Presented in part at the European Calcified Tissue Symposium, Maastricht, The Netherlands, 1999 and the American Society for Bone and Mineral Research, St. Louis, MO, USA, 1999.
Intracapsular Hip Fracture and the Region-Specific Loss of Cortical Bone: Analysis by Peripheral Quantitative Computed Tomography†
Article first published online: 1 JUL 2001
Copyright © 2001 ASBMR
Journal of Bone and Mineral Research
Volume 16, Issue 7, pages 1318–1328, July 2001
How to Cite
Crabtree, N., Loveridge, N., Parker, M., Rushton, N., Power, J., Bell, K. L., Beck, T. J. and Reeve, J. (2001), Intracapsular Hip Fracture and the Region-Specific Loss of Cortical Bone: Analysis by Peripheral Quantitative Computed Tomography. J Bone Miner Res, 16: 1318–1328. doi: 10.1359/jbmr.2001.16.7.1318
- Issue published online: 2 DEC 2009
- Article first published online: 1 JUL 2001
- Manuscript Accepted: 30 OCT 2000
- Manuscript Revised: 18 OCT 2000
- Manuscript Received: 16 MAY 2000
- hip fracture;
- peripheral quantitative computed tomography;
- cortical bone;
- human femoral neck;
- site-specific bone loss
Generalized bone loss within the femoral neck accounts for only 15% of the increase in intracapsular hip fracture risk between the ages of 60 and 80 years. Conventional histology has shown that there is no difference in cancellous bone area between cases of intracapsular fracture and age and sex-matched controls. Rather, a loss of cortical bone thickness and increased porosity is the key feature with the greatest change occurring in those regions maximally loaded during a fall (the inferoanterior [IA] to superoposterior [SP] axis). We have now reexamined this finding using peripheral quantitative computed tomography (pQCT) to analyze cortical and cancellous bone areas, density, and mass in a different set of ex vivo biopsy specimens from cases of intracapsular hip fracture (female, n = 16, aged 69-92 years) and postmortem specimens (female, n = 15, aged 58-95 years; male, n = 11, aged 56-86 years). Within-neck location was standardized by using locations at which the ratio of maximum to minimum external diameters was 1.4 and at more proximal locations. Cortical widths were analyzed using 72 radial profiles from the center of area of each of the gray level images using a full-width/half-maximum algorithm. In both male and female controls, cancellous bone mass increased toward the femoral head and the rate of change was gender independent. Cancellous bone mass was similar in cases and controls at all locations. Overall, cortical bone mass was significantly lower in the fracture cases (by 25%; p < 0.001) because of significant reductions in both estimated cortical area and density. These differences persisted at locations that are more proximal. The mean cortical width in the cases was significantly lower in the IA (22.2%; p = 0.002) and inferior regions (19%; p < 0.001). The SP region was the thinnest in both cases and controls. These data confirm that a key feature in the etiology of intracapsular hip fracture is the site-specific loss of cortical bone, which is concentrated in those regions maximally loaded during a fall on the greater trochanter. An important implication of this work is that the pathogenesis of bone loss leading to hip fracture must be by a mechanism that varies in its effect according to location within the femoral neck. Key candidate mechanisms would include those involving locally reduced mechanical loading. This study also suggests that the development of noninvasive methodologies for analyzing the thickness and estimated densities of critical cortical regions of the femoral neck could improve detection of those at risk of hip fracture.
THE FRAGILITY of an individual bone is dependent on changes in its material properties and its distribution as well as the amount of bone tissue it contains. Age-related increases in hip fracture risk are related to the simultaneous decreases in bone mass. However, the risk of hip fracture increases 14-fold between the ages of 60 and 80 years but decreases in bone mineral density (BMD) only account for a 2-fold increase in risk.(1) Although the increased incidence of falls accounts for a further 1.1- to 1.5-fold(1, 2) increase in risk, this still leaves a 4-fold increase in risk that probably is related to changes in the structural and perhaps the material properties of the remaining bone.(3) The importance of these last two factors is highlighted by the recent trials with alendronate (an antiresorptive bisphosphonate) in patients with low bone mass. Although bone mass was increased by similar amounts in subjects with low or very low bone density, protection against hip fracture was only afforded to those with severe osteopenia before the start of therapeutic intervention.(4) There was little or no reduction in hip fracture rates in those subjects with mild to moderate osteopenia despite a similar increase in bone mass. An alternative explanation is that there may be underlying structural changes that are masked by such volume-averaged measurements as BMD.
