Osteoporos Int (2008) 19:537–545 DOI 10.1007/s00198-007-0478-x
ORIGINAL ARTICLE
Does thoracic or lumbar spine bone architecture predict vertebral failure strength more accurately than density? E.-M. Lochmüller & K. Pöschl & L. Würstlin & M. Matsuura & R. Müller & T. M. Link & F. Eckstein
Received: 23 January 2007 / Accepted: 4 September 2007 / Published online: 3 October 2007 # International Osteoporosis Foundation and National Osteoporosis Foundation 2007
Abstract Summary Trabecular bone microstructure was studied in 6 mm bone biopsies taken from the 10th thoracic and 2nd lumbar vertebra of 165 human donors and shown to not differ significantly between these sites. Microstructural parameters at the locations examined provided only marginal additional information to quantitative computed tomography in predicting experimental failure strength. Introduction It is unknown whether trabecular microstructure differs between thoracic and lumbar vertebrae and whether it adds significant information in predicting the mechanical strength of vertebrae in combination with QCTbased bone density.
E.-M. Lochmüller : K. Pöschl : L. Würstlin : M. Matsuura Department of Gynecology I, LMU Munich, Munich, Germany K. Pöschl : L. Würstlin : M. Matsuura Musculoskeletal Research Group, Institute of Anatomy, Ludwig-Maximilians-Universität Munich, Munich, Germany R. Müller Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland T. M. Link Department of Radiology, University of California, San Francisco, CA, USA F. Eckstein (*) Institute of Anatomy and Musculoskeletal Research, Paracelsus Medical University (PMU) Salzburg, Strubergasse 21, A 5020 Salzburg, Austria e-mail:
[email protected]
Methods Six mm cylindrical biopsies taken at mid-vertebral level, anterior to the center of the thoracic vertebra (T) 10 and the lumbar vertebra (L) 2 were studied with micro-computed tomography (μCT) in 165 donors (age 52 to 99 years). The segment T11-L1 was examined with QCT and tested to failure using a testing machine. Results The correlation of microstructural properties was moderate between T10 and L2 (r ≤0.5). No significant differences were observed in the microstructural properties between the thoracic and lumbar spine, nor were sex differences at T10 or L2 observed. Cortical/subcortical density of T12 (r2 =48%) was more strongly correlated with vertebral failure stress than trabecular density (r2 =32%). BV/TV (of T10) improved the prediction by 52% (adjusted r2) in a multiple regression model. Conclusion Microstructural properties of trabecular bone biopsies displayed a high degree of heterogeneity between vertebrae but did not differ significantly between the thoracic and lumbar spine. At the locations examined, bone microstructure only marginally improved the prediction of structural vertebral strength beyond QCT-based bone density. Keywords Failure stress . Mechanical strength . Micro computed tomography . Quantitative computed tomography (QCT) . Spine . Trabecular microstructure
Introduction Spinal fractures are considered a hallmark of osteoporosis [1] and high incidence rates have been reported for North America [2, 3] and Europe [4]. Vertebral fractures dramatically reduce the quality of life by causing pain, physical deformity, functional deficits [5–7], and are also associated
538
with increased mortality [8–11]. Vertebral fractures are known to most commonly occur in the thoracic spine or at the thoracolumbar junction [12–14], but women with vertebral deformities at the lumbar spine are more likely to suffer pain than those with deformities of the thoracic spine [14]. The type of vertebral deformity varies between the thoracic and lumbar region: wedge deformities are found to represent the most frequent type of vertebral fracture and tend to cluster at the mid-thoracic and thoracolumbar region [15]. The same is observed for crush fractures, but biconcave deformities are more frequent in the lumbar spine. Furthermore, contrary to wedge and crush fractures, they do not decline in frequency at lower lumbar vertebral levels [15]. The biomechanical and morphological basis for these different distribution patterns of vertebral fractures types, however, is currently unknown. Amling and coworkers [16] found the bone volume fraction (BV/TV) to be highest in the cervical spine and lower in the thoracic and lumbar spine using histomorphometry. In osteoporotic subjects, they reported an overall reduction in BV/TV, but the same distribution pattern throughout the thoracolumbar spine. Micro-computed tomography (μCT) has been used to characterize the vertebral microstructure of lumbar vertebra 2 (L2) and 4 (L4) [17] and to compare bone microstructure between different anatomical sites in the human body [17, 18], but no systematic comparison of trabecular microstructure has been previously performed between the thoracic and lumbar spine of human donors. Previous experimental studies have indicated that between 50 to 65% of the variability in vertebral bone strength is predicted by bone density as measured by either dual energy X-ray absorptiometry (DXA) or quantitative computed tomography (QCT) [19–22]. It has been reported that trabecular microstructure (assessed with μCT or high resolution CT) improves the prediction of mechanical bone strength of trabecular biopsy samples of vertebral bone [23, 24], and that the trabecular microarchitecture of vertebral bodies [25, 26] or other skeletal sites [26, 27] (using clinical imaging technology) improves the discrimination between subjects with and without spinal fractures. One experimental study in a small number of human vertebrae indicated that microstructural parameters may improve the prediction of structural strength of whole vertebrae over bone mass alone using clinical QCT [28]. Another study in normal, ovariectomized and ibandronate-treated monkeys found that inclusion of μCT-based microstructural indices increased the prediction of whole bone mechanical strength to 88% (r2) from 67% by bone mass alone [29]. However no experimental study so far has investigated the relationship between the human trabecular bone microstructures of thoracic and lumbar vertebrae and their correlation with the mechanical strength of the thoracolumbar spine. The current study was thus designed to address the following
Osteoporos Int (2008) 19:537–545
specific questions: 1) Are there differences in microstructural bone properties between the thoracic (thoracic vertebra 10) and the lumbar spine (lumbar vertebra 2) using bone biopsy samples, how do they correlate, and do sex differences in microstructural properties differ between the thoracic and lumbar region? 2) Can the structural strength (strength) of the thoracolumbar spine be more accurately predicted when combining densitometric information (from QCT) with microstructural information (from μCT) and, if yes, are microstructural properties of the thoracic or the lumbar spine more predictive?
