SEARCH

SEARCH BY CITATION

Keywords:

  • osteoporosis;
  • teriparatide;
  • bone strength;
  • CT;
  • finite element analysis

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Monitoring of osteoporosis therapy based solely on DXA is insufficient to assess antifracture efficacy. Estimating bone strength as a variable closely linked to fracture risk is therefore of importance. Finite element (FE) analysis–based strength measures were used to monitor a teriparatide therapy and the associated effects on whole bone and local fracture risk. In 44 postmenopausal women with established osteoporosis participating in the EUROFORS study, FE models based on high-resolution CT (HRCT) of T12 were evaluated after 0, 6, 12, and 24 mo of teriparatide treatment (20 μg/d). FE-based strength and stiffness calculations for three different load cases (compression, bending, and combined compression and bending) were compared with volumetric BMD (vBMD) and apparent bone volume fraction (app. BV/TV), as well as DXA-based areal BMD of the lumbar spine. Local damage of the bone tissue was also modeled. Highly significant improvements in all analyzed variables as early as 6 mo after starting teriparatide were found. After 24 mo, bone strength in compression was increased by 28.1 ± 4.7% (SE), in bending by 28.3 ± 4.9%, whereas app. BV/TV was increased by 54.7 ± 8.8%, vBMD by 19.1 ± 4.0%, and areal BMD of L1–L4 by 10.2 ± 1.2%. When comparing standardized increases, FE changes were significantly larger than those of densitometry and not significantly different from app. BV/TV. The size of regions at high risk for local failure was significantly reduced under teriparatide treatment. Treatment with teriparatide leads to bone strength increases for different loading conditions of close to 30%. FE is a suitable tool for monitoring bone anabolic treatment in groups or individual patients and offers additional information about local failure modes. FE variables showed a higher standardized response to changes than BMD measurements, but further studies are needed to show that the higher response represents a more accurate estimate of treatment-induced fracture risk reduction.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

The deficiencies of DXA as a tool to monitor osteoporosis treatment are well established. Especially for antiresorptive medication, only a small proportion of the reduced fracture incidence during and after treatment can be explained by the changes in areal BMD.(1,2) A similar situation is found for anabolic medication. Changes in DXA-based areal BMD explain only ∼30% of the antifracture efficacy of teriparatide.(3) Therefore, it is necessary to develop diagnostic methods that are more sensitive indicators of the treatment-induced changes in fracture risk. Bone strength, the maximum force the bone can bear, is the most important determinant of fracture risk. It can be assessed in vivo by a simulated mechanical test based on CT images using the method of finite element analysis (FEA).

FEA is a widely used engineering tool to simulate the mechanical behavior of complex structures.(4) It has been used for some time to assess bone strength in ex vivo basic research, in animal models, but more recently also in in vivo clinical trials. Applying FEA to medical problems with large numbers of patients or specimens requires automated procedures for image segmentation and model generation. The ex vivo comparison of FEA with biomechanical testing has shown good agreement of the ultimate strength at different skeletal sites (e.g., vertebrae,(5) proximal femur,(6) or distal radius).(7) In several studies, FEA was applied in vivo (e.g., in a recent study by Imai et al.(8) targeting fracture discrimination at the lumbar spine). In these studies, FEA-derived bone strength was shown to be a better predictor of experimentally measured bone strength than DXA-based BMD. Keaveny et al. used FEA for comparing teriparatide and alendronate treatments in the FACT trial, both at the spine(9) and the hip.(10)

Most studies for the spine use a uni-axial compression model only, which represents one typical loading situation. On the other hand, Homminga et al.(11) showed that osteoporotic vertebrae are still adjusted to this loading situation but not as well to other less typical ones, so called “error loads.” Therefore, in this study the vertebrae were tested in compression and bending as well as in a combination of these two loading conditions.

A density-based, continuum nonlinear FE model specifically designed to meet the biomechanical behavior of bone to consistently depict bone strength changes under teriparatide treatment in vivo was used. Besides providing estimates of whole bone strength, FEA also permits to study the pattern of loads and strains throughout the volume of the vertebral body. This provides insights into the spatial pattern of strong and weak parts of the trabecular network along with local changes induced by teriparatide treatment. Specifically, it can be assessed whether teriparatide treatment reduces the size of subregions that are particularly weak and at highest risk of local failure.