The traditional view of osteoporosis and associated fractures is that it is primarily a disease of cancellous bone. Although it may be true that alterations in bone mass are most evident in trabecular bone or at predominantly trabecular sites, much of the strength of whole bones resides in the cortex. This is particularly true in long bones such as the femur and recently it has been argued that extremity fractures including those of the femoral neck originate in cortical bone rather than the spongiosa.(5) This shift in emphasis is supported by both experimental evidence(6) and finite element analysis(7) indicating that the femoral neck cortex supports at least 50% of the load borne by the proximal femur. During normal gait, peak compressive stresses occur in the inferior neck cortex. These change during a sideways fall onto the greater trochanter; and peak compressive and tensile stresses at impact occur in the superoposterior (SP) and inferoanterior (IA) cortex, respectively.(7–9) Bone that is loaded habitually in compression (such as the IA cortex) is only half as strong when it is loaded in tension,(10, 11) suggesting that specific loss of cortical bone in such regions is a plausible mechanism to explain the disproportionate rise in the risk of hip fracture with aging.
To test this hypothesis, we recently have used a histomorphometric approach to analyze cortical and cancellous bone in biopsy specimens, consisting of the bone between the plane of prosthesis insertion during hemiarthroplasty and the fracture site. These were compared with similar biopsy specimens from age and sex-matched postmortem controls who had no evidence of bone disease. There were no differences in cancellous bone between groups but significant cortical thinning in the SP and IA regions was present in the fracture cases.(12) Other regions were less affected. This localized cortical thinning was accompanied by a marked increase in cortical porosity,(13) which, when combined with the reduction in cortical thickness, resulted in a predicted 60% reduction in bone strength in those very regions maximally loaded during a sideways fall onto the greater trochanter.
These histological studies clearly indicated that within a single plane of the femoral neck, fracture patients had lost cortical bone in localized but structurally important areas. However, intracapsular fractures often occur more proximally so it is important to establish whether these findings were representative of the femoral neck at potential fracture sites along its length. A further consideration is that the clinical utility of these findings currently is limited by the fact that the routine two-dimensional (2D) analysis of bone density by dual-energy X-ray absorptiometry (DXA) cannot distinguish between cortical and cancellous bone or identify local areas of bone loss within specific cortical regions. Quantitative computed tomography (QCT), given a sufficiently high resolution, is capable of making quantitative and separate assessments of cortical and cancellous bone mass.(14, 15) Peripheral QCT (pQCT) recently has been used to study the femoral neck either in cadaveric specimens(16, 17) or in vivo in nonhuman primates.(18)
The aim of this study was 2-fold: (1) to use pQCT to establish whether the previously observed pattern of cortical bone loss and the lack of any appreciable change in cancellous bone was representative of the whole femoral neck biopsy and (2) to initiate studies into the feasibility of using 3D images of the femoral neck for the development of noninvasive diagnostic approaches to the assessment of hip fracture risk.
MATERIALS AND METHODS
Subjects recruited for this study had suffered an intracapsular fracture of the femoral neck (16 females, aged 69-92 years). Of these patients, 2 denied having a fall before sustaining a fracture. All 16 cases had Garden 3/4 type fractures, which were operated on 0-6 days (mean, 1.1 days) later. The nonfractured femoral necks were collected sequentially from routine hospital postmortem cases with no previous history of diseases such as carcinoma or usage of drugs like glucocorticoids known to affect bone metabolism (15 females, aged 58-95 years; 11 males, aged 56-86 years). Postmortem subjects were not included if they had been admitted to the hospital for more than 14 days before death or if they were admitted from other hospitals or residential care. Written informed consent for participation in the study, which was approved by the local ethics committee, was obtained from all live subjects or the relatives of the postmortem controls. Of the biopsy specimens used in the current study, 4 female cases (25%), 6 female controls (37%), and 8 male controls (73%) were used in our previous study on cortical bone thickness.(12)
Femoral neck biopsy specimens from cases of intracapsular fractures were taken during standard hemiarthroplasty with minimal thermal and mechanical damage.(19) The biopsy specimens were the bones between the line of fracture and the cut face prepared for the insertion of the prosthesis at the base of the femoral neck (Fig. 1A). For the control material portion of the femur was removed from the neck at the same site. The biopsy specimens were fixed in 80% ethanol and then embedded in methylmethacrylate without prior decalcification.(12, 13, 20) After embedding, the cut face of the biopsy specimen (i.e., nearest the trochanter) was trimmed and all subsequent analyses (whether histomorphometry or pQCT) were assessed from this cut face toward the fracture site (or the femoral head in the control biopsy specimens). Orientation of the sections was dependent on the identification of the thick inferior cortex and whether the biopsy specimen has been taken from the left or right femur.
Histological analysis of cortical widths was done in 33 of the biopsy specimens (10 female controls; 15 female cases; 8 male controls) for comparison with that derived from pQCT. Briefly, 10-μm sections, parallel to the face of the biopsy specimen from the trochanteric side, were cut with a Jung Polycut microtome (Leica, Milton Keynes, UK) and stained using the von Kossa protocol. Gray level images of the whole cross-section (1 pixel = 50 μm) were thresholded and the regional cortical widths were analyzed using proprietary image analysis software described previously.(12) For each biopsy specimen the data were averaged from three sections approximately 60 μm apart.(12) These were compared with the regional cortical widths derived from the first 1-mm usable pQCT slice.
