Comparison of Dual-Energy X-Ray Absorptiometry and Dual Photon Absorptiometry for Bone Mineral Measurements of the Lumbar Spine

Comparison of Dual-Energy X-Ray Absorptiometry and Dual Photon Absorptiometry for Bone Mineral Measurements of the Lumbar Spine HENZ W. WAHNER, M.D., WLLAM L. DUNN, M.Sc, MANUEL L. BROWN, M.D., Section of Diagnostic Nuclear Medicine; RCHARD L. MORN, Ph.D., Department of Diagnostic Radiology; B. LAWRENCE RGGS, M.D., Division of Endocrinology/Metabolism and nternal Medicine A new x-ray-based (dual-energy x-ray absorptiometry [DEXA]) instrument for measurement of bone mineral in the spine and hips (QDR-1000, Hologic, nc.) was compared with a commercial dual photon absorptiometry (DPA) instrument that uses a 153 Gd source (DP3, Lunar Radiation Corporation). Measurements were made on phantoms and lumbar spines of patients to study accuracy, precision, limitations, and compatibility of results between instruments. Both instruments measure bone mineral of integral bone in terms of area bone density with an entrance exposure of less than 5 mr. For spinal bone mineral measurements, the DEXA instrument had a shorter scanning time and higher resolution images than the DPA system. The DEXA instrument also showed better precision in a spine phantom and reduced influence of thickness for patient measurement. For bone mineral content, accuracy was about equal for both instruments; for measurements of the area of the region of interest, accuracy was better with the DEXA instrument. With both instruments, fat had little effect on bone mineral density in bone phantom studies. Measurements on both instruments were influenced by the location of a bone phantom within the photon beam. Results in patients showed good correlation (r = 0.988) for bone mineral density. Measurements of bone mineral density in patients were consistently lower with the DEXA instrument because of better accuracy in area measurements. The new x-ray-based instrument is a major advance in bone mineral absorptiometry and provides improved, yet less expensive, measurements in research and clinical applications. During the past 2 decades, numerous procedures has been reviewed previously. 3 Even though physihave been proposed for the nontraumatic mea- cians have only recently become familiar with surement of bone mineral in different portions of results from these instruments, another modifithe skeleton. 1 Of these procedures, dual photon cation for bone mineral measurements in vivo is absorptiometry (DPA) for bone mineral measure- now available a new densitometer based on dualments of the axial skeleton has become a rela- energy x-ray absorptiometry (DEXA). This new tively common tool in the diagnosis and manage- technology necessitates reexamination of normal ment of patients with osteoporosis, 2 and several ranges and may create problems in comparing different commercial instruments are available, previously acquired data with results from the The principle of operation of DPA instruments new measurements. DEXA, however, introduces several improvements over the conventional DPA procedure based on isotope sources that make it particularly attractive for clinical applications Address reprint requests to Dr. H. W. Wahner, Section of Diagand Justify the inconveniences that accompany nostic Nuclear Medicine, Mayo Clinic, Rochester, MN 55905. a change of methods. We foresee a gradual re- Mayo Clin Proc 63:1075-1084,1988 1075

1076 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE Mayo Clin Proc, November 1988, Vol 63 placement of the DPA procedure with DEXA in the future. n this study, we compared a commercial instrument based on DEXA (QDR-1000, Hologic, nc., Waltham, Massachusetts) dedicated to measurements of bone mineral of the lumbar spine and hip with a commercial isotope DPA instrument (DP3, Lunar Radiation Corporation, Madison, Wisconsin), also used for spine and hip measurements. The variables we studied were accuracy and linearity, reproducibility in phantoms and patients, dosimetry, and effects of patient thickness and body composition on results of measurements. Results of measurements made on both instruments in the same patients were compared, and the limits within which data on patients can be made comparable when they are obtained with either instrument were determined. MATERAL AND METHODS nstruments. The general features of the DEXA and DPA instruments are compared in Table 1. The new bone densitometer based on DEXA uses an x-ray tube as the radiation source. This source provides a manyfold increase in photon flux in comparison with the 1-Ci gadolinium source typically used in DPA. As with the DPA method, the DEXA system scans the lumbar spine or hips in a rectilinear fashion and records separate lowand high-energy transmitted photon intensity values on a pixel-by-pixel