good primary stability, it is also clear that introducing a interference fit has caused a clinically significant increase in intra-operative femoral canal fractures (Cameron, 2004; Meek et al., 2004), an effect which has also been demonstrated during in vitro testing (Jasti et al., 1993; Monti et al., 2001). The appropriate range of interference fit that ensures primary stability without risking femoral fracture is not well understood. There has been very little work on the effect of interference fit and except for a few papers (Ramamurti et al., 1997; Viceconti et al., 2000; Shultz et al., 2006), previous FE studies do not include an interference fit. Therefore, the aim of this paper was to study the effect of interference fit in a realistic model of a hip stem and to establish the optimal range of interference fit that ensures primary stability without risking femoral fracture. Current surgical techniques specify under-reaming of 3% of the major cross-sectional dimension and values of interference fits of 0.3–0.5 mm are typically mentioned (Otani et al., 1995; Ramamurti et al., 1997). However, the achieved interference fit is probably much lower as removal of the reamer and insertion of a rough coated implant is likely to remove and crush further material resulting in a larger cavity than indicated by the size of the reamer (Shirazi-Adl et al., 1994). Furthermore, the surgeon is guided by visual and auditory clues when deciding if the implant is ‘‘firmly’’ seated and also visco-elastic effects in the bone will reduce the interference fit. As a consequence of all these effects, even a ball-park estimate of the effective interference fit is fundamentally unknown. There are in principle two parts to this study. In order to get a rough idea of the interference fit introduced using current surgical practise, in the first part of this study, finite element predictions were correlated with in vitro micromotion measurements. The aim of this was to enable back calculation of the real interference fit introduced by the surgeon during the in vitro experiment. In the second part of the study, the effect of a range of interference fits on micromotion predictions was investigated using finite element models of a more physiologically realistic loading scenario than was possible during the first part of the study.
2.1. In vitro experimental set-up The experiment was designed for direct comparison of micromotion values between experiment and FE analyses. Four cadaver femurs and Alloclassic (Zimmer GmbH, Winterthur, Switzerland) hip stems were used, and two points, one in the proximal part and another in the distal part of the stem (Fig. 1), were chosen for micromotion measurement. In order to avoid damaging the stem–bone interface during drilling action, the two points on the implant were drilled before implantation. A guide jig ensured that the bone, subsequent to stem insertion, was drilled in the position matching these same two points on the stem. Finally, steel pegs were glued into the holes in the stem and protruding through the bone (Fig. 1, right). A linear variable differential transducer (LVDT Model DFg5, DC Miniature series, Solartron Metrology, UK), was rigidly fixed to the outside of the femur (Fig. 1, right). The connecting rod of the LVDT core rested on the free-end of the steel peg. When the implant was loaded, the implant and hence the peg moved relative to the bone and the LVDT measured the axial movement of the peg relative to the transducer, thus providing an estimate of the relative axial movement between bone and stem. Implantation was carried out by an experienced orthopaedic surgeon (D.L.). The neck of the femur was first resected, and the femur was then reamed with firm impaction using a series of reamers to open the canal. A femoral stem was then implanted in the femur. The femur was sectioned 250 mm distal to the lesser trochanter and its distal end fixed inside a cylindrical metal container using polymethylmethacrylate (PMMA). These were then placed onto the table bed of a universal materials testing machine (Instron 5565, Instron Corp., Canton, MA). The specimen was adjusted so that the long axis of the stem was coaxial to the direction of loading. A cyclical axial compression load of 0–2 kN and triangular waveform was applied to the shoulder of the stem for 50 cycles at a rate of 1 kN/min using a 5 kN load cell. Micromotion readings via the LVDT were taken manually at maximum load of 2 kN and when fully unloaded at each cycle.
2.2. Finite element methodology for correlation study A 3D model of a hip stem (Alloclassic, Zimmer GmbH) was constructed from CAD files received from the manufacturer (Fig. 2). In the correlation part of the study, the finite element model needs to be as accurate a representation of the experimental set-up as possible. Hence, the FE simulations of this part of the study were based on CT scans of the specific bones used in the experiments. There were two sets of scans: one
proximal and distal peg
2. Methods In the first part of the study, the finite element models were based on CT scans from the specific bones used in the experiment. In the second part of the study, the CT scans from the visible human dataset were used. Also in the first part of the study, the purpose was simply to compare finite element predictions and experiments and to simplify the experiments, a simple load configuration was chosen. In the second part of the study, physiological loads including muscle loads were used.