The trimmed biopsy specimens were taped to a jig so that they were central in the field of view of the pQCT (Densiscan 1000; Scanco Medical Ag, Zurich, Switzerland).(21, 22) This is a second-generation translational and rotational CT scanner that uses a narrow fan beam and a highly collimated detector system to reduce scattering and increase resolution (maximum, 0.192 mm). The radiation source has a maximum voltage of 50 kVp and an effective energy of 39.5 keV with the X-rays filtered to a narrow energy band. Quality assurance was performed using the Institute for Biomedical Engineering of ETH (IBT) phantom and interim measurements were made with the European forearm phantom(21) to ensure machine stability (CV over 1 year: area, 0.096%; linear attenuation coefficient, 0.086%).
Scientific software was used to allow the preparation of a user-defined control file, which was used for all the specimens. Initially, a scout view of the junction between the jig and the biopsy specimen was acquired to enable the operator to place accurately a reference line at this point. From this reference point, tomograms (1 mm thick; 10 cm scan diameter) were then made at 0.5-mm intervals for a total distance of up to 1 cm in the controls (20 slices) or 0.5 cm in the cases (10 slices). The first tomogram, which included the biopsy specimen and the end of the jig, was discarded.
Bone areas and density
Images were reconstructed using a convolution and backprojection technique with a Shepp and Logan filter. Where possible, all subsequent slices were analyzed using a semiautomatic approach. The operator drew one region of interest (ROI) that encompassed the whole specimen and a second ROI that approximately corresponded to the apparent endocortical boundary. These ROIs were then adjusted by an automated iterative peeling procedure, which sought the delineation between bone and soft tissue based on their linear attenuation coefficient (bone, 1.421 cm−1; marrow, 0.241 cm−1). The average linear attenuation coefficient of methylmethacrylate surrounding the biopsy specimen was close to that for marrow (0.281 cm−1). Reducing the number of iterations for the internal ROI allowed the operator to have more control on placement of the endocortical boundary. This was necessary because of increased variation in bone density on the endocortical surface. In some of those slices nearest the fracture site it was not possible to determine the endocortical boundary because regions of the cortex were missing.
Cortical and cancellous bone areas (mm2) were measured by summing the number of pixels between the two ROIs (cortical area) or within the internal ROI (cancellous area). The bone densities for each of the two regions were computed by the Densiscan software. For the measurement of bone density, the average linear attenuation coefficient within each of the regions was calculated and converted into a density (g/cm3) using the known values for bone and marrow and assuming a compact bone density of 2.2 g/cm3. Cortical and cancellous bone masses (mg) for each slice were calculated by multiplying area, slice width, and density.
Image analysis of cortical widths
TIFF images of each slice were imported into the public domain National Institutes of Health (NIH) image program v1.61 (this software was developed at the U.S. NIH and is available on the Internet by anonymous FTP from zippy.nimh.gov or on floppy disk from the National Technical Information Service, Springfield, VA, USA [part number PB95-500195GEI]). An initial macro was used to invert the image and subtract the background gray level, which was set as the mean + 2 SD of the gray levels associated with the methylmethacrylate (Figs. 1B and 1C). A second macro was used to measure the gray levels in 72 radial profiles (every 5°) from the center of area (as determined during the pQCT analysis), which was then transferred to Excel 98 for further analysis (macros are available from the authors).
For each profile, the distances from the center of area to the endosteal and periosteal surfaces were calculated using a full-width/half-maximum algorithm and the resultant cortical widths were determined. The mean cortical width for each of the eight octants around the femoral neck was determined by averaging the values from each of the nine profiles for that octant (Fig. 1B).
To help assess the location of the biopsy specimens along the axis of the neck, the maximum and minimum external diameters of the each of the pQCT slices were measured. For the control biopsy specimens, the ratio of these dimensions declined from 1.79 near the trochanter to 0.96 in the subcapital region.
Density corrected bone cross-sectional area and estimated DXA BMD
Based on the assumption that differences in cortical density were more likely to be caused by differences in porosity rather than mineralization, cortical areas were adjusted by multiplying by the ratio of measured cortical density to the assumed density of fully mineralized compact bone. To provide an estimate of the BMD that would be provided by a DXA scanner, the sum of the cortical and trabecular masses (mc and mt) were used in the following formula:
where dmax is the maximum outer diameter, w is the CT slice width, and the last term is the ratio of DXA BMD(23) to bone tissue density.
For the analysis of differences at a single plane, the data are presented as the mean ± SEM of the individual patients within each sample group (i.e., female fracture, female control, or male control) and analyzed using a single analysis of variance (ANOVA; JMP v3.2.2; SAS Institute, Cary, NC, USA). This was followed by the Tukey-Kramer Honest Significance Difference (HSD) test if there were more than two sample groups.
