- Email: [email protected]
Mechanical strength of the thoracolumbar spine in the elderly: prediction from in situ dual-energy X-ray absorptiometry, quantitative computed tomography (QCT), upper and lower limb peripheral QCT, and quantitative ultrasound
Bone Vol. 31, No. 1 July 2002:77– 84
Mechanical Strength of the Thoracolumbar Spine in the Elderly: Prediction From In Situ Dual-energy X-ray Absorptiometry, Quantitative Computed Tomography (QCT), Upper and Lower Limb Peripheral QCT, and Quantitative Ultrasound ¨ LLER,1 D. BU ¨ RKLEIN,1,2 V. KUHN,1,2 C. GLASER,3 R. MU ¨ LLER,4 C. C. GLU ¨ ER,5 and E.-M. LOCHMU 2 F. ECKSTEIN 1
Frauenklinik Innenstadt, Ludwig-Maximilians-Universita¨t (LMU), Mu¨nchen, Germany Musculoskeletal Research Group, Institute of Anatomy, LMU, Mu¨nchen, Germany 3 Institute for Clinical Radiology, LMU, Mu¨nchen, Germany 4 Orthopedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, USA 5 Medical Physics, Diagnostic Radiology, Christian-Albrechts-Universita¨t, Kiel, Germany 2
Key Words: Vertebral failure; Mechanical strength; Spine; Biomechanical testing; Densitometry; Ultrasound.
The objective of this study was to compare the ability of clinically available densitometric measurement techniques for evaluating vertebral strength in elderly individuals. Measurements were related to experimentally determined failure strength in the thoracic and lumbar spine. In 127 specimens (82 women and 45 men, age 80 ⴞ 10 years), dual-energy X-ray absorptiometry (DXA) was performed at the lumbar spine, femur, radius, and total body, and peripheral-quantitative computed tomography (pQCT) at the distal radius, tibia, and femur under in situ conditions with intact soft tissues. Spinal QCT and calcaneal ultrasound parameters were performed ex situ in degassed specimens. Mechanical failure loads of thoracic vertebrae 6 and 10 (T-6 and -10), and lumbar vertebra 3 (L-3) were determined in axial compression on functional three-segment units. In situ anteroposterior DXA and QCT of the lumbar spine explained approximately 65% of the variability of thoracolumbar failure. A combination of cortical and trabecular density (QCT) provided the best prediction in the lumbar spine. However, this was not the case in the thoracic spine, for which lumbar cortical density (QCT) and DXA provided significantly better estimates than trabecular density (QCT). pQCT was significantly less correlated with the strength of lumbar and thoracic vertebrae (r2 ⴝ 40%), but was equivalent to femoral or radial DXA. pQCT measurements in the lower limb showed no advantage over those at the distal radius. Ultrasound explained approximately 25% of the variability of vertebral failure strength and added independent information to spinal QCT, but not to spinal DXA. These experimental results advocate site-specific assessment of vertebral strength by either spinal DXA or QCT. (Bone 31:77– 84; 2002) © 2002 by Elsevier Science Inc. All rights reserved.
Introduction Vertebral fractures represent the classic hallmark of osteoporosis1 with an annual incidence of approximately 500,000 patients in North America.11,34 These fractures substantially reduce the quality-of-life in the elderly,26 and the costs involved have been estimated to be 750 million $US per year in the US.39 Reductions in bone mass and strength have been identified as the main causes for these fractures in the elderly,25,35 with only a small proportion resulting from traumatic events.11 Because treatment is available to slow the age-related reduction of bone strength,27,31 accurate measurements of bone properties (as surrogates of vertebral fracture risk) are required to reliably identify patients that require therapeutic intervention. A wide range of densitometric techniques is currently available for determining bone properties at different skeletal sites.16,18 Dual-energy X-ray absorptiometry (DXA) is the method used most widely and has the advantage that measurements can be obtained at or close to the site of interest.16 A disadvantage is that DXA cannot discriminate between cortical and trabecular bone, and that the surrounding soft tissues may introduce relevant measurement errors.29,30,42 In addition, conventional anteroposterior (AP) measurements of the lumbar spine include the posterior elements, which are frequently subject to osteoarthritic changes, particularly in the elderly. Quantitative computed tomography (QCT) is less affected by surrounding soft tissue errors, eliminates the posterior elements, and permits analysis of trabecular and cortical bone separately. However, QCT involves relatively high X-ray doses and relies on expensive equipment.16 Peripheral QCT (pQCT) is less costly, uses less radiation, and X-rays are applied to less vulnerable (peripheral) sites. However, because of skeletal heterogeneity,2,3,20 it is questionable as to whether peripheral measurements reflect bone properties of the axial skeleton. A relevant relationship of ultrasonic properties has been reported with both trabecular micro-
Address for correspondence and reprints: Dr. E.-M. Lochmu¨ller, Universita¨tsfrauenklinik, Maistrasse 11, D-80337 Mu¨nchen, Germany. E-mail: [email protected] © 2002 by Elsevier Science Inc. All rights reserved.
77
8756-3282/02/$22.00 PII S8756-3282(02)00792-5
78
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
Bone Vol. 31, No. 1 July 2002:77– 84
Table 1. Descriptive statistics of the mechanical and densitometric data
Population data Body height (cm) Body weight (kg) Mechanical data Failure load (kN) T-6 T-10 L-3 Failure stress (MPa) T-6 T-10 L-3 Site-specific densitometric data DXA (in situ) BMD (g/cm2) at L-3 Spinal QCT (ex situ) Trab density (g/cm3) at L2–4 Cort density (g/cm3) at L2–4
Men
Women
Difference
170 ⫾ 8 63 ⫾ 13
155 ⫾ 7 52 ⫾ 12
⫺9%b ⫺17%b
3.19 ⫾ 1.48 4.21 ⫾ 1.85 4.09 ⫾ 1.69
2.06 ⫾ 0.87 2.57 ⫾ 1.09 2.45 ⫾ 1.02
⫺35%b ⫺39%b ⫺40%b
426 ⫾ 167 394 ⫾ 174 300 ⫾ 132
346 ⫾ 143 300 ⫾ 122 213 ⫾ 81.8
⫺19%a ⫺24%b ⫺29%b
1.17 ⫾ 0.25
0.96 ⫾ 0.25
⫺18%b
70.2 ⫾ 32.7 300 ⫾ 71.1
58.9 ⫾ 32.3 233 ⫾ 68.3
⫺16% n.s. ⫺23%b
KEY: DXA, dual-energy X-ray absorptiometry; BMD, bone mineral density; QCT, quantitative computed tomography; trab, trabecular; cort, cortical; n.s., difference not statistically significant. a p ⬍ 0.01. b p ⬍ 0.001 (unpaired Student’s t-test).
structure and mechanical properties.23,37 This has stimulated hope that quantitative ultrasound (QUS) can provide significant information for predicting bone strength, in addition to bone mass or density. The attractiveness of calcaneal QUS also lies in the complete lack of ionizing radiation and in the relatively low costs involved. The objective of the current study was therefore to determine which measurement technique in which region is best suited for evaluating vertebral strength in elderly individuals. Densitometric measurements were related to mechanical strength of the thoracic and lumbar spine. Previous experimental studies have examined bones ex situ (except for two studies5,28). This may have biased the results in favor of DXA, because of the lack of soft tissue artifacts. Moreover, experimental studies have generally investigated the relationship of bone densitometry and mechanical failure at the lumbar spine, compared with calcaneal QUS.9,28 However, most clinical fractures occur in the thoracic spine.12,21 Studies may thus have overestimated the value of spinal DXA and QCT in predicting vertebral fracture risk, given the heterogeneity of bone properties within the spine.3 No previous experimental study has compared in situ spinal DXA with QCT, and with non-site-specific techniques, including QUS and pQCT. The following specific questions were addressed: 1. Does QCT show a higher correlation with mechanical strength of vertebrae than spinal (anteroposterior) in situ DXA, because it eliminates artifacts from posterior elements and soft tissues, and because it permits to assess trabecular and cortical properties separately? 2. Do lumbar DXA and QCT display a higher correlation with vertebral strength than non-site-specific (peripheral) measurements (DXA, pQCT, QUS), even when mechanical testing is performed in the thoracic spine? 3. Does pQCT at the lower limb display a higher correlation with vertebral strength than pQCT at the upper limb, because of the involvement of the lower limb in weight-bearing? 4. Is calcaneal QUS capable of adding relevant independent information to site-specific measurements of bone status in
multiple regression models, because of its relationship with trabecular microstructure and mechanical properties? Materials and Methods Study Sample The study sample initially comprised 140 formalin-fixed cadavers with intact skin and soft tissues from a course of macroscopic dissection. The only criterion of inclusion was a testamentary decree to the institute several years prior to death. The sample can therefore be assumed to be a representative selection of the elderly resident population of Bavaria. Bone biopsies were harvested from the iliac crest for routine histomorphometric analysis.22 Specimens with bone disease other than osteopenia or osteoporosis were excluded (three with malignancy and seven with renal osteopathy). One specimen was discarded due to an osteolysis on the QCT scans, and two due to fracture of lumbar vertebrae 2– 4 (L-2– 4). Of the remaining 127 specimens 82 were women (age 82 ⫾ 9 years) and 45 were men (age 77 ⫾ 11 years). The age range was 47–98 years. The men were significantly larger and heavier than the women (Table 1). Bone Densitometry DXA (DPX-L, Lunar, Madison, WI) was used to determine the bone mineral content (BMC) and areal bone mineral density (BMD, in grams per centimeter) under in situ conditions (with intact skin and soft tissues), analyzing the total skeleton, lumbar vertebra 3 (L-3) in the anteroposterior (AP) direction, total proximal femur, femoral neck, greater trochanter, and distal radius. In cases where L-3 was fractured (nine cases), spinal DXA (but not the other) measurements were discarded. We have shown previously that DXA is as reproducible in cadavers as it is in vivo, and that values are not affected by 10 months of formalin fixation.30 pQCT was also performed in situ. An XCT 2000 scanner was used at the distal radius (4% and 20% from the wrist), and an
Bone Vol. 31, No. 1 July 2002:77– 84
