Although bone mass is a contributory risk factor for hip fracture, its distribution about the femoral neck is also important. Femoral neck biopsies were obtained from 13 females with intracapsular hip fracture (fracture: mean age 74.3 ± 2.3 years [SEM]) and 19 cadaveric samples (control: 9 males and 10 females 79.4 ± 1.7 years) and the areas of cortical and cancellous bone were quantitated in octants. In the control group, although males had larger bones than females, the proportions of cortical and cancellous bone were not different (p > 0.05) between the genders. The total amount of bone, as a proportion of bone + marrow, was significantly reduced in the fractures compared with the female controls (%Tt.Ar: fracture 27.83 ± 1.18, female control 33.62 ± 1.47; p = 0.0054). Reductions in cortical bone area occurred in all regions but particularly in the inferior, inferoanterior, and anterior octants (p < 0.05). There were no differences between cases and controls in the regional amount of cancellous bone (all regions, p > 0.178). Marked reductions in mean cortical bone width between the fracture and female control group occurred in the anterior, inferoanterior (31%), and superoposterior (25%) regions. Representing cortical widths as simple Fourier functions of the angle about the center of area (R2adj = 0.79) showed in the cases that there was preservation of the cortical bone in the inferior region, with the proportional loss of cortical bone being greatest in the inferoanterior and superoposterior regions. It is concluded that loss of cortical, rather than cancellous, bone predominates in cases of femoral neck fracture. This loss occurs primarily along the inferoanterior to superoposterior axis. As this axis bears the greatest strain during a fall, it is hypothesized that specific thinning of the cortex in these regions leads to an exaggerated propensity to fracture in those so affected, above that resulting from an equivalent general decrease in bone mass.
Determinants of hip fracture risk include bone mass and structure, a tendency to fall and other related factors such as postural instability and impaired visual acuity.(1,2) Fractures are usually classified as either intracapsular (femoral neck) or extracapsular (trochanteric) and are a significant cause of morbidity and mortality in the elderly. Recently, it has been suggested that intracapsular fractures tend to be associated with a family history of hip fracture and have a weaker relationship with generalized bone loss.(3) The decline in bone mass with age in this region is known to be a contributory factor to fracture risk.(4) However, De Laet et al.(5) estimated that the risk of hip fracture increases 13-fold from age 60–80 years, but the decrease in bone density with age during this period would only contribute to a doubling of the fracture risk.(5) Hence, changes in other factors such as the structure of bone, its distribution and quality are also considered to be important(6) and need to be investigated at this clinically relevant site.(7)
Despite the pathological importance of femoral neck fracture only a few investigators have examined the bone structure and histology of the femoral neck.(8-16) From these studies, it is apparent that there are changes in the structural features of the femoral neck associated with aging and fracture. Eventov et al.(10) have demonstrated osteopenia in the trabecular microanatomy of femoral neck biopsies taken superiorly. Cortical remodeling changes resulting in a reduction in the amount of cortical bone and increased porosity due to expansion of the Haversian canal area have also been identified.(11-12,14-16) However, the information gained from these studies has been limited. Some of the studies have only examined fracture cases(8) or femoral necks from autopsy samples.(14) Other studies have compared fractures with osteoarthritic cases(15) instead of age-matched cadaveric material, femoral neck parameters to those of the iliac crest,(9,10) or have only studied a selected region of the femoral neck.(11,12)
In a previous study, an image analysis system to assess the proportions and distribution of cortical and cancellous bone in the femoral neck was evaluated.(17) From this initial evaluation, various modifications to the package have been incorporated, and in the current study the effect of these changes on reproducibility have been determined. In this study, this software has now been utilized to examine two hypotheses related to the role of cancellous and cortical bone in fracture risk. These were: that the loss of bone mass in the fracture cases is due to localized rather than a generalized loss of cortical and/or cancellous bone, and that such losses are associated with the plane of greatest stress following a fall increasing the propensity to fracture. Thus, the amounts of cancellous and cortical bone and the regional variation in cortical width have been assessed in femoral neck cross-sections of biopsies taken from intracapsular hip fractures and compared with age-matched cadaveric material.
MATERIALS AND METHODS
Subjects recruited to the study had suffered a nontraumatic intracapsular fracture of the femoral neck resulting from a fall from less than standing height. The 13 fracture cases were females aged 60–91 years (fracture: mean = 75.4 ± 2.1 years [SEM]) and the biopsies were taken 0–6 days postfracture (mean = 1.5 ± 0.4 days; Table 1). Written informed consent to participation in the study, which was approved by the local ethics committee, was obtained for all subjects. Fractures were categorized as transcervical (n = 4) or subcapital (n = 9) by an expert radiologist.