Bone area, density, and mass for an individual subject were obtained by averaging the data from each pQCT slice over a distance for which approximately equal numbers of slices were present in the sample groups being analyzed. This represented a distance of ∼4 mm for the comparison of cases and controls and 7 mm for the comparison of male and female controls. The data are presented as the mean ± SEM of individual subjects within each sample group and analyzed by unpaired two-tailed t-test.
To analyze the effect of distance along the femoral neck on bone areas, density, and mass, multivariate analyses (least squares regression) were used. For these, subject was nested within sample group (i.e., female fracture cases, female controls, and male controls). Sample group, distance, and any interaction between sample group and the distance along the femoral neck were considered as independent variables.
To investigate differences in cortical widths between female fracture cases and female controls, the data for each region from an individual subject was averaged initially over the 4 mm. Least squares regression analysis was used to model the data. Cortical thickness was the dependent continuous variable and subject nested within sample group (fracture and control), sample group, region, and any interaction between sample group and region were considered as independent variables. To assess any possible effect of distance along the neck on cortical thickness within individual regions, the data from each region was modeled separately by least squares regression. Subject nested within sample group, sample group, distance, and any interaction between sample group and the distance along the femoral neck were considered as independent variables.
To account for both the differences in location and the variation in the number of complete slices for each biopsy specimen, the pQCT parameters for individual slices were adjusted to a standard maximum-to-minimum ratio. This was done by initially estimating the distance of each slice from the point where the ratio would have been 1.4 using the slope of the change in the ratio with distance along the neck in the female cases and controls (0.04998/mm). Each parameter was then adjusted by the product of this distance and the change per millimeter in that parameter.
Ratio of external diameters
Among the female fracture cases and controls, the mean ratio of the maximum-to-minimum external diameters decreased at very nearly the same rate with increasing distance from the cut face toward the femoral head (fractures: 0.051/mm, p < 0.0001, and adjusted r2 = 0.984; controls: 0.048/mm, p < 0.0001, and adjusted r2 = 0.991). However, the mean value at the cut face was significantly higher for the female controls (female controls: 1.535 ± 0.037; male controls: 1.429 ± 0.043; female fractures: 1.413 ± 0.029; p < 0.05 Tukey-Kramer HSD). In male controls the ratio decreased more slowly with distance toward the head (0.041/mm, p < 0.0001, and adjusted r2 = 0.994). The difference in ratio between female cases and controls was interpreted as an effect of positioning of the cuts made by the orthopedic surgeon (cases) or the mortuary technician (controls) to separate the specimen from the more distal part of the femur. To compensate for differences in this position along the femoral neck, subsequent analyses were done starting at locations where the mean ratio of maximum-to-minimum external diameters was similar between the sample groups. This corresponded to 2.5 mm and 0.5 mm toward the femoral head in the case of female and male controls, respectively (Fig. 2).
Sex differences in cortical and cancellous bone in control femoral necks
As expected (Table 1) total bone and cancellous and cortical bone areas were significantly different in the smaller bones of females compared with males. Similarly, total bone mass and estimated DXA BMD also were lower in females. Cortical and cancellous densities and cortical masses were not significantly different although female cancellous mass was lower than that of males (−32.6%; p = 0.007). After correction for density differences, the cortical areas, the relative amounts of cortical bone area, and cortical masses in the cross-section were no longer significantly different (p > 0.05).
With regard to changes along the neck, there were minimal (<0.7%) but statistically significant changes in total bone area along the axis of the neck, which were different between males and females (Table 2). Cortical bone area, density, and mass decreased toward the femoral head (area, −1.6%/mm; density, −2.5%/mm; mass, −3.9%/mm) but only the change in area was influenced markedly by gender with the females having a greater rate of change per unit distance (+23%; p = 0.004). This is likely because of the relatively shorter femoral neck in the females. Although cancellous bone area (Table 2) increased slightly but significantly along the neck axis (+1.4%/mm; p < 0.001), the marked change in cancellous bone density (+4.5%/mm; p < 0.001) resulted in an increase in cancellous bone mass (+6.4%/mm; p < 0.001). Cancellous bone density increased slightly more in females (+0.6%/mm compared with males; p < 0.001) but the increases in cancellous bone area (p = 0.44) and mass (p = 0.11) were unaffected by gender. This effect on cancellous bone was interpreted as a size scaling difference between the sexes.
