Measurement of Periprosthetic Bone Density in Hip Arthroplasty Using Dual-energy X-ray Absorptiometry Reproducibility of Measurements

Transcription

1 The Journal of Arthroplasty Vol. 11 No Measurement of Periprosthetic Bone Density in Hip Arthroplasty Using Dual-energy X-ray Absorptiometry Reproducibility of Measurements R. C. Smart, BSc, MSc, PhD,* S. Barbagallo, ANMT,* G. L. Slater, MB, BS,t R. S. Kuo, MB, BS, t S. R Butler, MB, BS, MA, BSc, FRACP,* R. P. Drummond, MB, BS, FRCS (Edin), FRCS (Eng), FRACS (Ortho),t and R. Sekel, MB, BS, FRCSE, FRACS (Ortho)t Abstract: To define the precision (reproducibility) of measurement of periprosthetic bone mineral density and bone mineral content, dual-energy x-ray absorptiometry scans were obtained on 45 randomly selected patients who had had a unilateral total hip arthroplasty within the previous 3 years. The coefficients of variation of the bone mineral density in the proximal Gruen zones were 5.0 and 5.3%, corresponding to errors of 0.07 and 0.11 g/cm a. The coefficients of variation of the bone mineral density for the distal zones averaged 2.8%, with an error of 0.08 glcm 2. The coefficients of variation of the bone mineral content were 4.8 and 2.9% for the proximal and distal zones. The contralateral femur was also scanned in 32 of the patients. For the contralateral femur bone mineral density, the coefficients of variation were 5.0% for the proximal zones and 4.8% for the distal zones. The bone mineral content was 6.0% for the contralateral regions. These results imply that differences in bone mineral density greater than 0.16 g/cm 2 (2 standard errors) can be reliably measured. Dualenergy x-ray absorptiometry therefore provides a highly reproducible technique for quantitatively monitoring the changes in bone density that occur after total hip arthroplasty. Key words: bone density, hip arthroplasty, reproducibility. Total hip arthroplasty (THA) is a relatively common procedure for the treatment of degenerative joint disease of the hip; however, resorptive bone remodeling secondary to proximal femoral stress shielding may occur. The pattern of bone loss depends, among other things, on the stiffness and size ot the stem of the prosthesis [1] and the extent of the porous coating of the implant [2]. The pattern From the Departments of *Nuclear Medicine and ]-Orthopaedics, St. George Hospital, Sydney, New South Wales, Australia. Reprint requests: Dr. R. C. Smart, Department of Nuclear Medicine, St. George Hospital, Kogarah, NSW 2217, Australia. and extent of bone resorption cannot be predicted in the individual patient, and numerous techniques have been proposed to evaluate the degree of remodeling of the periprosthetic bone. Traditionally, bone resorption has been assessed by visual interpretation of radiographs; however, quantitative studies employing x-ray densitometry have demonstrated that changes in density as large as 20% can be introduced due to differences in film response, exposure variations, and positional inaccuracies [3]. More recently, other imaging modalities including computed tomography [4], ultrasound [51, and dual-energy x-ray absorptiometry (DEXA) [6] have been used to assess bone resorption. 445

2 446 The Journal of Arthroplasty Vol. 11 No. 4 June 1996 Dual-energy x-ray absorptiometry was developed to measure the bone mineral density (BMD) of the lumbar spine and femoral neck [7-9], with minimal radiation exposure to the patient [i0]. The technique has recently been extended to the measurement of periprosthetic bone density by the development of suitable software algorithms for the detection of the bone-prosthesis interface. Early experience with this technique was limited to either two or four regions of interest (ROIs) positioned over the proximal bone area [11,12]. These studies reported reproducibility measurements (coefficients of variation) in the range 1.8 to 7.5%. Modifications to the analysis software have recently become available enabling the BMD to be measured in the seven standard Gruen zones [13]. Trevisan et al. reported coefficients of variation of 1.8 to 6.8% for the seven Gruen zones in a limited study of 14 women [14]. In this study, we measured the reproducibility of this technique using the Gruen zones to define the ROIs in a larger group of 45 patients, to establish the minimum statistically detectable change in BMD between successive measurements. The minimum detectable change was defined as two standard errors from the regression analysis. The BMD was assessed in a 5- mm band, immediately adjacent to the prosthesis. It is probable that changes in BMD in this band will significantly precede radiographic changes associated with loosening or bone remodeling. Materials and Methods Forty-five patients were included in the study. They were randomly chosen from the patients of two surgeons (R.P.D. and R.S.) and were included in the study if they had had a unilateral THA within the previous 3-year period. The plain radiographs of the femur were reviewed on all the patients, and none of the patients showed visible radiolucent lines at the bone-implant interface at the time of the DEXA scan. The mean time in situ for the prosthesis was 13.3 months (range, months). There were 23 women and 22 men. The mean age was 65 years (range, years). In all cases, the contralateral femur was