We studied the migration of 58 cemented Hinek

Transcription

1 MIGRATION OF CEMENTED FEMORAL COMPONENTS AFTER THR ROENTGEN STEREOPHOTOGRAMMETRIC ANALYSIS J. KISS, D. W. MURRAY, A. R. TURNER-SMITH, J. BITHELL, C. J. BULSTRODE From the Nuffield Orthopaedic Centre, Oxford, England We studied the migration of 58 cemented Hinek femoral components for total hip replacement, using roentgen stereophotogrammetric analysis over four years. The implants migrated faster during the first year than subsequently, and the pattern of migration in the second period was very different. During the first year they subsided, tilted into varus and internally rotated. After this there was slow distal migration with no change in orientation. None of the prostheses has yet failed. The early migration is probably caused by resorption of bone damaged by surgical trauma or the heat generated by the polymerisation of bone cement. Later migration may be due to creep in the bone cement or the surrounding fibrous membrane. The prosthesis which we studied allows the preservation of some of the femoral neck, and comparison with published migration studies of the Charnley stem suggests that this decreases rotation and may help to prevent loosening. J Bone Joint Surg [Br] 1996;78-B: Received 18 May 1994; Accepted after revision 9 February 1996 Early clinical assessment of the performance of a new design of total hip replacement is notoriously unreliable: full evaluation of function requires a clinical trial of at least ten years (Murray, Carr and Bulstrode 1993; Owen et al 1994). By then, if the implant is unsatisfactory, many J. Kiss, MD, Senior Lecturer in Orthopaedic Surgery Department of Orthopaedics, Semmelweis University of Medicine, Budapest, Karolina ut 27, 1113 Hungary. D. W. Murray, MD, FRCS Orth, Consultant Orthopaedic Surgeon C. J. Bulstrode, MD, FRCS Orth, Clinical Reader in Orthopaedic Surgery Nuffield Department of Orthopaedic Surgery, University of Oxford, Nuffield Orthopaedic Centre, Windmill Road, Headington, Oxford OX3 7LD, UK. A. R. Turner-Smith, DPhil, Senior Lecturer in Rehabilitation Engineering King s College School of Medicine and Dentistry, Denmark Hill, London SE5 9RS, UK. J. F. Bithell, MA, DPhil, Senior Lecturer, Department of Statistics University of Oxford, 1 South Parks Road, Oxford OX1 3TG, UK. Correspondence should be sent to Mr D. W. Murray British Editorial Society of Bone and Joint Surgery X/96/51230 $2.00 patients will have suffered. A method of predicting implant performance as soon as possible after implantation is needed, and roentgen stereophotogrammetric analysis (RSA) has this potential. It has been shown for some types of implant that RSA can detect abnormalities before there are clinical or radiological signs of loosening (Green et al 1983; Mjöberg, Hansson and Selvik 1984b; Chafetz et al 1984; Nistor et al 1991; Kärrholm and Snorrason 1993), and that there is a good correlation between increased migration and early loosening (Mjöberg et al 1985; Snorrason and Kärrholm 1990; Franzén, Mjöberg and Önnerfält 1992; Kärrholm and Snorrason 1993; Freeman and Plante- Bordeneuve 1994). There is little information from RSA studies, however, about the rate or pattern of migration which predicts early loosening of the cemented femoral components (Kärrholm et al 1994). RSA can measure the migration and rotation of components relative to bone in tenths of millimetres and degrees. Since its introduction in 1974 by Selvik (1989), a number of different types of cemented and non-cemented acetabular and femoral components have been studied in different centres, but details of the early rate and direction of migration and rotation of different parts of cemented femoral components have not been reported. We aimed to determine the three-dimensional migration and rotation of one type of cemented femoral component at different times after implantation in order to provide a base-line with which other implants could be compared. PATIENTS AND METHODS We replaced 58 osteoarthritic hips in 56 patients using cemented Hinek components (Corin Medical, Cirencester, UK) (Fig. 1). The average age of the patients was 70 years (44 to 87); 31 were female and 25 male. An anterolateral approach was used and at the time of surgery 1 mm stainless-steel marker balls were inserted into the bone around the femoral implant using an injector gun. These were positioned in pairs at the tip of the greater trochanter, at the anterior margin of the lesser trochanter, and 1 cm distal to the tip of the femoral component. The patients were mobilised fully weight-bearing on the second or third postoperative day. The first RSA examination was at about ten days. This was repeated at six months, 12 months and annually; at 796 THE JOURNAL OF BONE AND JOINT SURGERY