Cortical and cancellous bone in female cases and controls
The available length of femoral neck biopsy material varied somewhat among the fracture cases between the cut face (toward the trochanter) and the fracture plane. To determine whether the length of specimen influenced the results, biopsy specimens from the fracture cases initially were categorized into those <2 mm long (10/16) and those 2-4 mm long (6/16). The ratio of maximum-to-minimum diameter of the first slice was not different between these groups (<2 mm, 1.39 ± 0.05; 2-4 mm, 1.45 ± 0.03; p = 0.33), suggesting that these categories represented groups where the site of the fracture was either closer to the femoral head (length, 2-4 mm) or those where the fracture was closer to the middle of the femoral neck (length, <2 mm) rather than bone from different regions along the femoral neck. Although cancellous bone mass was higher in the shorter biopsy specimens (<2 mm, 163.2 ± 8.7; 2-4 mm, 120.0 ± 13.2; p = 0.013) in neither group was this parameter different from the control biopsy specimens (143.5 ± 12.7). Because there were no significant differences in any cortical parameters or in cancellous bone area or density, these subgroups of fracture cases were subsequently combined and treated as a single group in further analyses.
As shown in the comparison of female fracture cases and controls in Table 3, there were no differences in the mean total area, but total bone mass (−15%; p = 0.013) and estimated DXA BMD were reduced in fracture cases (−12%; p = 0.025). However, these differences were in cortical bone not cancellous bone; the mean cancellous bone area, cancellous bone density, and mass did not differ between control subjects and the fracture cases. Adjusting the parameters in each individual slice to a maximum-to-minimum ratio of 1.4 accentuated the changes in cortical bone (Table 3). In the fracture cases mean cortical bone area was 15% lower than in the female controls (p = 0.01) and mean cortical bone density was reduced by 13% (p < 0.001) in the cases compared with the female controls. This resulted in a marked reduction in the mean cortical bone mass (25%; p < 0.001). Cortical areas in the cases were markedly smaller after adjusting for density (−26%; p < 0.001) or total area (−21% smaller; p = 0.008). Reanalysis of the data after exclusion of the two cases in which falling was denied did not affect these changes (e.g., cortical mass: controls, 235.2 ± 13.2, and cases, 179.2 ± 8.0, p = 0.001; cancellous mass: controls, 143.5 ± 12.6, and cases, 147.9 ± 7.4, p = 0.77).
The differences in the cortical bone mass in the female cases compared with the female controls were maintained over the whole of the 4 mm between the cut face and the fracture plane (Fig. 3A). Least squares regression analysis (adjusted r2 = 0.971) showed that although cortical bone mass was variable between subjects (p < 0.001) it was significantly reduced in the fracture cases (p < 0.001). Cortical bone mass decreased by 12.4 ± 0.63 mg/mm, moving proximally along the neck (p < 0.001) but this rate of change was not different between cases and controls (p = 0.5820). Cancellous bone mass increased as cortical bone mass decreased (Fig. 3B). Least squares regression analysis (adjusted r2 = 0.976) showed that cancellous bone mass also was variable between subjects (p < 0.001) but was not significantly reduced in the fracture cases (p = 0.4380). Cancellous bone mass increased by 7.5 ± 0.48 mg/mm proximally along the femoral neck (p < 0.001) but this was not different between cases and controls (p = 0.32).
Regional cortical widths in cases and controls
There was excellent agreement between the cortical widths at the cut face of each biopsy specimen as measured by histomorphometry and pQCT. This was true for all sample groups (female cases: adjusted r2 = 0.846, p < 0.001, and CT width = 0.22 + 0.8 × histological width; female controls: adjusted r2 = 0.933, p < 0.001, and CT width = 0.05 + 0.96 × histological width; male controls: adjusted r2= 0.877, p < 0.001, and CT width = 0.11 + 0.87 × histological width). Over 66% of the root mean square error in width estimation within each group could be accounted for by the effect of the resolution of the pQCT (0.192 mm).
The mean cortical widths (over a distance of 4 mm) for each of the eight regions of the femoral neck are shown in Fig. 4A. In both cases and female controls, the SP region was thinnest but was not different between cases and controls. Both the IA and the inferior (I) regions were thinner in the cases (IA: −22.2% and p = 0.002; I: −18.8% and p < 0.001). Reanalysis of the data after exclusion of the two cases in which falling was denied did not markedly affect these changes (IA: −18.9% and p = 0.012; I: −17% and p < 0.001). The cortical widths in the inferior and IA regions decreased at around 5%/mm over the 4 mm toward the femoral head but this rate of change was similar for both cases and controls (Figs. 4B and 4C).
Osteoporosis and associated fractures often are viewed as a failure of cancellous bone so insufficient recent attention has been given to the possible role of cortical bone. Cortical bone strength is dependent not only on bone mass but also on its degree of mineralization and porosity(24, 25) and the distribution of bone (geometry) around the axis through which it is loaded.(26–28) Some of these material and geometric factors are evident in measurements of bone mass, although not always in an easily interpretable manner. Cortical bone strength is clearly important in cases of intracapsular hip fracture in which we have reported histological evidence of increased porosity(13) and cortical thinning(12) with only minor changes in cancellous bone.
