SEARCH

SEARCH BY CITATION

Keywords:

  • bone instability;
  • sexual dimorphism;
  • bone geometry;
  • fracture risk;
  • BMD;
  • buckling ratio

Abstract

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

We studied HSA measurements in relation to hip fracture risk in 4806 individuals (2740 women). Hip fractures (n = 147) occurred at the same absolute levels of bone instability in both sexes. Cortical instability (propensity of thinner cortices in wide diameters to buckle) explains why hip fracture risk at different BMD levels is the same across sexes.

Introduction: Despite the sexual dimorphism of bone, hip fracture risk is very similar in men and women at the same absolute BMD. We aimed to elucidate the main structural properties of bone that underlie the measured BMD and that ultimately determines the risk of hip fracture in elderly men and women.

Materials and Methods: This study is part of the Rotterdam Study (a large prospective population-based cohort) and included 147 incident hip fracture cases in 4806 participants with DXA-derived hip structural analysis (mean follow-up, 8.6 yr). Indices compared in relation to fracture included neck width, cortical thickness, section modulus (an index of bending strength), and buckling ratio (an index of cortical bone instability). We used a mathematical model to calculate the hip fracture distribution by femoral neck BMD, BMC, bone area, and hip structure analysis (HSA) parameters (cortical thickness, section modulus narrow neck width, and buckling ratio) and compared it with prospective data from the Rotterdam Study.

Results: In the prospective data, hip fracture cases in both sexes had lower BMD, thinner cortices, greater bone width, lower strength, and higher instability at baseline. In fractured individuals, men had an average BMD that was 0.09 g/cm2 higher than women (p < 0.00001), whereas no significant difference in buckling ratios was seen. Modeled fracture distribution by BMD and buckling ratio levels were in concordance to the prospective data and showed that hip fractures seem to occur at the same absolute levels of bone instability (buckling ratio) in both men and women. No significant differences were observed between the areas under the ROC curves of BMD (0.8146 in women and 0.8048 in men) and the buckling ratio (0.8161 in women and 0.7759 in men).

Conclusions: The buckling ratio (an index of bone instability) portrays in both sexes the critical balance between cortical thickness and bone width. Our findings suggest that extreme thinning of cortices in expanded bones plays a key role on local susceptibility to fracture. Even though the buckling ratio does not offer additional predictive value, these findings improve our understanding of why low BMD is a good predictor of fragility fractures.


INTRODUCTION

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

Osteoporosis is a systemic skeletal disease that reduces the mechanical strength of bones so that fractures occur under minimal trauma conditions at diverse skeletal locations (including vertebrae, long bones, and ribs). In terms of morbidity and costs, hip fractures are considered the dominant complication of osteoporosis in both men and women.(1) Most of these fractures occur in women because they have a higher incidence of fracture at any given age and because of their higher life expectancy.(2)

Because strength is difficult to measure in vivo, the disease is often characterized by surrogate markers such as low BMD or microarchitectural deterioration of bone tissue. Although not in itself a mechanical property, the easily and widely measured areal BMD (aBMD) by DXA has long been considered a good surrogate measure of bone strength. Strong relationships between aBMD measures and the probability of fracture have been documented in large prospective studies.(3,4) We and others have shown that each SD decrease in femoral neck aBMD increased the age-adjusted risk of hip fracture two to three times in postmenopausal women.(3,5–7) For diagnostic purposes, osteoporosis in women is defined by a WHO-based T-score of BMD less than –2.5.(8)