basis. Both methods, DPA and DEXA, measure integral (trabecular and Table 1. Comparison of Features of Dual Photon Absorptiometry (DP3) and Dual-Energy X-Ray Absorptiometry (QDR-1000) nstruments Feature DP3 QDR-1000 Method Source Energy Source collimator (mm) Detector collimator (mm) Scan speed (mm/s) Line increment (mm) Scanning time (min) Processing time (min) Table detector distance (cm) Daily calibration Transmission scan 153 Gd 44,100 kev 3 8,13 2.5 4.5 35 5 30 3-block phantom Transmission scan X-ray 70,140 kvp 2 None 60 1 8 5 40 Spine phantom compact) bone mineral, and the data are expressed as area density (in grams per square centimeter) or bone mineral density (BMD). Separate information on bone mass (in grams) within the designated region of interest, referred to as bone mineral content, and area (in square centimeters) of bone that constitutes the region of interest is available. The ratio of soft tissue absorption coefficients at the two photon energies, a number that provides information about the composition of body tissue surrounding the bone, is available only on the DPA instruments. The data printout is similar for both instruments (Fig. 1). The high photon flux for DEXA allows the use of a smaller radiation source collimator and a faster scanning speed with more scan passes per unit area of bone. The principles of operation of the instrument have been described in detail elsewhere. 4,5 The x-ray bone densitometer developed by Hologic and tested in this study uses a tungsten stationary anode x-ray tube, pulsed alternately at 70 kvp and 140 kvp and operated at a tube current of about 2-mA peak. Hologic measured R ; :\ ;Hf '. ^ ; * : ί - 4fc, <- ί»,." *-. i, ' *'»" * - - *έ, L1 L2 L3 L4 LS BMD (g/cm 2 ) L2 0.962 0.854 L3 0.991 0.866 L4 0.987 0.905 L2-4 0.981 0.878 Fig. 1. ntensity-modulated bone mineral images from two instruments the isotope-based dual photon absorptiometry (Lunar DP3) instrument (left) and the dual-energy x-ray absorptiometry (Hologic QDR-1000) instrument (right). Bone mineral density (BMD) is shown for individual vertebrae and the clinically useful region of L2-4. The algorithm determines the bone edges, as outlined on the images (stippled lines). L

Mayo Clin Proc, November 1988, Vol 63 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE 1077 effective beam energies at 43 and 110 kev. The resulting dual-energy x-rays are collimated to yield a 2-mm beam. X-rays are detected by a cadmiumtungstate scintillator, coupled to a photomultiplier tube. The latter is operated in the current mode, and photon counting techniques are not used. This approach avoids problems related to dead time. The instrument has a patented, novel internal reference system that compensates for drifts in measurements due to fluctuations in x-ray beam characteristics and detector response. The beam passes through a rotating wheel that contains segments representing air, tissue, and bone. The wheel rotates synchronously with the x-ray pulse. 4 Experiments. Several experiments were performed to analyze and compare the accuracy, precision, and bone mineral results obtained with the DEXA and DPA instruments. The experiments are summarized in Table 2. The 8-mm detector collimator was used in all studies with the DP3 instrument. Accuracy was measured in two phantoms of different complexity. A threestep block phantom (Fig. 2) with well-defined edges, a constant bone area, and three different thicknesses of triphosphate hydroxyapatite was used for evaluation of the performance of the bone edge-detection algorithm at different levels of bone mineral, ranging from values in patients with severe osteoporosis (0.6 g/cm 2 ) to those in young male subjects (1.6 g/cm 2 ). A spine phan- Table 2. Summary of Experimental Analysis of Dual-Energy X-ray Absorptiometry (DEXA) and Dual Photon Absorptiometry nstruments Variable Experimental method Accuracy Ashed bone in 20 cm of water Spine phantom (DEXA; QDR-1000, Hologic, nc.) Precision Short-term Spine phantom, ashed bones, patients Long-term Spine phantom, ashed bones Effect of absorber Ashed bone in different water levels (patient) thickness (10-30 cm) Effect of bone Ashed bones in 20 cm of water with distance from varied distance of specimen from imaging table table (radiation source) Effect of fat content Ashed bone in 20 cm of waterin bone surrounding vegetable oil mixture (0-60% oil) media Effect of absorber Ashed bone in different levels of (patient) thickness various water-vegetable oil mixtures and fat content (obesity) Fig. 2. Three-step block phantom (left) and spine phantom (Hologic, nc.) (right) used to compare accuracy and precision of dual photon absorptiometry and dual-energy x-ray absorptiometry instruments. Phantoms are made of calcium hydroxyapatite mixed with an epoxy resin. The models are then encased in a block