Metal-on-metal hip resurfacing is increasingly being used as an alternative to conventional total hip replacement in a selected group of recipients of primary hip implants worldwide. The distinguishing feature of resurfacing arthroplasty is that it preserves the bone of the femoral neck. Improvements in metal-on-metal bearing technology and interest in bone-conserving devices, coupled with reports of early clinical success1-6, have contributed to the growth in the number of procedures being performed.
The advantages of hip resurfacing are accompanied by risks that differ from those associated with conventional arthroplasty and are related to the loading of the femoral neck. The most frequently reported complication leading to revision of modern metal-on-metal hip resurfacing is femoral neck fracture1,2,4,7,8, followed by femoral component loosening1. The prevalence of femoral neck fracture has ranged from 0% to 2.1% in larger series7-9. Several clinical factors have been observed in association with fracture of the femoral neck after resurfacing. These factors include patient characteristics, such as low bone-mineral density and cystic changes, and mechanical abnormalities, such as coxa vara or coxa breva. Surgical technique variables associated with femoral component failure have been ascertained by radiographic analysis and retrieval analysis of failed implants. These surgical variables include varus placement of the component8,10-12, incomplete seating of the component1,7-9, and notching of the superior part of the femoral neck during implantation8,9,11,12.
Three-dimensional finite element analysis has suggested that resurfacing the femoral head can be expected to change trabecular bone strain and density in the region underlying and distal to the implant13, trabecular bone stress at the implant-bone interface14, and cortical stress and strain adjacent to and distal to the rim of the implant15. Although malpositioning of the femoral implant has been implicated clinically as a risk factor for implant failure, there are no data available from the direct measurement of the changes in femoral neck loading associated with component position after resurfacing. The mean time to fracture after hip resurfacing has been reported to be between fifteen and sixteen weeks after surgery8, implying that the stresses in the bone in the first weeks after implantation are important8,16. These stresses may also be critical factors in longer-term femoral remodeling after hip resurfacing.
The hypotheses of this study were that placement of a femoral resurfacing component alters femoral neck loading and that the cortical strain pattern reflecting this loading is directly related to the spatial orientation of the resurfacing component. An additional hypothesis was that notching of the superior aspect of the neck during implantation results in decreased resistance to neck fracture under axial loading. The specific aims of the study were to quantify changes in cortical strain in the femoral neck associated with femoral head resurfacing, with use of component position as a variable, and to determine the magnitude of the decrease in the ultimate load to failure of the femoral neck if a cortical notch in the neck is created during implantation.
Materials and Methods
Sixty-four fresh-frozen human femora stored at −18°C were used in this study. The donor ages ranged from twenty-two to eighty-two years, with a mean age of sixty years. Forty femora were from male donors, and twenty-four were from female donors. Anteroposterior radiographs were made of each femur, and the absence of degenerative changes or lytic lesions was verified. Strain analysis of the intact femora under the loading conditions described below was performed first, followed by implantation of a resurfacing component and repeat measurement of surface strain under the same loading conditions. Load-to-failure testing was then performed.
Strain Analysis of Intact Femora
Cortical shear strain in the femoral neck was quantified in forty-eight intact femora before femoral head resurfacing arthroplasty was performed. These femora comprised both single, unmatched specimens and members of matched pairs (Table I). A photoelastic coating was applied to the neck of each femur with use of a previously described technique17. Briefly, this entails bonding a strain-sensitive clear plastic coating to the bone with reflective cement. When the bone is subsequently elastically loaded, the resulting stresses in the bone are transferred to the plastic coating and the shear strain at any point may be quantified with use of a polariscope. These strains have been shown to be proportional to the tensile and compressive stresses within the bone17.
To facilitate strain documentation, the femoral neck surface was partitioned into four quadrants outlined by thin ink lines: anterior superior, anterior inferior, posterior superior, and posterior inferior. The dividing lines between the quadrants were located along the anterior, posterior, superior, and inferior midlines of the neck. Evenly distributed strain measurement points were marked randomly on each quadrant with use of ink dots, with the number of points dictated by the available surface area. On the average, eleven points (range, six to seventeen points in the anterior aspect and six to sixteen points in the posterior aspect) were located on each of the superior quadrants and seventeen points (range, twelve to twenty-two in the anterior aspect and eleven to twenty-five in the posterior aspect) were located on each of the inferior quadrants, for an average total of fifty-six measurement points (range, thirty-six to seventy-six points) around the entire femoral neck circumference.