XCT 3000 (Stratec Medizintechnik, Pforzheim, Germany) at the tibia (8% and 50% from the ankle) and distal femur (8% and 30% from the knee). Technical details and measurement precision have been described previously.20 QCT was not performed in situ, because it was not possible to transport the cadavers to the clinical CT unit. QUS measurements were also obtained ex situ, because, due to supination contractions, the feet could not be brought into the required position in the scanner, and also because there was concern that gas bubbles in the soft tissue may alter the measurements. After the dissection course, the thoracolumbar spines were excised and radiographed15 using a Polyphos 30 M X-ray system (Siemens, Erlangen, Germany) and SR-H film (35 ⫻ 42 cm, Konica, Hohenbrunn, Germany). Four films were obtained in each specimen (two in AP and two in lateral projection), one focusing on the thoracic and one on the lumbar spine. In vertebrae with deformity suggestive of a fracture (11 cases in L-3, 7 cases in thoracic vertebra 10 [T-10], and 12 cases in T-6) the densitometric and biomechanical data of these specific vertebrae were discarded. Specimens with osteoarthritic changes were not excluded. Single-energy QCT was obtained ex situ in degassed specimens, then sealed within water in thin polyethylene bags. We used a Somatom Plus 4 CT scanner at 80 kV (Siemens) applying standard settings of the manufacturer (195 mA, rotation time 0.75 sec, kernel SP90, slice thickness 10 mm, field of view 216 mm, matrix 512 pixels). A phantom with equivalents of 0 and 200 mg/cm3 hydroxylapatite was used to convert Hounsfield units into density values. Images were obtained at the midvertebral level of L-2, L-3, and L-4, with average values of L2– 4 being used in further analyses. Trabecular and cortical density were determined with the software. As the cortical shell is only several hundred microns thick,40 an apparent cortical/subcortical density (but not a true cortical density) was determined. The root-mean-square (RMS) average coefficient of variation (CV%)17 of repeated measurements (four repetitions in 14 specimens with repositioning, obtained on different days) was 2.9% for trabecular, and 3.1% for cortical density. In addition, the cross-sectional areas of the vertebral bodies were determined at the midvertebral level, with the precision (RMS average CV%) being 1.2%. To assess whether a combination of trabecular and cortical properties could improve the prediction of vertebral strength, we calculated the individual percentage deviation from the mean value (of all individuals) for both parameters. The mean values of these percent deviations were then used as a combined measure. QUS of the calcaneus was obtained under ex situ conditions. The calcanei were dissected clean of the surrounding tissues and degassed. Measurements were performed with an Achilles Plus scanner (Lunar) sealed for underwater use. The calcanei were measured in a temperature-controlled water bath (37°C) in a defined position.38 The precision for speed-of-sound (SOS), broadband ultrasound attenuation (BUA), and stiffness index (SI)23 was found to be in the range of in vivo measurements.38 Although a prolonged period of formalin fixation (⬎1 year) was shown to cause a decrease of SOS and SI, the values were highly correlated with those before fixation.38 Please note that the so-called stiffness index in QUS does not reflect stiffness in mechanical terms. Mechanical Testing Three segments of the thoracolumbar spine (T5–7, T9 –11, L2– 4) were tested as functional units with ligaments and inter-
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
79
vertebral disks, but without posterior elements.7,10,36,41 The upper and lower vertebral bodies were embedded planoparallel to the endplates of the central target vertebra, and segments were compressed axially, as described elsewhere.7 To compute the failure stress (MPa), the cross-sectional area was estimated for each vertebra by multiplying the anteroposterior and mediolateral diameter (measured with calipers at the midvertebral level), and by a scaling factor. The factor was derived from a comparison between the actual cross-sectional areas of the vertebrae (determined from prints in 65 other specimens) and the two diameters. A scaling factor of 0.86 produced no systematic difference between measured and computed cross-sectional area, with r ⫽ 0.96 between both methods. Statistics To determine average failure load and stress of the spine from the three segments, the deviation from the mean (of all individuals) was derived (in percent) for each individual for each vertebral level (T-6, T-10, and L-3). The average (without data from fractured vertebrae) was then computed as a representative value for the thoracolumbar spine. The association between failure loads and stress with the densitometric data was assessed by linear regression analysis, using STATVIEW 4.5 software (Abacus Concepts, Berkeley, CA). With DXA, the bone mineral content (BMC; grams) was used for correlation with failure load, and the bone mineral density (BMD; grams per square centimeter) for that with failure stress. With QCT, the density values (milligrams per cubic centimeter) were used for the correlation with failure stress, and the density multiplied by the mean cross-sectional areas of the vertebral bodies for correlation with failure loads.4,6 With pQCT, the total bone mineral content (milligrams) was used for correlation with failure loads and total density (milligrams per cubic centimeter) for that with failure stress. Fisher’s Z-test was used to test whether the correlations were significantly different from site-specific DXA. To determine whether QUS measurements at peripheral sites could add significant, independent information, we used a stepwise multiple regression analysis (forward mode). Results Men displayed a significantly higher failure load and stress (load per unit area) than women at all spinal levels (Table 1). This was also the case after adjusting for age. The correlation of failure stress between T-6 and T-10 (r ⫽ 0.78) was significantly higher (p ⬍ 0.01) than that between T-6 and L-3 (r ⫽ 0.60) and that between T-10 and L-3 (r ⫽ 0.62). The correlation of mechanical strength in the lumbar spine (L-3) with densitometry is given in Table 2. The highest correlation with failure load was observed for cortical density (QCT), multiplied by the cross-sectional area (r ⫽ 0.78; standard error of the estimate [SEE] 31.1%), but the coefficients for trabecular density (QCT: r ⫽ 0.75; SEE ⫽ 33.0%) and BMC of L-3 (in situ DXA: r ⫽ 0.73; SEE ⫽ 34.2%) were not significantly different. A combined measure of trabecular and cortical density (QCT) provided a significantly (p ⬍ 0.05) better prediction of failure load (r ⫽ 0.82; SEE ⫽ 28.8%) than spinal DXA (p ⬍ 0.05). DXA at other sites, and pQCT provided significantly lower (p ⬍ 0.05– 0.01) correlations (r ⫽ 0.55– 0.62). Coefficients for pQCT at the lower limb were not significantly different from those of the distal radius (Table 2). The correlation of QUS (r ⫽ 0.48; SEE ⫽ 43.7%) was significantly lower (p ⬍ 0.01) than spinal DXA and QCT, but not significantly different from non-sitespecific DXA or pQCT. The best predictor of L-3 failure stress was cortical density (QCT: r ⫽ 0.76; SEE ⫽ 30.0%). Trabecular
80
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
Bone Vol. 31, No. 1 July 2002:77– 84
Table 2. Correlation between failure load and failure stress at the lumbar spine (L-3) with densitometric data (N ⫽ 116)
DXA (in situ): for load, BMC (g); for stress, BMD (g/cm2) L-3 Neck Trochanter Femur Radius Total body
Load
Stress
0.73a 0.61a 0.55a 0.62a 0.59a 0.69
0.68 0.53a 0.49a 0.53a 0.49a 0.60
Spinal QCT 2–4: for load, CNT (mg/cm); for stress, bone density (mg/cm3) Trabecular L2–4 0.75 Cortical L2–4 0.78 Combination L2–4 0.82c Peripheral QCT: for load, CNT (mg/mm); for stress, bone density (mg/cm3) Radius 4% 0.53b 20% 0.62a Tibia Distal 0.57a Shaft 0.60a Femur Distal 0.53b Shaft 0.60a Calcaneal ultrasound SOS (m/sec) BUA (dB/MHz) Stiffness index
0.45b 0.47b 0.48b
0.67 0.76 0.76
0.35b 0.35b 0.49a 0.41b 0.51a 0.39b 0.44b 0.44b 0.46b
KEY: Cnt, bone mineral content; SOS, speed of sound; BUA, broadband ultrasound attenuation. Other abbreviations identical to those in Table 1. All correlations statistically significant (p ⬍ 0.01). a Correlation significantly (p ⬍ 0.05) lower than that with in situ DXA of L-3. b Correlation significantly (p ⬍ 0.01) lower than that with in situ DXA of L-3. c Correlation significantly (p ⬍ 0.05) higher than that with in situ DXA of L-3.
density (QCT: r ⫽ 0.67; SEE ⫽ 33.5%) and spinal DXA (r ⫽ 0.68; SEE ⫽ 33.4%) were, however, not significantly different. Both non-site-specific DXA and pQCT displayed a lower (p ⬍ 0.05– 0.01) correlation with failure stress (r ⫽ 0.35– 0.53). QUS (r ⫽ 0.46; SEE ⫽ 39.7%) was not significantly different from non-site-specific DXA or pQCT. The correlation of mechanical strength in the thoracic spine (T-10) with densitometry is given in Table 3. The coefficients obtained for DXA, pQCT, and QUS were not significantly different from those with L-3. Lumbar DXA displayed an identical correlation with failure load when comparing T-10 and L-3 (r ⫽ 0.73; SEE ⫽ 34.2%). However, trabecular density (QCT) displayed a significantly lower correlation with T-10 failure load (p ⬍ 0.01) and stress (p ⬍ 0.05) than with L-3 failure load and stress. The cortical density also showed a lower correlation, but the difference was only significant for the failure stress (p ⬍ 0.05). Spinal DXA and cortical density (QCT) provided similar estimates of thoracic strength, but trabecular density displayed a lower correlation (p ⬍ 0.05). The combination of cortical and trabecular density did not improve the correlation over that of cortical density alone. Non-site-specific DXA, pQCT, and QUS displayed lower correlations (p ⬍ 0.05– 0.01) than site-specific densitometry. The correlation of densitometric variables with average failure strength of the thoracolumbar spine is shown in Table 4. Cortical density (QCT) and in situ spinal DXA (Figure 1a,b) provided the best estimate of vertebral failure loads (r ⫽ 0.80) and stress (0.73/0.72). Trabecular density (QCT) displayed a lower correlation with the difference being significant (p ⬍ 0.05) for failure load. A combination of cortical and trabecular density
(QCT) did not improve the prediction beyond that of the cortical density alone. Non-site-specific DXA and pQCT displayed significantly (p ⬍ 0.01) lower correlations with thoracolumbar failure load and stress than spinal DXA and QCT, except for total body DXA. pQCT at the lower limb (r ⫽ 0.52– 0.64) did not offer an advantage over the distal radius (r ⫽ 0.57– 0.66; Figure 1c). The correlation of QUS parameters with strength (Figure 1d) was lower (p ⬍ 0.01) than for spinal DXA and QCT, but not significantly lower than for non-site-specific DXA or pQCT. None of the QUS parameters added significant, independent information to spinal DXA. However, BUA and stiffness index improved the prediction of failure stress in combination with femoral or radial DXA, and the prediction of failure load in combination with radial (but not femoral) DXA. In combination with trabecular or cortical density (QCT), all QUS parameters improved the prediction of failure load and stress in multiple regression models, but the gain in correlation was relatively small. Discussion In this study we sought to determine which measurement technique in which region is best suited for evaluating vertebral strength in elderly individuals. A clinical study design would have been problematic in this context, because of the cumulative X-ray exposure involved in multiple measurements, and because of the difficulty in objectively diagnosing the incidence of vertebral fractures, which often pass unnoticed. A large number of individuals and very long observation periods are required in prospective clinical studies to obtain statistically meaningful data.