Fig. 1. The jig used to position the holes in the bone and the pegs in the implant, respectively (left). The implant–bone specimen with LVDT attached to the femur loaded in compression in the mechanical testing machine (right).
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
SL=0.02
589
SL=0.1
SL=0.5
µm 200 150 100 50 0
Fig. 3. Contour plots of micromotion over the surface of the Alloclassic stem under stairclimbing loads and for different values of the SL parameter (SL describes the non-linear friction characteristics of the interface). Fig. 2. The hip stem used in the study indicating the FE mesh used (left) and the implant inserted in the femur (right).
scan prior to inserting the implant in the femur and a subsequent scan after implantation. The first set of scans was used to derive bone geometry and material properties from the Houndsfield units of the scan, while the second set of scans was used to ensure that the implant position and orientation in the FE model precisely matched the implant position within the femur in the experiment. The reason for this two-step procedure is that it would be inappropriate to use the CT datasets from the implanted femur for bone property assignment due to artefacts in these datasets caused by the metal stem. The construction of 3D models of the hip was done using AMIRA software (Mercury Computer Systems, Inc., San Diego, CA). Segmentation was compiled automatically using the software’s marching cubes algorithm which generates a 3D triangular surface mesh. The completed model was then converted to solid linear tetrahedral elements using Marc.Mentat (MSC.Software, Santa Ana, CA) software. The mesh was inspected to ensure it was reasonably shaped throughout. The Marc finite element software package was used in this study. Material properties for the bone were assigned based on the grey-scale value of the CT images on an element-by-element basis. The grey-level of the CT images was related to the apparent density using a linear correlation (Cann and Genant, 1980; McBroom et al., 1985). This allowed for the transformation of the spatial radiological description into the description of bone density. The modulus of elasticity of individual elements was then calculated from the assigned apparent densities using the cubic relationship proposed by Carter and Hayes (1977). The material properties were assumed to be linear elastic and isotropic with Poisson’s ratio set to 0.35. The FE model was loaded at the centre of the shoulder of the stem with 2 kN, the stem being coaxial to the direction of loading, hence, matching the loading configuration in the experiment. Mesh convergence is a standard issue in any finite element analysis and in a contact analysis, there are many other numerical parameters that affect the predicted micromotions. The default contact strategy in Marc is a ‘‘direct constraint’’ algorithm (MSC.Marc-Manual, 2004) which most importantly requires the input of a ‘‘contact zone size’’ (CZS). Furthermore, Bernakiewicz and Viceconti (2002) described the importance of the convergence tolerance (CTol) in non-linear analyses. They also suggested that the appropriate parameter settings should be such that the
resultant change in predicted micromotion between models with different parameter settings should be small relative to 150 mm. A sequential sensitivity analysis involving mesh density, CZS and CTol was carried out and a model with 12,078 nodes, CZS ¼ 0.025 mm and CTol ¼ 1% was found to be sufficient for an accurate solution. We also chose a Coulomb friction model which in Marc requires the input of the friction coefficient (m) as well as a parameter (SL). The Marc software has introduced the parameter SL, which describes a smoothing of the step-function of the Coulomb model, only in order to deal with an otherwise numerically difficult to handle discontinuity. However, not only does this parameter dramatically affect the predicted micromotion (Fig. 3) it also has an important physical interpretation. Shirazi-Adl et al. (1993) showed that the bone–implant interface friction curve is highly non-linear, exhibiting micromotion on the order of 150 mm (that is in the order of the critical level for osseointegration) before the slip load predicted by the Coulomb model is reached. The implication of Shirazi-Adl et al.’s work is that adopting the ideal Coulomb model is inadequate. However, the SL parameter can be interpreted and used to represent this non-linear behaviour. To establish the appropriate setting of the SL parameter, we simulated Shirazi-Adl et al.’s relatively simple experiment consisting of a bone cube exposed to normal and tangential loads moving on a metal plate. In Fig. 4 is shown Shirazi-Adl’s experimental curve of tangential load versus tangential displacement. The tangential load that would initiate slip according to the Coulomb model is 30.6. The finite element predicted curves for various settings of SL is also shown and a setting of SL ¼ 0.1 predicts the experimental curve well. Hence, in the rest of this study, this setting was used. The effect of friction coefficient on micromotion is relatively minor for friction coefficients higher than 0.15 (Kuiper and Huiskes, 1996). Viceconti et al. (2000) found that a friction coefficient between 0.2 and 0.5 led to the best correlation with experiments. Rancourt et al. (1990) measured friction coefficients experimentally and found a coefficient of 0.4. Based on these previous studies, a friction coefficient of 0.4 was used in this study. The objective of this study was to estimate the effective interference fit. Hence, we varied the interference fit in the finite element models. The predictions were then compared to the experimentally measured values to estimate which level of interference best matched the experiment.