clinically normal, with no evidence of osteoarthritis (ie, no radiologic evidence of a decrease in joint space, subchondral sclerosis, cyst formation, or osteophyte formation). There were 10 cemented prostheses and 35 uncemented prostheses. The details of the uncemented group are given in Table 1. The cemented group included five Mueller (Zimmer, Warsaw, IN), three Centralign (Zimmer), one International (Smith & Nephew Surgical, Memphis, TN), and one Harris Pre-coat (Zimmer). The patients who had received the cemented prostheses were significantly older than those who had uncemented prostheses (mean ages, 73.4 and 60.2 years, P <.001 t-test) and had had the prostheses in situ for a shorter period (mean time of 6.4 months compared with a mean time of 15.1 months, P <.01 t-test). The patients were scanned using a Lunar DPX-L bone densitometer and analyzed using the Lunar Orthopaedic Software Package, Version 1.2 (Lunar Corporation, Madison, WI). The patients were positioned supine on the scan table with the knee of the leg to be scanned supported on a foam positioner. The patient's foot was strapped to a foot brace so that the foot was held vertical, with the patient's thigh parallel to the long axis of the scan table. This ensured that the position of the patient's leg could be precisely reproduced on successive measurements. A bag containing rice, which is approximately tissue equivalent, was placed on the outside of the patient's thigh to prevent the x-ray beam from scanning into air. To assess the reproducibility of the BMD measurements, each patient was scanned a second time, on the same occasion, after being repositioned on the scan table. The scan acquisition commenced approximately 15 mm below the distal end of the prosthesis and continued until approximately 15 mm above the greater trochanter. The scans were obtained using a current of 3 ma and a pixel size of 0.6 x 1.2 mm (transverse longitudinal). Scans typically took 4 to 5 minutes to complete. The contralateral leg was also scanned in 32 patients. In these patients, the Table 1. Details of the Prostheses Used in the Uncemented THA No. of Extent of Prosthesis Manufacturer Patients Porous Coating Alloy Anatomic Zimmer, Warsaw, IN 20 Proximal Titanium ABG Howmedica, Rutherford, NJ 5 Proximal Titanium and hydroxyapatite Autophore Osteo AG, Selzach, Switzerland 6 Full Cobalt-chrome Omnifit Osteonics, Allendale, NJ 3 Proximal Titanium Precision Osteolock Howmedica 1 Proximal Titanium and hydroxyapatite

3 Periprosthetic Bone Density Smart et al. 447 same length of normal femur was scanned as had been scanned for the leg containing the prosthesis. All patients were scanned and the scans analyzed by one of two operators. The Lunar software employed an edge-detection algorithm to define the edge of the prosthesis and the edge of the bone, and enabled the seven Gruen zones to be automatically defined from the length of the prosthesis (Fig. 1). With the exception of zone 4 (at the tip of the prosthesis), all zones were of equal length. The length of Gruen zone 7, the proximomedial zone, was manually adjusted so that the top of the zone coincided with the top of the bone in the scan. The length of zone 4 was standardized at 10 mm for all patients, except in studies where there was less than 10 mm of bone in the scan image below the tip of the prosthesis. The BMD and bone mineral content were calculated for a 5-mm-wide region of bone immediately adjacent to the surface of the prosthesis, for each of the seven Gruen zones (Fig. 1). The Lunar software allowed the ROIs from the prosthesis and bone edges to be superimposed on the image of the contralateral femur after they have been automatically mirrored about the axis of the leg. The operator manually aligned the bone ROI with the image of the intact femur, using the greater and lesser trochanter and the femoral shaft as anatomic landmarks. Bone mineral density was then calculated for the same 5-mm-wide Gruen zones (Fig. 2). The reproducibility was calculated as the coefficient of variation from the two scans on each patient. This was calculated using the formula [15] Ed )n (1) cv / = (:~,-+~x~--~, 1oo 2 where n is the number of paired observations and d is the ditference between two paired measurements, x1 and x2. In addition, the standard error of the estimate from linear regression was calculated for the paired patient results. The patient results were subdivided into several categories and the mean BMD and coefficient of variation calculated for each category: 1. For the uncemented prostheses, the 6 fully coated prostheses were compared with the 29 proximally coated prostheses. 2. For the uncemented prostheses, the 9 patients who had had their prosthesis in situ longer than 24 months were compared with the 26 patients who had had their prosthesis less than 24 months. 3. The 10 cemented prostheses were compared with the 35 uncemented prostheses. The significance of the differences in the results for the various categories was tested using the t-test. This project was approved by the Ethics Committee of the Southern Sydney Area Health Service. Fig. 1. Bone density scan of a femur containing a THA, together with the computer-generated Gruen zones. Fig. 2. Bone density scan of the contralateral femur, with the image of the prosthesis and regions of interest superimposed.