Bone Vol. 31, No. 1 July 2002:77– 84
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
81
Table 3. Correlation between failure load and failure stress in the thoracic spine (T-10) with densitometric data (N ⫽ 120)
DXA (in situ): for load, BMC (g); for stress, BMD (g/cm2) L-3 Neck Trochanter Femur Radius Total body
Load
Stress
0.73 0.56a 0.51b 0.55b 0.57a 0.65
0.62 0.46a 0.45a 0.44a 0.45a 0.55
Spinal QCT 2–4: for load, CNT (mg/cm); for stress, bone density (mg/cm3) Trab density (g/cm3) L2–4 0.57a Cort density (g/cm3) L2–4 0.71 Combination L2–4 0.68 Peripheral QCT: for load, CNT (mg/mm); for stress, bone density (mg/cm3) Radius 4% 0.52b 20% 0.61a Tibia Distal 0.58a Shaft 0.56a Femur Distal 0.42b Shaft 0.58a Calcaneal ultrasound SOS (m/sec) BUA (dB/MHz) Stiffness index
0.38b 0.47b 0.45b
0.53 0.62 0.61
0.33b 0.34b 0.45a 0.31b 0.48a 0.37b 0.37b 0.44a 0.43a
Abbreviations identical to those in Tables 1 and 2; all correlations statistically significant (p ⬍ 0.01). a Correlation significantly lower (p ⬍ 0.05) than that with in situ DXA of L-3. b Correlation significantly lower (p ⬍ 0.01) than that with in situ DXA of L-3.
A strength of the study is that failure was not only determined in the lumbar, but also in the thoracic spine, that DXA was performed under in situ conditions, and that non-site-specific techniques were included, such as radial and femoral DXA, pQCT of the upper and lower limb, and calcaneal QUS. In particular, a large sample size permits determination of whether certain techniques perform significantly better than others. A potential limitation of this study is the use of fixed specimens. Calabrisi et al.8 and McElhaney et al.33 observed a small decrease, but Greenberg et al.19 and Edmondston et al.14 noted an increase in bone strength with fixation. Edmondston et al.14 showed, however, that fixation does not alter the correlation between vertebral bone mass and strength. We have previously demonstrated that long-term fixation has no significant effect on in situ DXA,30 and a predictable effect on QUS.38 Our data are in good agreement with those reported on failure loads and their correlations with DXA, QCT, and calcaneal QUS in fresh specimens (e.g., see Cheng et al.9). We therefore believe that the results of this study are not biased by the use of fixed specimens. Another limitation is that no lateral DXA of the spine could be performed under in situ conditions. However, lateral DXA was obtained ex situ in a subset of specimens.7 These measurements displayed a significantly (p ⬍ 0.05) higher correlation with failure loads of L-3 than in situ AP DXA (r ⫽ 0.85 vs. 0.73), but not with T-10 (r ⫽ 0.69 vs. 0.73) or T-6 (0.61 vs. 0.68). Because ex situ lateral DXA did not improve the prediction of thoracic failure, we assume that in situ lateral DXA is even less capable of improving the prediction of failure strength of the thoracic spine over AP scans. For logistical reasons, QCT could only be examined under ex situ conditions. It cannot be ruled out entirely that the lack of surrounding soft tissue spuriously improved the correlation of
Figure 1. Correlation of thoracolumbar failure loads with densitometric measurements. The regression plots show failure loads throughout the thoracic spine (in percent deviation from mean values) as correlated with: (a) bone mineral content (BMC) of lumbar vertebra 3 (L-3), measured using in situ DXA; (b) cortical density (multiplied by cross-sectional area) of lumbar vertebrae 2– 4 (L2– 4), measured using spinal QCT; (c) total bone mineral content of the distal radius (20% measurement site), measured using in situ pQCT; and (d) broadband ultrasound absorption (BUA) of the calcaneus, measured using ex situ quantitative ultrasound.
82
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
Bone Vol. 31, No. 1 July 2002:77– 84
Table 4. Correlation between average failure load and failure stress throughout the thoracolumbar spine (T-6, T-10, L-3) with densitometric data (N ⫽ 127)
DXA (in situ): for load, BMC (g); for stress, BMD (g/cm2) L-3 Neck Trochanter Femur Radius Total body
Load
Stress
0.80 0.64b 0.57b 0.63b 0.63b 0.72
0.72 0.58a 0.53b 0.55a 0.53b 0.65
Spinal QCT 2–4: for load, CNT (mg/cm); for stress, bone density (mg/cm3) Trabecular density (g/cm3) L2–4 0.67a Cortical density (g/cm3) L2–4 0.80 Mean (trab/cort) L2–4 0.78 Peripheral QCT: for load, CNT (mg/mm); for stress, bone density (mg/cm3) Radius 4% 0.57b 20% 0.66b Tibia Distal 0.64b Shaft 0.63b Femur Distal 0.52b Shaft 0.64b Calcaneal ultrasound SOS (m/sec) BUA (dB/MHz) Stiffness index
0.47b 0.53b 0.52b
0.64 0.73 0.73
0.37b 0.41b 0.52b 0.40b 0.56a 0.42b 0.47b 0.52b 0.52b
Abbreviations identical to those in Tables 1 and 2. All correlations statistically significant (p ⬍ 0.01). a Correlation significantly lower (p ⬍ 0.05) than that with in situ DXA of L-3. b Correlation significantly lower (p ⬍ 0.01) than that with in situ DXA of L-3.
QCT with mechanical failure vs. DXA. However, the major source of error with single energy techniques is the bone marrow, which was preserved in our study. It is also worth noting that ex situ AP DXA was not able to better predict mechanical failure than in situ AP DXA,7 suggesting that the comparison between DXA and QCT was not biased by the in situ vs. ex situ measurement conditions. Cortical and trabecular density (QCT, multiplied by the cross-sectional area) were both able to predict about 60% of the variability in failure loads in the lumbar spine, and their combination improved the prediction to 67% (vs. 55% for DXA). The high predictive ability of cortical density (QCT) is surprising, particularly because the cortical shell cannot be delineated accurately by QCT.40 Our results differ from those of Cheng et al.9 who found no significant correlation of cortical density with failure strength, although they used a different scanner and software and their specimens were considerably younger (average 68 years). It may be that only with a substantial loss of trabecular bone at an older age will the cortical/subcortical compartment become a relevant prerdictor of vertebral failure. This interpretation is in agreement with Rockoff et al.,41 who reported an important change in vertebral strength when removing the cortex, but it is in contrast with McBroom et al.,32 who observed only a small effect after removing the cortical shell. In testing planoparallel sections of vertebral bodies (L-3), Ebbesen et al.13 found a higher correlation between mechanical failure and site-specific densitometry (r2 ⫽ 0.75– 0.86). In contrast, we have tested functional spinal units, including the intervertebral disks and endplates. Although this test adds variability to the mechanical data, it has the distinct advantage that the mode of clinical failure is simulated more realistically. When predicting failure in the thoracic spine, trabecular density (QCT) displayed a significantly lower association (r2 ⫽
32%) than cortical density (QCT) or in situ spinal DXA (r2 ⫽ 50%–55%), and the combination of trabecular and cortical density (QCT) was unable to improve the prediction vs. cortical density alone. It has been demonstrated previously by histomorphometry that trabecular bone density displays considerable heterogeneity between adjacent vertebrae.3 A higher consistency of cortical properties throughout the spine may explain the higher predictive ability of cortical density (QCT) or spinal DXA (including the cortical posterior elements) in predicting thoracic failure strength. This may also explain why DXA is relatively successful in predicting vertebral fracture risk clinically, despite its inability to separate cortical from trabecular bone, and despite inclusion of the posterior elements. Non-site-specific DXA, and pQCT of the radius, tibia, and femur displayed a lower correlation with vertebral failure load (r2 approximately 40%) than lumbar QCT or DXA, and this also applied for mechanical tests in the thoracic spine. This is in contrast to the experimental results of Bjanarson et al.,5 who reported a higher correlation of in situ femoral vs. in situ lumbar DXA with spinal failure. It also contradicts the theoretical considerations of Hassager et al.24 who reasoned that vertebral failure can be more accurately predicted from radial DXA, because soft tissue errors are smaller at this site. It has also been suggested that measurement of bone properties in the lower limb are advantageous, because it is subjected to normal weightbearing, whereas the radius is not. We found, however, no difference between non-site-specific DXA and pQCT sites at the lower extremity and distal radius. As a clinical consequence for peripheral measurement techniques, there appears to be no advantage in measuring the lower vs. the upper limb. QUS parameters are shown to predict only approximately 25% of the variability in vertebral strength, which is somewhat
Bone Vol. 31, No. 1 July 2002:77– 84
more than reported in two previous studies.9,28 In multiple regression models we found QUS not to add significant, independent information to spinal DXA, but to slightly improve the prediction of failure load and stress by QCT. Supposedly, bone properties throughout the skeleton are sufficiently heterogeneous, so that measurement at another site cannot provide relevant significant information in addition to site-specific (vertebral) bone mineral content. We have, however, no explanation as to why QUS contributes to strength prediction in combination with QCT, but not with DXA. Conclusions Our results show that for predicting mechanical strength of the thoracolumbar vertebrae: 1. QCT is not superior to AP in situ spinal DXA, although it eliminates artifacts from the posterior elements and soft tissues and also permits assessment of trabecular and cortical properties separately. This is despite the fact that the individuals were relatively old and that no effort was undertaken to exclude subjects with osteoarthritic changes. 2. Non-site-specific (peripheral) measurements are inferior to spinal DXA and QCT, and also for predicting strength in the thoracic spine. This is most likely due to the smaller degree of skeletal heterogeneity within the spine, than between the spine and other sites. 3. Assessment by pQCT at the weight-bearing lower limb does not have an advantage over that at the distal radius. 4. Calcaneal QUS cannot add significant, independent information to site-specific bone status. These experimental results clearly favor the site-specific assessment of vertebral strength in elderly individuals over that by non-site-specific or peripheral measurements.
Acknowledgments: The authors thank Jan Grimm (Medizinische Physik, Klinik fu¨ r Diagnostische Radiologie, Kiel, Germany) for reading the spinal X-rays, Gu¨ nther Delling and coworkers (Abteilung Osteopathologie, Universita¨ tsklinikum Eppendorf, Hamburg, Germany) for the histomorphometric analyses, and Rainer Barkmann (Medizinische Physik, Klinik fu¨ r Diagnostische Radiologie, Kiel, Germany) for helpful discussion. Gudrun Goldmann, Nadine Krefting, Boriana Barth, Oliver Groll, and Hans-Ju¨ rgen Becker (Musculoskeletal Research Group, Institute of Anatomy, Mu¨ nchen) are acknowledged for their help with the radiographic, DXA, QCT, pQCT, and QUS examinations. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG LO 730/2-1).