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
590
Fn=45N FT
bone Metal
30.6N: Coulomb prediction
Tangential force, FT (N)
35 30 25 20
SL=0.5 15 10 SL=0.1
5
SL=0.02
experimental
0 0
100 200 Micromotion (microns)
300
Fig. 4. Tangential load versus tangential displacement of bone cube sliding on metal plate. The finite element predicted non-linear friction behaviour for different levels of the parameter SL is shown as well as the experimental curve reported by Shirazi-Adl et al. (1993). The critical value at which sliding would initiate according to an ideal Coulomb friction model is also indicated.
Table 1 Applied joint and muscle loads corresponding to peak loads during stairclimbing loads (Heller et al., 2005) Force (Percentage of body weight) Hip contact force Abductor Ilio-tibial tract, proximal part Ilio-tibial tract, distal part Tensor fascia lata, proximal part Tensor fascia lata, distal part Vastus lateralis Vastus medialis
251 114 17 17 6 7 137 270
2.3. Finite element methodology for parametric study on the effect of interference fit In this part of the study, geometry and material characterisation was based on the CT scans available from the visible human project (VHP) dataset. The hip stem model was aligned inside the femur according to the recommendations of the manufacturer (Alloclassic surgical technique, Zimmer Ltd., Warsaw, IN) and an experienced surgeon (D.L.) inspected the resulting configurations and considered them appropriate. The models were restrained distally and loaded with physiological stair-climbing loads including all relevant muscle forces. Stair-climbing loads were applied as this loading scenario has been shown to be more critical than other activities (Kassi et al., 2005). Similarly, Kassi et al. (2005) showed that it is essential to include muscle loads although this issue has been debated (Cristofolini and Viceconti, 2006). This load configuration was based on the extensive work by Bergmann (2001) and Heller et al. (2005) in which load directions and muscle attachments are described. The magnitudes of the loads in percentage body weight are shown in Table 1 and a body weight of 82 kg was used. All other aspects of the model were as described in Section 2.2.
3. Results 3.1. In vitro micromotion measurements and correlation with model predictions During the experiments, the stem initially subsided into the bone but after a sufficient number of cycles the stem settled. Even in this relatively stable state, there continued to be low levels of reversible motion at the stem–bone interface as a result of the continued load–unload cycle. It is high levels of this continued disruption of the interface that is thought to prevent osteogenic cells from bonding to the surface of the stem (Pilliar et al., 1986). Hence, in terms of evaluating the ingrowth potential of an arthroplasty, it is the reversible micromotion rather than subsidence which is the relevant constituent of the overall relative motion between bone and implant. The reversible micromotion during a load cycle was estimated as the difference between the total micromotion measured at maximum load and the total micromotion when the specimen was unloaded. For the remainder of this study, we will refer to this quantity as reversible micromotion or just micromotion. In Fig. 5 is shown the experimentally determined reversible micromotion for each of the four specimens plotted against load cycles for the distal and proximal parts, top and bottom, respectively. During the initial cycles, this micromotion was high but then stabilised at a lower value. Relatively high levels of micromotion during the first few times a patient exposes a joint to loading, are probably not critical. It is the long term or stabilised value of reversible micromotion that will continue to disrupt the implant–bone interface and prevent osseointegration. Hence, it is the converged values of Fig. 5 which are relevant. Based on the data of Fig. 5, the converged average value in the distal and proximal regions were 1872 and 1975 mm, respectively. The results of the FE analyses using different levels of interference fit and simulating the experiment are shown in Fig. 6. The figure shows that with just 1 mm of interference, the level of micromotion is predicted