Jortrnul of Orthopnedic Research 1083M44 Raven Press, Ltd., N e w York 0 1992 Orthopaedic Research Society

Determination of Bone Mineral Density by Dual X-Ray Absorptiometry in Patients with Uncemented Total Hip Arthroplasty B. Jenny Kiratli, John P. Heiner, Andrew A. McBeath, and "Michael A. Wilson Division of Orthopedic Surge17 and *Department of Nitclear Medicine, University qf Wisconsin-Madison, Madison, Wiscon3 in, U .S .A.

Summary: Bone remodeling is an expected sequela with total hip arthroplasty (THA). Although there are several methods of estimating bone response in THA patients from radiographs, there are no accurate and generally accepted methods for quantitative determinations in vivo. In this study, we describe an application of dual x-ray absorptiometry (DXA) for measuring bone mineral content and bone mineral density in the proximal femur following THA. DXA is a noninvasive technique with minimal radiation exposure (30% to be observed with certainty (2,28,34). West et al. (35), using quantitative radiographic densitometry, examined the magnitudes of error for variables such as film lot, exposure setting, target distance, field variability, and femoral rotation. Small changes in each of these variables resulted in densitometric differences of 9-2196, and these erReceived May 28, 1991; accepted June 5 . 1992.

Address correspondence and reprint requests to Dr. B. J. Kiratli at VA Medical Center (128), 3801 Miranda Avenue, Palo Alto CA 94304, U.S.A.

836

837

BONE MINERAL DENSIII’Y A N D T H A in visible bone density (6). Various scoring scales have been dcveloped to improve the precision of and discrimination between observations. Theqe variables indicate gross bone response (ucually scored as prcsentiabsent), which is adequate only to give percent occurrence in grouped data. However, they do not provide an accurate quantification of bonc density. and hence do not allow determination of severity of response or individual responsc. Furthermore, the clinical Fignificance of these variables is uncertain. In comparisons of postoperative symptomatology with radiographic evidence (13,16,17), the radiographic features did not correlate well with clinical outcome. This may reflect a true lack of correlation or might result from poor precision and accuracy of radiographic assessment. In another study, radiographic data, determined from Singh and Engh indices, were not correlated with histologic data in THA patients (33). A method has been derived by Engh and coworkers (6) that uses the number (frequency), in grouped patients, of recorption sites detectable on two-year radiographs compared with those on immediate postoperative radiographs as an estimation of severity of bone resorption. However, thc authors notc that their method is an approximation. and comment that resorption below the threshold of detection (30% change) certainly occurs. Improvemcnt in techniques for monitoring bone remodeling responses in patients after THA will improve diagnostic and, perhaps, prognostic decision making. In this study, we describe an application of the technique, dual x-ray absorptiomctry (DXA), for determination of regional bone mineral density in patients undergoing THA. The conventional application of DXA is for diagnosis and monitoring of metabolic bone disease, and the primary regions of interest are the femoral neck and the lumbar spine. We have used special orthopedic software supplied by the manufacturer which identifies sites (in the proximal femur) with clinical importance for patients with hip arthroplasty, and we have established a protocol to ensure reliable measurements at these sites. This orthopedic application of DXA will allow investigation of the magnitude and rate of the bone response in follow-up of THA patients, METHODS Description of Dual X-Ray Absorptiometry Dual x-ray absorptiometry (DXA) is the most recent development in bone densitometry (3 1). This

technique uses x-ray absorption (i.e., no administered radionuclide or contrast) to determine the amount of bone in specified skeletal regions. Bone densitometry is commonly used for monitoring bone changes related to aging, drug therapies, exercise intervention, metabolic disorders, etc. We used a DPX densitometer (LUNAK Corp., Madison, WI, U . S . A . ) in this study. This instrument uses a narrowly collimated x-ray beam which is filtered to allow two energies to be emitted (38 and 70 keV). These two energies attenuate differentially in soft tissue and bone, and the transmitted radiation is measured by a scintillation detector. The beam is passed through the proximal femur region in a rectilinear scan pattern caudad to cephalad. The amount of bone in the beam path is calculated as bone mineral content (BMC) in grams. BMC is then divided by the projected area of the region scanned, and this is reported as the bone mineral density (BMD) in gicm’. BMD thus provides an “area density” representing bone concentration in a given region, corrected for size of that region. The DXA output is therefore similar to the AP projection in conventional radiography where a three-dimensional structure is imaged in two dimensions. The radiation dose to the patient is small, less than 5 mrem per scan, allowing repeat measurements with little hazard. A more detailed description of this equipment has bccn given elsewhere (20,21,32). Scan Acquisition: THA Protocol The THA protocol involved the orthopedic software package provided by the manufacturer for both acquisition and analysis of data. This software allowed determination of BMC in small regions adjacent to the implant and reduced, as compared with the standard software, effects of the metal prosthesis on bone results. Acquisition settings were standard for the DPX scanner (750 PA, 1.68 mm collimation). However, the transverse speed was two times slower than standard (1/32 s sample interval) and the longitudinal step half as great, giving a pixel size of 0.6 x 0.6 mm and a fourfold increase in resolution over the standard application. The smaller pixel siLe improved precision for the relatively small regions used in this THA protocol. The subject is positioned supine on the padded scan table with the leg in a neutral rotation. The knee is placed in a foam positioning device in slight flexion to further stabilize the leg in the neutral position and to alleviate knee and back discomfort.