The aim of this study was to show the feasibility of using nonlinear FE analysis to monitor teriparatide treatment and the additional information gained in comparison with DXA.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Study design

The EUROFORS high-resolution CT (HRCT) addendum is a substudy of a pan-European, controlled, randomized, open-label clinical trial of postmenopausal women with severe osteoporosis.(12) The patients differed in their osteoporosis medication history; the majority had received extensive pretreatment with antiresorptive medication, which had to be discontinued at study start. All participants of the substudy underwent HRCT scans of the T12 vertebra and DXA scans of L1–L4 at baseline and after 6, 12, and 24 mo of treatment. Only those participants of the HRCT substudy who were assigned to receiving teriparatide over the complete study duration of 2 yr were included in this analysis. Patients were excluded from this analysis if they had (1) an invalid baseline HRCT scan or (2) more than one invalid HRCT follow-up scans. If one of the follow-up scans was missing, the last valid scan was used instead (last observation carried forward [LOCF]). This procedure was not applied to the baseline visit because this study aims at judging changes under teriparatide treatment. Details of the pretreatment regimen and study design have been described previously.(13)

HRCT scan protocols

The participants of this substudy were recruited in seven centers in Spain and Germany using different CT scanners from Siemens and GE. Similar but not identical scan protocols were implemented on the different scanner types to obtain optimized HRCT data. All scans were taken at 120 kV and 360 mAs, with a reconstructed slice thickness between 300 and 500 μm and an in-plane voxel size of 156 or 188 μm. The reconstruction kernel varied between the different manufacturers, but in all cases, a high-resolution kernel was selected. This paper focuses on a longitudinal analysis, which is less affected by the scanner differences. Special care was taken to assure identical scan procedures for each of the follow-up scans for each patient. This resulted in a relatively large number of excluded patients.(13)

HRCT image processing

During the HRCT scans, a solid calibration phantom (Mindways, San Francisco, CA, USA) was positioned under the patient. All scans were reconstructed twice with identical parameters except for the size of the field of view (FOV). The first reconstruction was large enough to contain the calibration phantom. It was used to calibrate the gray values of the second reconstruction obtained with a smaller FOV size, resulting in the pixel size given above. The second reconstruction was scaled to mineral density; it was manually rotated to align the endplates axially. The entire vertebra with the exception of the posterior processes was segmented from the background using a semiautomatic active shape algorithm. First, a 3D template vertebra was registered to the image. To speed up this process, the registration was based on three orthogonal projections of both the template and the image. The template served as a starting point for the 3D active shape targeting the high gray values of the cortical shell. To include the entire cortical shell, the active shape was afterward expanded from the center of the vertebra stopping in regions of high-density gradients.

A subregion of trabecular bone was generated by peeling off the outer 3 mm of the total volume of interest (VOI). In this trabecular subregion, volumetric BMD was calculated as well as “apparent bone volume fraction” (app. BV/TV) after thresholding with a constant value of 250 mg/ml. App. BV/TV had shown the largest response to teriparatide treatment among a set of structural variables.(13)

FE model generation

The entire segmented vertebra was resampled to isotropic voxels of either 0.7 (linear model) or 1.3 mm (nonlinear model). Bone volume fraction was calculated as the BMD of each voxel divided by an assumed density of a fully mineralized voxel of 1200 mg/ml. The values of the bone volume fraction were limited to the interval [0,1]. The image voxels were directly converted into hexahedral elements. Transverse isotropic mechanical properties were assigned according to a power law based on the bone volume fraction, with the symmetry axis along the inferior–superior axis of the vertebrae. A virtual layer of bone cement (polymethyl methacrylate [PMMA]) was added to the upper and lower endplate, creating two parallel planes similar to those used in ex vivo compression tests (Fig. 1). PMMA was modeled as an isotropic, elastic material with a Young's modulus of 3000 MPa. Details of the modeling and its ex vivo validation can be found elsewhere.(14) All FE calculations were solved with ABAQUS Standard V 6.6 (Dassault Systèmes).

thumbnail image

Figure Figure 1. Schematic overview of the three loading modes. The vertebra is shown in white and the PMMA embedding in dark gray. The bottom endplate was always fixed, whereas the top endplate was displaced vertically (top), rotated around the most posterior and superior point (center), or a combination of both (bottom).

Download figure to PowerPoint

Linear FE analysis

A uni-axial compression test was simulated by moving the upper PMMA layer in the direction of the lower plane by an arbitrary displacement. The lower PMMA plane was fully restricted in all degrees of freedom. Linear vertebral stiffness was defined as the reaction force of the vertebra divided by the applied displacement.