to be in the range of 20–30 mm. With 2 mm of interference, this drops to 10–20 mm. Comparing this to the experimental values of 18 and 19 mm also shown in the figure, this implies that the interference fit introduced by the surgeon is only 1 or 2 mm. This seems perhaps unrealistically low. Shultz et al. (2006) considered an interference fit of 100 mm to cause bone interface damage and reported this level of interference as a threshold value. Therefore, we included an interference fit of 100 mm in one of the finite element models and inspected the resulting tensile hoop stresses (Fig. 7). This model was not exposed to any other loads. As can be seen from the figure, interference induced hoop stresses are on the order of 50 MPa on the surface of the bone (internally the stresses are somewhat higher). Comparing this stress level with the transverse tensile strength of cortical bone of approximately 50 MPa (Reilly and Burstein, 1975), it would seem that 100 mm represents
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
591
Distal micromotion 60 Spec 1
Micromotion (micrometer)
Spec 2
50
Spec 3 Spec 4
40 MPa 30
50
20
40 30
10
20 0 0
10
20
30
40
50
Cycles
10 0
Proximal micromotion 45 Spec 1
Micromotion (micrometer)
40
Spec 2 Spec 3
35
Spec 4
Fig. 7. Hoop stresses in femoral bone caused by an interference fit of 100 mm. No external loads are applied in this model.
30 25 20 15 10 5 0 0
10
20
30
40
50
Cycles Fig. 5. Distal micromotion (top) and proximal micromotion (bottom) results from the experiment.
no interference
1 micron
2 microns
5 microns
the critical level of effective interference fit above which the femoral canal will fracture. The location of high hoop stresses towards the distal end of the implant seen in Fig. 7 also matches the location of 77% of intra-operative fractures (Meek et al., 2004). Considering that femoral canal fractures are not infrequently occurring intra-operatively (Cameron, 2004; Meek et al., 2004), it would seem that surgeons are introducing close to the critical level of interference fit of 100 mm. Assuming that surgeons are able to control the insertion process within a factor of 2, perhaps a realistic range of interference fit can be argued to be in the range of 50–100 mm. In summary, this first part of the study indicates that the range of realistic interference fits may be within a range of very low levels (just a few microns) and up to 100 mm. 3.2. The effect of interference fit on micromotion
µm
19 ± 5µm experiment
40 30 20 10 0
experiment 18 ± 2µm
Fig. 6. Contour plots of micromotion over the surface of the stem under an axial load of 2 kN, using interference fits of (from left to right) 0, 1, 2 and 5 mm, respectively. The experimentally determined proximal and distal micromotion is also indicated.
Fig. 8 shows contour plots of predicted micromotion over the stem surface under stairclimbing loads and for four different levels of interference fit. Fig. 9 shows the change in micromotion with levels of interference fit for the two points labelled P (proximal) and D (distal) shown on the left model of Fig. 8. Also in Fig. 9 is indicated, by the grey-coloured region, the threshold range of micromotion above which soft tissue formation will be predicted and below which osseointegration would be expected. From these two figures, it is clear that the interference fit had a very large effect on micromotion predictions. In the case of no interference fit, the entire surface area of the implant was in or above the grey area indicating that the
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
592
No interference
5 microns
25 microns
50 microns
P
µm 200 150 100 50 0
D
Fig. 8. Contour plots of micromotion over the surface of the stem under stairclimbing loads and with interference fits of (from left to right) 0, 5, 25 and 50 mm, respectively.
micromotion (microns)
200
150 micromotion at 'P' 100
50 micromotion at 'D' 0 0
25
50 75 interference (microns)
100
Fig. 9. Micromotion at points P (proximal) and D (distal) as a function of the level of interference fit. Locations of point P and D are shown in Fig. 8 (left). The grey area indicates the range of the critical micromotion threshold. Above this level, fibrous tissue formation would be expected; below, osseointegration is anticipated.