of water-simulant epoxy, 17.8 by 15.2 by 14.7 cm in dimension. torn (Fig. 2) made of the same material with known mineral content and dimensions was used to test the performance of the instruments. This empirical setting duplicates the clinical situation but omits the tissue that surrounds the spine within patients. Linearity of results at various BMD values was evaluated with four bones of different mineral content. Phantoms and ashed bone samples were measured in 20 cm of water (tissue equivalent). For assessment of long-term (4- to 6-month) precision, measurements of the spine phantom were performed once or twice daily with the DEXA instrument and once weekly with the DPA instrument. n addition, weekly measurements were made of a bone specimen in 20 cm of water with the DPA instrument. For estimation of short-term precision in patients, duplicate measurements on the lumbar spine were made in 15 volunteers with the QDR- 1000 (DEXA) instrument and in 5 volunteers with the DP3 (DPA) instrument. For correlation of measurements, the lumbar spine of 15 subjects was scanned on both instruments. All subjects signed informed-consent forms. The protocols were approved by our nstitutional Review Board. The effect of various absorber (patient) thicknesses on BMD measurements as encountered when patients of various abdominal diameters are scanned was evaluated by studying ashed bone in water with levels that varied from 10 to

1078 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE Mayo Clin Proc, November 1988, Vol 63 30 cm. The effect on BMD of the position of a bone in the radiation beam was evaluated by scanning an ashed bone in 20 cm of water but at different distances from the bottom of the tank (distance between radiation source and object scanned). The normal curvature of the spine or spinal deformities can result in a measurable deviation from a constant radiation source-bone distance when the lumbar spine is scanned. Ashed bones were scanned in 20 cm of a mixture of water and vegetable oil (0 to 60% oil) to study the effect of different fat content in the medium surrounding the bone. This technique simulates changing fat content of tissue surrounding the spine. Lastly, an ashed bone was scanned in different levels of a varied oil and water mixture to study changes associated with obesity, characterized by a concomitant increase in absorber thickness and fat content. Dosimetry determinations were made on cadavers and patients with calcium fluoride thermoluminescent detector chips, and radiation measurements were performed with an MDH (Monrovia, California) system using 6- and 60-cc ionization chambers. RESULTS Accuracy. Table 3 summarizes the results from measurements of the three-block phantoms that were designed to test accuracy of bone mineral and area determinations. The bone mass (bone mineral content) differences between the two instruments were small for all three steps of the block. The difference between the mean of three measurements and the actual mineral mass in the blocks ranged from -2.4 to +3.2% for the QDR- 1000 (DEXA) instrument and from -1.4 to +1.2% for the DP3 (DPA) instrument. For bone area, the difference was more substantial for the DP3 instrument, ranging from -11.6 to +0.7%, whereas it was only between -3.3 and +3.4% for the QDR- 1000 instrument. The difference between actual and measured BMD was greater for the DP3 than for the QDR-1000 instrument (-2.0 to +14.9% in comparison with -0.9 to +0.3%, respectively). The greatest error was found in the small block when measured with the DP3 instrument. Results from measurements on the more complex spine phantom designed to test accuracy and precision in a simulated real object are summarized in Table 4. Again, the accuracy error was Table 3. Accuracy of Bone Mineral and Area Measurements by Dual Photon Absorptiometry (DPA) and Dual-Energy X-Ray Absorptiometry (DEXA) in Three-Block Phantom* Region of phantom Small Medium Large Measurement Bone mineral content Mass (g) CV(%) Area RO (cm 2 ) CV (%) Bone mineral density Density (g/cm 2 ) CV(%) Block phantom (Hologic) Mass (g) Area (cm 2 ) Bone mineral density (g/cm 2 ) DPA 10.94 3.3 +1.2 15.47 4.6-11.6 0.71 6.2 +14.9 10.81 17.5 0.618 DEXA 10.55 0.9-2.4 16.93 0.3-3.3 0.62 1.1 +0.3 DPA 18.09 2.6 +0.4 16.94 6.9-3.2 1.07 4.9 +4.0 L8.01 L7.5 ] ] ] L.029 DEXA 18.03 1.3 +0.1 17.74 0.5 +1.4 1.02 0.8-0.9 DPA 28.01 1.2-1.4 17.62 2.3 +0.7 1.59 1.2-2.0 28.41 17.5 1.623 DEXA 29.33 0.3 +3.2 18.10 0.7 +3.4 1.62 0.8-0.2 ""Values are means of triplicate measurements. The DPA instrument used was the DP3 (Lunar Radiation Corporation), and the DEXA instrument used was the QDR-1000 (Hologic, nc.). CV = coefficient of variation; RO = region of interest. fdifference between measured result and actual data from phantom, expressed as percent of actual data from phantom.