2 MIGRATION OF CEMENTED FEMORAL COMPONENTS AFTER THR 797 Fig. 1 RSA radiograph of a Hinek THR showing the calibration grids and the marker balls in the greater and lesser trochanters and distal to the tip. Part of the femoral neck has been preserved to resist rotation. each review the patient completed a clinical questionnaire. During the RSA examination, the patients stood fully weight-bearing within a calibration frame. Two X-ray beams were angulated at 60 to each other, and two X-ray films were placed perpendicular to the beams. To avoid fogging from radiation scatter the two exposures were separated by 0.7 s, during which time radio-opaque shutters moved to allow each film to be exposed in turn. The films were digitised on an electromagnetic digitising tablet (GTCO Corporation, California) using a CCD image sensor and an image capture board (Primagraphics, Guildford, UK) connected to a SUN workstation (SUN, California). The images of the calibration frame were used to determine the position of the X-ray sources and films in space. The positions of the marker balls implanted into the bone were determined, as was the centre of the head and the tip and the shoulder of the femoral component. The position of the centre of the femoral component was determined from the positions of the head, shoulder and tip. Special software was developed for the three-dimensional reconstruction and for the analysis of migration and rotation (Turner-Smith and Bulstrode 1993; Kiss et al 1995). Accuracy of RSA. We assessed the accuracy of the system with a test object and in selected patients, determining the differences between two sets of measurements. This method is used in most RSA studies, but alternatively, and perhaps more logically, accuracy could be determined from the standard deviation of a series of measurements, in which case it would be 2 times smaller. We expressed accuracy as two standard deviations of the differences which we found. The test object, which contained 13 marker balls and a femoral component, was imaged in nine different orientations. The system could measure the relative position of bone and implant to better than 0.11 mm in three dimensions. To determine the clinical accuracy we reanalysed the 28 film pairs of seven patients, and showed that the migration of the different parts of the component in three dimensions could be determined with an accuracy of between 0.25 and 0.50 mm/year and the rotation of the component with an accuracy of between 0.2 and 0.8 /year. Analysis of results. The four-year follow-up was divided into two periods, early (the first year) and late, since it is known that migration is faster at first and then becomes slower and continuous (Mjöberg 1986; Nistor et al 1991; Kärrholm and Snorrason 1993; Søballe et al 1993). For each period we calculated the migration rate in three directions at each of four landmarks (head, shoulder, tip and centre) and the rotation rate around the centre using linear regression. We determined the mean and SEM for migration and rotation rates. Hips were included in the analysis for each period if they had at least two satisfactory sets of radiographs during that time. Multivariate analysis (Hotelling s T 2 ) was used for testing significance to analyse migration and rotation data in different directions and at different points (see statistical appendix). The data combined for significance testing were: a) migration in three directions at each landmark; b) rotation in three directions, which assesses changes in orientation; and c) rotation and migration of the centre, which assesses the six degrees of freedom of movement. Both absolute displacements and differences in displacement between the two groups of implants were analysed in the same manner. When comparisons were made between the early and late period, the hips analysed in the early period were excluded from the late period to maintain independence. RESULTS There were no implant failures during the study, but technical problems with the radiographs meant that only 14 of the VOL. 78-B, NO. 5, SEPTEMBER 1996