In this study we have taken considerable care to ensure that analyses were done at similar locations for all the sample groups. This was important because the proportion of the total bone mass that is cancellous bone increases markedly between the neck adjacent to the trochanter (17%) and that adjacent to the femoral head (71%).(16) The best available method of adjusting the relative position along the femoral neck is to use the ratio of the external maximum-to-minimum diameter; this declines from 1.7 at the trochanter to around unity at the femoral head. This approach is based on the assumption that the external cross-sectional shape (not the size) of the femoral neck is similar between cases and controls. There is no evidence that differences in this ratio might occur between cases and controls although cases and controls may differ moderately in some other geometric proportions such as the hip axis length. After adjusting the distance along the femoral neck so that the ratios were the same (1.4) the proportion of total bone mass that was cancellous bone in the control biopsy specimens (both male and female) was 36 ± 10%.(16) This is very close to the “midpoint” of the femoral neck as defined by Kuiper et al.(16) based on the smallest total bone volume in sequential QCT slices. At this location, the proportion of cancellous bone was 34 ± 10%. At a location 6 mm further toward the femoral head, the proportion of cancellous bone mass was 49 ± 11% in the current study and 54 ± 11% in the Kuiper study.(16)
As expected, averaged measures of bone mass such as pQCT total bone mass and estimated DXA BMD were reduced in fracture cases compared with controls. However, these differences were caused by cortical bone rather than cancellous bone. The current study adds strong support to the hypothesis that losses in cortical bone are the predominant feature related to osteoporosis in the fractured femoral neck. Throughout the biopsy specimens examined there was no evidence for any marked reduction in cancellous bone mass. Although the proportion of cancellous bone mass increased at a rate of 5.5%/mm in nonfractured femoral necks as the plane of analysis moved toward the femoral head, a similar rate of change was seen in the fracture cases (5.2%/mm; p = 0.33).
In contrast, cortical bone mass was reduced by a quarter in cases of intracapsular hip fracture. This was a result of a combination of cortical thinning, as shown by the reduction in cortical area, and a decrease in the density of the cortical bone. This decrease in cortical bone density may represent either an increase in cortical porosity or a decrease in the mineralization density. Generally, changes in cortical porosity would be difficult to detect with QCT because the limit of resolution (approximately 1 mm) is too high to detect changes in Haversian canal diameter (mean, approximately 0.05 mm). However, in cases of femoral neck fracture the major contributor to increased cortical porosity is the increased proportion of canals with diameters exceeding 0.35 mm rather than a more generalized increase in overall canal diameters.(13) In the current study, the pQCT images have a resolution below 0.2 mm so very large canals such as these would be detected; nevertheless, most would not and a larger proportion of canals would appear as reduced cortical density as seen here (Table 1). It is also likely that the reduced cortical density is at least partially the result of a partial volume effect. It has been reported (based on scanners with a 0.3-mm resolution) that QCT-measured cortical density falls linearly with cortical thickness when the cortical width is below 2 mm.(29) When we examined the relationship of CT cortical thickness to that measured with histological methods the slope was 1.00 with an r2 = 0.90. However, when the comparison was restricted to cortical thicknesses <1 mm, the slope was reduced to 0.3 and r2 to 0.15. Because the CT method tended to overestimate thickness of thinner cortices, it is likely that partial volume averaging is responsible for at least some of the apparent reduction in cortical density among fracture cases. When the cortical area was adjusted to a constant density to eliminate the effects of porosity, the adjusted cortical area was 25% lower in the fracture cases; moreover, the cortical proportion of the total cross-section became significantly smaller. This observation would suggest significantly higher bending stresses under a given load because such stresses are predominantly in the cortex. The pattern of reduction in cortical bone areas (apparent), density, and mass in the fracture cases (Fig. 3) continued along the femoral neck despite the fact that the cortical shell becomes thinner as the proportion of cancellous bone increases.
We used the gray-level images derived from the pQCT analysis to assess regional cortical widths. This confirmed our previous report(12) that the cortex of fracture cases was preserved in the superior region of the femoral neck, which is loaded in tension during normal gait.(7) Similarly, we confirmed that the cortical thickness of fracture cases in the IA region (which is maximally loaded in tension during a sideways fall) is reduced by over 20%. Unlike our previous report, we could not confirm that the SP cortex, which is loaded in compression during a fall, was also thinner in fracture cases. This may have been related to the poorer resolution of the pQCT approach, which has a resolution of 0.192 mm against a cortical thickness of the SP region of only 0.65 mm. The inferior cortex also was thinner in the fracture cases but other regions showed no significant reductions in cortical width. Nevertheless, the fact that cortical densities were lower in fracture cases suggests that there is proportionately less bone in the cortex, which should result in higher stresses in a sideways fall. Overall, the partial volume effect probably led to underestimates of the difference in regional cortical thicknesses between cases and controls.
We also examined the differences between male and female control specimens and found some important differences relevant to this study. The bigger bones of males had significantly larger total, cortical, and cancellous areas and bone masses. However, after adjustment for cortical density, the sex difference in cortical area became nonsignificant. Estimated DXA BMD was higher in male controls as well, but cortical and cancellous densities were indistinguishable and there was no apparent difference in the proportion of the cross-section occupied by cortical bone. These differences appear to be largely scaling differences, rather than differences in the proportions of cortical and trabecular bone as seen when comparing female fracture cases with controls.