References 1. Albright, F., Smith, P. H., and Richardson, A. M. Postmenopausal osteoporosis. JAMA 116:2465–2474; 1941. 2. Amling, M., Herden, S., Po¨ sl, M., Hahn, M., Ritzel, H., and Delling, G. Heterogeneity of the skeleton: Comparison of the trabecular microarchitecture of the spine, the iliac crest, the femur, and the calcaneus. J Bone Miner Res 1:36 –45; 1996. 3. Amling, M., Posl, M., Ritzel, H., Hahn, M., Vogel, M., Wening, V. J., and Delling, G. Architecture and distribution of cancellous bone yield vertebral fracture clues. A histomorphometric analysis of the complete spinal column from 40 autopsy specimens. Arch Orthop Trauma Surg 115:262–269; 1996. 4. Biggemann, M., Hilweg, D., and Brinckmann, P. Prediction of the compressive strength of vertebral bodies of the lumbar spine by quantitative computed tomography. Skel Radiol 17:264 –269; 1988. 5. Bjarnason, K., Hassager, C., Svendsen, O. L., Stang, H., and Christiansen, C. Anteroposterior and lateral spinal DXA for the assessment of vertebral body strength: Comparison with hip and forearm measurement. Osteopor Int 6:37– 42; 1996.
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
83
6. Brinckmann, P., Biggemann, M., and Hilweg, D. Prediction of the compressive strength of human lumbar vertebrae. Spine 14:606 –610; 1989. 7. Bu¨ rklein, D., Lochmu¨ ller, E.-M., Kuhn, V., Grimm, J., Barkmann, R., Mu¨ ller, R., and Eckstein, F. Correlation of thoracic and lumbar vertebral failure loads and prediction by in-situ vs. ex-situ dual X-ray absorptiometry. J Biomech 35:579 –587; 2001. 8. Calabrisi, P. and Smith, F. C. The effects of embalming on the compressive strength of a few specimens of compact human bone. Naval Medical Research Institute Memorandum Report 51-2 NM 000 018.07.02; 1952. 9. Cheng, X. G., Nicholson, P. H. F., Boonen, S., Lowet, G., Brys, P., Aerssens, J., Vander Perre, G., and Dequeker, J. Prediction of vertebral strength in vitro by spinal bone densitometry and calcaneal ultrasound. J Bone Miner Res 12:1721–1728; 1997. 10. Cody, D. D., Goldstein, S. A., Flynn, M. J., and Brown, E. B. Correlations between vertebral regional bone mineral density (rBMD) and whole bone fracture load. Spine 16:146 –154; 1991. 11. Cooper, C., Atkinson, J., O’Fallon, W. M., and Melton, J. L. Incidence of clinically diagnosed vertebral fractures: A population-based study in Rochester, Minnesota, 1985–1989. J Bone Miner Res 2:221–227; 1992. 12. De Smet, A. A., Robinson, R. G., Johnson, B. E., and Lukert, B. P. Spinal compression fractures in osteoporotic women: Patterns and relationship to hyperkyphosis. Radiology 166:497–500; 1988. 13. Ebbesen, E. N., Thomsen, J. S., Beck Nielsen, H., Nepper Rasmussen, H. J., and Mosekilde, L. Lumbar vertebral body compressive strength evaluated by dual energy X ray absorptiometry, quantitative computed tomography, and ashing. Bone 25:713–724; 1999. 14. Edmondston, S. J., Singer, T. P., Day, R. E., Breidahl, P. D., and Price, R. I. Formalin fixation effects on vertebral bone density and failure mechanics: An in-vitro study of human and sheep vertebrae. Clin Biomech 9:175–179; 1994. 15. Genant, H. K. and Osteoporosis Research Group. Guidelines for Quality Assurance in Radiographic Examinations of the Spine. San Francisco, CA: Regents of the University of California; 1996. 16. Genant, H. K., Engelke, K., Fuerst, T., Glu¨ er, C. C., Grampp, S., Harris, S. T., Jergas, M., Lang, T., Lu, Y., Majumdar, S., Mathur, A., and Takada, M. Noninvasive assessment of bone mineral and structure: State of the art. J Bone Miner Res 11:707–730; 1996. 17. Glu¨ er, C. C., Blake, G., Lu, Y., Blunt, B. A., Jergas, M., and Genant, H. K. Accurate assessment of precision errors: How to measure the reproducibility of bone densitometry techniques. Osteopor Int 5:262–270; 1995. 18. Grampp, S., Genant, H. K., Mathur, A., Lang, P., Jergas, M., Takada, M., Glu¨ er, C. C., Lu, Y., and Chavez, M. Comparisons of noninvasive bone mineral measurements in assessing age related loss, fracture discrimination, and diagnostic classification. J Bone Miner Res 12:697–711; 1997. 19. Greenberg, S. W., Gonzalez, D., Gurdjian, E. S., and Thomas, L. M. Changes in Physical Properties of Bone Between the In Vivo, Freshly Dead and Embalmed Condition. Society of New York: Automotive Engineers; 1968. 20. Groll, O., Lochmu¨ ller, E.-M., Bachmeier, M., Willnecker, J., and Eckstein, F. Precision and intersite correlation of bone densitometry at the radius, tibia and femur with peripheral quantitative CT. Skel Radiol 28:696 –702; 1999. 21. Ha¨ rma¨ , M., Heliovaara, M., Aromaa, A., and Knekt, P. Thoracic spine compression fractures in Finland. Clin Orthop 205:188 –194; 1986. 22. Hahn, M., Vogel, M., and Delling, G. Undecalcified preparation of bone tissue: Report of technical experience and development of new methods. Virchows Arch A Pathol Anat Histopathol 418:1–7; 1991. 23. Hans, D., Fuerst, T., and Duboeuf, F. Quantitative ultrasound bone measurement. Eur Radiol 7(Suppl. 2):S43–S50; 1997. 24. Hassager, C., Jensen, S. B., Gotfredsen, A., and Christiansen, C. The impact of measurement errors on the diagnostic value of bone mass measurements: Theoretical considerations. Osteopor Int 1:250 –256; 1991. 25. Hayes, W. C. and Bouxsein, M. L. Biomechanics of cortical and trabecular bone: Implications for assessment of fracture risk. Mow, V. C. and Hayes, W. C., eds. Basic Orthopaedic Biomechanics. 2nd Ed. Philadelphia: LippincottRaven; 1997. 26. Leidig-Bruckner, G., Minne, H. W., Schlaich, C., Wagner, G., Scheidt-Nave, C., Bruckner, T., Gebest, H. J., and Ziegler, R. Clinical grading of spinal osteoporosis: Quality of life components and spinal deformity in women with chronic low back pain and women with vertebral osteoporosis. J Bone Miner Res 12:663–675; 1997. 27. Liberman, U. A., Weiss, S. R., Broll, J., Minne, H. W., Quan, H., Bell, N. H., Rodriguez-Portales, J., Downs, R. W., Dequeker, J., and Favus, M. Effect of oral alendronate on bone mineral density and the incidence of fractures in postmenopausal osteoporosis. The Alendronate Phase III Osteoporosis Treatment Study Group. N Engl J Med 333:1437–1443; 1995.
84
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
28. Lochmu¨ ller, E.-M., Eckstein, F., Kaiser, D., Zeller, J. B., Landgraf, J., Putz, R., and Steldinger, R. Prediction of vertebral failure loads from spinal and femoral dual-energy X-ray absorptiometry, and calcaneal ultrasound: An in situ analysis with intact soft tissues. Bone 23:417–424; 1998. 29. Lochmu¨ ller, E.-M., Miller, P., Bu¨ rklein, D., Wehr, U., Rambeck, W., and Eckstein, F. In situ femoral DXA related to ash weight, bone size, and density, and its relationship with mechanical failure loads of the proximal femur. Osteopor Int 11:361–367; 2000. 30. Lochmu¨ ller, E.-M., Krefting, N., Bu¨ rklein, D., and Eckstein, F. Effect of fixation, soft-tissues and scan projection on bone mineral measurements with dual energy X-ray absorptiometry (DXA). Calcif Tissue Int 68:140 –145; 2001. 31. Lufkin, E. G., Wahner, H. W., O’Fallon, W. M., Hodgson, S. F., Kotowicz, M. A., Lane, A. W., Judd, H. L., Caplan, R. H., and Riggs, B. L. Treatment of postmenopausal osteoporosis with transdermal estrogen. Ann Intern Med 117:1–9; 1992. 32. McBroom, R. J., Hayes, W. C., Edwards, W. T., Goldberg, R. P., and White, A. A. Prediction of vertebral body compressive fracture using quantitative computed tomography. J Bone Jt Surg 67A:1206 –1214; 1985. 33. McElhaney, J., Fogle, J., Byars, E., and Weaver, G. Effect of embalming on the mechanical properties of beef bone. J Appl Physiol 19:1234 –1236; 1964. 34. Melton, L. J. III, Kan, S. H., Frye, M. A., Wahner, H. W., O’Fallon, W. M., and Riggs, B. L. Epidemiology of vertebral fractures in women. Am J Epidemiol 129:1000 –1011; 1989. 35. Melton, L. J., Thamer, N. F., Ray, J. K., Chesnut, C. H., Einhorn, T. A., Johnston, C. C., Raiz, L. G., Silverman, S. L., and Siris, E. S. Fractures attributable to osteoporosis: Report from the National Osteoporosis Foundation. J Bone Miner Res 12:16 –23; 1997. 36. Moro, M., Hecker, A. T., Bouxsein, M. L., and Myers, E. R. Failure load of thoracic vertebrae correlates with lumbar bone mineral density measured by DXA. Calcif Tissue Int 56:206 –209; 1995.
Bone Vol. 31, No. 1 July 2002:77– 84 37. Nicholson, P. H., Mu¨ ller, R., Cheng, X. G., Ruegsegger, P., Van Der Perre, G., Dequeker, J., and Boonen, S. Quantitative ultrasound and trabecular architecture in the human calcaneus. J Bone Miner Res 16:1886 –1892; 2001. 38. Po¨ pperl, G., Lochmu¨ ller, E.-M., Becker, H.-J., Mall, G., Steinlechner, M., and Eckstein, F. Ex-situ determination of calcaneal ultrasound properties: Reproducibility, effects of storage, formalin fixation, maceration, and changes in anatomic measurement site. Calcif Tissue Int 65:192–197; 1999. 39. Ray, N. F., Chan, J. K., Thamer, M., and Melton, L. J. Medical expenditures for the treatment of osteoporotic fractures in the United States in 1995: Report from the National Osteoporosis Foundation. J Bone Miner Res 12:24 –35; 1997. 40. Ritzel, H., Amling, M., Po¨ sl, M., Hahn, M., and Delling, G. The thickness of human vertebral cortical bone and its changes in aging and osteoporosis: A histomorphometric analysis in the complete spinal column from thirty-seven autopsy specimens. J Bone Miner Res 12:89 –95; 1997. 41. Rockoff, S. D., Sweet, E., and Bleustein, J. The relative contribution of trabecular and cortical bone to the strength of human lumbar vertebrae. Calcif Tissue Res 3:163–175; 1969. 42. Svendsen, O. L., Hassager, C., Skodt, V., and Christiansen, C. Impact of soft tissue on in vivo accuracy of bone mineral measurements in the spine, hip, and forearm: A human cadaver study. J Bone Miner Res 10:868 –873; 1995.