J Orthop Rrs, Vol. 10,N u . 6 , I992

838

B. J . KIRATLI ET AL.

The foot is anchored to a rigid positioning device with toes pointed up. The leg is positioned parallel to the long axis of the table so that the transverse beam path will be perpendicular to the femoral shaft. The orientation of the patella and femoral condyles is used to verify this position. The neutral leg position was chosen for two reasons. First, the lesser trochanter is clearly visible in most scans. Second, this position was comfortable during the preoperative and immediately postoperative scans, ensuring minimal patient movement. Regions of Interest

Four regions of interest (ROIs) are delineated with the midpoint of the lesser trochanter as the reference point (Fig. 1). This anatomic landmark was chosen as the reference point, rather than an implant-based landmark, in order to eliminate measurement error which might arise in subsequent scans with changes in implant position (due to rotation, displacement, subsidence). The lesser trochanter is the most invariate feature of the image, and all regions are therefore located relative to this anatomic reference. Each ROI is 2 cm in height and includes all the bone points between the implant edge and the periosteal bone edge. Preliminary studies showed that smaller regions (1 cm height) were difficult to relocate with precision. BMD and BMC are determined for all points defined by the analysis algorithm as “bone” (based on known absorption characteristics) within the designated ROIs. The implant is automatically isolated

FIG. 1. Regions of interest. A-D: The midpoint of the lesser trochanter is the reference point; all regions are 2 cm in height.

J Ortitop Hes, Vol. 10, hro. 6, 1992

by the software, using proprietary algorithms, through recognition of extreme density outside the normal range for bone. The software provides isolation for a range of prosthesis types, differing in both material and design. Additional processing is done to define the edge of the implant and to minimize the effect of extreme density change (“partial volume effect”) at the bone-implant interface. The bone edges and reference point are used to establish a region for determination of soft tissue baseline. An iterative process is used to converge on a soft tissue value within a region lateral to the bone edge, with a total area similar to that included in the bone ROTS. The selection of soft tissue baseline is crucial to calculation of bone values (22). Measurement Error Assessments were made of both accuracy (similarity between the measured value and actual value) and precision (similarity among repeated measurements on the same subject). Potential sources of error, including both data acquisition and analysis, may be machine based (hardware and/or software) or operator based (i.e., subject placement and determination of reference point). Accuracy was assessed by scanning bone phantoms of known densities. The phantoms were constructed of simulated bone mineral of constant material density (RMI Corp., Middleton WI, U.S.A.) cut into four blocks of different thickness. Calibration of the DXA unit has previously been done against sample5 of hydroxyapatite submerged in water and also against the European COMAC phantom (23). The mineral content of these blocks was determined to be 1.520, 1.205, 0.834, and 0.471 gicm’, respectively. All measurementc were made with a prosthesis overlying the bone phantom and both submerged 16 cm in water (which served as soft tissue equivalent). ROls for each scan were chosen to include (a) bone phantom only and (b) phantom-plus-implant for each scan (Fig. 2). We made measurements of two femoral prostheses of different shapes and material: chrome-cobalt dual lock (DePuy Corp., Warsaw IN, U.S.A.) and titanium alloy (TiAl,V,) Wisconsin Hip (Biomet Inc., Warsaw IN, U.S.A.). Initial statistical analysis involved comparison of sets of analyses (with and without implant) by paired t test. Then, two-way analysis of variance (ANOVA) was used to determine the amount of variance attributable to implant type and/or mate-

BONE MINERAL DENSITY AND THA

FIG. 2. Determination of mineral content on a bone phantom.

rial density of the phantom. Finally, comparisons were made between the measured and actual values. Thus, we were able to assess (a) the effect of a prosthetic component in the scan, (b) the effects of two types of implant design, and (c) the accuracy error at different bone densities. Precision was assessed under three conditions: (a) bane phantoms (described above) constructed of a relatively homogeneous material arranged in a constant thickness and density distribution, (b) cadaver femora, implanted ex vivo, remeasured with no repositioning between scans, and (c) patients who had undergone THA, remeasured with repositioning between scans. In this way, successively more variables were added to the testing condition, thereby separating “machine-based” error from “operator-based” error. Phantom