3 798 J. KISS, D. W. MURRAY, A. R. TURNER-SMITH, J. BITHELL, C. J. BULSTRODE Table I. The mean migration rate (mm/year ± SEM) of three landmarks on the femoral component Landmark First year After first year Difference Centre of head Medial 0.48 ± ± 0.04 Distal 0.54 ± ± 0.04 Posterior 0.61 ± ± 0.11 Overall p value (0.001) 0.15 (0.15) <0.001 (0.00) Shoulder Medial 0.20 ± ± 0.14 Distal 0.39 ± ± 0.06 Anterior 0.16 ± ± 0.18 Overall p value (0.051) 0.24 (0.19) (0.00) Distal tip Lateral 0.05 ± ± 0.05 Distal 0.04 ± ± 0.05 Anterior 0.13 ± ± 0.14 Overall p value 0.63 (0.59) (0.021) 0.52 (0.50) the p values obtained by simulation are given in parentheses, see statistical appendix Table II. Mean migration (mm/year ± SEM) and rotation ( /year ± SEM) around the centre of the femoral component (the p values obtained by simulation are given in parentheses, see statistical appendix) First year After first year Difference Migration Medial 0.21 ± ± 0.05 Distal 0.29 ± ± 0.04 Posterior 0.10 ± ± 0.07 Overall p value (0.066) (0.014) (0.01) Rotation Anterior ± ± 0.08 Internal 1.17 ± ± 0.29 Abduction 0.32 ± ± 0.04 Overall p value 0.13 (0.11) 0.67 (0.66) (0.01) Migration and rotation combined p value (0.092) (0.011) Fig. 2 Fig. 3 investigated hips had suitable data sets for detailed analysis in the first year, while 49 had suitable sets for the later period. The migration patterns for the first postoperative year and subsequently are summarised in Figures 2 to 4 and Tables I and II. There were significant levels of migration and rotation in both periods, but they were substantially higher during the first year. The pattern of migration was also very different during the two time periods. During the first year the head of the prosthesis had the largest overall rate of migration (0.94 mm/year, p = 0.002). This was a combined distal, medial and posterior movement, with similar rates of migration in each direction. The tip of the femoral implant had the slowest migration, and this was not statistically significant (0.15 mm/year, p = 0.63). The rates of migration of the shoulder and centre were intermediate between the head and tip (0.47 mm/year, p=0.05 and 0.37 mm/year, p = 0.08 respectively); the Fig. 4 Migration rates during the first year (left) and subsequently (right). Figure 2 Coronal plane. Figure 3 Sagittal plane. Figure 4 Transverse plane. THE JOURNAL OF BONE AND JOINT SURGERY

4 MIGRATION OF CEMENTED FEMORAL COMPONENTS AFTER THR 799 direction was predominantly medial and distal. During this period the prosthesis was moving into varus and internal rotation around its distal part which tended to subside slowly. After the first year, all the measured points on the prosthesis appeared to migrate slowly distally. The migration rate at the centre was 0.15 mm/year (p = 0.017), and there was no significant change in rotation (p = 0.67), showing that its orientation was not changing. The centre of the prosthesis showed significant differences in migration and rotation between the first and subsequent years (p = 0.01, p = 0.01). The migration rate was two-and-a-half times greater in the first year than subsequently and the rotation rate was six times greater in the first year. Both the rate and pattern of migration were different in the two periods, with a greater change in pattern than in rate. These differences in migration were also highly significant for the head and shoulder (p < 0.001, p = 0.001) but were not significant at the tip (p = 0.05). This is because the implant was rotating around its tip in the first year, causing only a small amount of movement at the tip, which was similar to the movement at the tip after the first year. DISCUSSION The substantially larger migration of the femoral component in the first year confirms other reports (Mjöberg 1986; Nistor et al 1991; Kärrholm and Snorrason 1993; Søballe et al 1993; Önsten et al 1995), but our study has also shown that the pattern or direction of migration changed with time, and that this change in pattern is larger than the change in rate. This difference suggests that different mechanisms underly the migration at different periods. The rapid early migration is probably caused by resorption of the bone layer which has been injured by surgical trauma and the heat of polymerisation of PMMA cement (Feith 1975; Huiskes 1980; Mjöberg et al 1984a). This migration occurs at the cement-bone interface and is limited by the extent of bone damage. The direction of migration during the first postoperative year was a combination of internal rotation and varus tilt (Figs 1 to 3), with the largest