The authors state that they have no conflicts of interest.
Alterations of Cortical and Trabecular Architecture Are Associated With Fractures in Postmenopausal Women, Partially Independent of Decreased BMD Measured by DXA: The OFELY Study†
Article first published online: 18 DEC 2006
Copyright © 2007 ASBMR
Journal of Bone and Mineral Research
Volume 22, Issue 3, pages 425–433, March 2007
How to Cite
Sornay-Rendu, E., Boutroy, S., Munoz, F. and Delmas, P. D. (2007), Alterations of Cortical and Trabecular Architecture Are Associated With Fractures in Postmenopausal Women, Partially Independent of Decreased BMD Measured by DXA: The OFELY Study. J Bone Miner Res, 22: 425–433. doi: 10.1359/jbmr.061206
- Issue published online: 4 DEC 2009
- Article first published online: 18 DEC 2006
- Manuscript Accepted: 13 DEC 2006
- Manuscript Revised: 9 NOV 2006
- Manuscript Received: 27 JAN 2006
- bone microarchitecture;
- postmenopausal women;
We assessed the role of low aBMD and impaired architecture—assessed by an HR-pQCT system—in a case-control study of postmenopausal women with fractures. Vertebral and nonvertebral fractures are associated with low volumetric BMD and architectural alterations of trabecular and cortical bone, independent of aBMD assessed by DXA.
Introduction: Alterations of bone architecture and low BMD both contribute to skeletal fragility, but the contribution of cortical and trabecular architecture, independently of areal BMD (aBMD), to the risk of fracture in postmenopausal women has not been thoroughly evaluated. We assessed the role of impaired architecture and low BMD in postmenopausal women with fractures.
Materials and Methods: A matched case-control study in women from the OFELY cohort was performed after 13 years of follow-up. One hundred one women (mean, 73.7 ± 8 years) who sustained a fragility fracture during the follow-up of the study were age-matched with one control who never had a fracture. Density and architecture at the distal radius and tibia were measured with high-resolution pQCT (HR-pQCT) using an XTreme CT (Scanco Medical AG, Bassersdorf, Switzerland). aBMD at the total hip and ultradistal radius was measured by DXA.
Results: There were 80 peripheral fractures in 72 women, 44 vertebral fractures in 34 women, and both types of fractures in 5 women over the 14 years of follow-up. At the distal radius, women with fractures had lower volumetric total (D tot) and trabecular (D trab) BMDs, BV/TV, cortical thickness (Cort Th), trabecular number (TbN), and trabecular thickness (TbTh) and higher trabecular separation (TbSp) and distribution of trabecular separation (TbSpSd) than controls without fractures. In a logistic model, each SD decrease of volumetric total and trabecular densities was associated with a significantly increased risk of fracture at both sites (ORs ranged from 2.00 to 2.47). After adjusting for aBMD measured by DXA at the ultradistal radius, differences between cases and controls remained significant for D trab, and there was a similar trend for TbN, TbSp, and TbSpSd, with adjusted ORs ranging from 1.32 to 1.50. At the distal tibia, before and after adjusting for total hip aBMD, differences between cases and controls remained significant for D tot, D trab, Cort Th, and TbTh, with adjusted ORs ranging from 1.80 to 2.09.
Conclusions: In postmenopausal women, vertebral and nonvertebral fractures are associated with low volumetric BMD and architectural alterations of trabecular and cortical bone that can be assessed noninvasively and that are partially independent of aBMD assessed by DXA.
Osteoporosis is a systemic skeletal disease characterized by low bone mass and microstructural deterioration of bone tissue with a consequent increase in bone fragility.(1)
Until recently, only bone mass and mineral density were routinely assessed in patients. DXA is currently the most frequently used technique to measure areal BMD (aBMD) to assess the risk of fragility fracture. Nevertheless, several limitations are associated with aBMD measurements: (1) DXA does not measure true volumetric BMD; (2) DXA cannot distinguish between cortical and trabecular bone compartments; and (3) DXA does not have an adequate resolution to measure cortical and trabecular architecture. However, the relevance of trabecular microstructure to bone strength has been well documented,(2,3) and the evaluation of both microarchitecture and BMD may improve estimation of the risk of fracture.