Cortical shear strain in the femoral neck was measured at each of the described points on each of the forty-eight intact femora while the femur was subjected to loading designed to simulate single-leg stance. The axial load was applied with use of a previously described and validated17 hemipelvis loading jig that imparted both joint reaction forces and abductor musculature forces (Fig. 1). The femoral shaft was oriented in 12° of varus alignment and the abductor force angle was 15° from vertical, consistent with published parameters for single-limb stance18. The femoral head articulated with an acetabular component with a diameter that was 2 to 4 mm greater than both the native femoral head and the resurfacing components. The distal end of the femur was embedded at the level of the femoral condyles in a container attached to a universal joint to allow unrestricted bowing of the femur under load. Load was applied by a servohydraulic materials testing machine (model 1321; Instron, Canton, Massachusetts) such that the hip joint force was 2060 N, three times the body weight of a 70-kg individual (687-N), directly measured by a load cell that was integral with the loading jig. Cortical shear strain was quantified at each marked measurement location on the femoral neck under the described loading conditions with use of a reflection polariscope (model 040; Vishay Measurements Group, Raleigh, North Carolina).
Determination of the Neck Axis and the Center of the Neck
Before the resurfacing of each femoral head, the locations of the anatomic neck axis and the geographic center of the neck were established from the anteroposterior radiograph of the femur. The neck-shaft angle was determined by establishing the midline of the proximal femoral diaphysis and measuring the angle of its intersection with a line tangential to the inferior cortex of the neck at the apex of its curvature (Fig. 2). Because the curve of the inferior aspect of the femoral neck is not a consistent single radius, the orientation of the tangential line was a subjective determination, made by a visual assessment of the femoral neck anatomy as represented by the radiograph. The anatomic axis of the neck in the coronal plane was defined by a line parallel to the previously established tangential line and traversing the midpoint between the inferior and superior cortices of the neck at its narrowest point. The geographic center of the neck was located along this line, halfway between the intertrochanteric line and a line delineating the head-neck intersection.
Alignment of the Implant Stem
A goniometer set at the measured neck-shaft angle was used to position the guidewire for the preparation of each specimen. An orthogonal plane was determined with use of the midline drawn on the superior aspect of the neck, traversing the center of the head and the center of the neck at its narrowest point. Calipers were used to transfer the appropriate dimension from the radiograph to locate the midpoint of the neck on both neck midlines. A customized aiming jig aided direct guidewire insertion through the geographic center of the neck in the desired orientation, either aligned with the anatomic neck axis or at the desired angle of coronal or transverse plane deviation (Fig. 3). The aiming jig consisted of two arms joined by a lockable pivot. One arm was rigidly clamped parallel to the femoral shaft, and the other arm, carrying a series of locking swivel clamps and rods and the guidewire sleeve, was adjusted to achieve the desired guidewire alignment (anatomic, 10° of anteversion, 10° of retroversion, or 10° of varus) relative to the neck axis. Anatomic implant positioning was defined as occurring when the head of the implant covered the femoral head and the stem of the implant was oriented parallel to the anatomic axis of the neck. The entry point on the femoral head for the instrumentation guidewire was determined from extension of the anatomic neck axis to its intersection with the superior aspect of the femoral head.
Implantation of the Resurfacing Component
A cemented Articular Surface Replacement femoral component (ASR; DePuy Orthopaedics, Warsaw, Indiana) was implanted with use of instrumentation supplied by the manufacturer and following the manufacturer's recommended surgical technique. All implantations were performed by the same person (D.E.D.), who designed the jig and who was trained in the implantation process prior to performing the experimental implantations. Of the forty-eight femora, in which neck strain had been measured in the intact condition, twenty-four, all of which were from different donors and had been randomly selected with regard to left or right side, were implanted with the stem of the resurfacing component aligned with the anatomic axis of the neck as defined above. The remaining twenty-four intact femora, also all from different donors, were divided into three groups. These femora were implanted with the stem of the resurfacing component deviated into 10° of varus alignment relative to the anatomic neck axis (eight femora), 10° of anteversion (eight femora), or 10° of retroversion (eight femora) (Table I). In all cases, the size of the resurfacing component was selected on the basis of the diameter of the femoral neck, which minimized the risk of notching. The nominal implant sizes ranged from 39 to 53 mm, corresponding to implant inside diameters of 32 to 46 mm. No contact was allowed between the implant and the photoelastic coating. After implantation, radiographs of the specimens were made to confirm that the desired component alignment had been achieved.