primary stability of the implant is at risk. In contrast, with 50 mm of interference, all but the most proximal part of the implant was predicted to osseointegrate. Interestingly, increasing the level of interference beyond 50 mm had negligible effect. Also, it is clear that the effect of the interference fit was most dramatic at low levels of interference. Including just 5 mm of interference causes almost a 50% reduction in micromotion and including more interference only has a relatively small effect. 4. Discussion This study has shown that modelling the interference fit characteristic of hip stems is crucial for quantitative predictions of micromotion. Ignoring the interference fit will probably lead to an under-estimation of the stability of
the stem. In contrast, ignoring the non-linear friction behaviour reported by Shirazi-Adl et al. (1993) and reproduced in Fig. 4, will probably to lead to too optimistic predictions of stem stability. The magnitude of interference fit is fundamentally unknown and may be the reason most previous works have omitted this parameter from their finite element analyses. Indeed, during this study it became clear just how difficult it is to estimate this parameter. Nevertheless, this study demonstrates the importance of the interference fit as including only a small level of interference changed the evaluation of the investigated stem from that of an unstable stem to that of a stable stem. Our predictions showed high levels of micromotion distally and proximally while micromotion at the stem midsection was lower (Fig. 8, left). This is qualitatively consistent with the finite element predictions by Keaveny and Bartel (1993). Keaveny and Bartel did not include an interference fit and predicted very high absolute values of micromotion (0–550 mm). Keaveny and Bartel simulated a cylindrical stem which is likely to be less resistant to torsional loads and that may explain the higher levels of micromotion as compared to our results. Viceconti et al. (2000) did simulate a press-fit although it is not possible to quantify this press-fit in a manner that allows a direct correlation with our results. Viceconti et al. predicted micromotions ranging from 17 to 49 mm across the surface of the implant which is reasonably consistent with our results simulating an interference fit of 25 mm (Fig. 8). The results of Fig. 6 indicate that surgeons introduce very low interference fits, on the order of 1–2 mm. Apart from any aspects of the model that may cause inaccurate predictions, it is of course also possible that the experimental results are inaccurate. Notably, our experiment, like the vast majority of other experimental micromotion studies, does not measure the actual interface micromotion but instead measures the motion between the LVDT fixation point on the bone and the point of the peg insertion on the implant. The motion measured, therefore, includes other flexibilities such as bone deformation and will tend to overestimate micromotion (Bu¨hler et al., 1997). If these flexibilities are substantial compared to the true interface micromotion, it would cause our methodology to predict very small levels of interference which is of course what seems to be the case. In connection with Fig. 7, we proposed that surgeons are in fact more likely to introduce interference fits of 50–100 mm. Shultz et al. (2006) predicted that with an interference fit of 100 mm, the hoop stresses in the bone would visco-elastically relax by approximately 50%. In other words, if a surgeon introduces an interference fit of 100 mm, this would relax and represent an effective interference of 50 mm. Shultz reported that interference fits lower than 100 mm would relax less than 50%. Therefore, even if a surgeon only achieves the lower range of the 50–100 mm interference, we have estimated, there should be
ARTICLE IN PRESS M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
at least 25 mm of effective interference left after relaxation, well above the 1–2 mm estimated from the experiment. We have no evidence to explain the small levels of interference fit predicted from the experiments but we are inclined to believe that the experiment overestimated the micromotion, for the reasons noted above. We have assumed a uniform interference fit over the entire surface of the implant. Accordingly, the press-fit (pressure) varied considerably from the proximal cancellous femur to the cortical distal femur as modelled through the variation in the local Young’s modulus of the bone adjacent to the implant. This variation in press-fit between the proximal and distal region is undoubtedly qualitatively correct. However, our study was not set up to investigate variation in interference fit. This was not included due to the practical difficulty in quantifying the variation and generalising such variation that is likely to vary between implants. It is also probable, given the very small interferences calculated, that surgeons cannot create implant cavities with uniform interference across the interface area, so that clinical cases would include variations from the micromotions predicted. The effects of a more realistic scenario are not yet known. The results of this study support the suggestion made earlier (Shirazi-Adl et al., 1994) that the cavity that is created in the femur is larger than is indicated by the nominal interference of 0.3–0.5 mm (Otani et al., 1995; Ramamurti et al., 1997); such a large interference would cause the femur to fracture, according to our results. Perhaps the most important result of the study and the result with direct clinical relevance relates to Figs. 7 and 9. Fig. 7 predicts that surgery is safe against femoral canal fracture at interference fits lower than 100 mm. Fig. 9 predicts that the stem would osseointegrate at interference levels of 50 mm. Therefore, the recommendation is for the surgeon to err on the side of a low interference fit during surgery as only 50 mm is enough to achieve stability and provides a safety factor of 2 against femoral canal fracture. If considering a stem likely to be successful as long as just the distal part of the stem (embedded in the strong cortical bone) osseointegrates, Fig. 9 indicates that just 10 mm of interference fit is necessary for stability and provides a safety factor of 10 against femoral canal fracture. Of course, our computational predictions should be further investigated before being applied in clinical practise. It is likely, that stems with different geometry or material will behave differently. The Alloclassic stem in this study, for example, has a rectangular cross-section, which might be advantageous in resisting torsional loading during the stairclimbing simulated. Nevertheless, the predictions clearly indicate a recommendation to modify surgical practise thereby reducing or even eliminating the 7% intra-operative femoral canal fractures during primary hip surgery reported by Cameron, (2004) and the 6–50% fracture rates reported by Meek et al. (2004) in connection with revision hip surgery.
