SEARCH

SEARCH BY CITATION

Keywords:

  • HIGH-RESOLUTION PERIPHERAL QUANTITATIVE COMPUTED TOMOGRAPHY;
  • FINITE ELEMENT ANALYSIS;
  • MEN OSTEOPOROSIS;
  • MICROARCHITECTURE;
  • FRACTURE

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Few studies have investigated bone microarchitecture and biomechanical properties in men. This study assessed in vivo both aspects in a population of 185 men (aged 71 ± 10 years) with prevalent fragility fractures, compared to 185 controls matched for age, height, and weight, from the Structure of the Aging Men's Bones (STRAMBO) cohort.

In this case-control study, areal BMD (aBMD) was measured by DXA, bone microarchitecture was assessed by high resolution (HR)-pQCT, and finite element (µFE) analysis was based on HR-pQCT images of distal radius and tibia. A principal component (PC) analysis (PCA) was used to study the association of synthetic PCs with fracture by computing their odds ratio (OR [95%CI]) per SD change. Specific associations with vertebral fracture (n = 100), and nonvertebral fracture (n = 85) were also computed.

At both sites, areal and volumetric BMD, cortical thickness and trabecular number, separation, and distribution were significantly worse in cases than in controls, with differences ranging from −6% to 15%. µFE-derived stiffness and failure load were 8% to 9% lower in fractures (p < .01). No difference in load distribution was found between the two groups. After adjustment for aBMD, only differences of µFE-derived stresses, stiffness, and failure load at the tibia remained significant (p < .05).

PCA resulted in defining 4 independent PCs, explaining 83% of the total variability of bone characteristics. Nonvertebral fractures were associated with PC1, reflecting bone quantity and strength at the radius (tibia) with OR = 1.64 [1.27–2.12] (2.21 [1.60–3.04]), and with PC2, defined by trabecular microarchitecture, with OR = 1.27 [1.00–1.61]. Severe vertebral fractures were associated with PC1, with OR = 1.56 [1.16–2.09] (2.21 [1.59–3.07]), and with PC2, with OR = 1.55 [1.17–2.06] (1.45 [1.06–1.98]).

In conclusion, microarchitecture and biomechanical properties derived from µFE were associated with all types of fractures in men, showing that radius and tibia mechanical properties were relatively representative of distant bone site properties. © 2011 American Society for Bone and Mineral Research.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Although osteoporosis is less frequent in men than in women, 25% to 30% of fragility fractures occur in men.1 The diagnosis, however, is less well recognized in men than in women because the WHO definition of osteoporosis has been established for postmenopausal white women. The diagnosis, based on densitometric measurement (DXA) of areal bone mineral density (aBMD), is defined by a sex-specific T-score < −2.5. However, this score identifies only half of the future osteoporotic fractures that will occur in postmenopausal women.2, 3 In men, this predictive value is even smaller because only 20% of fractures in elderly men are observed among those with a low T-score.4

During the last decade, others aspects of bone quality have been investigated with various techniques in order to better understand the pathophysiology of osteoporosis. Indeed, aBMD measured by DXA reflects mainly the quantity of bone mineral and does not capture other parameters such as geometry, microarchitecture, or biomechanical properties, which also contribute to bone fragility and thus to fracture risk.5

The assessment of bone microarchitecture has been possible in vivo since the recent development of a three-dimensional imaging device that can achieve high resolution.6–9 In the cross-sectional study of a sample of the Rochester, MN, population, men had decreasing trabecular thickness with age, whereas in women the trabecular number diminished,9 as previously observed with histomorphometry of iliac crest biopsies.10 The improvements in resolution of 3D imaging have made possible the use of micro-finite element analysis (µFEA) in order to assess quantitatively the biomechanical properties that result from the bone microarchitecture.11–13 We have shown in previous case-control fracture studies conducted in women that HR-pQCT images of radius and tibia could be used to determine microarchitecture14 and—in combination with µFEA—biomechanical parameters that were associated with prevalent fracture.12, 15 These associations have also been observed by others.16–18

Studies of men are scarce, but a recent investigation of the study of osteorporotic fractures in men (MrOS) cohort, using QCT imaging of the hip, demonstrated the ability of biomechanical properties assessed by FEA to prospectively predict hip fractures in men, independently from aBMD.19

In this study we sought to investigate whether microarchitecture or biomechanical properties of radius and tibia in men, obtained from HR-pQCT images, could be associated with prevalent fractures.