Strain Analysis of Resurfaced Femora
Loading of each femur and strain measurements were performed again as described above after the femoral heads were resurfaced. Shear strain magnitudes measured under the intact and resurfaced conditions at each measurement point on the twenty-four femora were compared with use of generalized estimating equations19. This method accounts for the different variance across measurements made from different femoral specimens, which is in contrast to usual statistical tests that take each measurement as an independent observation. The generalized estimating equations method is also robust to measurements without a normal distribution. All p values were calculated, with a level of <0.05 considered significant. The mean strain magnitude in each femoral neck quadrant was expressed in absolute terms as microstrain (10,000 microstrain, or μ∈, is the equivalent of a 1% change in dimension). The change in strain magnitude attributable to resurfacing was expressed as the percent change from strain in intact femora with use of the means of the strain magnitudes in the respective quadrants.
Strength Testing of the Femoral Neck
After completion of the photoelastic strain analysis, three groups of matched pairs of femora were subjected to load-to-failure testing of the femoral neck to determine the influence of implant anteversion, implant retroversion, and notching of the superior aspect of the neck on the strength of the neck. The first two groups comprised the femora that had previously been used in the strain analysis and that were members of pairs (specifically, the bones that had received an anteverted or retroverted implant and the contralateral femur) (Table I). Prior to neck strength testing, the photoelastic coating was removed from these specimens by carefully prying the material from the bone with an osteotome. The third group consisted of eight additional matched pairs of femora that had not been used previously. Thus, in the first group (eight pairs), the resurfacing component was anteverted 10° in one femur from each pair and was implanted with the stem aligned with the anatomic axis in the contralateral femur. In the second group (eight pairs), the resurfacing component was retroverted 10° in one femur and was implanted in alignment with the anatomic axis in the contralateral femur. In the third group (eight pairs), the resurfacing component was implanted in alignment with the anatomic axis of the neck, but the cylindrical reamer was intentionally advanced until a 4-mm-deep notch was created in the superior cortex of the neck in one femur from each pair.
For the destructive testing of the femoral neck, the proximal one-half of each femur was embedded in a metal container attached to a base plate with a universal joint. The femur was then mounted in a servohydraulic testing machine (model 510.21C; MTS Systems, Eden Prairie, Minnesota) with the shaft angled 15° to 20° from vertical. The femoral head was mated with an acetabular cup mounted to a load cell, which was in turn attached to the actuator of the loading device (Fig. 4).
Displacement was applied at a rate of 10 cm/min until the femoral neck fractured. The ultimate load to failure was documented, and comparisons between femora with anatomically aligned implanted components and the contralateral femora that were notched or had an anteverted or retroverted component were performed with use of Wilcoxon matched-pairs signed-rank tests. An alpha level of 0.05 was considered significant for these comparisons.
Effects of Resurfacing on Neck Strain
Preimplantation and postimplantation shear strain data were obtained from a total of 273 measurement points on the anterior-superior neck quadrant, 271 points on the posterior-superior quadrant, 394 points on the anterior-inferior quadrant, and 397 points on the posterior-inferior quadrant. In all four quadrants, the mean shear strain decreased significantly (p < 0.001 in all instances) after resurfacing, indicating stress-shielding of the femoral neck when the implant was placed with the stem in alignment with the anatomic axis of the neck. The magnitude of change in the mean strain was −17% in the anterior-superior quadrant, −19% in the posterior-superior quadrant, −6% in the anterior-inferior quadrant, and −12% in the posterior-inferior quadrant (Table II).
10° of Varus Alignment
Strain data from the eight femora in which the resurfacing implant was placed in 10° of varus alignment provided 108 pairwise comparisons of strains, before resurfacing and after resurfacing, in the anterior-superior neck quadrant, 103 in the posterior-superior quadrant, 163 in the anterior-inferior quadrant, and 164 in the posterior-inferior quadrant. Shear strain increased significantly (p < 0.001) in both the anterior-superior (23%) and posterior-superior (19%) quadrants after resurfacing (Table II).