Mayo Clin Proc, November 1988, Vol 63 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE 1079 Table 4. Accuracy of Bone Mineral and Area Measurements by Dual Photon Absorptiometry (DPA) and Dual-Energy X-Ray Absorptiometry (DEXA) in Spine Phantom* Measurement DPA DEXA Bone mineral contentf Mass (g) CV (%) Area RO (cm 2 ) CV (%) Bone mineral densityw Density (g/cm 2 ) CV(%) 61.95 1.4 +2.6 51.20 1.2-17.1 1.21 0.14 +23.5 58.00 0.2-4.0 56.13 0.5-9.1 1.03 0.5 +5.1 * Values are means ofthree measurements. The DPA instrument used was the DP3 (Lunar Radiation Corporation), and the DEXA instrument used was the QDR-1000 (Hologic, nc.). CV = coefficient of variation; RO = region of interest. festimated calcium hydroxyapatite = 60.4 g. ^Difference between measured result and estimated data from phantom, expressed as percent of estimated data from phantom. Estimated area = 13 by 4.75 cm = 61.75 cm 2. \\Estimated density = 0.98 g/cm 2. greater for area than for bone mineral content measurements in both instruments. The DP3 overestimated the BMD by 23.5%. This finding and also the differences in the results of measurements on the block phantom probably were attributable to differences in the edge-detection algorithms and in the calibration of the two instruments the DP3 instrument is calibrated to an ashed bone standard, whereas the QDR- 1000 instrument is calibrated to a triphosphate hydroxyapatite standard similar to the phantom in use. Both instruments showed a linear response to increasing bone mineral content, as found in experiments in which ashed bones of different thickness were scanned under water (Fig. 3). The regression equation for the QDR-1000 instrument was as follows: measured mass (g) = -1.168 + 1.080x (bone mass), with a correlation coefficient (r) of 0.999. For the DP3 instrument, the regression equation was as follows: measured mass (g) = +0.135 + 1.049x (bone mass); r = 0.999. Long-term precision was studied in repeated scans of the spine phantom for the QDR-1000 instrument and in scans of an ashed bone and a spine phantom for the DP3 instrument. The scans were performed approximately daily with the QDR-1000 and weekly with the DP3 instrument for a period of 4 to 6 months; the coefficient of variation for BMD was 0.4% for the QDR-1000 (N = 214) and 1.5% for the DP3 instrument (standard bone, N = 20). Weekly measurements of the spine phantom for a period of 140 days are compared in Figure 4. The slope of the linear regression was not significant for either instrument. The coefficient of variation was 0.7% on the DP3 instrument and 0.4% on the QDR-1000 instrument (BMD, N = 16). n vivo short-term precision was evaluated in duplicate scans (obtained on the same day) in patients. The mean percentage difference between the two scans was 1.0% (range, 0 to 3.1%; N = 15) for the QDR-1000 and 1.7% (range, 0.2 to 4.4%; N = 5) for the DP3 instrument. Radiation Dose. Scan dosimetry for both instruments was performed with calcium fluoride thermoluminescent dosimeters in cadaver and patient studies. For the QDR-1000, the mean entrance exposure was 3.4 mr in studies of five patients. n a study of a male cadaver, the gonadal dose was not above background. For the DP3 instrument with a 1-Ci source, the mean 10 20 30 40 Ashed bone mass, g Lunar DP3X±1SD y=+0.135 + 1.049x;r = 0.999 o Hologic QDR-1000 7± 1SD y = -1.168 + 1.08x; r=0.999 Fig. 3. Accuracy of bone mass (bone mineral content) measurements obtained by scanning ashed bones in 20 cm of water on dual photon absorptiometry (Lunar DP3) and dualenergy x-ray absortiometry (Hologic QDR-1000) instruments. For both instruments, measurements are linear for range used in clinical studies. Standard deviations fall within confines of dots. 50