Material and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Subjects

The Structure of the Aging Men's Bones (STRAMBO) study is a single-center prospective cohort study of the skeletal fragility and its determinants in 1169 men aged 20 to 85 years, which has already been described in a previous publication20 and has been approved by an independent ethical committee. Briefly, all participants replied to an epidemiological questionnaire that was administered by an interviewer and covered lifestyle factors and health status. All men able to give informed consent, to answer the questionnaire, and to participate in the diagnostic exams were included. No specific exclusion criteria were used.20

The present analysis was carried out in a subset of 370 men from the STRAMBO cohort: 185 men (mean age, 71 ± 10 years) with a history of fragility fractures and 185 healthy controls without history of fracture, randomly matched for age, height, and weight. Both groups underwent measurement of density and microarchitecture at the distal radius and tibia at the baseline visit.

Fracture evaluation

Peripheral fragility fractures were self-reported in the questionnaire and were not further ascertained. All fracture sites were included except the head, toes, and fingers.

Vertebral fractures were assessed on the lateral scans of the thoracic and lumbar spine obtained in the dorsal decubitus position by Vertebral Fracture Assessment software (VFA, Hologic, Bedford, MA, USA) using the Hologic Discovery A device equipped with the C-arm (Hologic). Using this method, vertebral bodies of the vertebrae from Th7 to L4 were assessable for all the patients, whereas the vertebrae Th5 and Th6 were not assessable in 10% to 11% of participants and the Th4 vertebral body could not be assessed in 23% of men. The vertebral fractures were assessed using the semiquantitative method of Genant,21 modified for men by Szulc et al.22 Specifically, vertebrae were considered fractured if vertebral height was diminished by 20% at the lumbar spine and 15% at the thoracic spine. Vertebral fractures were identified in 100 men who were classified according to their most severe fracture, resulting in 18 men as grade 1 (20–25% reduction of vertebral height), 62 as grade 2 (25–40% reduction of vertebral height) and 20 as grade 3 (>40% reduction of vertebral height).21 We used slightly modified cutoffs for wedge deformities at the level of dorsal kyphosis (T6–T9): grade 1, 25–30%; grade 2, 30–40%; and grade 3, >40%.22

Men who reported vertebral fractures sustained after major trauma were excluded from this analysis. Particular attention was paid to excluding mild vertebral deformities supposedly related to conditions other than osteoporosis (such as osteoarthritis, Scheuerman's disease). The reproducibility of the diagnosis of vertebral fracture was assessed using the simple kappa score [95% CI] per fracture (yes vs no) and per grade (no fracture, grade 1, grade 2, grade 3). The intraobserver agreement scores were κ = 0.93 [0.84–0.94] and κ = 0.89 [0.84–0.94], respectively. The interobserver agreement scores were κ = 0.88 [0.81–0.95] and κ = 0.87 [0.82–0.92], respectively. PS read the VFA images.

Measurement of bone mineral density and bone microarchitecture

As already described in earlier work on the STRAMBO study, ultradistal radius and total hip areal bone mineral densities (aBMD, g/cm2) were measured using DXA (Hologic, Bedford, MA, USA). Moreover, volumetric bone mineral densities and microarchitecture were measured at the distal radius and tibia using a HR-pQCT device (XtremeCT, Scanco Medical AG, Brüttisellen, Switzerland) that acquires a stack of 110 parallel CT slices with an isotropic voxel size of 82 µm.8 Methods used to process the CT data have been described previously.8 The standard HR-pQCT parameters that were used were bone cross-sectional area (CSA, mm2); volumetric bone density (g/cm3) for total (Dtot) and trabecular (Dtrab) bone; cortical thickness (CTh, µm); and trabecular number (TbN, mm−1), separation (TbSp, µm), and intra-individual distribution of separation (TbSpSD, µm).7

Both DXA and HR-pQCT scans were acquired on the nondominant forearm and nondominant hip and tibia. If there was a history of fracture on that limb, the nonfractured limb was measured.