10° of Anteversion
Strain data from the eight femora in which the resurfacing implant was placed in 10° of anteversion provided ninety-two pairwise comparisons of strains before resurfacing and after resurfacing in the anterior-superior neck quadrant, eighty-eight in the posterior-superior quadrant, 123 in the anterior-inferior quadrant, and 133 in the posterior-inferior quadrant. Shear strain decreased significantly (p < 0.001) in the posterior-inferior quadrant (−22%) and increased (p = 0.033) in the anterior-inferior quadrant (13%) after resurfacing (Table II). There was no significant difference in shear strain in the two superior neck quadrants.
10° of Retroversion
Strain data from the eight femora in which the resurfacing implant was placed in 10° of retroversion provided seventy-six pairwise comparisons of strains before resurfacing and after resurfacing in the anterior-superior neck quadrant, seventy-seven in the posterior-superior quadrant, 124 in the anterior-inferior quadrant, and 125 in the posterior-inferior quadrant. Shear strain decreased significantly (p < 0.001) in the anterior-inferior quadrant (−14%). Shear strain increased significantly (p < 0.001) in the posterior-inferior quadrant (22%) after resurfacing (Table II).
Effects of Superior Notching, Anteversion, and Retroversion on Neck Strength
Failure occurred in the neck region of all twenty-four femora. In the group in which necks with and without notching were compared, there were fifteen transcervical fractures (Fig. 5) and one intertrochanteric fracture. The mean peak force before failure was 8754 N for femora without a notched neck and 6915 N for femora with a superior notch, a mean decrease of 21% attributable to the notch (p = 0.008, Table III). No difference in neck-breaking strength was demonstrable between femora in which the resurfacing component was implanted in approximate alignment with the anatomic femoral neck axis and those in which the implant was anteverted or retroverted (p = 0.742 and p = 0.195, respectively; Table III).
This study demonstrates in a cadaver model that implantation of a femoral head resurfacing device with a short femoral stem results in stress-shielding in the femoral neck when the device is aligned with the anatomic femoral neck axis. The data suggest that variations in the position of the implant of 10° of varus alignment, anteversion, or retroversion result in significant changes in the loading pattern of the femoral neck. Most notable from a clinical perspective is the observation that both the anterior-superior and posterior-superior quadrants of the femoral neck had increased surface strain when the component was placed into varus alignment relative to the axis of the neck. Exaggerated malpositioning of >10° from the ideal position was not studied, and the model did not allow an analysis of the effect of incremental changes in component position. Additionally, this study did not determine whether a valgus position may have the opposite effect of a varus position. The findings in this cadaver model reflect the strain pattern that might be present immediately after implantation, and they do not reflect in vivo stress relaxation changes or effects of bone remodeling that might occur later in vivo.
The effect of a 4-mm notch in the superior aspect of the neck resulted in a significant decrease of 21% in the load to failure of the neck under the described loading conditions. This observation suggests that notching of the superior aspect of the neck contributes to a substantial decrement in the femoral neck strength within the tested loading parameters, supporting the clinical observation that notching can be associated with femoral neck fracture after hip resurfacing. We acknowledge that the factors contributing to neck fracture after resurfacing are probably multiple, including bone quality, patient activity, disruption of the extraosseous blood supply, failure to adequately seat the femoral resurfacing implant, inadvertent neck notching, and implant malpositioning. Given that neck fracture has been clinically associated with varus positioning of the resurfacing implant8,10-12 and notching of the cortex in the superior aspect of the neck during reaming8,9,11,12, the data serve to provide some quantitative assessment of how a notch might by itself decrease the resistance to a femoral neck fracture.
The photoelastic strain measurement technique is a powerful tool that, unlike strain gauges, allows full-field assessment of bone surface strain. Nevertheless, there are some limitations to the technique when only absolute strain values are reported; namely, a small structural reinforcement effect and errors due to variations in the coating thickness17. Sequential measurements of strains at the same points on the same femora (preimplantation and postimplantation) and the reporting of strain changes as percentages avoided these limitations by controlling for all but the implantation condition. In this manuscript, we reported the mean strain magnitudes in each neck quadrant as a way to convey the magnitude of the strains involved, but all statistical comparisons were performed in a pairwise manner for each specific measurement point, consistent with our previously published reports with use of the photoelastic method17.