Since its introduction by Feldkamp et al.,(4) high-resolution μCT has become an important tool to quantify the morphology of the trabecular structure of bones with a 3D evaluation in vitro. In recent years, noninvasive imaging methods have been developed, enabling a 3D evaluation of bone architecture in vivo. MRI technology has the advantage of a nonionizing radiation technique and can provide both cortical and trabecular bone structure evaluation. μMRI has been performed in vivo at peripheral sites including the distal radius,(5–8) calcaneus,(9,10) distal tibia,(11) and, more recently, the femoral neck.(12) Although MRI is widely available, these measurements of bone architecture require special equipment and software and are still research procedures that cannot be performed routinely. CT-based evaluation of trabecular architecture has been performed in vivo both at peripheral sites, distal radius, and tibia, and, to some extent, at the lumbar spine.(13,14) At peripheral sites, improvements in spatial resolution have enabled 3D determination of trabecular bone architecture with sufficient accuracy.(15–18) It has been shown that measures of bone structure obtained from conventional histological 2D sections and 3D μCT data are highly correlated, indicating that 3D μCT data sets can be used as a substitute for conventional histological sections for bone structure evaluation.(19,20)
Recently, a high-resolution HR-pQCT system that permits in vivo assessment of cortical and trabecular architecture at the distal radius and tibia with a voxel size of 82 μm3 has been developed.(21,22) In a preliminary study performed in a small population of osteopenic women, some trabecular parameters were different between osteopenic women with and without a history of fracture.(21) However, the independent contribution of aBMD and cortical and trabecular architecture to the risk of fracture in postmenopausal women, whether they are normal, osteopenic, or osteoporotic according to the WHO definition, has not yet been evaluated.
The aim of this study was to determine if radial and tibial bone structure evaluated with a HR-pQCT system is able to discriminate postmenopausal women who have sustained a fragility fracture from controls without history of fracture, independent of aBMD evaluated by DXA at the same site.
MATERIALS AND METHODS
We performed a matched case-control study from the OFELY cohort. OFELY is a prospective study of the determinants of bone loss in 1039 volunteer women, 31–89 years of age, recruited between February 1992 and December 1993, randomly selected from the affiliates of a large health insurance company (Mutuelle Générale de l' Education Nationale) from the Rhône district with an annual follow-up. The OFELY cohort has been described elsewhere.(23,24) The first 101 women (mean age, 73.7 ± 8 years) who sustained a fragility fracture during the 13 years follow-up of the study and who had measures of density and architecture at the distal radius and tibia at the visit performed at the end of the 13 years were randomly age-matched at the same visit with one control from the same cohort who never had a fracture.
Previous fragility fractures were those annually registered during the 13-year follow-up, all confirmed by radiographs (all sites were included, except head, toes, and fingers). Only low-trauma fractures (i.e., those occurring after falls from standing height or less) were taken into account. Vertebral fractures were identified on lateral X-ray films of the thoracic and lumbar spine according to the semiquantitative method of Genant et al.(25) The clinical vertebral fractures (radiographically confirmed) were collected in the annual questionnaire, whereas vertebral fractures that did not reach clinical attention were assessed on the X-ray films performed every 4 years. We excluded vertebral fractures that occurred because of major trauma.