Finite Element Analysis

Finite element models of the radius and the tibia were created directly from the segmented HR-pQCT images using software delivered with the HR-pQCT device (IPL v1.12, Scanco Medical AG). A custom script specifying a minimum cortical thickness of 6 voxels, without filling the porosity (ie, without adding any bone voxel), was used to differentiate cortical and trabecular bone. A voxel-conversion procedure was used to convert each bone voxel tissue into an equally sized brick element,11 thus creating µFE models that could represent the actual trabecular architecture in detail. The models contained approximately 2 million elements at the radius and 5 million elements at the tibia. All models could be solved in approximately 3 and 5 hours, respectively. Material properties chosen were homogeneous, isotropic, and elastic for all voxels. Cortical bone elements were assigned a Young's modulus of 20 GPa, whereas trabecular tissue elements were assigned a Young's modulus of 17 GPa,23 and for all elements a Poisson's ratio of 0.3 was specified.

The µFE simulation for the radius bone has been previously described in detail.12 Briefly, a compression test was simulated in which a displacement in the longitudinal direction was applied at one end while the other end was fully constrained, to simulate a fall from standing height on the outstretched hand,24 representing the most common trauma associated with Colles fractures. The boundary conditions are the same as those used in the validation study of Pistoia et al.25 In that study, high correlations between measured and calculated failure loads were found when using a criterion that states that failure is expected when 2% of the bone tissue is strained beyond a critical limit of 7000 µstrain (R2 = 0.75; p < .001). However, because Pistoia et al. used elastic properties of 10 GPa for all tissues, in our study we had to set the critical strain at 3500 µstrain in order to get comparable values for the failure load. The µFE models were the same for radius and tibia analyses. We have previously shown in postmenopausal women that this model, despite its validation at the radius only, provided similar prediction of fracture when used at the tibia and at the radius.15

In addition to failure load [N], µFEA-derived variables used in our study also included stiffness [kN/mm], the percentage of load carried by the trabecular bone at the distal and proximal surface of the volume of interest (% load trab distal, and % load trab proximal, respectively), and the average and SD values of the von Mises stress in the trabecular and cortical bone (Trab av stress and Cort av stress; Trab SD stress and Cort SD stress, all in [MPa]).

Statistical analysis

Descriptive statistics were standardized by means and SDs. Depending on the distribution of variables, Wilcoxon signed rank tests or Student t-tests were used to compare men with and without fractures for densities, microarchitecture, and mechanical parameters. The differences between the two groups were expressed in percentages and in Z-scores using means and SDs of the controls. Correlations between all the variables were assessed by Pearson or Spearman correlation, depending on the distribution of the variable.

Correlations among architectural parameters measured by HR-pQCT and between architectural parameters with aBMD measured by DXA at the radius have already been reported in women.8, 14 These correlations, also found in men (|r| = 0.65 to 0.98), revealed multicolinearity issues for many parameters.

To address this issue, we conducted a principal components analysis (PCA) among the 370 men for the radius and the tibia separately. All parameters of biomechanical properties, volumetric densities, microarchitecture, age, height and weight, and aBMD either at the ultradistal radius or the total hip were introduced, after standardization, in the PCA.26 Trabecular thickness and volumetric cortical density were not included in this analysis, because they are strongly affected by partial volume effects.

PCA is a statistical technique that transforms a set of correlated variables into a substantially smaller set of uncorrelated parameters, defined as the principal components (PCs). Those PCs are linear combinations of the original parameters, which summarize most of the information (or variance) of the original dataset: the first principal component accounts for as much of the data variability as possible, with the remaining variance being explained decreasingly by the following PCs. In this study, the optimization of information carried by each PC was operated by a varimax rotation step that terminated the PCA.26 Finally, only PCs with eigenvalues > 1.0 were selected for further analyses. Each PC can be interpreted on the basis of the initial variables' weight. Associations between PCs and fracture were evaluated by conditional logistic regression analysis and expressed as odds ratio (OR [95% CI]) per SD change.

A subgroup analysis was performed to explore the relation between µFEA and fracture types with a logistic regression, specifically for vertebral and nonvertebral fractures. The associations between PCs and fracture type were computed by multinomial logistic regression where cases were compared to the entire control pool, with adjustment for age, height, and weight.

Statistical analyses were performed using SPSS software (SPSS v16.0, SPSS Inc., Chicago, IL, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Among the 185 men with fracture, 168 vertebral fractures were identified in 100 men and 121 nonvertebral fractures were identified in 108 men at the wrist (n = 35), ribs (n = 21), ankle (n = 11), tibia (n = 10), proximal humerus (n = 9), fibula (n = 9), hip (n = 8), metatarsus (n = 5), elbow (n = 5), clavicula (n = 2), omoplate (n = 2), calcaneum (n = 3) and pelvis (n = 1).