1080 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE Mayo Clin Proc, November 1988, Vol 63 1.15 -» 1.10 - Spine phantom τ^«- 60 80 Time, days AHologic QDR-1,000 Lunar DP3 -»«- - *- -^w^ Fig. 4. Long-term precision of dual photon absorptiometry (Lunar DP3) and dual-energy x-ray absorptiometry (Hologic QDR-1000) instruments, assessed by weekly scan of a spine phantom in 20 cm of water. For both instruments, no significant difference from a zero slope was noted in a regression of all points against time (N = 16). BMD = bone mineral density. entrance exposure was 2.6 mr in a study of 17 patients (slow scan speed). The cadaver ovarian exposure was 0.13 mr. Free in-air exposure measurements demonstrated a radiation exposure of 2.1 mr at 2.5 cm above the tabletop. At the side of the table, the rate of exposure for the QDR- 1000 instrument was less than 0.5 mr/h. Absorber Thickness and Fat Content. The effect of absorber thickness on measured bone mineral was studied by using a 36.6-g bone specimen with an area density of 1.84 g/cm 2. Three scans of the sample were performed in 10-, 15-, 20-, 25-, and 30-cm depths of water. The measured mean BMD values obtained at each depth were normalized to the mean BMD obtained at the 20-cm water depth, and the results are plotted in Figure 5. n 22 women seen consecutively in our laboratory for bone mineral measurements, the midabdominal thickness in the supine position was 20 ± 5.3 cm (mean ± 2 SD). Referenced to 20 cm of water in the QDR-1000 instrument, the measured BMD was 2.7% low at 10 cm and 1.5% high at 30 cm. For the DP3 instrument, the BMD was 0.7% low at 10 cm and 7.7% high at 30 cm. The useful distance between detector and tabletop is 30 cm for the DP3 and 40 cm for the QDR- 1000 instrument. The effects of changing the distance of the bone specimen above the scanning table were investigated by using the same bone specimen in 20 cm of water. n this study, the absorber thickness remained constant, but the height of the bone above the table was varied. This analysis plays a role in scanning patients with spinal lordosis or a spinal deformity. Two to three scans were performed with the bone at each of the following heights above the scanning table: 0, 5, 10, and 15 cm. n actual patients, the spine (level of transverse process) is about 5 cm above the table, and a range of 3 to 8 cm can be anticipated for this distance. Figure 6 details the results. Again, the measured BMD values obtained at each height were normalized to the zero-height BMD. The measured BMD showed less dependence on the distance between the radiation source and bone in the DP3 than in the QDR-1000 instrument. Both instruments showed little effect of fat in studies of a bone specimen scanned in water and oil, with the fat content ranging from 0 to 60% by volume (Fig. 7). The QDR-1000 system showed a 2% error in BMD at 40% fat, whereas BMD remained unchanged with the DP3 instrument. Little change was noted with either instrument when a layer of 1 to 3 cm of margarine was placed on top of the bone to imitate bone marrow fat. 25 30 10 15 Water depth, cm Fig. 5. Dependence of bone mineral measurements on absorber thickness (patient abdomen), determined by scanning a bone phantom in various levels of water. Both dual photon absorptiometry (Lunar DP3) and dual-energy x-ray absorptiometry (Hologic QDR-1000) instruments are stable in the clinically useful range from 15 to 25 cm of water. BMD = bone mineral density.