Nonlinear FE analysis

In the nonlinear case, an elastic plastic damage constitutive law developed to describe the postyield behavior of bone was used.(15) The elastic constants were identical as in the linear model, but additional parameters were necessary for describing the plastic behavior, associated with permanent deformation, as well as damage accumulation in bone. Both permanent deformation and damage occurred in highly deformed regions. The damage values ranged from 0 (undamaged) to 1 (failed), with a deterioration of the mechanical properties of the element with rising damage.(16) The damage variable is interpreted as an indication for the risk of local mechanical failure in the given subregion. That means that a fracture of the complete vertebra might likely initiate in mechanically weak subregions with high damage values.

Because computation time of the nonlinear model was substantially longer, the resolution of the model had to be reduced in comparison with the linear analysis.

Three different load cases were tested with the nonlinear model (Fig. 1): pure compression, pure bending, and a combination of bending and compression. The setup for the compression test was identical to the one described above, with the displacement applied in initial steps of 0.05 to 1 mm. For anterior bending, a transversal axis was defined through the most posterior point of the upper PMMA and incremental rotations were prescribed up to a maximum of 0.06 radians. The combined load case was carried out proportionally until the yield points of both the individual bending and compression load cases were reached.

An initial stiffness for each of the load cases (called initial stiffness compression, bending, and combined, respectively) was defined as the slope of the initial step of the force-displacement, moment-angle, and force-angle curves, respectively. The maximum force and moment were derived from the maxima of the load-displacement curves for pure compression and bending only. An example of the global force-displacement diagram resulting from an FE compression test is given in Fig. 2.

thumbnail image

Figure Figure 2. Force-displacement diagrams showing the force-displacement characteristic of the nonlinear compression test for all visits in one typical patient. The dotted line indicates the lowest maximum force of all visits and the corresponding displacement in other visits. The damage variable was compared at these displacements.

Download figure to PowerPoint

Statistical analysis

Statistical analysis was performed using JMP 6.0.0 (SAS Institute, Cary, NC, USA). Baseline variables were compared using Pearson correlation coefficients. Significance of the observed changes was assessed by a one sample, two-sided t-test. The volume of highly damaged bone did not follow a normal distribution and was analyzed by the Wilcoxon signed rank test.

Increases in different variables were expressed in absolute values scaled to the SD of the baseline visit. This method allows the comparison of differently scaled variables. It expresses changes in a variable as a change relative to the variation in the study population, which is a combination of true variability and the accuracy error. Compared with percent changes, which adjust for scaling effects, it also reduces random errors as well as additive bias.

The long-term precision error was estimated as the root mean square error (RMSE) of a quadratic regression over time for the individual patients. The relevant precision error for this treatment-oriented study was a standardized precision. Standardized precision is derived from long-term precision errors by multiplication with the treatment response rate of DXA (as the reference technique) divided by the treatment response rate of the variable under study. In addition, the monitoring time interval (MTI) was calculated as the time needed for the expected increase to exceed the least significant change (2.8 × RMSE).(17)

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Of the 119 participants enrolled in the baseline visit, 54 were excluded from this analysis because of CT protocol violations.(13) A further 15 participants did not receive teriparatide over the entire study duration. Six participants were not eligible for FE analysis, because their vertebral bodies were not depicted completely in the HRCT image. In total, 44 participants 68.0 ± 6.7 yr of age at baseline were included in the study. Their bone-specific baseline characteristics are listed in Table 1. Of the 44 patients, a total of 37 (84%) were pretreated with antiresorptive medication before study start. Two patients had no valid HRCT scan at the 6-mo visit, one patient missed the 12-mo visit, and two patients missed the 24-mo visit. Because the baseline data were not carried forward, the 6-mo visit contains only 42 data points, whereas for the later visits, the missing data were replaced by the corresponding earlier visit.

Table Table 1.. Absolute Baseline Values and Relative and Standardized Changes Under Treatment (Mean ± SD) of FE Variables, Trabecular BV/TV, and BMD
Thumbnail image of

A typical force-displacement diagram of the nonlinear compression tests at all four visits for one patient is shown in Fig. 2. These tests were carried out to a maximum displacement of 1 mm, at which all the simulations went far beyond the yield point of bone.