migration at the head in a medial, distal and posterior direction. The posterior migration is explained by the posterior component of the joint contact force, which is particularly large during stair-climbing, straight-leg raising and rising from a chair (Rydell 1966; Davy et al 1988; Phillips, Messieh and McDonald 1990). The medial element of migration, when the joint contact force is usually distal or distal and lateral (Bergmann, Graichen and Rohlmann 1993), can be explained by our finding that the tip is the point with the least migration. The implant is therefore most securely fixed near the tip and will tend to rotate about this point. As a result vertical components of the joint contact force will tend to cause the component to tilt into varus and the head to migrate medially. After the first year, when the initial settling is complete, the implants tend to migrate slowly distally. This is the only direction in which they can migrate without substantial bone loss and probably results from a combination of creep in the cement allowing the implant to sink within its mantle and the gradual remodelling of bone and fibrous tissues around the cement. Migration at the cement-implant interface is also possible, but this is unlikely to be large since the prosthesis which we studied has a series of ridges in the metaphyseal region designed to prevent such movement. We determined the average migration rates in each direction and at four landmarks on the implant, since we believed that this would give the best information. Our results are presented in diagrams for easy understanding and we used multivariate analysis for statistical comparisons. We believe that this approach is more useful than the determination of maximal total point motion (MTPM) or the number of implants which migrate more than a certain amount (Green et al 1983; Chafetz et al 1984; Mjöberg et al 1984b, c, 1985, 1986; Wykman, Selvik and Goldie 1988; Snorrason and Kärrholm 1990; Franzén et al 1992; Søballe et al 1993). MTPM is derived from the magnitude of the migration of each landmark on the prosthesis in each of the three axes, and is the single largest measurement. This ignores direction, and may give misleading information because it compounds errors and makes the magnitude of migration inappropriately large. The proportion of implants that migrate more than a certain distance, usually defined as the accuracy of the system, also ignores direction and does not help in the understanding of the mechanism of migration. The most useful role for RSA is to compare different implants and design features, but unfortunately most other studies are not directly comparable with ours because of the different method of analysis. Absolute migration cannot be directly compared, but relative migrations in different directions can be correlated. It has been demonstrated that the head of a Charnley total hip replacement migrates between two and three times as fast in a posterior direction as it does in a distal or mediolateral direction (Önsten et al 1995). The Hinek femoral component which we used extended into the greater trochanter and allowed for preservation of the calcar and most of the femoral neck in order to reduce rotation, which is an important mode of failure (Mjöberg et al 1984b,c; Freeman 1986; Nistor et al 1991). In contrast to results for the Charnley femoral component, posterior migration of the head of the Hinek implant was not appreciably more than that in a distal or mediolateral direction. This suggests that the design of the Hinek implant does prevent rotation, and may help to prevent loosening. The migration of cemented femoral components is a complex combination of rotation and translation in three dimensions, and can therefore be fully assessed only by RSA. We have given a more detailed description of this migration than has been available, and consider that this VOL. 78-B, NO. 5, SEPTEMBER 1996