Bone microarchitectural measurement
The nondominant forearm (or the nonfractured forearm in the case of prior fracture) and distal tibia were scanned using a high-resolution 3D pQCT device (XTreme CT; Scanco Medical AG, Bassersdorf, Switzerland). This system uses a 2D detector array in combination with a 0.08-mm point-focus X-ray tube, enabling the simultaneous acquisition of a stack of parallel CT slices with a nominal resolution (voxel size) of 82 μm. The following settings were used: effective energy of 60 kVp, X-ray tube current of 95 mA, and matrix size of 1536 × 1536. Attenuation data were converted to equivalent hydroxyapatite (HA) densities. A phantom scan was performed every day. The details of the acquisition and analysis has been described previously(21) and is summarized here. At each skeletal site, 110 CT slices were obtained, thus delivering a 3D representation of ∼9 mm in the axial direction. The entire volume of interest was automatically separated into a cortical and trabecular region. Thus, we obtained volumetric total (D tot), trabecular (D trab), and cortical (D cort) BMDs in milligrams of HA per cubic centimeter. Mean cortical thickness (Cort Th) was defined as the mean cortical volume divided by the outer bone surface. Trabecular bone volume fraction (BV/TV, %) was derived from trabecular density, assuming fully mineralized bone to have a mineral density of 1.2 g HA/cm3. Among the trabecular structural parameters, trabecular number (TbN, 1/mm) was measured, whereas trabecular thickness (TbTh, μm) and separation (TbSp, μm) were derived from BV/TV and TbN using standard methods from histomorphometry [i.e., TbTh = (BV/TV)/TbN and TbSp = (1 − BV/TV)/TbN].(26) Distance transformation techniques also enable the calculation of the distribution of trabecular separation (TbSp SD, μm), reflecting the heterogeneity of the trabecular network.(7) The in vivo precision error of density (total, trabecular, and cortical) measurements, expressed as the coefficient of variation, ranged from 0.7% to 1.5%. The reproducibility of structural parameters was slightly lower, with CVs ranging from 0.9% to 4.4%.(21)
aBMD was measured at the same visit by DXA, at the radius and the total hip (QDR 1000+ and 4500; Hologic, Waltham, MA, USA). For the radius, we analyzed specifically the ultradistal site because the bone area measured is larger than but includes the area analyzed with HR-pQCT (Fig. 1). The in vivo precision error of DXA, expressed as the coefficient of variation, was 1.2% for ultradistal and 1% for total hip. A control phantom was scanned every day, and all DXA measurements were performed by the same experienced operator.
Clinical evaluation and physical examination
Women completed a questionnaire at the same visit and at each prior annual follow-up, as described previously.(27) Data related to the use of treatments affecting bone metabolism (bisphosphonates, hormone replacement therapy, selective estrogen receptor modulator, tibolone, or aromatase inhibitors) were collected every year and considered in this study if the duration of treatment was ≥1 year at the time (or within 6 months before) of architectural evaluation. Height and body weight were recorded at the same visit.
χ2 and paired t-tests were used to compare characteristics between women with and without fractures. Correlation analysis between density and architecture parameters assessed by HR-pQCT and aBMD assessed by DXA was performed with the test of Spearman, because most of the variables were not normally distributed. Diagnostic tests of colinearity have been assessed according to the method of Besley et al.(28) The association between measurement values and fracture status was calculated by conditional logistic model analysis to take into account the matched file structure of our data and expressed as ORs (with 95% CIs) per 1 SD change from the control group. All statistical analyses were performed using the Statistical Analysis Software (SAS V8; SAS Institute, Cary, NC, USA).
Among the first 101 women (mean age, 73.7 ± 8 years) assessed at the visit performed at the end of 13 years who sustained a fragility fracture during the previous 13 years of follow-up during the study, there were 80 peripheral fractures at the wrist (n = 34), ankle (n = 18), ribs (n = 9), metatarsus (n = 6), proximal humerus (n = 6), patella (n = 3), elbow (n = 2), and hip (n = 2) occurring in 72 women and 44 vertebral fractures occurring in 34 women. Five women had both types of fractures. The duration between the occurrence of the last fracture and the architectural measurements was a median of 5.0 years (IQR, 1.6–9.3).
As shown in Table 1, women with fractures did not differ from age-matched controls without fracture for height, weight, and years since menopause. As expected, aBMD of the hip and the forearm was lower in women with fractures. The use of treatments affecting bone metabolism, bisphosphonates (n = 42), hormone replacement therapy (n = 16), selective estrogen receptor modulator (n = 5), tibolone (n = 2), or aromatase inhibitors (n = 3) used ≥1 year at the time (or within 6 months before) of this study was not significantly more frequent in women with fractures.
The correlation between aBMD and architectural parameters was high, especially at the radius (Table 2). The results of the tests of colinearity showed that both ultradistal aBMD and radius architecture parameters could be added in the same model (data not shown).