Relationship between FE variables

Linear correlation coefficients of the absolute values of all FE variables to each other are given in Table 2. The absolute values of maximum force and moment were well correlated to the respective initial stiffness in both compression and bending. Maximum force was better correlated to maximum moment than the corresponding initial stiffness in bending and compression. The combined loading seemed to be dominated by compression. The stiffness as computed by the linear model correlated closely to initial stiffness of the nonlinear compression test but less so to maximum force. Correlations after 24 mo were similar but slightly better than at baseline.

Table Table 2.. Linear Correlation Coefficients Between FE and Densitometry Variables at Baseline and After 24 mo
Thumbnail image of

Relationship of FE variables to vBMD and app. BV/TV

The correlation between FE variables and vBMD and app. BV/TV was weak at baseline but improved after 24 mo (Table 2). These correlations were lower in bending compared with compression. Initial stiffness and vBMD correlated poorly over the whole treatment period for all load cases.

DXA L1–L4 correlated weakly to all FE variables at baseline; after 2 yr of treatment, no significant correlations to any of the FE results remained.

Response to teriparatide treatment

All FE variables, trab. BV/TV and vBMD, as well as DXA L1–L4, showed highly significant increases after 6, 12, and 24 mo. Table 1 lists changes from baseline in percent and scaled to SD of baseline. These standardized increases were used to compare the different variables; results after 24 mo are shown in Fig. 3. Only increases in variables of the compression test simulation were significantly larger than those of both densitometry measures. The largest response of all FE variables at the end of treatment was observed in linear stiffness, although this was not significantly larger than either maximum force or initial stiffness compression. App. BV/TV showed the largest response to treatment of all variables, as also observed in the investigation of microstructural variables after 12 mo of treatment.(13) The standardized increase of app. BV/TV was not significantly larger than those of any of the compression test variables.

thumbnail image

Figure Figure 3. Responsiveness to treatment for different variables (mean ± SE), expressed in standardized increases after 24 mo scaled to SD of the baseline visit. Asterisks indicate a significantly larger increase than DXA: *p < 0.05, **p < 0.01, ***p < 0.001. The same comparison with vBMD always resulted in a higher degree of significance. There was no significant difference between app. BV/TV and results from the FE compression tests.

Download figure to PowerPoint

Long-term precision

The standardized long-term precision error using DXA-based areal BMD as a reference technique was in the range of 3.2% (linear stiffness) to 4.5% (initial stiffness bending) for the FE variables (Table 1). Precision for DXA-based areal BMD was 2.2%, for vBMD was 3.3%, and for app. BV/TV was 2.3%. The MTI for all FE variables with the exception of maximum moment and initial stiffness bending was below the treatment duration of 24 mo, linear stiffness and maximum force with MTI of 18 mo being the FE variables with the best longitudinal sensitivity. Both DXA and app. BV/TV had an MTI of 12 mo.

Damage

An example of the distribution of the damage within the vertebrae is given in Fig. 4 showing focused areas of high damage. Although there are differences between the two load cases, damage seems to be concentrated in similar areas under both loadings. The patient shown here is the same as in Fig. 2 and had baseline values of maximum force and moment comparable to the group average as well as average increases after 24 mo in maximum force and moment of 23.0% and 17.8%, respectively, slightly below the group averages.

thumbnail image

Figure Figure 4. Distribution of the damage variable in the same patient shown in Fig. 2. The wire frame voxels on top and bottom of the vertebrae show the size of the virtual PMMA layer. The left column shows compression and the right one shows bending. Top to bottom are baseline and 6, 12, and 24 mo.

Download figure to PowerPoint

The damage variable at baseline, given as median (10th and 90th percentiles) at the same force or moment in all visits (Fig. 5) was 0.134 (0.086, 0.165) in compression and 0.113 (0.053, 0.14) in bending. The damage variable was significantly lower at all follow-up visits (p < 0.0001): 0.092 (0.060, 0.144) and 0.066 (0.045, 0.116) at 6 mo; 0.074 (0.040, 0.122) and 0.057 (0.031, 0.077) at 12 mo; and 0.067 (0.036, 0.134) and 0.046 (0.026, 0.104) at 24 mo for compression and bending, respectively. Because regions of high damage are of special interest for the assessment of fracture risk, Fig. 5 shows the volume with damage >0.75. This volume was also significantly reduced under treatment; from 2.7% (0.2–4.9%) to 0.1% (0.0–2.3%) after 24 mo in compression and from 2.2% (0.1–3.9%) to 0.02% (0.0–1.4%) after 24 mo in bending.