4 448 The Journal of Arthroplasty Vol. 11 No. 4 June 1996 Results Bone mineral density was measured in the seven Gruen zones around 45 prostheses, Mean BMD (in gtcm2), mean bone mineral content (in g), mean area of the ROIs (in cm2), coefficient of variation, and standard error from the linear regression analysis for each Gruen zone are given in Table 2. In five patients, the BMD could not be accurately measured in zone 4 as there was insufficient length of bone below the tip of the prosthesis on the scan image. The bone density of the proximal zones (1 and 7) was 48 and 72%, respectively, of that of the distal zones. The coefficients of variation of the BMD of the proximal zones were 5.0 and 5.3%, corresponding to standard errors of 0.07 and 0.11 g/cm 2. The coefficients of variation of the BMD for the distal zones (2, 3, 5, and 6) averaged 2.8%, with a standard error of 0,08 g/cm x. As an example, Figure 3 illustrates the correlation between the BMD measured from the first and repeat scans for Gruen zone 3. These results imply that a change in BMD of 0.16 g/cmx or greater (2 standard errors) between successive measurements is necessary for the observed change to be statistically significant. The coefficients of variation for bone mineral content were similar to those measured for BMD (4.9% for the proximal zones and 2.9% for the distal zones). The variation in the area of the ROIs was considerably less, 2.0% and 3.5% for proximal zones 1 and 7, and 1.3% for the distal zones. The reproducibility of the area of zone 4 was poor, with a coefficient of variation of 10.6%, The results for the subgroups of patients are presented in Table 3. The reproducibility of the measurements (coefficient of variation) was similar for all subgroups. The mean BMD was lower for all seven zones tor the cemented prostheses when compared with the uncemented prostheses (akhough this did not reach statistical significance); however, this was not unexpected due to the older age of the patients who had received the uncemented THA. It must, however, be noted that no difference in density between the cement and the cancellous bone could be distinguished on the images of the cemented prostheses. The magnitude of the error that this would introduce is unknown. There was no significant difference between the mean BMDs of those THAs involving fully coated prostheses and those with proximally coated prostheses, nor between those that had been in situ longer than 24 months and those that had been in place less than 24 months. Bone mineral density and bone mineral content were measured in the area corresponding to the periprosthetic zones in the normal contralateral femur and are tabulated in Table 4 together with the area of the ROIs. For the contralateral femur, the coefficients of variation of the BMD of the proximal zones were 4.3 and 5.7%, corresponding to standard errors of 0.07 and 0.14 g/cmx. The coefficients of variation for the distal zones (2, 3, 5, and 6) averaged 4,8%, with a standard error of 0.14 g/cm 2. A comparison of the differences in BMD between the operated femur and the contralateral femur (Table 5) indicated that for all seven zones, the mean BMD for the prosthetic femur was lower than that for the unoperated femur. The relative decrease in BMD ranged from 2 to 11%; however, only the difference in BMD in zone 1 reached statistical significance at the 5 % level. Discussion Bone remodeling around THAs plays a major role in long-term clinical outcome of the prosthesis. This bone loss, due to alterations in stress in the bone, Table 2. Reproducibility of Bone Mineral Density, Bone Mineral Content, and Region-of-interest Area in the Seven Periprosthetic Gruen Zones in the Operated Femur Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 No, of hips 45 Mean bone mineral density (g/cm a) 0.98 CV % 5.0 Standard error (g/cm ~) 0.07 Mean bone mineral content (g) 2.33 CV % 4.5 Standard error (g) 0.15 Mean area (cm 2) 2.40 CV % 2.0 Standard error (cm 2) 0, ,91 2,07 1, ,91 1, , ,10 0,08 0, ,38 4, ,4 2, ,8 3,0 5, ,18 0,49 0, , , ,63 1,2 1,3 10,6 1, ,04 0, ,05 0, CV, coefficient of variation.