From a practical standpoint, there is uncertainty about how this definition should be used across sexes, in particular if the same well-studied BMD paradigm devised for women should be applied to men.(9,10) We have previously provided evidence that the use of sex-specific T-scores for the diagnosis of osteoporosis (corresponding to absolute femoral neck aBMD thresholds of 0.74 g/cm2 in men and 0.68 g/cm2 in women) would capture the same proportion of fractures in men and women, but would not be a good threshold for defining the same absolute fracture risk in both sexes.(11) We also showed that, even though the average aBMD in men who fracture their hip is higher than in women, the relative risk for hip fracture per SD decrease in aBMD was similar at the same BMD level in men and women.(11) This similarity in hip fracture risks for a given aBMD level of elderly men and women is remarkable considering the well-described sex differences in body and bone size, for example, as a consequence of differential periosteal and endosteal apposition during modeling and remodeling men have overall thicker cortices and bigger and stronger bones than women.(12) Such sexual dimorphism of bone (sex differences in bone structure) is thought to explain at least some of the differences in fracture rates between men and women. Assuming that aBMD is a good surrogate of bone strength, it is not clear why men who fracture seem on average to do so with “stronger” bones (higher BMD) than women who fracture.

Furthermore, from a mechanical perspective, it is also unclear why BMD is a good predictor of fragility. Certainly a bone containing less material will fail under smaller loads. However, aBMD seems to decline with age not only by bone loss but also by the associated expansion of the periosteal envelope.(13) In elderly women of the SOF study, femoral neck BMD declined at a rate of 0.7%/yr, of which nearly one third of the decline was caused by femoral neck expansion and not bone loss.(13) A still conflicting issue recently noted by Heaney,(14) is that from the common symbolic expression BMD = BMC/area follows that, a bone with more mass (mineral content) has higher BMD, but also implies counter intuitively that BMD would be inversely related to bone size (because region area is a rough measure of bone size). Regardless of the influence of the depth projection (volumetric dimension) on the aBMD measurement, the question here is how an increase in the diameter of a bone containing the same amount of mass instead of improving strength may actually weaken the bone.

The aim of this study is to address the questions of (1) why men seem on average to fracture with “stronger” bones (higher BMD) than women and (2) how wider bones instead of improving strength may actually increase the risk of fracture. We do this by studying the mechanical characteristics underlying the DXA aBMD measurement of the femoral neck and the relation with the risk of hip fracture in both elderly men and women. We applied a mathematical model previously validated for aBMD(11,15) to determine the 1-yr hip fracture risks and the distribution of hip fractures according to DXA-derived parameters of hip structure (geometry) in both sexes. In addition, to validate the modeled relations, we analyzed these hip structural parameters in relation to incident hip fractures occurring in elderly men and women of the prospective Rotterdam Study. Finally, we compared the operative characteristics and classificatory capacity of the hip structural parameters to those of femoral neck BMD to determine any possible gain in the predictive value for hip fracture.

MATERIALS AND METHODS

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

Subjects

Individuals were derived from the Rotterdam Study (n = 7983), a single-center prospective population-based study of determinants of chronic disabling diseases in the elderly (age, ≥55 yr). Written informed consent was obtained from every participant. The design and rationale of the study has been described earlier.(16) The current study included 4806 participants of both sexes (2740 women) with hip structural analysis (HSA) performed on DXA scans at baseline (1990–1993) and who were followed for the occurrence of fracture until January 1, 2002, with an average follow-up of 7.4 ± 3.3 (SD) yr. Fracture events in the studied individuals included 147 incident hip fracture cases (106 women and 41 men). Methodology on fracture follow-up has been described in detail previously.(6) In brief, computerized records of the general practitioners and hospital registries were regularly checked by research physicians who reviewed and coded the information. Only low-impact (osteoporotic) fractures were included.