Date Received: May 17, 2001 Date Revised: December 21, 2001 Date Accepted: February 14, 2002
Mechanical Strength of the Thoracolumbar Spine in the Elderly: Prediction From In Situ Dual-energy X-ray Absorptiometry, Quantitative Computed Tomography (QCT), Upper and Lower Limb Peripheral QCT, and Quantitative Ultrasound ¨ LLER,1 D. BU ¨ RKLEIN,1,2 V. KUHN,1,2 C. GLASER,3 R. MU ¨ LLER,4 C. C. GLU ¨ ER,5 and E.-M. LOCHMU 2 F. ECKSTEIN 1
Frauenklinik Innenstadt, Ludwig-Maximilians-Universita¨t (LMU), Mu¨nchen, Germany Musculoskeletal Research Group, Institute of Anatomy, LMU, Mu¨nchen, Germany 3 Institute for Clinical Radiology, LMU, Mu¨nchen, Germany 4 Orthopedic Biomechanics Laboratory, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, USA 5 Medical Physics, Diagnostic Radiology, Christian-Albrechts-Universita¨t, Kiel, Germany 2
Key Words: Vertebral failure; Mechanical strength; Spine; Biomechanical testing; Densitometry; Ultrasound.
The objective of this study was to compare the ability of clinically available densitometric measurement techniques for evaluating vertebral strength in elderly individuals. Measurements were related to experimentally determined failure strength in the thoracic and lumbar spine. In 127 specimens (82 women and 45 men, age 80 ⴞ 10 years), dual-energy X-ray absorptiometry (DXA) was performed at the lumbar spine, femur, radius, and total body, and peripheral-quantitative computed tomography (pQCT) at the distal radius, tibia, and femur under in situ conditions with intact soft tissues. Spinal QCT and calcaneal ultrasound parameters were performed ex situ in degassed specimens. Mechanical failure loads of thoracic vertebrae 6 and 10 (T-6 and -10), and lumbar vertebra 3 (L-3) were determined in axial compression on functional three-segment units. In situ anteroposterior DXA and QCT of the lumbar spine explained approximately 65% of the variability of thoracolumbar failure. A combination of cortical and trabecular density (QCT) provided the best prediction in the lumbar spine. However, this was not the case in the thoracic spine, for which lumbar cortical density (QCT) and DXA provided significantly better estimates than trabecular density (QCT). pQCT was significantly less correlated with the strength of lumbar and thoracic vertebrae (r2 ⴝ 40%), but was equivalent to femoral or radial DXA. pQCT measurements in the lower limb showed no advantage over those at the distal radius. Ultrasound explained approximately 25% of the variability of vertebral failure strength and added independent information to spinal QCT, but not to spinal DXA. These experimental results advocate site-specific assessment of vertebral strength by either spinal DXA or QCT. (Bone 31:77– 84; 2002) © 2002 by Elsevier Science Inc. All rights reserved.
Introduction Vertebral fractures represent the classic hallmark of osteoporosis1 with an annual incidence of approximately 500,000 patients in North America.11,34 These fractures substantially reduce the quality-of-life in the elderly,26 and the costs involved have been estimated to be 750 million $US per year in the US.39 Reductions in bone mass and strength have been identified as the main causes for these fractures in the elderly,25,35 with only a small proportion resulting from traumatic events.11 Because treatment is available to slow the age-related reduction of bone strength,27,31 accurate measurements of bone properties (as surrogates of vertebral fracture risk) are required to reliably identify patients that require therapeutic intervention. A wide range of densitometric techniques is currently available for determining bone properties at different skeletal sites.16,18 Dual-energy X-ray absorptiometry (DXA) is the method used most widely and has the advantage that measurements can be obtained at or close to the site of interest.16 A disadvantage is that DXA cannot discriminate between cortical and trabecular bone, and that the surrounding soft tissues may introduce relevant measurement errors.29,30,42 In addition, conventional anteroposterior (AP) measurements of the lumbar spine include the posterior elements, which are frequently subject to osteoarthritic changes, particularly in the elderly. Quantitative computed tomography (QCT) is less affected by surrounding soft tissue errors, eliminates the posterior elements, and permits analysis of trabecular and cortical bone separately. However, QCT involves relatively high X-ray doses and relies on expensive equipment.16 Peripheral QCT (pQCT) is less costly, uses less radiation, and X-rays are applied to less vulnerable (peripheral) sites. However, because of skeletal heterogeneity,2,3,20 it is questionable as to whether peripheral measurements reflect bone properties of the axial skeleton. A relevant relationship of ultrasonic properties has been reported with both trabecular micro-
Address for correspondence and reprints: Dr. E.-M. Lochmu¨ller, Universita¨tsfrauenklinik, Maistrasse 11, D-80337 Mu¨nchen, Germany. E-mail: [email protected] © 2002 by Elsevier Science Inc. All rights reserved.
77
8756-3282/02/$22.00 PII S8756-3282(02)00792-5
78
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
Bone Vol. 31, No. 1 July 2002:77– 84
Table 1. Descriptive statistics of the mechanical and densitometric data
Population data Body height (cm) Body weight (kg) Mechanical data Failure load (kN) T-6 T-10 L-3 Failure stress (MPa) T-6 T-10 L-3 Site-specific densitometric data DXA (in situ) BMD (g/cm2) at L-3 Spinal QCT (ex situ) Trab density (g/cm3) at L2–4 Cort density (g/cm3) at L2–4
Men
Women
Difference
170 ⫾ 8 63 ⫾ 13
155 ⫾ 7 52 ⫾ 12
⫺9%b ⫺17%b
3.19 ⫾ 1.48 4.21 ⫾ 1.85 4.09 ⫾ 1.69
2.06 ⫾ 0.87 2.57 ⫾ 1.09 2.45 ⫾ 1.02
⫺35%b ⫺39%b ⫺40%b
426 ⫾ 167 394 ⫾ 174 300 ⫾ 132
346 ⫾ 143 300 ⫾ 122 213 ⫾ 81.8
⫺19%a ⫺24%b ⫺29%b
1.17 ⫾ 0.25
0.96 ⫾ 0.25
⫺18%b
70.2 ⫾ 32.7 300 ⫾ 71.1
58.9 ⫾ 32.3 233 ⫾ 68.3
⫺16% n.s. ⫺23%b
KEY: DXA, dual-energy X-ray absorptiometry; BMD, bone mineral density; QCT, quantitative computed tomography; trab, trabecular; cort, cortical; n.s., difference not statistically significant. a p ⬍ 0.01. b p ⬍ 0.001 (unpaired Student’s t-test).
structure and mechanical properties.23,37 This has stimulated hope that quantitative ultrasound (QUS) can provide significant information for predicting bone strength, in addition to bone mass or density. The attractiveness of calcaneal QUS also lies in the complete lack of ionizing radiation and in the relatively low costs involved. The objective of the current study was therefore to determine which measurement technique in which region is best suited for evaluating vertebral strength in elderly individuals. Densitometric measurements were related to mechanical strength of the thoracic and lumbar spine. Previous experimental studies have examined bones ex situ (except for two studies5,28). This may have biased the results in favor of DXA, because of the lack of soft tissue artifacts. Moreover, experimental studies have generally investigated the relationship of bone densitometry and mechanical failure at the lumbar spine, compared with calcaneal QUS.9,28 However, most clinical fractures occur in the thoracic spine.12,21 Studies may thus have overestimated the value of spinal DXA and QCT in predicting vertebral fracture risk, given the heterogeneity of bone properties within the spine.3 No previous experimental study has compared in situ spinal DXA with QCT, and with non-site-specific techniques, including QUS and pQCT. The following specific questions were addressed: 1. Does QCT show a higher correlation with mechanical strength of vertebrae than spinal (anteroposterior) in situ DXA, because it eliminates artifacts from posterior elements and soft tissues, and because it permits to assess trabecular and cortical properties separately? 2. Do lumbar DXA and QCT display a higher correlation with vertebral strength than non-site-specific (peripheral) measurements (DXA, pQCT, QUS), even when mechanical testing is performed in the thoracic spine? 3. Does pQCT at the lower limb display a higher correlation with vertebral strength than pQCT at the upper limb, because of the involvement of the lower limb in weight-bearing? 4. Is calcaneal QUS capable of adding relevant independent information to site-specific measurements of bone status in
multiple regression models, because of its relationship with trabecular microstructure and mechanical properties? Materials and Methods Study Sample The study sample initially comprised 140 formalin-fixed cadavers with intact skin and soft tissues from a course of macroscopic dissection. The only criterion of inclusion was a testamentary decree to the institute several years prior to death. The sample can therefore be assumed to be a representative selection of the elderly resident population of Bavaria. Bone biopsies were harvested from the iliac crest for routine histomorphometric analysis.22 Specimens with bone disease other than osteopenia or osteoporosis were excluded (three with malignancy and seven with renal osteopathy). One specimen was discarded due to an osteolysis on the QCT scans, and two due to fracture of lumbar vertebrae 2– 4 (L-2– 4). Of the remaining 127 specimens 82 were women (age 82 ⫾ 9 years) and 45 were men (age 77 ⫾ 11 years). The age range was 47–98 years. The men were significantly larger and heavier than the women (Table 1). Bone Densitometry DXA (DPX-L, Lunar, Madison, WI) was used to determine the bone mineral content (BMC) and areal bone mineral density (BMD, in grams per centimeter) under in situ conditions (with intact skin and soft tissues), analyzing the total skeleton, lumbar vertebra 3 (L-3) in the anteroposterior (AP) direction, total proximal femur, femoral neck, greater trochanter, and distal radius. In cases where L-3 was fractured (nine cases), spinal DXA (but not the other) measurements were discarded. We have shown previously that DXA is as reproducible in cadavers as it is in vivo, and that values are not affected by 10 months of formalin fixation.30 pQCT was also performed in situ. An XCT 2000 scanner was used at the distal radius (4% and 20% from the wrist), and an
Bone Vol. 31, No. 1 July 2002:77– 84