Seven men in the control group and 18 in the fracture group were taking treatment known to affect bone metabolism: vitamin D (n = 22), calcium (n = 21), bisphosphonate (n = 13), and testosterone (n = 1). Results remained the same when those men were excluded from the analysis.

Age, height, and weight were similar in both groups: average values were 71 ± 10 years, 167 ± 4 cm, and 77 ± 10 kg, respectively. In men with fracture, total hip and radius aBMD were significantly lower compared to controls (0.895 ± 0.136 vs 0.961 ± 0.133 and 0.426 ± 0.069 vs 0.454 ± 0.071, respectively, with p < .01).

Microarchitecture

Twenty-three radius scans and 10 tibia scans were excluded because of poor quality (movement artefacts). Thus, there were 347 exploitable measurements at the radius and 360 at the tibia; ie, for pairwise comparisons, there were 162 and 175 pairs of complete data respectively at the radius and the tibia.

Both at the radius and the tibia, men with fracture had lower Dtot, Dtrab, and CTh (−8% to −11%; p < .01); the Dcort difference was not significant. TbN and TbSp were significantly impaired in cases (−4% to 8%; p < .05). TbSpSD was higher in cases at the radius (12%; p < .01), but not at the tibia (10%; p > .05) (Table 1).

Table 1. Microarchitecture and Biomechanical Characteristics (Mean ± SD) of Men With Fracture and Controls Matched for Age, Height, and Weight
 Distal radiusDistal tibia
ControlsMen with fractureDifferenceControlsMen with fractureDifference
(n = 168)(n = 159)%Z-Score(n = 173)(n = 169)%Z-Score
  • Differences between cases and controls:

  • *

    p < .05;

  • **

    p < .01

D tot (mg HA/cm3)291 ± 61268 ± 60**−8.19−0.39290 ± 55267 ± 58**−7.79−0.41
D trab (mg HA/cm3)173 ± 37158 ± 36**−8.78−0.40171 ± 35158 ± 36**−7.55−0.36
TbN (mm−1)1.86 ± 0.251.76 ± 0.26**−5.36−0.391.71 ± 0.291.64 ± 0.31*−3.99−0.23
TbSp (µm)469 ± 81507 ± 116**8.240.47519 ± 110556 ± 157*7.210.34
TbSpSD (µm)208 ± 57233 ± 93**12.040.43254 ± 92281 ± 14710.770.30
CTh (µm)690 ± 228619 ± 203**−10.29−0.311195 ± 2771076 ± 315**−9.94−0.43
CSA (mm2)385 ± 59391 ± 631.560.10823 ± 122830 ± 1210.780.05
% trab distal load (%)58 ± 8.457.6 ± 7.9−0.62−0.0452.6 ± 7.953.8 ± 10.32.380.16
% trab proximal load(%)18.3 ± 6.917.5 ± 6.8−4.05−0.1127.9 ± 6.528.3 ± 8.71.480.06
Trab average stress (MPa)5.6 ± 0.965.88 ± 0.94**5.070.292.51 ± 0.372.7 ± 0.53**7.600.51
Trab SD stress (MPa)3.67 ± 1.024.03 ± 1.02**9.730.351.32 ± 0.251.49 ± 0.41**12.450.65
Cort average stress (MPa)10.34 ± 2.4711.18 ± 2.28**8.090.344.39 ± 0.714.79 ± 0.99**9.260.57
Cort SD stress (MPa)3.22 ± 0.93.53 ± 1.01**9.540.340.94 ± 0.181.08 ± 0.38**14.590.76
Stiffness (kN/mm)176.5 ± 41.3160.1 ± 35.3**−9.29−0.40461.4 ± 76.8420.7 ± 77.6**−8.81−0.53
Estimated failure load (kN)4.17 ± 0.963.79 ± 0.80**−9.02−0.3910.85 ± 1.759.97 ± 1.76**−8.16−0.50

Finite Element Analysis

µFEA showed differences in load distribution between trabecular and cortical bone at the distal and proximal regions at the radius and tibia. At proximal regions cortical bone sustained more load than trabecular bone (72% to 82%). At distal regions, the share of load was more balanced, although more load was sustained by trabecular bone (53% to 58%). No differences were observed between cases and controls (Table 1).