BMD and bone geometry measurements

BMD measurements (g/cm2) of the right proximal femur were performed by DXA using a Lunar DPX-L densitometer (Lunar Radiation Corp., Madison, WI, USA) and analyzed with DPX-IQ v.4.7d software. Methods, quality assurance, accuracy, and precision issues of the DXA measurements have been described previously.(17) From the femoral neck BMD, sex-specific T-scores were calculated using the NHANES reference population.(18) Peak bone mass, as converted to the corresponding Lunar value, was 1.04 ± 0.12 g/cm2 for women and 1.13 ± 0.16 g/cm2 for men. The absolute BMD cut-off values for osteoporosis (T-score below −2.5) were 0.69 g/cm2 in women and 0.73 g/cm2 in men. We used the HSA software developed by Beck et al.(19) to measure hip bone geometry from the DXA scans of the narrow neck (NN) region across the narrowest point of the femoral neck. BMD, cross-sectional area of bone tissue (CSA), bone width (outer diameter), and cross-sectional moment of inertia (CSMI) were measured directly from mineral mass distributions using algorithms and precision properties described previously.(20,21) Section modulus (Z) was calculated as CSMI/dmax, where dmax is the maximum distance from the center of mass to the medial or lateral surface. In addition, estimates of cortical thickness and endocortical diameter were obtained by modeling the NN region as a circular annulus, which assumes a proportion of cortical/trabecular bone of 60/40. Buckling ratio was estimated as dmax divided by the mean cortical thickness estimate. Precision percentage CVs of the different parameters of the HSA method have been described previously.(21)

Mathematical model input data

We applied a previously validated mathematical model used for BMD(11) to calculate the hip fracture distribution by femoral neck BMD, BMC, bone area, and HSA parameters (cross-sectional area, section modulus, cortical thickness, narrow-neck width, and buckling ratio). The model combined the distribution of each parameter in the population with the hip fracture risk at each parameter level. We previously estimated the hip fracture risk by sex, age, and BMD, based on Dutch hip fracture registration data and the distribution of BMD.(15) The resulting risk functions (1-yr cumulative hip fracture incidence) were validated over almost 4 yr of follow-up in the prospective part of the Rotterdam Study.(15) In this study, the expected number of hip fractures at each parameter level was obtained by multiplying the proportion of the population at a specific parameter value with the corresponding 1-yr hip fracture risk as done previously for BMD.(11) When this was done across the whole range of parameter values, we obtained the distribution of hip fractures. For example, in women 70 yr of age, 3% of the population has a BMD of exactly 0.80 g/cm2; the 1-yr hip fracture incidence at this BMD is calculated at 0.2%; therefore, the calculated number of hip fractures at 70 yr of age in those women with a BMD of 0.80 g/cm2 is as follows: 3% × 0.2% = 0.00006 or (6/100.000). The sum of all these values across all BMD levels corresponds to the 1-yr cumulative incidence at that age. The model was applied using an MS-Excel spreadsheet. Calculations were made for several ages between 65 and 80 yr, but for ease of presentation, results are described in detail for subjects 70 yr of age. We chose this age because at 70 yr of age, hip fracture risk begins to increase exponentially.

Assumptions of the mathematical model

The baseline hip fracture risk functions by age, sex, and BMD were described previously.(2,11,15) These baseline risk functions show an exponential increase in hip fracture risk with lowering BMD and additionally an increase in risk with aging. In this study, we applied such functions described previously for BMD to BMC, femoral neck bone area, cross-sectional area, section modulus, cortical thickness, neck width, and buckling ratio parameters.

Fracture risk estimation

All analyses were performed for men and women separately. For the analysis of HSA (bone geometry) parameters, we used analysis of covariance (F-tests) to compare age-adjusted means between individuals with and without hip fracture. We estimated the relative risk for a first fracture associated with 1 SD decrease in femoral neck BMD using Cox proportional hazards models. These analyses were based on first fracture in individuals because multiple fractures do not contribute to independent observations. To account for confounding by age, we included age as a continuous variable in all models. Areas under the receptor operating characteristic (ROC) curves adjusted for age were computed from logistic regression models to compare the operative characteristics between BMD and buckling ratio measurements for the prediction of hip fracture using STATA V. 7.0. If not stated otherwise, all analyses were performed using SPSS V.11.0.

RESULTS

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

Table 1 compares baseline characteristics and hip fracture incidence rates between the study population who had HSA (n = 4806) and the entire Rotterdam Study population (n = 7983). On average, individuals with HSA were 3.4 yr younger and had a lower incidence (number) of hip fracture events. All other characteristics (including BMD levels) did not show significant differences.