Five replicate measurements were made of the bone phantoms as described above. For each density, coefficient of variation (CV) was determined as the standard deviation divided by the mean, expressed as a percent (%CV). Caduvers Five sets of measurements were made on six femora, righuleft pairs from three cadavers (1 female, 2 males, aged 79-83 years at death). Each femur was excised from the body, cleaned of surrounding soft tissue, and implanted with a titanium femoral component (Biomet Tnc.) using standard surgical tech-

839

nique. Replicate measurements were made with each femur surrounded by soft tissue equivalent material ( 18 cm rice and 3 cm Plexiglas). The femur was held in a neutral position with the distal condyles secured in a positioning device. There was no repositioning of the bone between scans, and all scans of a given femur were performed on the same day. The influence of rotation on precision was investigated by scanning each implanted femur rotated 5” medially and 5” laterally. These scans were performed on the samc day and under the same testing situation a5 those described above. Precision error was calculated as %CV of the five replicates of the six femurs for each of the four selected ROls (Fig. 1).

Patients Determination of precision in vivo was accomplished both by five measurements of five subjects and duplicate measurements of 30 subjects. Each patient was repositioned between scans. The set of five replicates on five patients were performed over several days; the duplicate sets of scans on 30 patients were obtaincd on the same day for each patient. Precision error was detcrmined for the five replicates of five patients as the %CV (described previously as SDimean) for each of the four ROls. Error in the duplicate scans of 30 patients was calculated by the formula (25, appendix): CV% = 100 x [(xd”/2nl’”/[(x, + xJ21 where d is the diffcrence between duplicate scans (xl and xJ for each patient and n is the total number of patients. A final component of precision error may arise during data processing. Although many efforts have been made to automate the analysis routines, certain decisions remain under the control of the operator, particularly, the choice of reference point. In order to assess this, four experienced operators independently processed the same set of 25 patient files (five replicates on five patients). The results of these sets of analyses were compared using repeated measures ANOVA, and between-operator and within-patient errors were examined. RESULTS Accuracy No differences were found between pairs of scans with the implant present versus implant absent from

f Orthop Res, Vol. 10, No. 6, 1992

840

B . J . KIRATLI ET AL.

the region of intcrest (Fig. 3 ) . Furthermore, there was no effect of implant type on measured BMD values at the four densities of material. The measured values were not different from expected values, with errors well below 1% for all densities, and there were no differences in accuracy among the different densities (Table 1).

TABLE 1. Arcuracy und precision error in BMD" determined in bone phantoms of four densitirs

1.520 1.205 0.834 0.471

'n Precision error from ten replicate measurements of the bone phantom at each of four densities was -0.5% (Table 1) and quite consistent across the four bone density levels (0.3 to 0.6%). These results demonstrate excellent precision in a controlled circumstance [homogeneous "bone" material in a homogeneous soft tissue equivalent (water) with no repositioning]. In a more realistic condition (cadaver bone, in a relatively homogeneous medium of rice, with no repositioning), the precision error was only slightly higher than that observed with bone phantoms. Mean coefficients of variation for BMD of ROIs A-D were 0.90%, 0.8996, 1.06%, and 1.4776, respectively. The precision was consistent across the four ROIs (-l%), with a slight increase in error in region D. The effect of rotation of 5' in either direction varied among the ROIs (Table 2). Data are presented here for BMC, BMD, and area in order to show how each component was influenced by the position

1.511 1.206 0.836 0.469 =

0.005 0.004 0.004 0.003

0.3 0.3 0.5 0.6

- 0.6

0.1 0.2 -0.4

10.

' Coefficient of variation (SDimean), expressed as a percentage. Determined as (measured - actua1)iactual expressed as a percentage.

change. Rotation had little effect on any of the variables for region A, and the error in BMD was equivalent to that of repeated scans in the same position. There was a slight decrease in precision for all variables in region B, but the error in BMD was only barely greater than 2%. However, precision decreased substantially in the medial regions C and D, with rotation, most obviously due to changes in the area (-9% error), which resulted in increased error in the BMD. In all regions, BMC was affected least (error always

Determination of bone mineral density by dual x-ray absorptiometry in patients with uncemented total hip arthroplasty.

Bone remodeling is an expected sequela with total hip arthroplasty (THA). Although there are several methods of estimating bone response in THA patien...
826KB Sizes 0 Downloads 0 Views