At the distal radius, we found that women with fractures had lower volumetric total (D tot) and trabecular (D trab) BMDs (−14% and −19%, respectively; p < 0.0001), BV/TV (−19%; p < 0.0001), cortical thickness (−12%; p < 0.001), trabecular number (−14%; p < 0.0001), and trabecular thickness (−6%; p = 0.02) and higher trabecular separation (+26%; p < 0.001) and distribution of trabecular separation (+31%; p = 0.004) than controls without fractures (Table 3). After adjusting for aBMD measured by DXA at the ultradistal radius, differences remained significant for D trab and BV/TV, and there was a trend for TbN, TbSp, and TbSpSd.
At the tibial site and after adjusting for total hip aBMD, we observed significantly lower values for D tot (p = 0.007), D trab (p = 0.03), Cort Th (p = 0.03), and TbTh (p = 0.001), but no significant difference for trabecular number and separation in women with fractures compared with controls (Table 3).
When the analysis was restricted to the 34 women who sustained a wrist fracture, both volumetric densities and all architectural parameters assessed on the other forearm were significantly different in women with fractures compared with controls (Table 4). After adjusting for ultradistal aBMD, differences remained significant for D tot, D trab, BV/TV, and TbN, and there was a trend for TbSp and TbSpSd. When we compared the 76 women who sustained a fracture at the most common osteoporotic sites (distal forearm, hip, proximal humerus, spine) with their age-matched controls, and after adjusting for ultradistal aBMD, differences remained significant for D trab, BV/TV, TbN, and TbSp (p < 0.05; data not shown). Figure 2 provides 3D reconstructions of images from a woman with a previous wrist fracture and her age-matched control without fracture. This shows visually the alterations of both cortical and trabecular architecture in the woman with fracture, whereas her aBMD was similar to that of her control.
In a conditional logistic model, each SD decrease of volumetric total and trabecular densities was associated with a significantly increased risk of fracture at the radius and tibia. The ORs were higher than those associated with each SD decrease of ultradistal radius and total hip aBMD (1.46 [1.16–1.83] and 2.03 [1.35;3.04], respectively). After adjusting for aBMD, each SD decrease of D tot and D trab remained significantly associated with an increased risk of fracture. Lower cortical thickness and lower trabecular thickness of the distal tibia remained associated with an increased risk of fracture after adjusting for aBMD. At the distal radius, a trend was observed for an increased risk of fracture for each SD change of TbN, TbSp, and TbSpSd after adjusting for aBMD (Table 5; Fig. 3). Adding the presence of treatment in the logistic model did not modify the results at both sites (data not shown).
Our study shows, with a noninvasive high-resolution method, that alterations of cortical and trabecular structure are associated with fragility fractures in postmenopausal women. This association is partially independent of decreased aBMD. aBMD measurement by DXA is widely used to estimate the risk of osteoporotic fractures. However, many fractures occur in women with BMD above the WHO threshold of osteoporosis (i.e., at a T score > −2.5). For example, in a previous study from the same cohort, we showed that 48% of postmenopausal women who sustained a fragility fracture over a 9-year follow-up had a baseline aBMD in the osteopenic range and 8% in the normal range.(29) Data from the National Osteoporosis Risk Assessment (NORA) population showed that 52% of women experiencing an incident osteoporotic fracture within 1 year had a T score measured peripherally between −1 and −2.5 and 82% had a T score > −2.5.(30) In the Rotterdam Study, only 44% of all nonvertebral fractures occurred in women with a T score below −2.5.(31) Part of the reason that aBMD does not better identify women with fractures, as described above, may be that it is insensitive to alterations of trabecular architecture that are likely to play an important role in skeletal fragility associated with osteoporosis.(2,3,32)
The structural changes associated with postmenopausal osteoporosis have been mainly documented on 2D analysis of histomorphometric sections of bone. Our findings are in agreement with previous ex vivo studies on iliac crest bone biopsies, showing lower trabecular bone volume (BV/TV), trabecular number, and higher trabecular separation in women with fractures.(3,26,33,34) We also found lower trabecular thickness that has been reported in some histological studies,(26,33,35) whereas others did not find trabecular thinning in women with fractures.(3,34,36) These findings led to the conclusion that the loss of entire elements because of trabecular perforation forms the main mechanism of structural changes in osteoporotic women. In agreement with histomorphometric studies,(37–39) we found lower cortical thickness values in women with previous fractures.