Table Table 1.. Characteristics of the Study Populations
  1. Data are unadjusted means (SD) or incidence rate per 1000 py (n).

  2. * Based on 6917 individuals with present anthropometric measurements.

  3. Based on 5826 individuals with present BMD measurements.

Thumbnail image of

In Table 2, individuals of both sexes with hip fracture were compared with those without using F-tests from covariance analysis. Compared with nonfractured individuals, hip fracture cases were significantly older and lighter (only in women), with lower BMD, sex-specific T-scores, and BMC values. HSA showed that hip fracture cases had significantly thinner cortices, greater neck widths (not significant in men, p = 0.10), lower cross-sectional areas, lower bending strength (section moduli), and greater bone instability (buckling ratios). No significant differences were observed in conventional region bone area in either sex. In both hip fracture cases and noncases, men were significantly taller, heavier, and had higher femoral neck BMD, T-scores, BMC, and bone area compared with women. HSA across sexes showed that, compared with women, men had thicker cortices, greater bone widths, increased cross-sectional areas, and greater bending strength (section modulus) in both fractured and nonfractured individuals. No significant differences were observed across sexes between the bone instability indexes (buckling ratios) of hip fracture cases.

Table Table 2.. Baseline Characteristics, DXA, and Hip Geometry in Relation to Incident Hip Fracture (n = 147)
  1. Estimates are (age-adjusted) means ± SE.

  2. NN, narrow-neck.

Thumbnail image of

Figure 1 shows, for femoral neck BMD, BMC, and bone area, the frequency distribution in the entire population, the 1-yr risk of fracture, and the hip fracture frequency distribution modeled for both sexes at 70 yr of age. The overall population frequency distribution of BMD, BMC, and bone area in men is consistently shifted to the right, showing as expected overall higher values than in women (Fig. 1A). The 1-yr risk of hip fracture (Fig. 1B) shows that the yearly risk is similar in men and women at any given BMD level, with an exponential increase at very low BMD levels (0.70 g/cm2). A different pattern is seen for BMC, where the 1-yr risk of fracture in men is higher than in women at all given BMC levels, also with an exponential increase at low BMC values (<4.0 g). No clear pattern across sexes is seen with bone area, where the 1-yr risk of fracture increases slightly with bone area only in men. The hip fracture frequency distribution obtained from the baseline parameter distribution and the 1-yr risk of fracture (Fig. 1C) shows that, overall, men fracture at higher BMD, BMC, and bone area levels than women. However, only the modeled hip fracture distribution by BMD levels shows clearly that more fractures occur in women than in men as observed in the prospective analysis.

thumbnail image

Figure Figure 1. Models of BMD, BMC, and bone area for men (gray lines) and women (black lines) 70 yr of age of the (A) population frequency distributions, (B) 1-yr fracture risk, and (C) hip fracture frequency distributions based on prospective data from the Rotterdam Study applying a previously validated mathematical model used for BMD.(11)

Download figure to PowerPoint

Figure 2 shows, for NN cortical thickness, bone width, and buckling ratio, the frequency distribution in the entire population, the 1-yr risk of fracture, and the hip fracture frequency distribution modeled for both sexes at 70 yr of age. The frequency distribution of cortical thickness and bone width in men is consistently shifted to the right, showing overall higher values than in women, whereas the frequency distribution of buckling ratio overlaps across sexes (Fig. 2A). The 1-yr risk of fracture at 70 yr of age (Fig. 2B) shows that the 1-yr risk of hip fracture in men is higher than in women at each given value of cortical thickness, with an apparent exponential increase with cortices thinner than 1.1 mm. In contrast, at all given dimensions of bone width, the 1-yr risk of fracture is higher in women than in men, with a moderate linear increase with bone width in both sexes. The 1-yr risk of hip fracture at any given level of buckling ratio is higher in women than in men, with an exponential increase in risk for both sexes at very high buckling ratio levels (>16). The hip fracture frequency distribution obtained from the baseline parameter distribution (Fig. 2C) shows that, overall, men fracture with thicker cortices and greater bone diameters than women. When these two parameters are combined in the buckling ratio, the frequency distribution of fractures in men is similar to that of women across all levels of buckling ratio values. In concordance to the observations found in the prospective data, a higher number of fractures occur in women than in men.