5 Periprosthetic Bone Density Smart et al E3 [ SEE = 0,08 g/crn2 1 = ~ == 2... '-...!... 1, ' o l... i , First Scan - BMD g/cm2 Fig. 3. Plot of the bone mineral density (BMD) in g/cm 2 calculated from the first patient scan against the BMD from the repeat scan, for the 5-ram-wide region adjacent to the surface of the prosthesis for Gruen zone 3. The line of identity is shown. CV, coefficient of variation; SEE, standard error of the estimate. has traditionally been assessed by serial plain radiographs. West et al., however, have dearly demonstrated the considerable errors that can arise from the analysis of sequential radiographs [3]. Using quantitative densitometric measurements, significant errors could arise from variations between batches of film, nonuniformity of exposure, variations in intensity of the radiation field, target distance, and variations in femoral rotations. Individual errors were as large as 20%. They also reported that, in a series of five radiographs taken over 8 weeks, the density of the radiograph varied by more than 50% in the lateral aspect of the femoral cortex. Dual-energy x-ray absorptiometry, however, offers the potential to be able to measure the periprosthetic bone density accurately with a high degree of precision. X-ray film is not used to record the variations in attenuation through the bone, thus removing one of the major causes of variation in standard radiography. Furthermore, a constantpotential x-ray generator is used, together with a cerium k-edge filter, to produce two broad bands of x-ray energies, with effective energies of 40 and 70 kev, enabling the soft tissue and bone densities to be calculated. Daily calibration and quality control ensure the high precision found in longitudinal phantom studies [7-9]. The orthopaedic software for the Lunar DPX-L allows the periprosthetic bone density to be measured for the seven standard Gruen zones [13]. In this study, the bone density was measured in a 5-mm region immediately adjacent to the prosthesis, in each of the zones. Loosening of the prosthesis, if it occurs, will be a result of bone resorption in the region abutting the prosthesis. Therefore, measurements over a 5-mm region may be more sensitive to differences in BMD than measurements averaged over the entire width of the bone. The patients in this study had BMD measurements over a wide range of values: the lowest Table 3. Reproducibility of Bone Mineral Density and Mean Bone Mineral Density for Various Subgroups of Patients No. of Hips Zonel Zone2 Zone3 Zone4 Zone5 Zone6 Zone7 Fully Coated Prostheses vs Proximally Coated Prostheses for the 35 Uncemented THAs Coating Mean BMD (g/cm 2) Fully Proximal CV % Fully Proximal Time in situ < 24 Months vs > 24 Months for the 35 Uncemented THAs Time in situ Mean BMD (g/cm 2) < 24 months i > 24 months CV % < 24 months > 24 months Cemented Prostheses vs Uncemented Prostheses Fixation method Mean BMD {g/cm a) Cemented Uncemented CV % Cemented Uncemented BMD, bone mineral density; CV, coefficient of variation.