This study may go some way to explain the relatively modest ability of areal BMD (aBMD), as determined by DXA, to predict the risk of hip fracture. This may be caused by the inability of this measurement to distinguish cortical from trabecular bone. Although epidemiological studies have shown that the risk of hip fracture increases around 14-fold between the ages of 60 and 80 years, the reduction femoral neck aBMD only accounts for a 2-fold increase in risk.(1) The increase in the risk of falling is only about 1.1- to 1.5-fold,(1, 2) so other factors such as bone quality and its distribution are likely to play an important role in the pathogenesis of hip fracture.
This study indicates that to develop improved diagnostic approaches for assessing hip fracture risk, it may be necessary to analyze bone properties using 3D images or to measure parameters in 2D that relate to IA cortical bone strength. Current routine DXA technologies cannot measure cortical and cancellous bone separately or identify changes within specific cortical regions. Furthermore, the current trend is to concentrate on the measurement in the total hip rather then the femoral neck because of the better long-term reproducibility of measurements at that site. Improved reproducibility, if it occurs at the expense of diagnostic discrimination, is not a sound choice when the measurement is being made for diagnosis. However, for reasons of low cost, low radiation dose, and ease of use, DXA is the most widely used method for assessing bone mass in osteoporosis. Other noninvasive imaging technologies such QCT and magnetic resonance imaging (MRI)(30) have particular advantages in their ability to assess selectively cortical bone, but so far there have been technical limitations to measuring the femoral neck in 3D with sufficiently good resolution. Measurement of components of bone strength associated with its architectural distribution from 2D images might be a complementary or alternative approach to providing relevant information not evident in the volume-averaged BMD measurement.(26, 27) Such measurements have shown potential in explaining geographical differences in hip fracture rates.(31)
There are number of limitations to the current study. First, it has involved only a moderate number of fracture cases and controls. However, 75% of the cases and 64% of the controls had not been analyzed in our previous histological study. Nonetheless, the results of the two studies are very similar, suggesting that our earlier results were not because of chance. Within the current study the majority of patients had suffered a subcapital fracture so it was not possible to investigate differences between this fracture type and those occurring in the midneck. This was primarily because the smaller quantity of bone between a transcervical fracture and the site of prosthesis insertion makes it difficult to prepare biopsy specimens that contain a complete or nearly complete cortical ring. There are technical difficulties with removing the femoral head and proximal neck for study because its quick removal usually is destructive of the bone architecture and for ethical reasons surgeons do not like to prolong operations in this vulnerable group of patients. However, cancellous bone is reported to contribute over 70% of the bone strength in the immediate subcapital region declining to around 50% at the midneck(7); therefore, we would predict that transcervical fractures would be even more dependent than subcapital fractures on the loss of cortical bone strength.
Given the fact that fractures occur at variable locations in the neck, we could not analyze equal numbers of slices for the cases and controls. This was particularly important in the measurement of cortical and cancellous bone mass because there were only five and three hip fracture cases in which biopsy specimens permitted a full series of analyses over distances of 3 mm and 4 mm, respectively. Although the differences were not significant, cancellous bone mass was lower and cortical mass was higher at the midneck position of these subjects. This suggests that if all the biopsy specimens had been long enough, the difference between cases and controls at planes closer to the femoral head would have been even greater. The pQCT machine and its associated algorithms have been designed for in vivo studies. This will have implications for the extrapolation of the data to the “real” in vivo values for bone mass. The correction for the effect of beam hardening on the linear attenuation coefficient was not optimized for in vitro studies. On the other hand, because the biopsy specimens from cases and controls were of equivalent sizes, any effects of beam hardening probably were similar between the two groups. Partial volume effects, even in this high-resolution scanner, appear to cause some overestimation of the width of the thinnest cortices. This also was apparent in the comparisons of cortical widths as determined by histology in the fracture cases. However, because the cortical widths in the cases are significantly thinner than in the controls such errors would systematically underestimate any differences between the two groups.
In conclusion, this study has confirmed that loss of cortical bone as opposed to cancellous bone is the predominant feature and a candidate contributing cause of intracapsular hip fracture. Overall, fracture cases have proportionately less cortical bone in the cross-section than their nonfractured controls. This study also shows that cortical bone loss is greatest in a region that is habitually loaded in compression but that bears the maximum tensile load during a sideways fall. Bone that is normally adapted to compression loading is only half as strong when loaded in tension(10, 11); so these observations have clear implications for understanding the pathogenesis of hip fracture. This study also highlights the importance of understanding the mechanisms associated with cortical bone remodeling as opposed to cancellous bone remodeling and the need to develop noninvasive methodologies that allow the discrimination of cortical bone from cancellous bone and enable the analysis of critical cortical regions of the femoral neck. Such approaches ultimately could provide better detection of those at risk of femoral neck fractures and thus better selection of subjects for subsequent therapeutic and/or lifestyle-based interventions.