Few studies have analyzed the density/architectural changes in fractured women, using a noninvasive method, showing that estimates of trabecular bone quality are significant predictors of nonvertebral(40,41) and vertebral fractures.(13,14) In a study comparing 107 women with and 107 women without prevalent wrist fracture and using pQCT at the radius, the trabecular volumetric BMC and BMD data seemed to discriminate fractured and nonfractured cases better than cortical area and vBMC.(40) In another cross-sectional study, 21 women with a recent wrist fracture were compared with a control group with no history of fracture and similar volumetric total BMD at the distal radius using pQCT. The average trabecular hole size was larger in women with fracture than in controls, but the connectivity index was not significantly different between both groups.(41) The resolution of the QCT systems used in those studies was relatively low (slice thickness of 2.5 mm and voxel dimension of 0.33 mm2) and did not permit evaluation of the structural changes at the level of trabeculae. A recent study using multidetector row CT (MDRCT) at the third lumbar vertebra with a higher spatial resolution (250 × 250 × 500 μm3) than conventional CT was performed in 82 postmenopausal women. The authors showed that both volumetric BMD and microstructure parameters (structure model index, bone volume fraction, Euler's number) were associated with a higher relative risk for prevalent vertebral fracture than did aBMD obtained by DXA.(14) In that study, the structural indices were not directly measured but were all calculated using 3D imaging methods. Moreover, cortical structure was not evaluated with that system. The use of high-resolution QCT for the analysis of architecture of the vertebrae in vivo is limited by the relatively high radiation dose, and most studies have been performed at peripheral sites. Although nonionizing, high-resolution MRI is also limited to peripheral skeletal sites because it needs a long acquisition time that could expose it to motion artifacts at the spine level. In a MRI study performed at the distal radius and the calcaneus with a 3D gradient-echo sequence (slice thickness of 500 μm and spatial resolution of 156 × 156 μm2), structural parameters could discriminate between women with and without hip fracture.(6) In another high-resolution MRI study at the radius, the authors showed that all structural parameters, including the distribution of trabecular separation, differed between women with nonfracture to those with vertebral fracture, except for trabecular thickness.(7) These measurements cannot be performed routinely on a standard MRI device.
In our study, we used a recently developed high-resolution pQCT system with a higher resolution (slice thickness of 82 μm, spatial resolution of 82 × 82 μm2) that permits in vivo assessment of both cortical and trabecular architecture at the distal radius and tibia. With that system and among trabecular parameters, trabecular number is directly assessed, whereas trabecular thickness and separation are derived from trabecular bone volume fraction and trabecular number using standard methods from histomorphometry.(26) Nevertheless, the higher spatial resolution of that system permits obtaining values of trabeculae thickness close to those obtained from histological studies, trabeculae varying in thickness from 80 to 250 μm.(26) We have previously found with this device that it showed a reproducibility similar to or better than previous pQCT systems that have a lower resolution.(21) In the latter study, 35 osteopenic postmenopausal women with a previous fracture were compared with 78 osteopenic women without fracture. At the distal radius, women with previous fractures had lower volumetric total and trabecular densities and a higher intraindividual distribution of separation than women without fracture. Conversely, no difference was observed at the distal tibia. In our study, using the same device but in a different cohort, we also found some discrepancies between both sites. Indeed, volumetric (total and trabecular) densities, cortical thickness, and trabecular thickness were lower in women with fractures at both sites. In contrast, trabecular number, trabecular separation, and distribution of trabecular separation discriminated fracture subjects from nonfracture patients only at the distal radius. The design of our age-matched case-control study and the higher number of women with previous fracture in this study could partly explain the slight discrepancies between both studies. Our findings are in agreement with those of Khosla et al.,(22) who used a similar device and showed a significant decrease with age in TbN and increase in TbSp at the wrist in women. In contrast to the tibia, the radius is not a weight-bearing bone and is not exposed to the same mechanical stimuli, a pattern that may explain the differences of alterations of architectural parameters in women with fractures between both sites.