thumbnail image

Figure Figure 5. Percentage of volume with a damage level >0.75, indicating a region at high risk of failure (median, 25th–75th percentiles). Damage was evaluated at the lowest maximum force of all visits as shown in Fig. 2. Asterisks indicate significantly reduced damage: *p < 0.001 vs. baseline; **p < 0.0001 vs. baseline; #p < 0.0001 vs. 6-mo visit.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

We present results on HRCT and FE modeling with nonlinear behavior specifically adapted to bone to study changes in patients over 24 mo of teriparatide treatment in vivo. The treatment was associated with significant improvements in densitometric variables and FE-assessed strength and stiffness both in compression and bending. The observed increases after 2 yr were in the order of 10% for aBMD, 20% for vBMD, and up to 30% for FE variables. This is in good agreement with results from the FACT trial, where Keaveny et al.(9) showed close to 20% increase in bending and compression stiffness after 18 mo of teriparatide treatment. Keaveny et al.(9) used a slightly different nonlinear FE model with comparable element size but based on lower-resolution QCT. The densitometric increases agree with clinical teriparatide trials reporting annual changes of 6% for DXA(18) and close to 13% for vBMD(19) at the lumbar spine.

The long-term precision of the FE variables was lower than for DXA or app. BV/TV but the MTI in compression and combined loading was still sufficient to assess changes in maximum force and linear stiffness over <18 mo for individual patients.

When comparing the linear and nonlinear model used for the compression test, both models resulted in similar responses to treatment. In addition, stiffness and strength in the nonlinear model were well correlated. The nonlinear model required several times more computing power, however, even at the lower spatial resolution. Therefore, it seems to be valid to use the simpler, linear models if only the global bone strength change has to be estimated. However, the correlation of linear stiffness with maximum force is weaker, which might be critical for fracture discrimination studies, because bone strength, the determinant of ultimate load, should be the better surrogate. In addition, the nonlinear model allows studying local damage processes, which might give valuable additional information to the fracture risk of the vertebrae.

After 24 mo of teriparatide treatment, all three load cases resulted in similar increases in initial stiffness of ∼25%, and maximum moment and maximum force also showed nearly identical changes. On the other hand, standardized increases for bending variables were lower than for compression, a result of the greater variability of the absolute values in bending at baseline. Correlations to vBMD and app. BV/TV were markedly lower for bending than for compression, indicating possible additional information in the FE models not captured by the global measures of BMD or bone volume fraction. However, the precision errors of the bending strength and stiffness were lower than those of the compression or combined load cases. This could be related to additional variability resulting from the definition of the axis of rotation. The axis was fixed at the most posterior and superior point of the segmented vertebra and therefore depended on the segmentation process for each visit. Because the length of the lever arm has a direct impact on the resulting load, even small variations in the segmentation could lead to unwanted changes in the bending results. A more highly automated segmentation could facilitate solving this aspect.

The damage variable allows depiction of likely fracture mechanisms within the vertebra and possibly the identification of regions at risk for local failure. A local failure carries a high risk of subsequent fracture of the entire vertebra. Strengthening these areas of local weakness thus may have a profound effect on reducing fracture risk, not only for the fracture modes studied here but also for more complex ones (i.e., during a fall). Investigating regions of high risk for local failure necessitated a damage risk threshold. In this study, the threshold was set to 0.75; similar observations were made when other thresholds were used. Ex vivo tests and high-resolution μCT imaging(20,21) have shown that fractures of vertebral trabecular bone start as a localized failure and not uniformly in the entire vertebrae. This indicates that peak values of damage are a better indicator of fracture risk than the average damage in the whole vertebra. The distribution of the voxels at high damage as visible in Fig. 4 seems consistent with the ex vivo study using high-resolution μCT and FEA by Eswaran et al.(21) Eswaran et al. reported defined regions at high fracture risk close to the endplates as also found in our study.

The drastically reduced volume of high damage at the same load over the treatment period shows the antifracture efficacy of teriparatide.

There are several limitations to this analysis.