This work was supported by a Medical Research Council (MRC) program grant 9321536 (N.L. and J.R.) and NIH (National Institutes of Arthritis and Musculoskeletal and Skin Diseases [NIAMS]) grant number RO1 AR44655 (T.J.B.).
- 11997 Bone density and the risk of hip fracture in men and women: Cross sectional analysis. BMJ 315:221–225., , , ,
- 21989 Perimenopausal risk of falling and incidence of distal forearm fracture. BMJ 298:1486–1488., ,
- 31998 Bone mass, bone fragility and the decision to treat. JAMA 280:2119–2120.
- 41998 Effect of alendronate on risk fracture in women with low bone density but without vertebral fractures: Results from the fracture intervention trial. JAMA 280:2077–2082., , , , , , , , , , , , ,
- 51995 Perspectives on osteoporosis research: Its focus and some insights from a new paradigm. Calcif Tiss Int 57:399–404., , , , , , ,
- 61983 The mechanical characteristics of cancellous bone at the upper femoral region. J Biomechan 16:971–983., , , ,
- 71995 Stress distribution within the proximal femur during gait and falls: Implications for osteoporotic fracture. Osteoporos Int 5:252–261., ,
- 81996 Impact direction from a fall influences the failure load of the proximal femur as much as age-related bone loss. Calcif Tiss Int 58:231–235., , , ,
- 91996 The effect of impact direction on the structural capacity of the proximal femur during falls. J Bone Miner Res 11:377–383., ,
- 101993 Mechanical implications of collagen fibre orientation in cortical bone of the equine radius. Anat Embryol 187:239–248., , , ,
- 111999 The development of microcracking and failure in bone depends on the loading mode to which it is adapted. J Exp Biol 202:543–552.,
- 121999 Structure of the femoral neck in hip fracture: Cortical bone loss in the inferoanterior to superoposterior axis. J Bone Miner Res 14:112–120., , , , , , ,
- 131999 Regional differences in cortical porosity in the fractured femoral neck. Bone 24:57–64., , , , ,
- 141999 Bone density reduction in various measurement sites in men and women with osteoporotic fractures of spine and hip: The European Quantitation of Osteoporosis Study. Calcif Tiss Int 64:191–199., , , , , , , , , , , , , , , , , , ,
- 151998 Assessment of bone mineral at appendicular sites in females with fractures of the proximal femur. Bone 22:395–402., , , , , , ,
- 161997 Distribution of trabecular and cortical bone related to geometry. A quantitative computed tomography study of the femoral neck. Invest Radiol 32:83–89., ,
- 171997 Assessment of the strength of proximal femur in vitro: Relationship to femoral bone mineral density and femoral geometry. Bone 20:213–218., , , , , ,
- 181999 Use of peripheral quantitative computed tomography for densitometry of the femoral neck and spine in cynomolgus monkeys (Macaca fascicularis). Bone 24:101–107.
- 191993 Scientific letter. Bone 14:183.,
- 201993 Methods for the histological study of femoral neck bone remodelling in patients with fractured neck of femur. Bone 14:249–255., , , , , , , , ,
- 211994 Europeran semi-anthropomorphic phantom for the cross-calibration of peripheral bone densitometers: Assessment of precision accuracy and stability. Bone Miner 27:109–120., , , , , , , , , , , , , , , , , , , ,
- 221994 The use of peripheral QCT in the evaluation of bone remodelling. Endocrinologist 4:167–176.
- 231984 Non-invasive measurement of long bone cross-sectional moment of inertia by photon absorptiometry. J Biomechan 17:195–201.,
- 241988 The effect of porosity and mineral content on the Young's modulus of elasticity of compact bone. J Biomechan 21:131–139.
- 251988 Stiffness of compact bone: Effects of porosity and density. J Biomechan 21:13–16.,
- 261998 Experimental testing of a dexa-derived curved beam model of the proximal femur. J Orthop Res 16:394–398., , , ,
- 271993 Age-related changes in female femoral neck geometry: Implications for bone strength. Calcif Tissue Int 53(Suppl 1):S41–S46., ,
- 281996 Curved beam model of the proximal femur for estimating stress using dual X-ray absorptiometry derived structural geometry. J Orthop Res 14:483–492., , , ,
- 291996 Evaluation of cortical bone by computed tomography. J Bone Miner Res 11:1518–1525.,
- 301998 Noninvasive assessment of bone density and structure using computed tomography and magnetic resonance. Bone 22:149S–153S., , , ,
- 312000 Hip geometry, bone density, and bone strength in European men and women: The EPOS study. Bone 27:161–168., , , , , , , , , , , , , , , , , , , , ,