Average stresses and their SDs were higher at both sites (5% to 15%) in men with fractures compared to controls (Table 1). Figure 1 and Fig. 2 illustrate the differences in the distribution of stresses in a control and in a case, respectively, at the radius and the tibia. Stiffness and estimated failure load were significantly lower in cases, both at the radius and the tibia (−8% to −9%; p < .01).

thumbnail image

Figure 1. Distribution of von Mises stresses in the bone tissue for a control28 and a fracture case (C,D) at the radius. The colors indicate the stress levels, with red for high stresses and blue for low stresses. (A,B) Distribution of von Mises stresses in the first distal (A) and last proximal (B) slices in a man who never had a fracture. (C,D) Distribution of von Mises stresses in the first distal (C) and first proximal (D) slices in a man who had fracture.

Download figure to PowerPoint

thumbnail image

Figure 2. Distribution of von Mises stresses in the bone tissue for a control28 and a fracture case (C,D) at the tibia. The colors indicate the stress levels, with red for high stresses and blue for low stresses. (A,B) Distribution of von Mises stresses in the first distal (A) and last proximal (B) slices in a man who never had a fracture. (C,D) Distribution of von Mises stresses in the first distal (C) and first proximal (D) slices in a man who had fracture.

Download figure to PowerPoint

Table 2. Odds Ratio for Associations Between Prevalent Fractures and Individual Parameters of aBMD, vBMD, Microarchitecture, and Biomechanical Parameters
 Distal radiusDistal tibia
OR [95% CI]Adjusted OR [95% CI]OR [95% CI]Adjusted OR [95% CI]
For 1 SD decrease
 aBMD (Radius/Hip)1.64 [1.28–2.10]1.82 (1.39–2.38]
 Dtot1.70 [1.29–2.24]1.36 [0.84–2.20]1.66 [1.28–2.17]1.32 [0.96–1.81]
 Dcomp1.40 [1.06–1.84]1.01 [0.70–1.44]1.50 [1.15–1.94]1.24 [0.93–1.66]
 Dtrab1.69 [1.29–2.22]1.31 [0.85–2.02]1.57 [1.21–2.04]1.20 [0.87–1.65]
 CTh1.59 [1.21–2.08]1.18 [0.78–1.79]1.68 [1.29–2.19]1.40 [1.03–1.89]
 TbN1.50 [1.16–1.92]1.14 [0.83–1.57]1.29 [1.01–1.63]1.02 [0.77–1.35]
 TbTh1.51 [1.16–1.97]1.10 [0.78–1.56]1.34 [1.06–1.71]1.20 [0.93–1.55]
 % load trab distal1.07 [0.83–1.39]1.05 [0.80–1.38]1.22 [0.97–1.54]1.15 [0.90–1.48]
 % load trab proximal1.21 [0.93–1.58]1.08 [0.82–1.43]1.12 [0.89–1.40]1.09 [0.85–140]
 Stiffness1.80 [1.37–2.38]1.81 [1.03–3.18]2.06 [1.53–2.79]1.80 [1.23–2.62]
 Failure load1.72 [1.35–2.34]1.66 [0.97–2.85]1.98 [1.47–2.65]1.68 [1.16–2.43]
For 1 SD increase
 CSA1.10 [0.88–1.39]1.13 [0.88–1.44]1.07 [0.87–1.33]1.06 [0.84–1.34]
 TbSp1.57 [1.17–2.09]1.15 [0.80–1.64]1.37 [1.06–1.77]1.01 [0.74–1.39]
 TbSpSD1.49 [1.09–2.03]1.11 [0.80–1.56]1.31 [1.00–1.71]1.02 [0.74–1.41]
 Trab av stress1.37 [1.08–1.73]1.06 [0.76–1.50]1.61 [1.23–2.10]1.33 [0.97–1.81]
 Trab SD stress1.62 [1.24–2.13]1.08 [0.65–1.82]2.21 [1.54–3.16]1.85 [1.20–2.85]
 Cort av stress1.57 [1.21–2.04]1.02 [0.60–1.73]1.96 [1.43–2.69]1.64 [1.12–2.41]
 Cort SD stress1.63 [1.22–2.19]1.15 [0.76–1.75]2.57 [1.68–3.95]2.04 [1.23–3.30]

Correlations

At both sites, µFE-derived stiffness was highly correlated with aBMD and vBMD (r = 0.65–0.88; p < .001), CTh (r = 0.66–0.75), as well as average and SD of cortical and trabecular stresses (|r| = 0.74–0.98) and failure load (r = 0.99) (data not shown). Its correlation with TbN and TbSpSD was weaker (|r| = 0.45–0.58; p < .001).