First, the EUROFORS cohort as a whole and the subset of patients in whom FEA was performed were both too small to allow for a statistically adequate fracture assessment. The motivation for FEA is its better performance compared with DXA in predicting failure load ex vivo and in cross-sectional studies. To adequately show that this would translate into improvements in treatment monitoring in vivo in a clinical setting would require the investigation of both methods in a large randomized controlled trial with a primary fracture endpoint. It is therefore not possible to conclude from this analysis that FEA predicts treatment success (i.e., fracture prevention) better than DXA. The greater numerical changes in standardized units indicate that FEA is the more responsive method and thus a worthwhile target for further investigation in a study with a fracture endpoint.

Second, this substudy of EUROFORS is limited by its lack of a control group not subjected to teriparatide treatment. However, the results of both densitometry and FE variables are in good agreement with other clinical studies of teriparatide.(9,18) The DXA L1–L4 increases in this substudy are similar to those observed in the complete EUROFORS study population.(12) Therefore, the patients included here seem to be a valid subsample.

A third limitation of the FE analysis was the artificially added PMMA layers on the top and bottom of the vertebra. Compared with intact spinal disks, they lead to a different load transmission to the vertebral body. Spinal disks have a complex mechanical behavior and are therefore difficult to include in a FE analysis. In addition, the status of the spinal disks in this elderly study population was not known and difficult to extract from the CT images. For these reasons, PMMA layers are commonly used in FE studies targeting osteoporosis.(5,8,9) A further assumption regarding the FE damage variable was that all vertebrae did not carry any microdamage at the start of the FE analysis (i.e., the damage variable was set to 0 in the whole vertebra). Microdamage can not be measured directly in vivo, so any assumed preexisting damage would have been speculation. Nevertheless, this results in a certain error in the damage calculation.

High-resolution CT as the imaging technique used for the FE models presented here requires a high X-ray dose of ∼3 mSv but offers the possibility to generate high-resolution models as well as analyze the bone structure of a central fracture site, the spine. A lower effective dose would be possible if dedicated peripheral CT was used. This technique allows for a better image quality, but only at peripheral sites such as the forearm. A recent study by Melton et al.(22) showed that structural variables acquired from Xtreme CT (Scanco, Zurich, Switzerland) scans of the forearm were not significant contributors to vertebral fracture discrimination. The best parameters with an OR >3 were the results of an FE model of the lumbar spine. This shows the importance of measuring directly at the central fracture sites even if this leads to reduced image quality.