6 450 The Journal of Arthroplasty Vol. 11 No. 4 June 1996 Table 4. Reproducibility of Bone Mineral Density, Bone Mineral Content, and Region-of-interest Area in the Seven Gruen Zones in the Contralateral Femur Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 No. of hips Mean bone mineral density (g/cm 2) CV % Standard error (g/cm 2) Mean bone mineral content (g) CV % Standard error (g) Mean area (cm 2) CV % Standard error (cm 2) CV, coefficient of variation. value, in zone 1, was 0.5 g/cm 2, and the highest was 3.2 g/cm 2 in zone 3. Despite this wide range, the standard error from the regression analysis was essentially constant in the six zones along the stem of the prosthesis (zones 1-3 and 5-7), averaging 0.08 g/cm 2. This implies that a change in BMD of 0.16 g/cm 2 or greater (2 standard errors) between successive measurements is necessary for the observed change to be significant. This corresponds to changes of 16% for zone l, 8% for zones 2, 3, 5, and 6, and I 1% for zone 7. The precision was worst for zone 4, at the tip of the prosthesis. This was due to the large variation in the ROI size for this zone. It is often difficult for the operator to estimate the starting point for the scan, resulting in too few lines of data prior to the tip of the prostheses in the image; however, the bone density in this zone is unlikely to be of clinical interest as it will be determined primarily by the amount of bone reamed during the surgical implantation of the prosthesis. The reproducibility of the size of the other ROIs was excellent, with coefficients of variation between 1.1 and 2.0% for zones 1, 2, 3, 5, and 6. Zone 7 was smaller than the others because its length was manually adjusted so that the top of the ROI coincided with the top of the bone in the image. The manual manipulation of this ROI was necessary to accurately select the correct area of bone but resulted in a less repro- ducible definition of the ROI size and placement (coefficient of variation of 3.5 %). The reproducibility of the bone mineral content was very similar to that found for the BMD due to the excellent reproducibility of the ROIs. A number of other centers have measured the precision (coefficient ot variation) of periprosthetic BMD measurements; however, most reported studies have included fewer than l0 patients. Their findings are summarized in Table 6. The range of coefficients of variation in this study on 45 prostheses ( %) is in agreement with that observed by other investigators, even though a wide variety of different ROIs were used. The study of Kiratli et al. is the most extensive of the earlier studies, and included duplicate measurements on 30 patients [16]. Their measured precision, 2.7 to 4.5%, is in close agreement with our findings. Trevisan et al. are the only previous group to define the ROIs using the seven Gruen zones {14]. Their study of only 14 women, all of whom received a single type of hip prosthesis (a custom computed tomographycomputer aided design), demonstrated a precision similar to that found in this larger series in which a range of prostheses were used. When the ROIs were superimposed on the patient's contralateral femur the reproducibility was poorer than for the periprosthetic regions. This is not a surprising finding, as the alignment of the Table 5. Comparison of Bone Mineral Density Around the Prosthesis and the Contralateral Femur Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 Mean difference in BMD (g/cm 2) -0.12, P = , NS -0.12, NS -0.04, NS -0.05, NS -0.16, NS -0.13, NS Mean percentage difference in BMD BMD, bone mineral density; NS, not significant.

7 Periprosthetic Bone Density Smart et al. 451 Table 6. Precision (Coefficient of Variation) of Bone Mineral Density Measurements in Previous Studies of Periprosthetic Bone Density, Together With the Results of This Study No. of Regions of Method of Coefficient of Author Patients Interest Assessing Precision Variation (%) Richmond et al. [12] 7 4 proximal zones 1 and 7, 4 scans on each patient nm~ in height, on the same day full width Kilgus et al. [6] 10 6 medial and 6 lateral, 4 scans on each patient mm in height, on the same day full width McCarthy et al. [17] 8 2 medial, 21 mm in height, 4 scans on each patient mm wide over 1 week Kiratli et al. [16] 30 4 proximal zones 1 and 7, 2 scans on each patient mm in height, on the same day full width Trevisan et al. [14] 14 7 Gruen zones, full width 2 scans on each patient week apart This study 45 7 Gruen zones, 5 mm wide 2 scans on each patient on the same day ROIs with the contralateral femur is performed manually and relies on only two anatomic landmarks, the greater and lesser trochanters. As the BMD varies by more than a factor of 2 down the length of the prosthesis, small differences in ROI position or rotation adversely affect the precision. Although other authors have compared the BMD values from the periprosthetic regions with those of the patient's contralateral femur, only Trevisan et al. reported the reproducibility of the