The load distribution variables correlated with cortical thickness moderately at the tibia (r = −0.58–−0.62; p < .001) and more weakly at the radius (r = −0.42; p < .05). So we chose to construct new parameters characterizing the distal radius and tibia through a PCA to avoid multicolinearity issues.

PCA

For PCA performed either at the radius or the tibia, 3 PCs explaining jointly 83% of the total variance were defined in similar ways at both sites. Table 3 presents each original parameter's correlation with each PC. These correlations indicate the degree and direction of each of the original variable's contribution to each component.

Table 3. Principal Components Loading Matrix
 Distal radiusDistal tibia
Component of VariablesBone strength and quantityTrabecular microarchitectureLoad distributionMorphologyBone strength and quantityTrabecular microarchitectureLoad distributionMorphology
  1. The highest weights (>0.6) are marked in bold.

  2. UD Radius and Total Hip aBMD were included in the PCA model at the radius and the tibia, respectively.

Age−0.28−0.090.50−0.21−0.370.180.35−0.31
Height0.060.23−0.100.710.110.52−0.030.47
Weight0.130.07−0.010.780.240.140.110.74
aBMD Radius/Hip0.830.41−0.090.080.630.42−0.100.11
CSA0.01−0.330.580.62−0.170.060.360.79
D tot0.750.47−0.36−0.210.820.30−0.23−0.36
CTh0.750.15−0.60−0.110.720.12−0.58−0.24
D trab0.630.690.20−0.130.750.460.33−0.23
TbN0.300.880.120.140.340.860.110.09
TbSp−0.34−0.90−0.10−0.14−0.43−0.87−0.11−0.03
TbSp SD−0.30−0.87−0.02−0.14−0.39−0.82−0.03−0.06
% Tb distal load0.030.180.88−0.01−0.070.030.950.15
% Tb proximal load0.060.280.880.03−0.030.080.940.20
Trab av stress−0.73−0.160.28−0.42−0.61−0.530.37−0.33
Trab SD stress−0.92−0.30−0.03−0.11−0.88−0.330.04−0.13
Cort av stress−0.92−0.23−0.12−0.18−0.89−0.33−0.02−0.17
Cort SD stress−0.89−0.220.09−0.05−0.81−0.260.30−0.10
Stiffness0.960.19−0.020.080.940.24−0.040.12
Percentage of variance50.4%16.2%10.0%7.6%50.2%17.8%8.3%6.3%

The first PC, called stiffness, was composed of quantitative bone variables (aBMD, vBMDs, and CTh) and of µFE-derived cortical and trabecular stresses, and µFE-derived stiffness. This PC may be referred to as the component of bone strength and quantity, and it explained by itself 50% of the total variance. The second PC, which included TbN and TbSpSD, may be called the trabecular microarchitecture PC, which explained up to 17% of the total variance. The third PC, which was mainly defined by the load distribution on distal and proximal regions, could be called the load distribution component; it explained 10% and 8% of variance for the radius and the tibia respectively. A fourth PC, correlated to height, weight, and bone CSA, was defined and can be called the morphology PC; it explained about 7% of variance at both sites.

Logistic regression analyses

Using conditional logistic regression, we computed the association between all individual variables and PCs with the fracture status (Table 2 and Fig. 3). At the radius, significant associations between individual microarchitecture parameters with fractures were obtained, but those were lost after adjustment for radius aBMD. Among biomechanical parameters only the stiffness was significantly associated with fractures after adjustment for aBMD (OR = 1.81 [1.03–3.18] for 1 SD decrease). At the tibia, associations were observed before adjustment for aBMD, but only associations between fractures and CTh (OR = 1.40 [1.03–1.89] for 1 SD decrease), stiffness (OR = 1.80 [1.23–2.62] for 1 SD decrease), failure load, or stress parameters remained significant after adjustment for hip aBMD.

thumbnail image

Figure 3. Association between PCs and fracture, expressed as ORs [95% CI] per SD change. Association with fracture was first computed for the 185 men with fracture compared to their age-matched control. Then a subanalysis was performed by grouping the different types of fractures, separating vertebral fractures (VFx) from nonvertebral fractures; in this case ORs were adjusted for age, height, and weight.