To conclude, nonlinear FE models were shown to offer valuable information with sufficient precision for monitoring teriparatide treatment in individual patients. The localized damage output characterizes possible subregions at higher risk of fracture; treatment substantially reduced the size of these regions. Linear FE models are able to predict the global strength increase but do not accurately reflect local damage processes. Both linear and nonlinear FE models showed a larger standardized response to treatment than did areal or volumetric density. In summary, teriparatide treatment over 2 yr resulted in substantial, statistically highly significant increases of densitometric and structural measures, as well as bone strength and stiffness under compression, bending, and their combination despite previous antiresorptive pretreatment in >80% of the study population.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES
  • 1
    Cummings SR, Karpf DB, Harris F, Genant HK, Ensrud K, LaCroix AZ, Black DM 2002 Improvement in spine bone density and reduction in risk of vertebral fractures during treatment with antiresorptive drugs. Am J Med 112: 281289.
  • 2
    Sarkar S, Mitlak BH, Wong M, Stock JL, Black DM, Harper KD 2002 Relationships between bone mineral density and incident vertebral fracture risk with raloxifene therapy. J Bone Miner Res 17: 110.
  • 3
    Chen P, Miller PD, Delmas PD, Misurski DA, Krege JH 2006 Change in lumbar spine BMD and vertebral fracture risk reduction in teriparatide-treated postmenopausal women with osteoporosis. J Bone Miner Res 21: 17851790.
  • 4
    Gilbertson LG, Goel VK, Kong WZ, Clausen JD 1995 Finite element methods in spine biomechanics research. Crit Rev Biomed Eng 23: 411473.
  • 5
    Crawford RP, Cann CE, Keaveny TM 2003 Finite element models predict in vitro vertebral body compressive strength better than quantitative computed tomography. Bone 33: 744750.
  • 6
    Cody DD, Gross GJ, Hou FJ, Spencer HJ, Goldstein SA, Fyhrie DP 1999 Femoral strength is better predicted by finite element models than QCT and DXA. J Biomech 32: 10131020.
  • 7
    Pistoia W, van Rietbergen B, Lochmuller EM, Lill CA, Eckstein F, Ruegsegger P 2004 Image-based micro-finite-element modeling for improved distal radius strength diagnosis: Moving from bench to bedside. J Clin Densitom 7: 153160.
  • 8
    Imai K, Ohnishi I, Yamamoto S, Nakamura K 2008 In vivo assessment of lumbar vertebral strength in elderly women using computed tomography-based nonlinear finite element model. Spine 33: 2732.
  • 9
    Keaveny TM, Donley DW, Hoffmann PF, Mitlak BH, Glass EV, San Martin JA 2007 Effects of teriparatide and alendronate on vertebral strength as assessed by finite element modeling of QCT scans in women with osteoporosis. J Bone Miner Res 22: 149157.
  • 10
    Keaveny TM, Hoffmann PF, Kopperdahl DL, Donley DW, Krohn K, Glass EV, Mitlak BH 2007 Comparison of the effects of teriparatide and alendronate on parameters of total hip strength as assessed by finite element analysis: Results from the Forteo and Alendronate Comparison Trial. J Bone Miner Res 22: S1 S26.
  • 11
    Homminga J, Van-Rietbergen B, Lochmuller EM, Weinans H, Eckstein F, Huiskes R 2004 The osteoporotic vertebral structure is well adapted to the loads of daily life, but not to infrequent “error” loads. Bone 34: 510516.
  • 12
    Eastell R, Hadji P, Farrerons J, Audran M, Boonen S, Brixen K, Melo Gomes J, Obermayer-Pietsch B, Avramidis A, Sigurdsson G, Gluer CC, Cleall S, Marin F, Nickelsen T 2006 Comparison of 3 Sequential Treatment Regimes of Teriparatide: Final Results from the EUROFORS study. J Bone Miner Res 21: S1 S70.
  • 13
    Graeff C, Timm W, Nickelsen TN, Farrerons J, Marin F, Barker C, Gluer CC 2007 Monitoring teriparatide-associated changes in vertebral microstructure by high-resolution CT in vivo: Results from the EUROFORS study. J Bone Miner Res 22: 14261433.
  • 14
    Chevalier Y, Pahr D, Allmer H, Charlebois M, Zysset P 2007 Validation of a voxel-based FE method for prediction of the uniaxial apparent modulus of human trabecular bone using macroscopic mechanical tests and nanoindentation. J Biomech 40: 33333340.
  • 15
    Garcia D, Zysset PK, Charlebois M, Curnier A 2009 A three-dimensional elastic plastic damage constitutive law for bone tissue. Biomech Model Mechanobiol 8: 149165.
  • 16
    Chevalier Y, Charlebois M, Pahr D, Varga P, Heini P, Schneider E, Zysset P 2008 A patient-specific finite element methodology to predict damage accumulation in vertebral bodies under axial compression, sagittal flexion and combined loads. Comput Methods Biomech Biomed Engin 11: 477487.
  • 17
    Gluer CC 1999 Monitoring skeletal changes by radiological techniques. J Bone Miner Res 14: 19521962.
  • 18
    Neer RM, Arnaud CD, Zanchetta JR, Prince R, Gaich GA, Reginster JY, Hodsman AB, Eriksen EF, Ish-Shalom S, Genant HK, Wang O, Mitlak BH 2001 Effect of parathyroid hormone (1-34) on fractures and bone mineral density in postmenopausal women with osteoporosis. N Engl J Med 344: 14341441.
  • 19
    McClung MR, San Martin J, Miller PD, Civitelli R, Bandeira F, Omizo M, Donley DW, Dalsky GP, Eriksen EF 2005 Opposite bone remodeling effects of teriparatide and alendronate in increasing bone mass. Arch Intern Med 165: 17621768.
  • 20
    Thurner PJ, Wyss P, Voide R, Stauber M, Stampanoni M, Sennhauser U, Muller R 2006 Time-lapsed investigation of three-dimensional failure and damage accumulation in trabecular bone using synchrotron light. Bone 39: 289299.
  • 21
    Eswaran SK, Gupta A, Keaveny TM 2007 Locations of bone tissue at high risk of initial failure during compressive loading of the human vertebral body. Bone 41: 733739.
  • 22
    Melton LJ III, Riggs BL, Keaveny TM, Achenbach SJ, Hoffmann PF, Camp JJ, Rouleau PA, Bouxsein ML, Amin S, Atkinson EJ, Robb RA, Khosla S 2007 Structural determinants of vertebral fracture risk. J Bone Miner Res 22: 18851892.