Mayo Clin Proc, November 1988, Vol 63 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE 1081 104 i- 104 BMD Φ P <z^- ^^*"^^ = (3 1001 ^s_ 96 92 ( A Hologic QOR-1,000 Lunar DP3 104 i- 1001 96 < ^ - >k - 5 10 15 0 5 > 5 " Sample height, cm s 10 10 Mass Area Fig. 6. Dependence of bone mineral measurements on distance of object from imaging table. Clinical measurements are made within 3 to 8 cm. Note greater dependency of dual-energy x-ray absorptiometry (Hologic QDR-1000) instrument than of dual photon absorptiometry (Lunar DP3) instrument. BMD = bone mineral density. The previously described experiments are simple models designed to test only a single variable. Differences in body composition and body habitus between patients generally involve more than one variable, and a neutralization of errors may result. n bone mineral measurements of the lumbar spine, body weight changes and obesity are the main variables, and variations in the distance of the spine from the imaging table play a minor role. Because with obesity both abdomi- 15 15 nal thickness and fat content in the tissue surrounding bone are increased, we have designed an experiment in which increased absorber thickness is achieved by adding successive layers of fat. The proportion of fat in the tissues surrounding the bone in women who had bone mineral measurements in the spine or hip was estimated by using the ratio of soft tissue absorption coefficients at the 44- and 100-keV photon energies, which is routinely obtained on the DP3 instruments and reflects fat-to-muscle (water) ratio in the tissue. 3 Figure 8 shows typical soft tissue absorption coefficient ratios at the hip and the spine in women who ranged in weight from 44 to 91 kg (middle and bottom panels). A standard curve (top panel) relates this number to a water and oil mixture of 20-cm thickness. Only a minimal decrease in BMD values and in precision was noted when water levels ranged from 20 to 30 cm and oil content ranged from 25 to 50% (Fig. 9). Outside this range, however, a significant error in BMD was recorded. Hence, major changes in a subject's weight between repeated measurements may obscure small changes in bone mass. As shown in Figure 9, for body composition of more than 50% fat, the error Water-fat mixture, 20 cm ue in water 104 102 8 8 8 g 'S "5 104 # g 102 100. 98-0 BMD (1.822 g/cm2) (Lunar DP3) BMD (1.839 g/crn2) (Hologic QDR-1,000) X Χ 20 40 60 Fat content, % Distribution of RST, hip Distribution of RST, spine Fig. 7. Dependence of bone mineral density (BMD) measurements on fat content in bone surrounding medium. Both dual photon absorptiometry (Lunar DP3) and dual-energy x-ray absorptiometry (Hologic QDR-1000) instruments show little change in the clinically useful range of 10 to 40% fat. Fig. 8. Range of ratio of soft tissue absorption coefficients (RST) in consecutive women who underwent bone mineral studies of hip or spine. RST represents the mean absorption coefficient ratio of the tissue around the bone. Top panel shows relationship to fat content of a water-vegetable oil mixture. Body weights of women ranged from 44 to 91 kg.

1082 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE Mayo Clin Proc, November 1988, Vol 63 Fig. 9. Effect on bone mineral measurements of concomitant variations in absorber thickness and fat content of material surrounding an ashed bone, imitating changes that may be expected with obesity. Percentages of fat and absorber thickness are shown in top panel, bone mineral determined by dualenergy x-ray absorptiometry is shown in middle panel, and corresponding distribution of fat content in patients is depicted in bottom panel. R = reference point for bone mineral measurements (100%). in BMD is about 8%; in obese patients with ratios of soft tissue absorption coefficients of less than 1.36 (generally more than 91 kg), a different instrument calibration is probably needed for accurate bone mineral measurements. We have obtained similar results previously with use of the DP3 instruments. 6 For a comparison of lumbar spine bone mineral results in patients, 15 patients and volunteers underwent spinal BMD measurements on both the QDR-1000 (Hologic) and the DP3 (Lunar) instrument. Figure 10 shows the linear regression of the results. The BMD measurements from the two instruments showed an excellent correlation (r = 0.988; S y., = 0.034 g/cm 2 ). The regression equation was as follows: Hologic BMD = -0.082 + 0.973 x Lunar BMD. The ratio of Lunar BMD to Hologic BMD for the mean BMD value was 1.113, but it ranged from 1.064 to 1.210 in individual patients. DSCUSSON Constraints on the clinical use of DPA instruments (such as the Lunar DP3 used in this study) arise from the relatively low and constantly decreasing intensity of the photon source used. This attrition leads to problems with the long-term stability of these instruments. Changes of several percent have been reported to occur when an old isotope source is replaced with a new source. 7,8 The magnitude of this error seems to vary between instruments and sources and is largest when the 13-mm detector collimator is used. A relatively long scanning time (35 minutes) and poor image resolution are additional drawbacks with the standard DPA method. Both instruments used in this study measure integral (that is, compact and trabecular) bone in the lumbar spine. For both instruments, the region of interest ranges from L-2 to L-4, including the disk spaces L2-3 and L3-4 and excluding the transverse processes. Bone mineral images on the QDR-1000 instrument are of superior resolution, which facilitates matching regions of interest and contributes to better precision. On the QDR-1000 instrument, the operator can display previous and recent bone mineral images concurrently on the screen for better comparison. The resulting improvement in precision increases the usefulness of this procedure for conducting longitudinal studies and for estimating the rate of bone loss. 9,10 n longitudinal studies, precision can be further improved by performing duplicate measurements at each measuring time; this approach is now possible because of the short scanning time and relatively low radiation dose. Scan- 1.40 i Patient comparison 5 ^ CO ' 100 -.^«A y^k A 0 L 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 DP3 BMD, g/cm 2 Fig. 10. Comparison of bone mineral densities (BMD), measured in 15 patients with dual photon absorptiometry (DP3) and dual-energy x-ray absorptiometry (QDR-1000) instruments.