contralateral leg measurements [14]. They found a better precision for the contralateral femur than for the femur containing the prosthesis. This difference between the results of Trevisan et al. and our results requires further investigation. It may relate to differences in the commercial DEXA scanners and software used: Trevisan used the Hologic QDR 1000/W (Waltham, MA) and this study was performed using the Lunar DPX-L. The mean BMD for the femur containing the prosthesis was lower than that for the contralateral femur for all seven Gruen zones, in agreement with other investigators [17]. This study was designed to measure the reproducibility of periprosthetic BMD rather than to assess the magnitude of differences in BMD between various types of prostheses or changes with time. Patients with nine different types of prostheses were included in the study and the prostheses had been in situ from as short as 7 days to as long as 3 years. It is therefore not surprising that no significant differences in reproducibility measurements were observed between the results of the various subgroups as shown in Table 3; however, it is probable that the periprosthetic BMD values are in error in those patients with cemented prostheses, as no difference in density between the cement and the bone could be distinguished on the images. Therefore, it is likely that the BMD value included a contribution from the cement interface. This would not be a serious problem in longitudinal studies of cemented prostheses, as relative changes in BMD could still be assessed by these studies. Conclusion This study has demonstrated that the periprosthetic BMD can be measured by DEXA with a precision of 3 to 5%. This implies that changes in BMD greater than 0.16 g/cm 2 can be reliably measured. We have recently commenced a longitudinal study of periprosthetic bone density to quantify the patterns of bone loss in specific types of prostheses and, it is hoped, to influence the design of future prostheses to minimize bone loss. References 1. Bobyn JD, Mortimer ES, Glassman AH et al: Producing and avoiding stress shielding: laboratory and clinical observations of noncemented total hip arthroplasty. Clin Orthop 274:79, Engh CA, Bobyn JD: The influence of stem size and extent of porous coating on femoral bone resorption after primary cementless hip arthroplasty. Clin Orthop 231:7, West JD, Mayor MB, Collier JP: Potential errors inherent in quantitative densitometric analysis of orthopaedic radiographs. J Bone Joint Surg 69A:58, ] Berry E, Langkamer VG, Jackson PC et ah An image processing technique for the identification of contact

8 452 The Journal of Arthroplasty Vol. 11 No. 4 June 1996 pixels applied to x-ray CT images of implanted hip prostheses. Phys Med Biol 38:323, Zimmerman MC, Meunier A, Katz JL et al: The evaluation of bone remodelfing about orthopaedic implants with ultrasound. J Orthop Res 7:607, I Kilgus D J, Shimaoka EE, Tipton JS, Eberle RW: Dual-energy x-ray absorptiometry measurement of bone mineral density around porous-coated cementless femoral implants: methods and preliminary results. J Bone Joint Surg 75B:279, Larnach TA, Boyd S J, Smart RC et al: Reproducibility of lateral spine scans using dual energy x-ray absorptiometry. Calcif Tissue Int 51:255, Mazess R, Collick B, Trempe Jet al: Performance evaluation of a dual-energy x-ray bone densitometer. Calcif Tissue Int 44:228, Orwoll ES, Oviatt SK, Nafarelin/Bone Study Group: Longitudinal precision of dual-energy x-ray absorptiometry in a multicenter study. J Bone Miner Res 6:191, Kalender WA: Effective dose values in bone mineral measurements by photon absorptiometry and computed tomography. Osteoporosis Int 2:82, Kiratli B J, Heiner JP, McKinley Net al: Bone mineral density of the proximal femur after uncemented total hip arthroplasty. Presented at the 37th Annual Meeting of the Orthopaedic Research Society, Anaheim, CA, March Richmond BJ, Eberle RW, Stulberg BN, Deal CL: DEXA measurement of periprosthetic bone mineral density in total hip arthroplasty. Presented at the 13th Annual Meeting of the American Society for Bone and Mineral Research, San Diego, CA, August Gruen TA, McNeice GM, Amstutz HC: "Modes of failure" of cemented stem-type femoral components: a radiographic analysis of loosening. Clin Orthop 141:17, Trevisan C, Bigoni M, Cherubini R et al: Dual x-ray absorptiometry for the evaluation of bone density from the proximal femur after total hip arthroplasty: analysis protocols and reproducibility. Catcif Tissue Int 53:158, Nilas L, Hassager C, Christiansen C: Long-term precision of dual-photon absorptiometry in the lumbar spine in clinical settings. J Bone Miner Res 3:305, Kiratli B J, Heiner JP, McBeath AA, Wilson MA: Deterruination of bone mineral density by dual energy x- ray absorptiometry in patients with uncemented total hip arthroplasty. J Orthop Res 10:836, McCarthy CK, Steinberg GG, Agren Met al: Quantifying bone loss from the proximal femur after total hip arthroplasty. J Bone Joint Surg 73B:774, 1991