We acknowledge that the femoral loading protocols used in the present study represent simplified approximations of how the hip is loaded in vivo. In the more complex loading used in the strain analysis (compared with the neck fracture testing), the gluteal attachment to the trochanter was modeled, but the iliopsoas attachment to the lesser trochanter and other muscle inputs were not. This simplified model of femoral neck loading provided a means of consistent and repeatable loading that allowed natural bowing of the femoral shaft under load. Therefore, we believed that it was a reasonable approach for this type of study, but not entirely representative of the in vivo loading condition. The assumption that the body weight of each specimen donor was 70 kg (687 N) in applying a standardized hip-joint force of three times body weight (2060 N) for the strain analysis is another simplification that we acknowledge could have some effect on the results. The failure test loading was performed after removing the distal half of the femur to ensure that failure would occur at the neck. Again, we acknowledge that it did not closely simulate in vivo loading, but it did provide a repeatable means of achieving fracture for a comparative study of neck strength.
The amount of component malpositioning from the anatomic axis (10°) was chosen because it approximates the degree of alignment error that might commonly be seen clinically. There is no doubt that, clinically, larger deviations in alignment might occur, yielding different patterns of femoral neck strain. In clinical terms, a 10° change in implant alignment equates to a 5 to 8-mm variation in the insertion point of the central stem of the resurfacing component, and it is dependent on the length of the femoral neck and the point within the neck about which the angle is measured. We elected to pivot the axis of the resurfacing implant stem about the geographic midpoint of the femoral neck to achieve reproducibility in the experimental method. In the clinical setting, misalignment may occur by pivoting the implant axis about points other than the geographic center within the neck or translating the component, potentially resulting in different alterations in strain.
The stress-shielding and areas of increased strain noted in the femoral neck after implantation compared with the intact femur may have implications as to how the bone remodels over time. Possible explanations for the decreased strain include load transfer conferred by the central stem of the implant or by the cement mantle. Our goal was to achieve a uniform cement mantle, and the controlled conditions of the laboratory setting maximized the likelihood of this, but the implanted femora were not sectioned so we cannot comment on how the cement may have played a role in this analysis. Using finite element modeling, Huiskes et al.14 investigated shear strain at the cement-bone interface of the Wagner femoral resurfacing component and concluded that bone stresses at the interface were of sufficiently great magnitude that they might contribute to femoral component loosening. Watanabe et al.15, in another finite element analysis, found stress-shielding in the anterosuperior aspect of the femoral neck beneath the McMinn resurfacing component, as well as elevated stress at the posteroinferior rim of the implant, and suggested that this may lead to femoral neck fracture in patients with compromised bone quality. Gupta et al.13 performed a finite element study designed to evaluate the short-term risk of neck fracture after resurfacing. They reported stress-shielding of the bone beneath the implant, along with elevated strain in the superior aspect of the femoral neck adjacent to the implant.
Notching the superior cortex at the head-neck border significantly reduced the maximum axial load that a resurfaced femur could withstand before neck fracture occurred compared with femora resurfaced without damage to the superior cortex. In addition to this mechanical phenomenon, another potential negative effect of superior notching not assessed in this study is the compromise of extraosseous blood flow to the femoral head20, which might compound the mechanical effect. The superior aspect of the femoral neck is at risk of notching during resurfacing because of the eccentricity of the head relative to the neck, with a tendency for degenerative hips to flatten and remodel into varus alignment5. Thus, the superior aspect of the neck becomes the most vulnerable location for impingement by a bone-milling device during the procedure.
In conclusion, the present study provides data from direct measurement of cortical strain after resurfacing of the femoral head, modeling the situation that might occur immediately after implantation. The implantation-related changes observed in this analysis might be used to model the initial loading conditions in the femoral neck after resurfacing and may serve to validate finite element and clinical observations.
NOTE: The authors thank Ricardo Pietrobon, MD, PhD, for his assistance with the statistical analysis.
Disclosure: In support of their research for or preparation of this work, one or more of the authors received, in any one year, outside funding or grants in excess of $10,000 from DePuy, Inc. In addition, one or more of the authors or a member of his or her immediate family received, in any one year, payments or other benefits of less than $10,000 or a commitment or agreement to provide such benefits from a commercial entity (DePuy, Inc.). Also, a commercial entity (DePuy, Inc.) paid or directed in any one year, or agreed to pay or direct, benefits in excess of $10,000 to a research fund, foundation, division, center, clinical practice, or other charitable or nonprofit organization with which one or more of the authors, or a member of his or her immediate family, is affiliated or associated.