Mayo Clin Proc, November 1988, Vol 63 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE 1083 ning time was significantly shorter with the QDR-1000 instrument than the DP3 instrument. The x-ray source of the QDR-1000 instrument is well shielded, and the instrument can be used without the special operator protection commonly used in radiography. The radiation exposure is low in both instruments (less than 5 mr). The new x-ray-based procedure has been referred to in previous communications as quantitative digital radiography, a name used by Hologic, nc., the company that introduced the technique, or as dual-energy radiography, as in our first reports on the performance of the Hologic instrument. The term "dual-energy x-ray absorptiometry" is probably the most appropriate, inasmuch as quantitative x-ray absorptiometry has been available for many years and thus, except for the use of two distinct energy x-rays, a time-honored approach is being used. Comparison of spinal BMD measurements obtained with both instruments in the same patients showed consistently lower BMD values for the DEXA (QDR-1000) instrument, and the regression line differed significantly from a line of identity. Hence, conversion of data in longitudinal studies is necessary when analyses originated with the DPA (DP3) instrument and are continued with the new QDR-1000 instrument. A first approach to this conversion is a regression equation obtained by scanning a large number of patients and normal subjects with both instruments. This approach is probably not suitable for longitudinal studies. Our formulas for data conversion obtained from phantom and from patient studies varied significantly. n our experience, depending on the BMD value, differences from 2 to 20% (mean percent difference) can be expected in individual patients. The regression formula for conversion between QDR-1000 and DP3 instruments provided by Hologic, nc. (technical note 1, preliminary information on QDR- 1000 normal data bases) differs from the data reported herein. An error of 1 to 5% (standard error of the estimate) in normal subjects is reported. The error is probably higher when patients with low bone mineral values are included. Hence, any conversion of data would result in an error that would exceed the normal bone loss expected to occur in 1 year in normal women or those with osteoporosis. Data conversion based on a regression curve obtained from a population of subjects and patients would, however, be sufficient for estimating the patient's risk for bone fracture on initial examination. f the instrument used in longitudinal studies is changed, all patients should be scanned with both instruments on at least one occasion, and an individual conversion factor should be calculated. The effects on BMD of varied absorber thickness and of the distance of the spine from the radiation source may be related to changes in the relative number of scattered photons detected with each system. 11 As already shown for the DP3 instrument, 11 in the QDR-1000 the effect of changes in fat in the surrounding tissue and in the bone marrow seems to be minimal within the range usually encountered in patient studies. Measurements of BMD with the QDR-1000 instrument showed more dependence on the position of the spine in the radiation beam than with the DP3 instrument. For both instruments, however, this factor probably has little influence on measurements made on the lumbar spine, where the distance between tabletop and spine varies minimally. During measurements, the patient is supine, and the physiologic lumbar scoliosis is flattened by elevation of the knees. With measurements of bone mineral in the hip, where the gluteal fat pad may vary substantially, and in patients with severe spinal deformity, falsely low bone mineral values may be obtained. On the basis of our studies of models, bone mineral measurements on the spine apparently can be performed without major errors in accuracy when the body thickness remains within 15 to 25 cm and when the ratio of soft tissue absorption coefficients remains more