thumbnail image

Figure Figure 2. Models of cortical thickness, neck width, and buckling ratio (index of bone instability) for men (gray lines) and women (black lines) 70 yr of age of the (A) population frequency distributions, (B) 1-yr fracture risk, and (C) hip fracture frequency distributions based on prospective data from the Rotterdam Study applying a previously validated mathematical model used for BMD.(11)

Download figure to PowerPoint

The frequency distribution at 70 yr of age of the indices of axial (cross-sectional area) and bending (section modulus) strength showed in the overall population consistently higher values in men than in women (data not shown). The 1-yr risk of fracture at age 70 of these indices showed, for CSA, a similar pattern as that observed for BMC and cortical thickness (data not shown). Men had a higher risk than women throughout all CSA values, with a consistent apparent exponential increase in risk at lower values of CSA (<1.8 cm2). In contrast, the 1-yr risk of fracture of the section modulus at age 70 showed a subtle (<1%) gradual increase in risk in both sexes (data not shown).

When comparing the operative characteristics and classificatory capacity of BMD and buckling ratio for hip fracture, we observed that the (age-adjusted) areas under the ROC curve were essentially the same for BMD (0.8146 in women and 0.8048 in men) and buckling ratio (0.8161 in women and 0.7759 in men). No additional classificatory capacity for hip fracture was obtained by combining the information from both the BMD and the buckling ratio measurements (area under ROC curve = 0.8199 in women and 0.8048 in men).

DISCUSSION

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

We showed that participants of the Rotterdam study who suffered incident hip fractures had significantly lower femoral neck BMD, lower BMC, thinner cortices, reduced bending strength (section modulus), and increased bone instability (buckling ratios) at baseline than nonfractured individuals. Femoral neck widths were wider in fracture cases of both sexes, although not quite significant in men (p = 0.10). In contrast, the buckling ratio (ratio of the femoral neck width to the average cortical thickness) was similar in men and women, suggesting a similar reduction of strength caused by bone instability across sexes. These results suggest that bone instability, as determined (crudely) by the buckling ratio, depicts the critical relation between cortical thickness and bone width, which seems to play a key role on local susceptibility to hip fracture. Considering that the predictive ability of the buckling ratio is essentially the same as that of BMD (as shown by the identical areas under the ROC curves for classifying hip fracture), we propose that the reason why femoral neck BMD is a good predictor of hip fracture is because it largely identifies bone instability.

Our findings have important implications for the interpretation of BMD in the bones of elderly men and women. One would not expect BMD to reflect strength differences in sexually dimorphic (hence dissimilar sized) bones. For example, if we compare femoral necks between men and women in nonfractured individuals of our study, the BMD differences indicate that male femoral necks are 10% stronger on average. However, they contain 27% more mass and are 15% wider in diameter, which translates to a 45% difference in bending strength (section modulus). From this, it can be concluded that BMD differences do not adequately reflect the difference in bone mass or strength across sexes.

Furthermore, bone size has a counterintuitive but subtle value. When body size is adjusted out, a wider bone is expected to be necessarily stronger because it increases the cross-sectional moment of inertia (or section modulus). Nevertheless, our results show that, in women, increased bone width is a subtle, yet independent predictor of hip fracture. In the absence of a change in material strength, the only reason why a wider bone would not be stronger is if cortical dimensions were thinned to the point where bone strength is lost because of instability. However, the relation between femoral neck width and the risk of fracture remains conflicting; whereas some find a smaller neck width a risk factor,(22–25) we and others find a wider neck as risk factor for hip fracture.(12,13,26–30) Considering the small changes in periosteal apposition during aging(31) and the low precision of DXA to determine bone diameter, these findings should be interpreted carefully.