5 800 J. KISS, D. W. MURRAY, A. R. TURNER-SMITH, J. BITHELL, C. J. BULSTRODE three-dimensional analysis contributes much more to the understanding of the mechanisms of failure than simple uniplanar, one-dimensional measurements of stem shrinkage. For uncemented femoral components, which show a large early migration, both uniplanar radiographs and RSA can be used to provide early predictions of implant performance (Freeman and Plante-Bordeneuve 1994). For cemented components, with much slower migration, uniplanar radiographs, without marker balls, may not have the necessary accuracy. We therefore believe that RSA is essential to predict the performance of cemented femoral stems within a few years of implantation. Different implants should be studied by RSA to determine their patterns of base-line migration and the effect of different design features. From this, RSA should be used to identify those features which enhance fixation and will improve implant design. RSA could also be used to compare various surgical techniques and even for the statutory testing of new implants. We wish to thank Mrs B. Marks, Dr T. Gunther, Mr W. J. Seymour, Mrs S. Whapham, Dr A. Polliack, Mr P. J. Burn and Mr J. Recht for their help, and the Department of Health and the Arthritis and Rheumatism Council for financial support. No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Statistical appendix. Various different approaches to the statistical analysis are possible. At each landmark migration is measured in three directions; to test them individually would ignore the way in which they are interrelated. The overall migration at each landmark could be studied but this may not reflect the greater importance of migration in one direction. A combination of tests relating to migrations in different directions or at different points could be used, but this would be difficult as these migrations are not statistically independent. Our solution was to use multivariate analysis, specifically Hotelling s T 2, which is a generalisation to several dimensions of Student s t-test, to which it is equivalent in the one-dimensional case (Chatfield and Collins 1980). The test comes in two versions, a onesample form to test the hypothesis that a vector of observations has expected value zero and a two-sample form to test the hypothesis that such a vector has the same theoretical expected value in each of two samples. The tests automatically allow for the correlation of the co-ordinates being considered and also, like Student s t-test in one dimension, for the fact that the variances and correlations must be estimated from the data. The tests may or may not be appropriate, however, for a given data set; in the two-sample case they also assume that the variances and correlations are the same in the two populations considered. Whether such assumptions are valid for multivariate data is hard to establish, particularly from moderately sized samples. We therefore checked the p values obtained using simulation to perform randomisation tests as follows. In the one-sample case, we randomly and independently changed the sign of each vector of co-ordinates and recomputed Hotelling s T 2 statistic. We repeated this exercise n times and counted the number of times s that the simulated value exceeded the observed value from the data. The ratio s/n is then an unbiased estimate of the p value for experiments in which the magnitudes and directions of the observation vectors are like those observed (not necessarily following a normal distribution); the correlation structure is thus preserved, but the sense of the displacement is symmetrical about zero. This is consistent with the null hypothesis that the mean displacement is zero. (n was chosen to be either 100, 500 or 1000 depending on the p value.) In the two-sample case, we put the two samples of n 1 and n 2 cases together and then randomly selected n 1 of them to serve as a new sample 1, with the remainder constituting a further new sample 2. We recomputed Hotelling s T 2 and repeated the exercise n times as before. This is consistent with the null hypothesis which specifies that there is no difference between the two populations concerned but preserves the correlation structure and other characteristics of the data. The p values obtained using simulation were, for the most part, consistent with the p values computed for Hotelling s test assuming multivariate normality (in Tables I and II, the simulated results are shown in parentheses). This indicates that for further analyses of this type it is probably reasonable to use Hotelling s test assuming multivariate normality. REFERENCES Bergmann G, Graichen F, Rohlman A. Hip joint loading during walking and running, measured in two patients. J Biomech 1993;26: Chafetz N, Baumrind S, Murray WR, Genant HK. Femoral prosthesis subsidence in asymptomatic patients: a stereophotogrammetric assessment. Invest Radiol 1984;19: Chatfield C, Collins AJ. An introduction to multivariate analysis. London: Chapman & Hall, Davy DT, Kotzar