than 1.36 (about 50% fat). These instrument-imposed limits are well within the range of body thickness and fat content commonly found in women with osteoporosis. A measurement of the ratio of soft tissue absorption coefficients, and therefore an estimation of fat content in the tissue surrounding the bone, is not available on the QDR-1000 instrument, and changes in fat between measurements cannot be monitored. We have added a simple device to the QDR-1000 instrument arm that allows measurement of patient thickness at the midabdomen before the scanning procedure, when this is deemed necessary. Patients with an abdominal thickness of more than 25 cm are flagged for special considerations. Changes in abdominal thickness of 10 cm or more have been noted between studies in patients with liver disease,

1084 BONE MNERAL MEASUREMENTS OF LUMBAR SPNE Mayo Clin Proc, November 1988, Vol 63 liver transplants, and renal dialysis, in whom the presence or absence of ascites may substantially change the abdominal thickness between measurements. Most patients on estrogen treatment gain some weight; however, the amount is usually insufficient to cause concern. Only occasionally have we noted weight gain that necessitated special calibration of the instrument. CONCLUSON With use of the QDR-1000 instrument, the small, yet repeatedly demonstrated improvement in measurement precision, 12 " 14 the faster scanning time, and the higher resolution of images should result in a cost reduction for BMD measurements and a wider clinical application in longitudinal studies. REFERENCES 1. Wahner HW, Dunn WL, Riggs BL: Assessment of bone mineral (in two parts). J Nucl Med 25:1134-1141; 1241-1253,1984 2. Riggs BL, Wahner HW: Bone densitometry and clinical decision-making in osteoporosis (editorial). Ann ntern Med 108:293-295,1988 3. Wahner HW, Riggs BL: Methods and application of bone densitometry in clinical diagnosis. CRC Crit Rev Clin Lab Sei 24:217-233,1986 4. Stein JA, Lazewatsky JL, Hochberg AM: Dual-energy x-ray bone densitometer incorporating an internal reference system (abstract). Radiology 165 (Suppl):313,1987 5. Stein JA, Lazewatsky JL, Hochberg AM: A dual energy x-ray bone densitometer incorporating an internal reference system (submitted for publication) 6. Wahner HW, Dunn WL, Mazess RB, Towsley M, Lindsay R, Markhard L, Dempster D: Dual-photon Gd-153 absorptiometry of bone. Radiology 156:203-206,1985 7. Dunn WL, Kan SH, Wahner HW: Errors in longitudinal measurements of bone mineral: effect of source strength in single and dual photon absorptiometry. J Nucl Med 28:1751-1757,1987 8. Lindsay R, Fey C, Haboubi A: Dual photon absorptiometric measurements of bone mineral density increase with source life. Calcif Tissue nt 41:293-294,1987 9. Wahner HW: Assessment of bone loss with repeated bone mineral measurements: application to measurements on the individual patient. Nuc Compact 18:7-10,1987 10. Wahner HW, Melton LJ, Dunn WL, Kan SH, Riggs BL: Assessment of normal bone loss and optimal interval required for prediction of lumbar spine bone loss by dual photon absorptiometry (DPA) (abstract). J Nucl Med 27:964,1986 11. Hauser MF, Wahner HW, Dunn WL: The effect of absorber thickness on bone mineral density. n Bone Mineral Measurements by Photon Absorptiometry: Methodological Problems. Edited by JV Dequeker, D Geusens, HW Wahner. Leuven, Belgium, Leuven University Press, pp 190-195 12. Sartoris DJ, Stein JA, Ramos E, Lambiase R, Ho C, Andre M, Resnick D: Quantitative dual-energy digital radiography of the spine: comparison to dual-photon absorptiometry and quantitative computed tomography. Presented at the Sixth nternational Workshop on Bone and Soft Tissue Densitometry, Buxton, Derbyshire, England, September 1987 13. Kelly T, Slovik D, Neer R: Accuracy of quantitative digital radiography (QDR) and dual photon absorptiometry (DPA) (abstract). J Bone Miner Res 3 (Suppl 1):S216,1988 14. Glueer CC, Steiger P, Selvidge R, Hayashi C, Genant HK: Performance of x-ray and isotope-based dual-energy bone densitometers (abstract). J Bone Miner Res 3 (Suppl 1):S126,1988