It is well established that the structural adaptations that occur in aging bones follow sex-specific patterns.(12,32–34) In Fig. 3, we summarize graphically how our proposed model of structural instability fits the sexual dimorphism of bone and the risk of fragility fracture. The age changes depicted in the figure are subtly different from those proposed by Duan et al.(12) Although Duan et al. reported that men with hip fractures had narrower femoral necks than age-matched controls, all their other findings were consistent with our study, including wider femoral necks in women with fractures and higher buckling ratios in both sexes. In a large sample of the U.S. adult white population, Looker et al.(35) suggested a similar rate of expansion in both men and women. In contrast, in a longitudinal study of elderly men and women in the United Kingdom, Kaptoge et al.(32) showed that the femoral neck and intertrochanteric regions expanded faster in women than in men. There is long standing evidence suggesting that periosteal expansion may be slower in younger women and that this occurs because the presence of estrogen from the onset of female puberty limits periosteal apposition and favors bone formation on the inner endocortical surfaces.(36) This process was postulated by Fuller Albright in the early 1940s to be an evolutionary response of the female skeleton to pack extra bone mineral that can be lost with minimal mechanical consequences but remain easily available for the needs of pregnancy and lactation. These slower rates of bone expansion in younger women do not seem to continue after the menopause.(32,33)

thumbnail image

Figure Figure 3. Proposed changes that cross-sections of the femoral neck will exhibit in aging men and women, showing the relation between inner (endosteal) and outer (periosteal apposition) bone expansion, with its consequences over bone density, strength, and bone instability (propensity of thinner cortices in wide diameters to buckle). Men have wider and thicker bones than women at all ages, yet the same instability thresholds apply for both sex.

Download figure to PowerPoint

One consequence of this “homeostatic” process of bone expansion in both sexes is that, as bone diameters increase, the nonlinear effect on section modulus ensures that bending strength may be preserved with progressively thinner cortices. Engineers use this principle in designing lightweight tubular structures but ensuring that tube walls are not thinned to the point where strength is lost because of local instability. Age-related expansion should both reduce BMD and increase the buckling ratio even while preserving the section modulus, but these adaptive changes seem not to be constrained in the bone (Fig. 3). Thus, in this model, men are expected to have, at all instances, higher BMD, thicker cortices, greater diameters, and greater bone strength than women. However, both sexes will be subject to the same instability threshold. As expansion progresses and cortices thin, eventually a point is reached where instability begins to reduce strength so that it is less than that predicted by the section modulus. Our study suggests this geometric threshold seems to be the same in both men and women, although it occurs at a higher BMD and greater absolute strength in men.

Narrow neck BMD is strongly correlated with both the buckling ratio (−0.90) and moderately with the section modulus (0.70). From this, we propose that a decline in BMD predicts fragility fracture because it is associated with a decrease in section modulus (bone strength) and an increase in buckling ratio (bone instability). The observation that section modulus is a poorer predictor of hip fracture than buckling ratio supports the view that structural instability (as detected by the buckling ratio) underlies the ability of BMD to predict fragility fractures of the hip.