GM, Brown RH, et al. Telemetric force measurements across the hip after total hip arthroplasty. J Bone Joint Surg [Am] 1988;70-A: Feith R. Side-effects of acrylic cement implanted into bone: a histological, (micro)angiographic, fluorescence-microscopic and autoradiographic study in the rabbit femur. Acta Orthop Scand 1975:Suppl 161. Franzén H, Mjöberg B, Önnerfält R. Early loosening of femoral components after cemented revision: a roentgen stereophotogrammetric study. J Bone Joint Surg [Br] 1992;74-B: Freeman MAR. Why resect the neck? J Bone Joint Surg [Br] 1986;68-B: Freeman MAR, Plante-Bordeneuve P. Early migration and late aseptic failure of proximal femoral prostheses. J Bone Joint Surg [Br] 1994;76-B: Green DL, Bahniuk E, Liebelt RA, Fender E, Mirkov P. Biplane radiographic measurements of reversible displacement (including clinical loosening) and migration of total joint replacements. J Bone Joint Surg [Am] 1983;65-A: Huiskes R. Some fundamental aspects of human joint replacement: analyses of stresses and heat conduction in bone-prosthesis structures. Acta Orthop Scand 1980:Suppl 185. Kärrholm J, Snorrason F. Subsidence, tip and hump micromovements of non-coated ribbed femoral prostheses. Clin Orthop 1993;287: Kärrholm J, Borssén B, Löwenhielm G, Snorrason F. Early micromotions in cemented femoral stems subsequently revised due to pain or osteolysis. Trans Orthop Res Soc 1994;19:246. Kiss J, Murray DW, Turner-Smith AR, Bulstrode CJ. Roentgen stereophotogrammetric analysis for assessing migration of total hip replacement femoral components. Proc Inst Mech Engn 1995;209: Mjöberg B. Loosening of the cemented hip prosthesis: the importance of heat injury. Acta Orthop Scand 1986;Suppl 221:1-40. Mjöberg B, Pettersson H, Rosenqvist R, Rydholm A. Bone cement, thermal injury and the radiolucent zone. Acta Orthop Scand 1984a;55: Mjöberg B, Hansson LI, Selvik G. Instability, migration and laxity of total hip prostheses: a roentgen stereophotogrammetric study. Acta Orthop Scand 1984b;55: Mjöberg B, Hansson LI, Selvik G. Instability of total hip prostheses at rotational stress: a roentgen stereophotogrammetric study. Acta Orthop Scand 1984c;55: Mjöberg B, Brismar J, Hansson LI, et al. Definition of endoprosthetic loosening: comparison of arthrography, scintigraphy and roentgen stereophotogrammetry in prosthetic hips. Acta Orthop Scand 1985;56: THE JOURNAL OF BONE AND JOINT SURGERY

6 MIGRATION OF CEMENTED FEMORAL COMPONENTS AFTER THR 801 Mjöberg B, Selvik G, Hansson LI, Rosenqvist R, Önnerfält R. Mechanical loosening of total hip prostheses: a radiographic and roentgen stereophotogrammetric study. J Bone Joint Surg [Br] 1986;68-B: Murray DW, Carr AJ, Bulstrode C. Survival analysis of joint replacements. J Bone Joint Surg [Br] 1993;75-B: Nistor L, Blaha JD, Kjellström U, Selvik G. In vivo measurements of relative motion between an uncemented femoral total hip component and the femur by roentgen stereophotogrammetric analysis. Clin Orthop 1991;269: Önsten I, Åkesson K, Besjakov J, Obrant KJ. Migration of the Charnley stem in rheumatoid arthritis and osteoarthritis: a roentgen stereophotogrammetric study. J Bone Joint Surg [Br] 1995;77-B: Owen TD, Moran CG, Smith SR, Pinder IM. Results of uncemented porous-coated anatomic total hip replacment. J Bone Joint Surg [Br] 1994;76-B: Phillips TW, Messieh SS, McDonald PD. Femoral stem fixation in hip replacement: a biomechanical comparison of cementless and cemented prostheses. J Bone Joint Surg [Br] 1990;72-B: Rydell NP. Forces acting on the femoral-head prosthesis: a study on strain gauge supplied prostheses in living persons. Acta Orthop Scand 1966;37 Suppl 88. Selvik G. Roentgen stereophotogrammetry: a method for the study of the kinematics of the skeletal system. Acta Orthop Scand 1989;60 Suppl 232:1-51. Snorrason F, Kärrholm J. Early loosening of revision hip arthroplasty: a roentgen stereophotogrammetric analysis. J Arthroplasty 1990;5: Søballe K, Toksvig-Larsen S, Gelineck J, et al. Migration of hydroxyapatite coated femoral stems: a roentgen stereophotogrammetric study. J Bone Joint Surg [Br] 1993;75-B: Turner-Smith AR, Bulstrode CJ. Stereoradiogrammetry for the prediction of hip replacement survival. In: Turner-Smith AR, ed. Micromovement in orthopaedics. Oxford: Clarendon Press, 1993: Wykman A, Selvik G, Goldie I. Subsidence of the femoral component in the noncemented total hip: a roentgen stereophotogrammetric analysis. Acta Orthop Scand 1988;59: VOL. 78-B, NO. 5, SEPTEMBER 1996