Materials and methods Study cohort and sample preparation A cohort of 168 embalmed human cadavers from a series of three macroscopic anatomy courses was investigated. Micro-CT analysis was performed at seven skeletal locations. The sex-differences of microstructural properties at the femoral neck and trochanter, distal radius, calcaneus, iliac crest and lumbar vertebra, two were reported previously for this sample [18]. The donors had agreed to dedicate their bodies to the institute several years prior to death, and we, therefore, assume that they constitute a representative selection of the elderly population resident in Bavaria. Bone biopsies were taken at the site of clinical transiliac biopsies from the right pelvis, embedded in methylmethacrylate, cut in 5 μm sections, stained with Goldner, Toluidin blue and von Kossa, and histologically evaluated by an experienced pathologist to exclude bone disease other than osteoporosis and osteopenia [18]. Three specimens with signs of malignancy were discarded from the study so that a total of 165 specimens were left for analysis (age range 52 to 99 years; 79 women aged 81.2±9.0 years, and 86 men aged 79.1±9.9 years). In all 165 subjects, the thoracolumbar spine (including of the segment between the thoracic vertebra 10 [T10] and the lumbar vertebra 2 [L2]) were harvested after the dissection courses. The spines were radiographed in two planes to exclude previous fractures using a Polyphos 30 M X-ray system (Siemens, Erlangen, Germany). Four films were obtained: two (anterior-posterior and lateral projection) focusing on the thoracic and two on the lumbar region. Spinal fractures were diagnosed by a musculoskeletal radiologist (T.M.L.) according to published guidelines [30], with vertebrae displaying deformities > grade 2 (> 25% reduction of height) being excluded from further analysis. T10 and L2 were separated from the T11 to L1 segment by cutting through the intervertebral discs. Cylindrical specimens were retrieved from T10 and L2 using an 8 mm diameter diamond drill (Salzmann, Munich, Germany) [18, 31]. One
Osteoporos Int (2008) 19:537–545
539
full length cylinder was obtained from each vertebra in superior-inferior direction at the middle along the transverse dimension (left to right) and at the transition of the anterior third (33%) to the posterior two thirds (66%) of the sagittal (anterior to posterior) vertebral dimension, to avoid the posterior venous plexus (Fig. 1). A 14 mm long specimen was obtained from the center (superior-inferior dimension) of the full length cylinder. The retrieval was standardized between vertebrae because of the known heterogeneity of the trabecular microstructure throughout the vertebral body [32, 33]. μCT scanning The μCT scans were acquired for the central 6 mm of the specimen (Fig. 2) using a μCT 20 scanner (Scanco Medical, Bassersdorf, Switzerland) as described previously [18, 31]. The resolution was set to 26 μm (isotropic) in the “medium” scan mode and with an integration time of 100 ms, giving a total scan time of 4.1 hours per sample. Within a defined volume of interest of 6 mm diameter and 6 mm length, the following 3D-structural parameters were determined using the following settings (Sigma 0.8; Support 1.0; Threshold 22% of maximum grey value) and the software provided by the manufacturer: 1) bone volume fraction (BV/TV) in %; 2) trabecular number (Tb.N) in
Fig. 1 Quantitative computed tomography (QCT) of thoracic vertebra 12 (1 cm thick section at the mid-vertebral level) showing regions of interest for analysis of trabecular (inside) and cortical/subcortical bone (outside). Hounsfield units (HU) were converted into density values using a calibration phantom. The bright small circle indicates where the trabecular biopsy was taken in T10 and L2, respectively
Fig. 2 Reconstruction of trabecular micro structure from thoracic vertebral body 10 (upper row) and lumbar vertebral body 2 (lower row) with high BV/TV (left), medium BV/TV (middle) and low BV/ TV (right)
1/mm; 3) trabecular thickness (Tb.Th) in mm; 4) trabecular separation (Tb.Sp) in mm; 5) structure model index (SMI) [34], a measure of plate- or rod-like trabecular architecture, 6) connectivity density (Conn.D) in 1/mm3 and 7) degree of anisotropy (DA). These parameters were computed in 3D without model assumptions required for 2D analysis [17]. Quantitative computed tomography The spinal segments T11 to L1 were degassed within a vacuum pump for at least 12 hours prior to imaging and being sealed in a saline solution in thin polyethylene bags (Fig. 1). QCT was performed using a clinical Somatom Plus 4 scanner (Siemens, Erlangen, Germany) and an osteodensitometry phantom provided by the manufacturer (Fig. 1). One-cm-thick sections of each T12 vertebra were obtained at the mid-vertebral level with 80 kVp und 146 mAs. Bone mineral density (BMD in