This study has some limitations intrinsic to the study design and most importantly to the technology used. The existence of some selection (survival) bias is imminent because some very elderly and severely ill individuals did not have DXA measurements done. Nevertheless, such bias is not expected to be of considerable importance because we restricted our conclusions to individuals in their 70s and avoided any modeling at very old ages. Another limitation of our study is the inability to discriminate between cervical and trochanteric hip fractures. However, because we limited our analysis to low-impact (osteoporotic) fractures, most fractures will be cervical considering that the majority of trochanteric fractures have been shown to occur at high failure loads.(37) Because risk factors have been shown to differ across these types of hip fractures, inclusion of this limited number of trochanteric fractures may have underestimated the effect sizes we report. Regarding the technology, one should not disregard the fact that bones are 3D structures and that DXA provides a 2D assessment. The densitometric image provides no direct information about material composition and can only provide geometry information extracted from mass distributions within the plane of the image.(38) Therefore, intrinsic limitations and necessary assumptions are part of the 2D methodology we have used. This way, the section modulus is measured directly from the mass distribution across the femoral neck, but it quantifies bending only in the image plane and is thus subject to positioning error. In addition, the algorithm used to compute the buckling ratio assumes that a fixed fraction of mineral mass is in the cortical shell, which is modeled as a uniform annulus. A study by Crabtree et al.(39) showed that fracture cases had disproportionate loss of cortical and not trabecular bone. Hence, our fixed fraction assumption (of 60% cortical and 40% trabecular bone) in our algorithm is likely to underestimate the actual instability occurring in fracture cases. In addition, a study by Zebaze et al.(41) showed that the femoral neck is elliptically shaped and that, even though DXA accurately measured femoral neck width (supero-inferior diameter), models assuming a circular cross-section overestimated femoral neck depth (antero-posterior diameter), volume, and geometric indices of bone strength. Therefore, our crude assumption of a uniform cortical annulus is undoubtedly more structurally stable than the irregular cortices of real femoral necks, especially in fracture cases.(41) This recent study using CT in cadaver femoral necks showed that cortices are quite nonuniform (particularly thinner at the upper than the lower femoral neck segments). Such structural adaptation of the femoral neck is proposed to occur in response to differential loading during walking and is thought to predispose to local cortical buckling and increased risk of hip fracture.(41) Another limitation of our biomechanical model is the inability of DXA to discriminate between cortical and trabecular bone. Therefore, the assumption of mass distribution used in HSA does not assess the internal mechanical support trabeculae place on cortices. Like BMD, buckling ratio is probably least useful in growing or young adult bones where buckling would be unlikely because of the presence of robust trabeculae. Nonetheless, trabecular support against buckling is largely lost with advancing age,(42) which is the reason why hip bending resistance would best be maintained by cortical adaptation in the presence of such trabecular loss. New biomechanical explorations using high-resolution 3D assessment with advanced engineering analyses are thus warranted to determine the strength of the bone structure as a whole.(43)

In line with various recent reports, we also showed (by comparing areas under the ROC curves) that the buckling ratio or bending strength indices crudely derived from the DXA measurement provide no further predictive ability useful in clinical practice for the prediction of fracture.(44–46) This is logical considering the high (inverse) correlation (−0.80) between femoral neck BMD and the buckling ratio. Therefore, it is our contention that local buckling phenomena are relevant to explain why BMD predicts hip fractures. Nevertheless, we would like to emphasize that hip structural/strength analyses may not constitute a replacement of the (less than perfect) fracture risk assessment using BMD. Alternatively, considering that BMD and BMC are not by themselves mechanical properties (because they can occur with similar values in bones with very different strengths), bone geometry measurements may be of preferable use in studies examining the direct effect of pharmacological interventions on bone strength(47–49) and/or the genetic determinants of bone structure.(50,51)

In summary, this large population-based study provides a mechanical rationale for why the BMD measurement, despite skeletal dimorphism, is a strong predictor of hip fracture risk in both men and women. The results of this study support the contention that fragility fractures of the hip are mainly caused by cortical instability and that BMD predicts those fractures because of its correlation with indices of bending strength and susceptibility to buckling. The mechanism indicates that bone strength and instability are a function of the thickness of the cortices relative to the diameter of the bone. However, this study also showed that the predictive ability of the instability index (buckling ratio) was essentially identical to that of BMD. Improved measurements that consider the actual complex 3D nature of these bone parameters should be pursued to achieve an enhanced prediction of osteoporotic fractures. These crude indexes of strength and bone instability should remain useful in research studies, but because of technological limitations of current DXA methods, they are not yet sufficiently precise or accurate for clinical use.

Acknowledgements

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

This work was supported by Netherlands Organization for Scientific Research (NWO) Project 014-90-001 and European GENOMOS Grant QLK6-CT-2002-02629.

REFERENCES

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