Investigation performed at the Department of Orthopaedic Surgery, University of California at San Francisco, San Francisco, California, and the Division of Orthopaedic Surgery, Duke University Medical Center, Durham, North Carolina
1. Amstutz HC, Beaulé PE, Dorey FJ, Le Duff MJ, Campbell PA, Gruen TA. Metal-on-metal hybrid surface arthroplasty: two to six-year follow-up study. J Bone Joint Surg Am. 2004;86:28-39.
2. Back DL, Dalziel R, Young D, Shimmin A. Early results of primary Birmingham hip resurfacings. An independent prospective study of the first 230 hips. J Bone Joint Surg Br. 2005;87:324-9.
3. Daniel J, Pynsent PB, McMinn DJ. Metal-on-metal resurfacing of the hip in patients under the age of 55 years with osteoarthritis. J Bone Joint Surg Br. 2004;86:177-84.
4. De Smet KA. Belgium experience with metal-on-metal surface arthroplasty. Orthop Clin North Am. 2005;36:203-13.
5. Schmalzried TP, Silva M, de la Rosa MA, Choi ES, Fowble VA. Optimizing patient selection and outcomes with total hip resurfacing. Clin Orthop Relat Res. 2005;441:200-4.
6. Treacy RB, McBryde CW, Pynsent PB. Birmingham hip resurfacing arthroplasty. A minimum follow-up of five years. J Bone Joint Surg Br. 2005;87:167-70.
7. Amstutz HC, Campbell PA, Le Duff MJ. Fracture of the neck of the femur after surface arthroplasty of the hip. J Bone Joint Surg Am. 2004;86:1874-7.
8. Shimmin AJ, Back D. Femoral neck fractures following Birmingham hip resurfacing: a national review of 50 cases. J Bone Joint Surg Br. 2005;87:463-4.
9. Little CP, Ruiz AL, Harding IJ, McLardy-Smith P, Gundle R, Murray DW, Athanasou NA. Osteonecrosis in retrieved femoral heads after failed resurfacing arthroplasty of the hip. J Bone Joint Surg Br. 2005;87:320-3.
10. Freeman MA. Some anatomical and mechanical considerations relevant to the surface replacement of the femoral head. Clin Orthop Relat Res. 1978;134:19-24.
11. Beaulé PE, Lee JL, Le Duff MJ, Amstutz HC, Ebramzadeh E. Orientation of the femoral component in surface arthroplasty of the hip. A biomechanical and clinical analysis. J Bone Joint Surg Am. 2004;86:2015-21.
12. Freeman MAR. The complication of double-cup replacement of the hip. In: Ling RSM, editor. Complications of total hip replacement. Edinburgh: Churchill Livingstone; 1984. p 172-200.
13. Gupta S, New AM, Taylor M. Bone remodelling inside a cemented resurfaced femoral head. Clin Biomech (Bristol, Avon). 2006;21:594-602.
14. Huiskes R, Strens PH, van Heck J, Slooff TJ. Interface stresses in the resurfaced hip. Finite element analysis of load transmission in the femoral head. Acta Orthop Scand. 1985;56:474-8.
15. Watanabe Y, Shiba N, Matsuo S, Higuchi F, Tagawa Y, Inoue A. Biomechanical study of the resurfacing hip arthroplasty: finite element analysis of the femoral component. J Arthroplasty. 2000;15:505-11.
16. Campbell P, Beaulé PE, Ebramzadeh E, LeDuff M, De Smet K, Lu Z, Amstutz HC. A study of implant failure in metal-on-metal surface arthroplasties. Clin Orthop Relat Res. 2006;453:35-46.
17. Glisson RR, Musgrave DS, Graham RD, Vail TP. Validity of photoelastic strain measurement on cadaveric proximal femora. J Biomech Eng. 2000;122:423-9.
18. McLeish RD, Charnley J. Abduction forces in the one-legged stance. J Biomech. 1970;3:191-209.
19. Hardin JW, Hilbe JM. Generalized estimating equations. Boca Raton, FL: Chapman and Hall/CRC; 2003.
Copyright 2008 by The Journal of Bone and Joint Surgery, Incorporated
20. Beaulé PE, Campbell PA, Hoke R, Dorey F. Notching of the femoral neck during resurfacing arthroplasty of the hip: a vascular study. J Bone Joint Surg Br. 2006;88:35-9.