mg/cm3) was determined for a central (trabecular) and peripheral (“cortical/subcortical”) region of interest (ROI), using the software provided by the manufacturer [35, 36]. To that end, the contours of the vertebral body and spinal canal were automatically detected using an edge detection algorithm. The costal and transverse processes were clipped by eliminating convex contour segments with > 135° over 9 mm. The vertical principle axis of the vertebra was computed, as well as a line perpendicular to that axis. The cortical ROI was defined as a 1.5-mm-thick ring inside the vertebral bone edge, and the trabecular ROI as an area inside a line 2.75 mm away from the cortical ROI. The trabecular and cortical ROIs were, however, designed to avoid the (posterior) vertebral venous plexus. To that end, a point was defined at the vertical principle axis of the vertebra half way between the anterior vertebral border and the center of the spinal canal. The posterior border of the regions was
540
then defined by lines intersecting this point and bilateral outer points, as defined by the intersection of the vertebral bone edge with the line perpendicular to the vertical principle axis (Fig. 1). Integral bone density was computed from measurements in both regions by weighing trabecular and cortical density within trabecular and cortical areas, respectively. Hounsfield units (HU) were converted into density values using the standard BMD calibration phantom provided by the manufacturer with equivalents of 0 and 200 mg/cm3 hydroxylappatite (Fig. 1). The same image was used to determine the cross-sectional area (CSA in cm2) of the vertebral body using the manufacturer’s software. Mechanical testing Specimens were kept moist during storage to avoid drying, and were degassed as described above to replace trapped air in the trabecular bone by fluid. The spinal segment T11 to L1 (with target vertebra T12) was then tested as a functional unit with intact ligaments and intervertebral discs as described previously [21, 37]. A functional unit was used to simulate physiological load transfer as closely as possible, but the vertebral arch and facet joints could not be maintained, because the vertebral canal was opened from posterior during the dissection course. The upper and lower vertebral bodies were embedded at mid-vertebral level plano-parallel to the endplates of the central target vertebra (T12). The units were then compressed axially with an uniaxial material testing machine (Zwick 1445, Ulm, Germany) using a 10 kN load cell. The test was stopped when the unit had been compressed to 25% of its original height and when the vertebra had reached the second load ascent. The first peak of the load-displacement curve (with a subsequent drop of > 10%) was taken as the mechanical failure load. To compute failure stress (MPa), the failure load was divided by the cross-sectional area of the vertebra as determined with CT. Statistical analysis For technical reasons, it was not possible to obtain samples from all vertebrae: five fractures were observed at T10, 18 at T12 and 11 at L2. In addition, some spines displayed too severe a deformity (scoliosis) or spondylosis (osteophytes) to be included for mechanical testing, and some trabecular samples disintegrated during the retrieval. The total number of specimens obtained from T10 was n=133 (66 from women and 67 from men), and n=134 from L2 (61 in women and 73 in men). In 118 specimens μCT data from both T10 and L2 were available; comparison between the two vertebrae was limited to this subsample. In 150 specimens μCT data from either T10 or L2 was available, and mechanical failure loads were determined in 134 of
Osteoporos Int (2008) 19:537–545
these (no fracture at T12). The correlation analysis with mechanical failure stress (see below) was, thus, assessed in this subsample. Differences and correlations (linear regression analysis) in bone microstructural properties between T10 and L2 were determined in the subsample of 118 specimens in which microstructural information from both T10 and L2 were available. Differences were tested for statistical significance using a paired T-test. Correlations of microstructural properties (μCT at T10 and L2) and QCT-based bone density (at T12) with failure load and stress of T12 was analyzed in the subsample for which failure loads and either microstructural data at T10 or L2 had been obtained (n=134). Stepwise multiple regression models (forward mode) were used to determine whether microstructural parameters from T10 or L2 were able to contribute significant information in addition to bone density in predicting failure stress at T12.
Results BV/TV obtained from μCT of the bone biopsies at T10 was only slightly higher than at L2 and the difference was not statistically significant (Table 1). The correlation between BV/TV of T10 and L2 was r=0.41 (p