Download figure to PowerPoint

Figure 3 presents the results of logistic regression OR results for each PC. At the radius, both the bone strength and quantity PC (OR = 1.64 [1.27–2.12] per SD decrease, resulting from decreases of BMD and failure load or increases of stress) and the trabecular microarchitecture PC (OR = 1.27 [1.00–1.61] per SD decrease, due to decrease in the trabecular number or increase in the distribution of trabecular separation) were significantly associated with fracture. The association between the load distribution PC and fractures was not significant (OR = 1.09 [0.82–1.45]). At the tibia, we found that only the bone strength and quantity PC was significantly associated with fracture (OR = 2.21 [1.60–3.04]). The second and third PCs were not associated with fractures (OR = 1.09 [0.86–1.37] and 1.10 [0.87–1.40], respectively, for trabecular microarchitecture PC and load distribution PC). No significant associations were found between the morphology PC and fractures.

Vertebral fractures vs nonvertebral fractures

When we explored separately by ANOVA men with at least one moderate or severe vertebral fracture (grade 2 and 3, n = 82), mild vertebral fracture (grade 1, n = 18), and nonvertebral fracture (n = 85), we found differences only for the load distribution component at the radius and hip aBMD after adjustment for age, height, and weight (p < .01). For other PCs or radius aBMD, no significant differences were observed (Table 4).

Table 4. Subanalysis by ANOVA With Different Types of Fracture
 p
ANOVANon VFx vs VFx1Non VFx vs VFx+VFx1 vs VFx+
  1. “Non VFx” stands for the non vertebral fracture group, “VFx1” for the group of mild vertebral fracture, and “VFx + ” for the group of moderate and severe vertebral fractures.

Radius aBMD0.2850.9630.3280.521
Hip aBMD0.0030.4470.0160.013
Radius PC1 bone strength and quantity0.1830.2930.3080.804
Radius PC2 trabecular microarchitecture0.0870.4890.2940.110
Radius PC3 load distribution0.0030.1100.0030.997
Tibia PC1 bone strength and quantity0.1150.9390.0970.684
Tibia PC2 trabecular microarchitecture0.0850.4770.2830.113
Tibia PC3 load distribution0.0790.3420.0930.975

Then each fracture group was compared to the group of 185 controls in a multinomial logistic regression model, adjusting for age, height, and weight, because age-matching was not possible here. At both the radius and tibia, the bone strength and quantity PC and the trabecular microarchitecture PC were associated with the presence of moderate-severe vertebral fractures, with ORs ranging from 1.45 [1.06–1.93] to 2.21 [1.59–3.07]. Mild vertebral fractures were significantly associated only with the bone strength and quantity PC at the radius, with OR = 1.82 [1.05–3.18]. No association was found with any PC at the tibia. Nonvertebral fracture was associated with bone strength and quantity component PC, with OR = 1.44 [1.07–1.93] and 1.83 [1.33–2.52], respectively, at the radius and the tibia. At the radius, trabecular microarchitecture PC and load distribution PC had association with fracture close to statistical significance (OR = 1.32 [0.98–1.79] and 1.37 [0.98–1.91], respectively). Such effect could not be observed at the tibia.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

In this nested case-control study carried out in a cohort of men, we found that microarchitectural bone parameters and µFE-derived biomechanical properties of the distal radius and tibia, based on in vivo HR-pQCT scans, were associated with all types of fractures. Moderate and severe vertebral fractures were associated with the bone strength and quantity PC, and the trabecular microarchitecture PC both at the distal radius and tibia.

PCA applied in this group of men defined PCs similar to those previously obtained in fracture case-control studies in women.12, 15 This consolidates our approach by using PCA for studying a set of bone characteristic parameters that are highly correlated,12, 15 in agreement with other authors' findings.17, 18, 27

In this study, men with moderate and severe vertebral fracture were clearly described by impaired aBMD at the hip or bone strength and trabecular microarchitecture components at both the radius and tibia. Hence, we cannot suggest that the weight-bearing function of the tibia helps to preserve its microarchitecture, as opposed to what we have observed in women.15

In women we have reported that vertebral fractures were associated with trabecular microarchitecture only at the radius, as well as with bone strength and quantity at the tibia.15 The difference observed could be due to a limited number of vertebral fracture cases in the previous study of women, because there were only 34 vertebral fractures cases, including 20 severe cases. Another study of the microarchitecture at the radius and tibia in 100 women with vertebral fracture reported a trend for a gradual impairment of the microarchitecture with the severity and number of vertebral fractures.28