XCT 3000 (Stratec Medizintechnik, Pforzheim, Germany) at the tibia (8% and 50% from the ankle) and distal femur (8% and 30% from the knee). Technical details and measurement precision have been described previously.20 QCT was not performed in situ, because it was not possible to transport the cadavers to the clinical CT unit. QUS measurements were also obtained ex situ, because, due to supination contractions, the feet could not be brought into the required position in the scanner, and also because there was concern that gas bubbles in the soft tissue may alter the measurements. After the dissection course, the thoracolumbar spines were excised and radiographed15 using a Polyphos 30 M X-ray system (Siemens, Erlangen, Germany) and SR-H film (35 ⫻ 42 cm, Konica, Hohenbrunn, Germany). Four films were obtained in each specimen (two in AP and two in lateral projection), one focusing on the thoracic and one on the lumbar spine. In vertebrae with deformity suggestive of a fracture (11 cases in L-3, 7 cases in thoracic vertebra 10 [T-10], and 12 cases in T-6) the densitometric and biomechanical data of these specific vertebrae were discarded. Specimens with osteoarthritic changes were not excluded. Single-energy QCT was obtained ex situ in degassed specimens, then sealed within water in thin polyethylene bags. We used a Somatom Plus 4 CT scanner at 80 kV (Siemens) applying standard settings of the manufacturer (195 mA, rotation time 0.75 sec, kernel SP90, slice thickness 10 mm, field of view 216 mm, matrix 512 pixels). A phantom with equivalents of 0 and 200 mg/cm3 hydroxylapatite was used to convert Hounsfield units into density values. Images were obtained at the midvertebral level of L-2, L-3, and L-4, with average values of L2– 4 being used in further analyses. Trabecular and cortical density were determined with the software. As the cortical shell is only several hundred microns thick,40 an apparent cortical/subcortical density (but not a true cortical density) was determined. The root-mean-square (RMS) average coefficient of variation (CV%)17 of repeated measurements (four repetitions in 14 specimens with repositioning, obtained on different days) was 2.9% for trabecular, and 3.1% for cortical density. In addition, the cross-sectional areas of the vertebral bodies were determined at the midvertebral level, with the precision (RMS average CV%) being 1.2%. To assess whether a combination of trabecular and cortical properties could improve the prediction of vertebral strength, we calculated the individual percentage deviation from the mean value (of all individuals) for both parameters. The mean values of these percent deviations were then used as a combined measure. QUS of the calcaneus was obtained under ex situ conditions. The calcanei were dissected clean of the surrounding tissues and degassed. Measurements were performed with an Achilles Plus scanner (Lunar) sealed for underwater use. The calcanei were measured in a temperature-controlled water bath (37°C) in a defined position.38 The precision for speed-of-sound (SOS), broadband ultrasound attenuation (BUA), and stiffness index (SI)23 was found to be in the range of in vivo measurements.38 Although a prolonged period of formalin fixation (⬎1 year) was shown to cause a decrease of SOS and SI, the values were highly correlated with those before fixation.38 Please note that the so-called stiffness index in QUS does not reflect stiffness in mechanical terms. Mechanical Testing Three segments of the thoracolumbar spine (T5–7, T9 –11, L2– 4) were tested as functional units with ligaments and inter-
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
79
vertebral disks, but without posterior elements.7,10,36,41 The upper and lower vertebral bodies were embedded planoparallel to the endplates of the central target vertebra, and segments were compressed axially, as described elsewhere.7 To compute the failure stress (MPa), the cross-sectional area was estimated for each vertebra by multiplying the anteroposterior and mediolateral diameter (measured with calipers at the midvertebral level), and by a scaling factor. The factor was derived from a comparison between the actual cross-sectional areas of the vertebrae (determined from prints in 65 other specimens) and the two diameters. A scaling factor of 0.86 produced no systematic difference between measured and computed cross-sectional area, with r ⫽ 0.96 between both methods. Statistics To determine average failure load and stress of the spine from the three segments, the deviation from the mean (of all individuals) was derived (in percent) for each individual for each vertebral level (T-6, T-10, and L-3). The average (without data from fractured vertebrae) was then computed as a representative value for the thoracolumbar spine. The association between failure loads and stress with the densitometric data was assessed by linear regression analysis, using STATVIEW 4.5 software (Abacus Concepts, Berkeley, CA). With DXA, the bone mineral content (BMC; grams) was used for correlation with failure load, and the bone mineral density (BMD; grams per square centimeter) for that with failure stress. With QCT, the density values (milligrams per cubic centimeter) were used for the correlation with failure stress, and the density multiplied by the mean cross-sectional areas of the vertebral bodies for correlation with failure loads.4,6 With pQCT, the total bone mineral content (milligrams) was used for correlation with failure loads and total density (milligrams per cubic centimeter) for that with failure stress. Fisher’s Z-test was used to test whether the correlations were significantly different from site-specific DXA. To determine whether QUS measurements at peripheral sites could add significant, independent information, we used a stepwise multiple regression analysis (forward mode). Results Men displayed a significantly higher failure load and stress (load per unit area) than women at all spinal levels (Table 1). This was also the case after adjusting for age. The correlation of failure stress between T-6 and T-10 (r ⫽ 0.78) was significantly higher (p ⬍ 0.01) than that between T-6 and L-3 (r ⫽ 0.60) and that between T-10 and L-3 (r ⫽ 0.62). The correlation of mechanical strength in the lumbar spine (L-3) with densitometry is given in Table 2. The highest correlation with failure load was observed for cortical density (QCT), multiplied by the cross-sectional area (r ⫽ 0.78; standard error of the estimate [SEE] 31.1%), but the coefficients for trabecular density (QCT: r ⫽ 0.75; SEE ⫽ 33.0%) and BMC of L-3 (in situ DXA: r ⫽ 0.73; SEE ⫽ 34.2%) were not significantly different. A combined measure of trabecular and cortical density (QCT) provided a significantly (p ⬍ 0.05) better prediction of failure load (r ⫽ 0.82; SEE ⫽ 28.8%) than spinal DXA (p ⬍ 0.05). DXA at other sites, and pQCT provided significantly lower (p ⬍ 0.05– 0.01) correlations (r ⫽ 0.55– 0.62). Coefficients for pQCT at the lower limb were not significantly different from those of the distal radius (Table 2). The correlation of QUS (r ⫽ 0.48; SEE ⫽ 43.7%) was significantly lower (p ⬍ 0.01) than spinal DXA and QCT, but not significantly different from non-sitespecific DXA or pQCT. The best predictor of L-3 failure stress was cortical density (QCT: r ⫽ 0.76; SEE ⫽ 30.0%). Trabecular
80
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
Bone Vol. 31, No. 1 July 2002:77– 84
Table 2. Correlation between failure load and failure stress at the lumbar spine (L-3) with densitometric data (N ⫽ 116)
DXA (in situ): for load, BMC (g); for stress, BMD (g/cm2) L-3 Neck Trochanter Femur Radius Total body
Load
Stress
0.73a 0.61a 0.55a 0.62a 0.59a 0.69
0.68 0.53a 0.49a 0.53a 0.49a 0.60
Spinal QCT 2–4: for load, CNT (mg/cm); for stress, bone density (mg/cm3) Trabecular L2–4 0.75 Cortical L2–4 0.78 Combination L2–4 0.82c Peripheral QCT: for load, CNT (mg/mm); for stress, bone density (mg/cm3) Radius 4% 0.53b 20% 0.62a Tibia Distal 0.57a Shaft 0.60a Femur Distal 0.53b Shaft 0.60a Calcaneal ultrasound SOS (m/sec) BUA (dB/MHz) Stiffness index
0.45b 0.47b 0.48b
0.67 0.76 0.76
0.35b 0.35b 0.49a 0.41b 0.51a 0.39b 0.44b 0.44b 0.46b
KEY: Cnt, bone mineral content; SOS, speed of sound; BUA, broadband ultrasound attenuation. Other abbreviations identical to those in Table 1. All correlations statistically significant (p ⬍ 0.01). a Correlation significantly (p ⬍ 0.05) lower than that with in situ DXA of L-3. b Correlation significantly (p ⬍ 0.01) lower than that with in situ DXA of L-3. c Correlation significantly (p ⬍ 0.05) higher than that with in situ DXA of L-3.
density (QCT: r ⫽ 0.67; SEE ⫽ 33.5%) and spinal DXA (r ⫽ 0.68; SEE ⫽ 33.4%) were, however, not significantly different. Both non-site-specific DXA and pQCT displayed a lower (p ⬍ 0.05– 0.01) correlation with failure stress (r ⫽ 0.35– 0.53). QUS (r ⫽ 0.46; SEE ⫽ 39.7%) was not significantly different from non-site-specific DXA or pQCT. The correlation of mechanical strength in the thoracic spine (T-10) with densitometry is given in Table 3. The coefficients obtained for DXA, pQCT, and QUS were not significantly different from those with L-3. Lumbar DXA displayed an identical correlation with failure load when comparing T-10 and L-3 (r ⫽ 0.73; SEE ⫽ 34.2%). However, trabecular density (QCT) displayed a significantly lower correlation with T-10 failure load (p ⬍ 0.01) and stress (p ⬍ 0.05) than with L-3 failure load and stress. The cortical density also showed a lower correlation, but the difference was only significant for the failure stress (p ⬍ 0.05). Spinal DXA and cortical density (QCT) provided similar estimates of thoracic strength, but trabecular density displayed a lower correlation (p ⬍ 0.05). The combination of cortical and trabecular density did not improve the correlation over that of cortical density alone. Non-site-specific DXA, pQCT, and QUS displayed lower correlations (p ⬍ 0.05– 0.01) than site-specific densitometry. The correlation of densitometric variables with average failure strength of the thoracolumbar spine is shown in Table 4. Cortical density (QCT) and in situ spinal DXA (Figure 1a,b) provided the best estimate of vertebral failure loads (r ⫽ 0.80) and stress (0.73/0.72). Trabecular density (QCT) displayed a lower correlation with the difference being significant (p ⬍ 0.05) for failure load. A combination of cortical and trabecular density
(QCT) did not improve the prediction beyond that of the cortical density alone. Non-site-specific DXA and pQCT displayed significantly (p ⬍ 0.01) lower correlations with thoracolumbar failure load and stress than spinal DXA and QCT, except for total body DXA. pQCT at the lower limb (r ⫽ 0.52– 0.64) did not offer an advantage over the distal radius (r ⫽ 0.57– 0.66; Figure 1c). The correlation of QUS parameters with strength (Figure 1d) was lower (p ⬍ 0.01) than for spinal DXA and QCT, but not significantly lower than for non-site-specific DXA or pQCT. None of the QUS parameters added significant, independent information to spinal DXA. However, BUA and stiffness index improved the prediction of failure stress in combination with femoral or radial DXA, and the prediction of failure load in combination with radial (but not femoral) DXA. In combination with trabecular or cortical density (QCT), all QUS parameters improved the prediction of failure load and stress in multiple regression models, but the gain in correlation was relatively small. Discussion In this study we sought to determine which measurement technique in which region is best suited for evaluating vertebral strength in elderly individuals. A clinical study design would have been problematic in this context, because of the cumulative X-ray exposure involved in multiple measurements, and because of the difficulty in objectively diagnosing the incidence of vertebral fractures, which often pass unnoticed. A large number of individuals and very long observation periods are required in prospective clinical studies to obtain statistically meaningful data.