593
Conflict of interest statement Each author has been involved in the design and interpretation of the study and writing of the manuscript. All authors have read and concur with the content of the manuscript. The under-signed warrant that the article is original, does not infringe upon any copyright or other proprietary right of any third party, is not under consideration by another journal and has not been published previously. Acknowledgements This work has been funded by the Government of Malaysia and partly funded by Zimmer Inc., Warsaw, IN, USA. References Bergmann, G. (Ed.), 2001. HIP98—Loading of the Hip Joint. Compact Disc. Free University of Berlin. Bernakiewicz, M., Viceconti, M., 2002. The role of parameter identification in finite element contact analyses with reference to orthopaedic biomechanics applications. Journal of Biomechanics 35, 61–67. Bu¨hler, D.W., Oxland, T.R., Nolte, L.P., 1997. Design evaluation of a device for measuring three-dimensional micromotions of press-fit femoral stem prostheses. Medical Engineering & Physics 19 (2), 187–199. Cameron, H.U., 2004. Intraoperative hip fractures: ruining your day. Journal of Arthroplasty 19 (4 (Suppl. 1)), 99–103. Cann, C., Genant, H., 1980. Precise measurement of vertebral mineral content using computed tomography. Journal of Computer Assisted Tomography 4, 493–500. Carter, D., Hayes, W., 1977. The compressive behaviour of bone as a twophase porous structure. Journal of Bone and Joint Surgery—American Volume 59, 954–962. Cristofolini, L., Viceconti, M., 2006. Comments on ‘‘Stair climbing is more critical than walking in pre-clinical assessment of primary stability in cementless THA in vitro’’ by Jean-Pierre Kassi, Markus O. Heller, Ulrich Stoeckle, Carsten Perka, Georg N. Duda, published in Journal of Biomechanics 2005; 38, 1143–1154. Journal of Biomechanics 2006; 39 (16), 3085–3087. Engh, C.A., Oconnor, D., Jasty, M., Mcgovern, T.F., Bobyn, J.D., Harris, W.H., 1992. Quantification of implant micromotion, strain shielding, and bone-resorption with porous-coated anatomic medullary locking femoral prostheses. Clinical Orthopaedics and Related Research, 13–29. Heller, M.O., Bergmann, G., Kassi, J.-P., Claes, L., Haas, N.P., Duda, G.N., 2005. Determination of muscle loading at the hip joint for use in pre-clinical testing. Journal of Biomechanics 38, 1155–1163. Jasti, M., Henshaw, R.M., O’Connor, D.O., Harris, W.H., 1993. High assembly strains and femoral fractures produced during insertion of uncemented femoral components. A cadaver study. Journal of Arthroplasty 8 (5), 479–487. Kassi, J.-P., Heller, M.O., Stoeckle, U., Perka, C., Guda, G.N., 2005. Stair climbing is more critical than walking in pre-clinical assessment of primary stability in cementless THA in vitro. Journal of Biomechanics 38, 1143–1154. Keaveny, T.M., Bartel, D.L., 1993. Effects of porous coating, with and without collar support, on early relative motion for a cementless hipprosthesis. Journal of Biomechanics 26, 1355–1368. Kim, Y.H., Oh, S.H., Kim, J.S., 2003. Primary total hip arthroplasty with a second-generation cementless total hip prosthesis in patients younger than fifty years of age. Journal of Bone and Joint Surgery—American Volume 85A, 109–114.