The results for groups with mild vertebral fractures and nonvertebral fractures are consistent with what was found in women with nonvertebral fractures,15 because the microarchitecture PC was not associated with fractures. In men the associations of fractures with microarchitecture or load distribution PC at the radius, however, were close to statistical significance, which is similar to our results obtained in women who had a wrist fracture.12

Our results regarding vertebral fracture in men confirm the possibility of estimating fracture at distant sites with HR-pQCT, which was already observed in prior studies.15, 16, 28 Also, we confirm the previous data indicating the association between the type of fracture and the type of bone microarchitectural impairment assessed by bone histomorphometry29 or HR-pQCT.15, 16

So far, bone fragility in men using QCT19, 30 or HR-pQCT9 imaging techniques have been poorly investigated. The microarchitecture investigation in a sample of the Rochester, MN, population has shown that if the decrease in trabecular density with age had the same magnitude in men and women, the effective bone loss was different: when women suffered loss of trabeculae and an increase of trabecular separation, men had their trabecular number remaining stable as they were thinning with time.9 The general observation that men have better bone microarchitecture is a finding we share with those authors when we compare the control men from this study to the control women of our previous study from the OFELY cohort.15 Consequently, men had better bone strength than women at both the radius and tibia. In the meantime, at the radius we could notice a different load balance between cortical and trabecular bone between the controls groups of both population; eg, cortical bone sustained 53% and 42% of the load at the distal end in women and men, respectively. Despite being thicker than in women, cortical bone has a lesser contribution than trabecular bone in men. This may be explained by the wider trabecular bone CSA and a higher Dtrab, which indicates a higher content of trabecular bone than in women, providing a greater influence on the load distribution balance. At the tibia, the differences in CSA and Dtrab between men and women had the same magnitude, but the load balance was similar in both populations, as cortical bone sustained 47% of load at the distal end.

In the longitudinal study investigating bone fragility in the MrOS cohort,19 Orwoll et al. have demonstrated that the use of the load-to-strength ratio, ie, the impact strength for a sideways fall over the FE estimation of the hip bone strength, could predict incident hip fractures independently from aBMD. Those authors have also shown that there was a threshold in bone strength (3000 N), which identified most of the men that suffered a hip fracture during follow-up. In our study, when we calculated the load-to-strength ratio on the radius bone, it was not possible to determine a threshold identifying people who had had a wrist fracture (data not shown).

Among the strengths of our work, we used techniques with good reproducibility of in vivo measurement of both the regular HR-pQCT measurements and of the biomechanical properties derivation by µFEA.8, 31

However, our study also has limitations. In addition to the case-control design of our study, nonvertebral fractures were self-reported by the patient without any clinical confirmation, unlike in our previous fracture studies in women. Indeed, in the OFELY cohort, fractures were recorded during follow-up and were clinically confirmed. Still, self-reported fractures in various studies such as the SOF or the EPOS study,32, 33 or in another study in a population of elderly,34 were considered to be quite accurate. Therefore, these reports support the use of self-reported fracture in our study. Also, the number of mild vertebral fractures is rather low compared to other groups of fracture. Moreover, the µFE model used in this study is linear and attributed homogeneous material properties to bone elements. The simplicity of this model, however, does not bias substantially the biomechanical outcomes, compared to more sophisticated models. Is it also more applicable than nonlinear models because it requires fewer computing resources and can be solved in a shorter time.31

In conclusion, areal and volumetric BMD, microarchitecture, and biomechanical properties derived from µFEA based on in vivo radius and tibia HR-pQCT images were associated with all types of fractures in men, showing that radius and tibia mechanical properties were relatively representative of distant bone site properties. Specifically, moderate and severe vertebral fractures were associated with both aspects of bone strength and quantity and trabecular microarchitecture PCs, whereas nonvertebral fractures were only associated with the bone strength and quantity PC. Prospective investigations are ongoing to confirm those results and determine whether µFEA can improve fracture risk prediction in men.

Disclosures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Dr van Rietbergen serves as a consultant for Scanco Medical AG. All other authors state that they have no conflicts of interest.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

We thank S. Giraud, C. Quillon and D. Foesser for valuable technical assistance. This study was supported in part by research grants from Roche, Agence Nationale de la Recherche, and Hospices Civils de Lyon.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References