Bone Vol. 31, No. 1 July 2002:77– 84
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
81
Table 3. Correlation between failure load and failure stress in the thoracic spine (T-10) with densitometric data (N ⫽ 120)
DXA (in situ): for load, BMC (g); for stress, BMD (g/cm2) L-3 Neck Trochanter Femur Radius Total body
Load
Stress
0.73 0.56a 0.51b 0.55b 0.57a 0.65
0.62 0.46a 0.45a 0.44a 0.45a 0.55
Spinal QCT 2–4: for load, CNT (mg/cm); for stress, bone density (mg/cm3) Trab density (g/cm3) L2–4 0.57a Cort density (g/cm3) L2–4 0.71 Combination L2–4 0.68 Peripheral QCT: for load, CNT (mg/mm); for stress, bone density (mg/cm3) Radius 4% 0.52b 20% 0.61a Tibia Distal 0.58a Shaft 0.56a Femur Distal 0.42b Shaft 0.58a Calcaneal ultrasound SOS (m/sec) BUA (dB/MHz) Stiffness index
0.38b 0.47b 0.45b
0.53 0.62 0.61
0.33b 0.34b 0.45a 0.31b 0.48a 0.37b 0.37b 0.44a 0.43a
Abbreviations identical to those in Tables 1 and 2; all correlations statistically significant (p ⬍ 0.01). a Correlation significantly lower (p ⬍ 0.05) than that with in situ DXA of L-3. b Correlation significantly lower (p ⬍ 0.01) than that with in situ DXA of L-3.
A strength of the study is that failure was not only determined in the lumbar, but also in the thoracic spine, that DXA was performed under in situ conditions, and that non-site-specific techniques were included, such as radial and femoral DXA, pQCT of the upper and lower limb, and calcaneal QUS. In particular, a large sample size permits determination of whether certain techniques perform significantly better than others. A potential limitation of this study is the use of fixed specimens. Calabrisi et al.8 and McElhaney et al.33 observed a small decrease, but Greenberg et al.19 and Edmondston et al.14 noted an increase in bone strength with fixation. Edmondston et al.14 showed, however, that fixation does not alter the correlation between vertebral bone mass and strength. We have previously demonstrated that long-term fixation has no significant effect on in situ DXA,30 and a predictable effect on QUS.38 Our data are in good agreement with those reported on failure loads and their correlations with DXA, QCT, and calcaneal QUS in fresh specimens (e.g., see Cheng et al.9). We therefore believe that the results of this study are not biased by the use of fixed specimens. Another limitation is that no lateral DXA of the spine could be performed under in situ conditions. However, lateral DXA was obtained ex situ in a subset of specimens.7 These measurements displayed a significantly (p ⬍ 0.05) higher correlation with failure loads of L-3 than in situ AP DXA (r ⫽ 0.85 vs. 0.73), but not with T-10 (r ⫽ 0.69 vs. 0.73) or T-6 (0.61 vs. 0.68). Because ex situ lateral DXA did not improve the prediction of thoracic failure, we assume that in situ lateral DXA is even less capable of improving the prediction of failure strength of the thoracic spine over AP scans. For logistical reasons, QCT could only be examined under ex situ conditions. It cannot be ruled out entirely that the lack of surrounding soft tissue spuriously improved the correlation of
Figure 1. Correlation of thoracolumbar failure loads with densitometric measurements. The regression plots show failure loads throughout the thoracic spine (in percent deviation from mean values) as correlated with: (a) bone mineral content (BMC) of lumbar vertebra 3 (L-3), measured using in situ DXA; (b) cortical density (multiplied by cross-sectional area) of lumbar vertebrae 2– 4 (L2– 4), measured using spinal QCT; (c) total bone mineral content of the distal radius (20% measurement site), measured using in situ pQCT; and (d) broadband ultrasound absorption (BUA) of the calcaneus, measured using ex situ quantitative ultrasound.
82
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
Bone Vol. 31, No. 1 July 2002:77– 84
Table 4. Correlation between average failure load and failure stress throughout the thoracolumbar spine (T-6, T-10, L-3) with densitometric data (N ⫽ 127)
DXA (in situ): for load, BMC (g); for stress, BMD (g/cm2) L-3 Neck Trochanter Femur Radius Total body
Load
Stress
0.80 0.64b 0.57b 0.63b 0.63b 0.72
0.72 0.58a 0.53b 0.55a 0.53b 0.65
Spinal QCT 2–4: for load, CNT (mg/cm); for stress, bone density (mg/cm3) Trabecular density (g/cm3) L2–4 0.67a Cortical density (g/cm3) L2–4 0.80 Mean (trab/cort) L2–4 0.78 Peripheral QCT: for load, CNT (mg/mm); for stress, bone density (mg/cm3) Radius 4% 0.57b 20% 0.66b Tibia Distal 0.64b Shaft 0.63b Femur Distal 0.52b Shaft 0.64b Calcaneal ultrasound SOS (m/sec) BUA (dB/MHz) Stiffness index
0.47b 0.53b 0.52b
0.64 0.73 0.73
0.37b 0.41b 0.52b 0.40b 0.56a 0.42b 0.47b 0.52b 0.52b
Abbreviations identical to those in Tables 1 and 2. All correlations statistically significant (p ⬍ 0.01). a Correlation significantly lower (p ⬍ 0.05) than that with in situ DXA of L-3. b Correlation significantly lower (p ⬍ 0.01) than that with in situ DXA of L-3.
QCT with mechanical failure vs. DXA. However, the major source of error with single energy techniques is the bone marrow, which was preserved in our study. It is also worth noting that ex situ AP DXA was not able to better predict mechanical failure than in situ AP DXA,7 suggesting that the comparison between DXA and QCT was not biased by the in situ vs. ex situ measurement conditions. Cortical and trabecular density (QCT, multiplied by the cross-sectional area) were both able to predict about 60% of the variability in failure loads in the lumbar spine, and their combination improved the prediction to 67% (vs. 55% for DXA). The high predictive ability of cortical density (QCT) is surprising, particularly because the cortical shell cannot be delineated accurately by QCT.40 Our results differ from those of Cheng et al.9 who found no significant correlation of cortical density with failure strength, although they used a different scanner and software and their specimens were considerably younger (average 68 years). It may be that only with a substantial loss of trabecular bone at an older age will the cortical/subcortical compartment become a relevant prerdictor of vertebral failure. This interpretation is in agreement with Rockoff et al.,41 who reported an important change in vertebral strength when removing the cortex, but it is in contrast with McBroom et al.,32 who observed only a small effect after removing the cortical shell. In testing planoparallel sections of vertebral bodies (L-3), Ebbesen et al.13 found a higher correlation between mechanical failure and site-specific densitometry (r2 ⫽ 0.75– 0.86). In contrast, we have tested functional spinal units, including the intervertebral disks and endplates. Although this test adds variability to the mechanical data, it has the distinct advantage that the mode of clinical failure is simulated more realistically. When predicting failure in the thoracic spine, trabecular density (QCT) displayed a significantly lower association (r2 ⫽
32%) than cortical density (QCT) or in situ spinal DXA (r2 ⫽ 50%–55%), and the combination of trabecular and cortical density (QCT) was unable to improve the prediction vs. cortical density alone. It has been demonstrated previously by histomorphometry that trabecular bone density displays considerable heterogeneity between adjacent vertebrae.3 A higher consistency of cortical properties throughout the spine may explain the higher predictive ability of cortical density (QCT) or spinal DXA (including the cortical posterior elements) in predicting thoracic failure strength. This may also explain why DXA is relatively successful in predicting vertebral fracture risk clinically, despite its inability to separate cortical from trabecular bone, and despite inclusion of the posterior elements. Non-site-specific DXA, and pQCT of the radius, tibia, and femur displayed a lower correlation with vertebral failure load (r2 approximately 40%) than lumbar QCT or DXA, and this also applied for mechanical tests in the thoracic spine. This is in contrast to the experimental results of Bjanarson et al.,5 who reported a higher correlation of in situ femoral vs. in situ lumbar DXA with spinal failure. It also contradicts the theoretical considerations of Hassager et al.24 who reasoned that vertebral failure can be more accurately predicted from radial DXA, because soft tissue errors are smaller at this site. It has also been suggested that measurement of bone properties in the lower limb are advantageous, because it is subjected to normal weightbearing, whereas the radius is not. We found, however, no difference between non-site-specific DXA and pQCT sites at the lower extremity and distal radius. As a clinical consequence for peripheral measurement techniques, there appears to be no advantage in measuring the lower vs. the upper limb. QUS parameters are shown to predict only approximately 25% of the variability in vertebral strength, which is somewhat
Bone Vol. 31, No. 1 July 2002:77– 84
more than reported in two previous studies.9,28 In multiple regression models we found QUS not to add significant, independent information to spinal DXA, but to slightly improve the prediction of failure load and stress by QCT. Supposedly, bone properties throughout the skeleton are sufficiently heterogeneous, so that measurement at another site cannot provide relevant significant information in addition to site-specific (vertebral) bone mineral content. We have, however, no explanation as to why QUS contributes to strength prediction in combination with QCT, but not with DXA. Conclusions Our results show that for predicting mechanical strength of the thoracolumbar vertebrae: 1. QCT is not superior to AP in situ spinal DXA, although it eliminates artifacts from the posterior elements and soft tissues and also permits assessment of trabecular and cortical properties separately. This is despite the fact that the individuals were relatively old and that no effort was undertaken to exclude subjects with osteoarthritic changes. 2. Non-site-specific (peripheral) measurements are inferior to spinal DXA and QCT, and also for predicting strength in the thoracic spine. This is most likely due to the smaller degree of skeletal heterogeneity within the spine, than between the spine and other sites. 3. Assessment by pQCT at the weight-bearing lower limb does not have an advantage over that at the distal radius. 4. Calcaneal QUS cannot add significant, independent information to site-specific bone status. These experimental results clearly favor the site-specific assessment of vertebral strength in elderly individuals over that by non-site-specific or peripheral measurements.
Acknowledgments: The authors thank Jan Grimm (Medizinische Physik, Klinik fu¨ r Diagnostische Radiologie, Kiel, Germany) for reading the spinal X-rays, Gu¨ nther Delling and coworkers (Abteilung Osteopathologie, Universita¨ tsklinikum Eppendorf, Hamburg, Germany) for the histomorphometric analyses, and Rainer Barkmann (Medizinische Physik, Klinik fu¨ r Diagnostische Radiologie, Kiel, Germany) for helpful discussion. Gudrun Goldmann, Nadine Krefting, Boriana Barth, Oliver Groll, and Hans-Ju¨ rgen Becker (Musculoskeletal Research Group, Institute of Anatomy, Mu¨ nchen) are acknowledged for their help with the radiographic, DXA, QCT, pQCT, and QUS examinations. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG LO 730/2-1).
References 1. Albright, F., Smith, P. H., and Richardson, A. M. Postmenopausal osteoporosis. JAMA 116:2465–2474; 1941. 2. Amling, M., Herden, S., Po¨ sl, M., Hahn, M., Ritzel, H., and Delling, G. Heterogeneity of the skeleton: Comparison of the trabecular microarchitecture of the spine, the iliac crest, the femur, and the calcaneus. J Bone Miner Res 1:36 –45; 1996. 3. Amling, M., Posl, M., Ritzel, H., Hahn, M., Vogel, M., Wening, V. J., and Delling, G. Architecture and distribution of cancellous bone yield vertebral fracture clues. A histomorphometric analysis of the complete spinal column from 40 autopsy specimens. Arch Orthop Trauma Surg 115:262–269; 1996. 4. Biggemann, M., Hilweg, D., and Brinckmann, P. Prediction of the compressive strength of vertebral bodies of the lumbar spine by quantitative computed tomography. Skel Radiol 17:264 –269; 1988. 5. Bjarnason, K., Hassager, C., Svendsen, O. L., Stang, H., and Christiansen, C. Anteroposterior and lateral spinal DXA for the assessment of vertebral body strength: Comparison with hip and forearm measurement. Osteopor Int 6:37– 42; 1996.