ARTICLE IN PRESS 594
M.R. Abdul-Kadir et al. / Journal of Biomechanics 41 (2008) 587–594
Knight, J.L., Atwater, R.D., Guo, J., 1998. Clinical results of the midstem porous-coated anatomic uncemented femoral stem in primary total hip arthroplasty: a five- to nine-year prospective study. The Journal of Arthroplasty 13, 535–545. Kuiper, J.H., Huiskes, R., 1996. Friction and stem stiffness affect dynamic interface in total hip replacement. Journal of Orthopaedic Research 14, 36–43. Mcbroom, R.J., Hayes, W.C., Edwards, W.T., Goldberg, R.P., White, A.A., 1985. Prediction of vertebral body compressive fracture using quantitative computed-tomography. Journal of Bone and Joint Surgery—American Volume 67A, 1206–1214. Meek, R.M., Garbuz, D.S., Masri, B.A., Gredanus, N.V., Duncan, C.P., 2004. Intraoperative fracture of the femur in revision total hip arthroplasty with a diaphyseal fitting stem. Journal of Bone and Joint Surgery 86-A (3), 480–485. Mont, M.A., Hungerford, D.S., 1997. Proximally coated ingrowth prostheses—a review. Clinical Orthopaedics and Related Research, 139–149. Monti, L., Cristofolini, L., Toni, A., Ceroni, R.G., 2001. In-vitro testing of the primary stability of the VerSys enhanced taper stem: a comparative study in intact and intraoperative cracked femora. Proceedings of the Institution of Mechanical Engineers, Part H—Journal of Engineering in Medicine 215 (1), 75–83. MSC.Marc-Manual, 2004. Vol. A Theory and User Information. MSC.Software Corporation. Otani, T., Whiteside, L.A., White, S.E., McCarthy, D.S., 1995. Reaming technique of the femoral diaphysis in cementless total hip arthroplasty. Clinical Orthopaedics and Related Research 311, 210–221. Petersilge, W.J., Dlima, D.D., Walker, R.H., Colwell, C.W., 1997. Prospective study of 100 consecutive Harris-Galante porous total hip arthroplasties-4- to 8-year follow-up study. Journal of Arthroplasty 12, 185–193.
Pilliar, R.M., Lee, J.M., Maniatopoulos, C., 1986. Observations on the effect of movement on bone ingrowth into porous-surfaced implants. Clinical Orthopaedics and Related Research, 108–113. Ramamurti, B.S., Orr, T.E., Bragdon, C.R., Lowenstein, J.D., Jasty, M., Harris, W.H., 1997. Factors influencing stability at the interface between a porous surface and cancellous bone: a finite element analysis of a canine in vivo micromotion experiment. Journal of Biomedical Materials Research 36, 274–280. Rancourt, D., Shirazi-Adl, A., Drouin, G., Paiement, G., 1990. Friction properties of the interface between porous-surfaced metals and tibial cancellous bone. Journal of Biomedical Materials Research 24, 1503–1519. Reilly, D.T., Burstein, A.H., 1975. The elastic and ultimate properties of compact bone tissue. Journal of Biomechanics 8, 393–405. Shirazi-Adl, A., Dammak, M., Paiement, G., 1993. Experimental determination of friction characteristics at the trabecular bone/ porous-coated metal interface in cementless implants. Journal of Biomedical Materials Research 27, 167–175. Shirazi-Adl, A., Dammak, M., Zukor, D.J., 1994. Fixation pull-out response measurement of bone screws and porous-surfaced posts. Journal of Biomechanics 27 (10), 1249–1258. Shultz, T.R., Blaha, J.D., Gruen, T.A., Norman, T.L., 2006. Cortical bone viscoelasticity and fixation strength of press-fit femoral stems: a finite element model. ASME Journal of Biomechanical Engineering 128 (1), 7–12. Szmukler-Moncler, S., Salama, H., Reingewirtz, Y., Dubruille, J.H., 1998. Timing of loading and effect of micromotion on bone-dental implant interface: review of experimental literature. Journal of Biomedical Materials Research 43, 192–203. Viceconti, M., Muccini, R., Bernakiewicz, M., Baleani, M., Cristofolini, L., 2000. Large-sliding contact elements accurately predict levels of bone–implant micromotion relevant to osseointegration. Journal of Biomechanics 33, 1611–1618.