Correlations among architectural parameters have been reported previously with HR-pQCT, showing that trabecular architectural measurements were strongly correlated to trabecular density at both the distal radius and tibia (r = 0.7–0.9) and that total density was strongly associated to both trabecular density (r = 0.8) and cortical thickness (r = 0.9).(21) In our study, we also found significant correlations between aBMD measured by DXA and architectural parameters, especially at the radius. However, after adjusting for aBMD, the differences between cases and controls—even if lower—remained significant at the tibial site for volumetric (total and trabecular) densities, cortical thickness, and trabecular thickness. At the distal radius, the adjustment for aBMD attenuated more strongly the differences between cases and controls, which remained significant for trabecular volumetric density with a trend for trabecular number, trabecular separation, and distribution of separation. Nevertheless, when the analysis is restricted to women who sustained a fracture at the four most common osteoporotic sites, differences between cases and controls remained significant for trabecular volumetric density, trabecular number, and trabecular separation. These data suggest that alterations of bone architecture are partially independent of BMD for the risk of fracture. Before this analysis, tests of colinearity were performed, permitting the addition of both aBMD and architecture parameters in the model. To our knowledge, there are only a few studies showing that BMD measured by DXA and architecture measured by pQCT are independently associated with fracture. In one study, the association between some architectural parameters and fracture status was found to be similar or higher than that observed with BMD, but the independence of both techniques was not analyzed.(7) Others showed that combining BMD with some microstructural parameters improved the discrimination capability for spinal fracture,(14) hip fracture,(6) and wrist fracture(42) over with BMD measurement alone. Another study showed that differences in trabecular structure parameters between subjects with and without vertebral fractures did not remain significant after adjustment for aBMD.(13)
Our study has strengths and limitations. Despite the cross-sectional design of this study, the OFELY study is a population-based cohort study, and all fragility fractures were prospectively assessed during a long follow-up and radiographically confirmed. The repetition of spine radiographs every 4 years allowed an optimal ascertainment of vertebral fractures because only a small proportion of them reach clinical attention. Another strength of this study is the concomitant assessment of radius aBMD obtained by DXA at the same site as the pQCT measurements, allowing evaluation of the ability of pQCT parameters to discriminate women with and without fractures, taking into account their aBMD. A limitation of HR-pQCT is that it is not applicable to the spine and hip, the two most common sites of osteoporotic fractures. BMD measurement at the radius, however, predicts all fragility fractures and measurement at the spine or hip, as shown in a meta-analysis.(43) Another technical limitation is that, whereas TbN is directly calculated without any plate model assumption, TbTh and TbSp are derived from TbN and BV/TV and with the plate model assumption. A limitation of the case-control design of our study is that a causal relationship between alterations of microarchitecture and fragility fracture cannot be established. Further longitudinal studies are needed in this regard. Another limitation is that some of women took treatments that could interfere with bone metabolism. Thus, bone architecture could have been modified by those treatments. Nevertheless, the number of treated women did not differ between both groups, and adding the presence of treatment in the logistic model did not modify the results.
In conclusion, our findings suggest that, in postmenopausal women, vertebral and nonvertebral fractures are associated with low volumetric BMD and architectural alterations of trabecular and cortical bone that can be assessed noninvasively by HR-pQCT. These alterations are partially independent of aBMD assessed by DXA. Prospective studies should be undertaken to determine if HR-pQCT can predict the risk of osteoporotic fractures with a better sensitivity/specificity than DXA.
The authors thank A Bourgeaud, B Vey-Marty, and W Wirane for excellent technical assistance. This work was supported by an unrestricted research grant from Eli Lilly to INSERM and from the Société Française de Rhumatologie.
- 281980 Regression Diagnostics: Identifying Influencial Data and Sources of Colinearity. John Wiley and Sons, New York, NY, USA., ,
- 371997 The thickness of human vertebral cortical bone and its changes in aging and osteoporosis: A histomorphometric analysis of the complete spinal column from thirty-seven autopsy specimens. J Bone Miner Res 12: 89–95., , , ,