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
83
6. Brinckmann, P., Biggemann, M., and Hilweg, D. Prediction of the compressive strength of human lumbar vertebrae. Spine 14:606 –610; 1989. 7. Bu¨ rklein, D., Lochmu¨ ller, E.-M., Kuhn, V., Grimm, J., Barkmann, R., Mu¨ ller, R., and Eckstein, F. Correlation of thoracic and lumbar vertebral failure loads and prediction by in-situ vs. ex-situ dual X-ray absorptiometry. J Biomech 35:579 –587; 2001. 8. Calabrisi, P. and Smith, F. C. The effects of embalming on the compressive strength of a few specimens of compact human bone. Naval Medical Research Institute Memorandum Report 51-2 NM 000 018.07.02; 1952. 9. Cheng, X. G., Nicholson, P. H. F., Boonen, S., Lowet, G., Brys, P., Aerssens, J., Vander Perre, G., and Dequeker, J. Prediction of vertebral strength in vitro by spinal bone densitometry and calcaneal ultrasound. J Bone Miner Res 12:1721–1728; 1997. 10. Cody, D. D., Goldstein, S. A., Flynn, M. J., and Brown, E. B. Correlations between vertebral regional bone mineral density (rBMD) and whole bone fracture load. Spine 16:146 –154; 1991. 11. Cooper, C., Atkinson, J., O’Fallon, W. M., and Melton, J. L. Incidence of clinically diagnosed vertebral fractures: A population-based study in Rochester, Minnesota, 1985–1989. J Bone Miner Res 2:221–227; 1992. 12. De Smet, A. A., Robinson, R. G., Johnson, B. E., and Lukert, B. P. Spinal compression fractures in osteoporotic women: Patterns and relationship to hyperkyphosis. Radiology 166:497–500; 1988. 13. Ebbesen, E. N., Thomsen, J. S., Beck Nielsen, H., Nepper Rasmussen, H. J., and Mosekilde, L. Lumbar vertebral body compressive strength evaluated by dual energy X ray absorptiometry, quantitative computed tomography, and ashing. Bone 25:713–724; 1999. 14. Edmondston, S. J., Singer, T. P., Day, R. E., Breidahl, P. D., and Price, R. I. Formalin fixation effects on vertebral bone density and failure mechanics: An in-vitro study of human and sheep vertebrae. Clin Biomech 9:175–179; 1994. 15. Genant, H. K. and Osteoporosis Research Group. Guidelines for Quality Assurance in Radiographic Examinations of the Spine. San Francisco, CA: Regents of the University of California; 1996. 16. Genant, H. K., Engelke, K., Fuerst, T., Glu¨ er, C. C., Grampp, S., Harris, S. T., Jergas, M., Lang, T., Lu, Y., Majumdar, S., Mathur, A., and Takada, M. Noninvasive assessment of bone mineral and structure: State of the art. J Bone Miner Res 11:707–730; 1996. 17. Glu¨ er, C. C., Blake, G., Lu, Y., Blunt, B. A., Jergas, M., and Genant, H. K. Accurate assessment of precision errors: How to measure the reproducibility of bone densitometry techniques. Osteopor Int 5:262–270; 1995. 18. Grampp, S., Genant, H. K., Mathur, A., Lang, P., Jergas, M., Takada, M., Glu¨ er, C. C., Lu, Y., and Chavez, M. Comparisons of noninvasive bone mineral measurements in assessing age related loss, fracture discrimination, and diagnostic classification. J Bone Miner Res 12:697–711; 1997. 19. Greenberg, S. W., Gonzalez, D., Gurdjian, E. S., and Thomas, L. M. Changes in Physical Properties of Bone Between the In Vivo, Freshly Dead and Embalmed Condition. Society of New York: Automotive Engineers; 1968. 20. Groll, O., Lochmu¨ ller, E.-M., Bachmeier, M., Willnecker, J., and Eckstein, F. Precision and intersite correlation of bone densitometry at the radius, tibia and femur with peripheral quantitative CT. Skel Radiol 28:696 –702; 1999. 21. Ha¨ rma¨ , M., Heliovaara, M., Aromaa, A., and Knekt, P. Thoracic spine compression fractures in Finland. Clin Orthop 205:188 –194; 1986. 22. Hahn, M., Vogel, M., and Delling, G. Undecalcified preparation of bone tissue: Report of technical experience and development of new methods. Virchows Arch A Pathol Anat Histopathol 418:1–7; 1991. 23. Hans, D., Fuerst, T., and Duboeuf, F. Quantitative ultrasound bone measurement. Eur Radiol 7(Suppl. 2):S43–S50; 1997. 24. Hassager, C., Jensen, S. B., Gotfredsen, A., and Christiansen, C. The impact of measurement errors on the diagnostic value of bone mass measurements: Theoretical considerations. Osteopor Int 1:250 –256; 1991. 25. Hayes, W. C. and Bouxsein, M. L. Biomechanics of cortical and trabecular bone: Implications for assessment of fracture risk. Mow, V. C. and Hayes, W. C., eds. Basic Orthopaedic Biomechanics. 2nd Ed. Philadelphia: LippincottRaven; 1997. 26. Leidig-Bruckner, G., Minne, H. W., Schlaich, C., Wagner, G., Scheidt-Nave, C., Bruckner, T., Gebest, H. J., and Ziegler, R. Clinical grading of spinal osteoporosis: Quality of life components and spinal deformity in women with chronic low back pain and women with vertebral osteoporosis. J Bone Miner Res 12:663–675; 1997. 27. Liberman, U. A., Weiss, S. R., Broll, J., Minne, H. W., Quan, H., Bell, N. H., Rodriguez-Portales, J., Downs, R. W., Dequeker, J., and Favus, M. Effect of oral alendronate on bone mineral density and the incidence of fractures in postmenopausal osteoporosis. The Alendronate Phase III Osteoporosis Treatment Study Group. N Engl J Med 333:1437–1443; 1995.
84
E.-M. Lochmu¨ ller et al. Spinal strength prediction from densitometry
28. Lochmu¨ ller, E.-M., Eckstein, F., Kaiser, D., Zeller, J. B., Landgraf, J., Putz, R., and Steldinger, R. Prediction of vertebral failure loads from spinal and femoral dual-energy X-ray absorptiometry, and calcaneal ultrasound: An in situ analysis with intact soft tissues. Bone 23:417–424; 1998. 29. Lochmu¨ ller, E.-M., Miller, P., Bu¨ rklein, D., Wehr, U., Rambeck, W., and Eckstein, F. In situ femoral DXA related to ash weight, bone size, and density, and its relationship with mechanical failure loads of the proximal femur. Osteopor Int 11:361–367; 2000. 30. Lochmu¨ ller, E.-M., Krefting, N., Bu¨ rklein, D., and Eckstein, F. Effect of fixation, soft-tissues and scan projection on bone mineral measurements with dual energy X-ray absorptiometry (DXA). Calcif Tissue Int 68:140 –145; 2001. 31. Lufkin, E. G., Wahner, H. W., O’Fallon, W. M., Hodgson, S. F., Kotowicz, M. A., Lane, A. W., Judd, H. L., Caplan, R. H., and Riggs, B. L. Treatment of postmenopausal osteoporosis with transdermal estrogen. Ann Intern Med 117:1–9; 1992. 32. McBroom, R. J., Hayes, W. C., Edwards, W. T., Goldberg, R. P., and White, A. A. Prediction of vertebral body compressive fracture using quantitative computed tomography. J Bone Jt Surg 67A:1206 –1214; 1985. 33. McElhaney, J., Fogle, J., Byars, E., and Weaver, G. Effect of embalming on the mechanical properties of beef bone. J Appl Physiol 19:1234 –1236; 1964. 34. Melton, L. J. III, Kan, S. H., Frye, M. A., Wahner, H. W., O’Fallon, W. M., and Riggs, B. L. Epidemiology of vertebral fractures in women. Am J Epidemiol 129:1000 –1011; 1989. 35. Melton, L. J., Thamer, N. F., Ray, J. K., Chesnut, C. H., Einhorn, T. A., Johnston, C. C., Raiz, L. G., Silverman, S. L., and Siris, E. S. Fractures attributable to osteoporosis: Report from the National Osteoporosis Foundation. J Bone Miner Res 12:16 –23; 1997. 36. Moro, M., Hecker, A. T., Bouxsein, M. L., and Myers, E. R. Failure load of thoracic vertebrae correlates with lumbar bone mineral density measured by DXA. Calcif Tissue Int 56:206 –209; 1995.
Bone Vol. 31, No. 1 July 2002:77– 84 37. Nicholson, P. H., Mu¨ ller, R., Cheng, X. G., Ruegsegger, P., Van Der Perre, G., Dequeker, J., and Boonen, S. Quantitative ultrasound and trabecular architecture in the human calcaneus. J Bone Miner Res 16:1886 –1892; 2001. 38. Po¨ pperl, G., Lochmu¨ ller, E.-M., Becker, H.-J., Mall, G., Steinlechner, M., and Eckstein, F. Ex-situ determination of calcaneal ultrasound properties: Reproducibility, effects of storage, formalin fixation, maceration, and changes in anatomic measurement site. Calcif Tissue Int 65:192–197; 1999. 39. Ray, N. F., Chan, J. K., Thamer, M., and Melton, L. J. Medical expenditures for the treatment of osteoporotic fractures in the United States in 1995: Report from the National Osteoporosis Foundation. J Bone Miner Res 12:24 –35; 1997. 40. Ritzel, H., Amling, M., Po¨ sl, M., Hahn, M., and Delling, G. The thickness of human vertebral cortical bone and its changes in aging and osteoporosis: A histomorphometric analysis in the complete spinal column from thirty-seven autopsy specimens. J Bone Miner Res 12:89 –95; 1997. 41. Rockoff, S. D., Sweet, E., and Bleustein, J. The relative contribution of trabecular and cortical bone to the strength of human lumbar vertebrae. Calcif Tissue Res 3:163–175; 1969. 42. Svendsen, O. L., Hassager, C., Skodt, V., and Christiansen, C. Impact of soft tissue on in vivo accuracy of bone mineral measurements in the spine, hip, and forearm: A human cadaver study. J Bone Miner Res 10:868 –873; 1995.
Date Received: May 17, 2001 Date Revised: December 21, 2001 Date Accepted: February 14, 2002