Table Table 1.. Summary of Subject Details and Parameters Studied
The nonfractured femoral necks were selected from routine postmortem cases with no previous history of disease such as carcinoma or usage of drugs such as glucocorticoids known to affect bone metabolism. In addition, subjects were not included if they had been admitted to hospital for more than 14 days prior to death (mean length of time: female 3.0 ± 0.7 days, male 5.3 ± 1.6 days; Table 1) or if they were admitted from other hospitals or residential care. There were 9 males and 10 females aged 62–89 years (control: female 79.9 ± 2.9 years, male 78.8 ± 1.7 years) at the time of death, and samples were obtained 1–5 days after death (mean = 3.1 ± 0.3 days; Table 1).
Femoral neck biopsies were taken during standard hemiarthroplasty with minimal thermal and mechanical damage.(18) For the control material, bone was removed from the neck at the same site. Biopsies were fixed in 80% ethanol and embedded in methylmethacrylate without prior decalcification.(13) Having identified the inferior cortex (which is always the thickest cortical region), the anterior region was identified on the basis of which side (i.e., left or right) of proximal femur the biopsy was taken. Ten-micron sections, parallel to the face of the biopsy, were cut from the trochanteric side of the biopsy with a Jung Polycut E microtome (Leica, Milton Keynes, U.K.) and stained using the von Kossa protocol.
Analysis of the femoral neck utilized software described previously.(17) Three von Kossa sections, a minimum of 60 μm apart, were quantitated using a segmental analysis that divides the neck into a variable number of regions with the center of area (calculated as the mean of the x and y coordinates of the pixels on the external edge of the cortical shell) as the central axis. The program requires a complete cortical shell before the analysis can be completed, so any gaps in the section due to an incomplete biopsy, vascular channels, or preparative artefacts must be “filled in.” Data for these regions are flagged, indicating possible inaccuracies and that data derived from flagged parts were not used in any further analysis. Table 1 indicates which of the biopsies were affected and which regions have been excluded due to them being incomplete.
For this study, the maximum and minimum diameters of the femoral neck cross-section were measured. Each cross-section was then divided into eight regions (Fig. 1). Briefly, the amount of bone was quantitated per octant. The measurements obtained were total area (bone plus marrow, Tt. Ar [%]), cortical area (Ct.Ar [%]), and cancellous bone area (cancellous bone area as a percentage of the area of trabecular bone and marrow space, Cn.Ar [%]).
To measure cortical width (minimum and mean cortical widths, Ct.Wi), the width of the cortex was measured automatically every 2.8° around the femoral neck (128 measurements) starting at the inferoposterior boundary. If the measurement occurred on a flagged part of the section, the width was censored to avoid any inaccuracy. For each region, the data presented are the mean and minimum width, providing at least 75% of the measurements in that octant were not flagged.
Modifications to package
Following our initial evaluation of the software,(17) various modifications were incorporated into the program. These changes were designed to minimize operator error during image capture in an attempt to improve the reproducibility between operators. In the current version of the software, the calculation(17) for the determination of threshold levels has been incorporated and the identification of cortical holes is automatic based on predetermined criteria. Each cavity is dilated by 50% of its diameter, the amount of bone within this boundary calculated, and if it is > 75%, then the cavity is included in the cortex. The modified package was evaluated by three of the original operators using six of the original archaeological samples.(17) In addition, intraoperator reproducibility was determined using six of the archaeological samples that were analyzed three times at a minimum of fortnightly intervals. Table 2 is a summary of the inter- and intraoperator variation in the analysis with the modified program.
Table Table 2.. Effect of the Modifications of the Package on Interoperator and Intraoperator Variability
We analyzed the regional cancellous and cortical bone distribution. To obtain the data for each individual, the measurements for the three sections were averaged to give one set of values for each biopsy. In the control group, data were analyzed to identify regional differences in the bone distribution between the females and males. The female control group was compared with the female fracture group to identify the effects of disease on the regional distribution of bone. The data are presented as mean ± SEM. Where data were normally distributed, single analysis of variance (ANOVA) was used to identify regional differences, with the differences between groups being analyzed by the Tukey–Kramer honestly significant difference (HSD) test (JMP statistical package; SAS Institute, Cary, NC, U.S.A.). Where data were not normally distributed, the differences were analyzed using the nonparametric Wilcoxon/Kruskal–Wallis test (rank sums).
Modeling strategy in the frequency domain
Because there are systematic changes in cortical width across individual octants, leading to a reduction in the power to separate cases and controls, it was decided to model the cortical width data in the frequency domain. Modeling cortical widths takes account of the subject variation, as well as location, biopsy type (i.e., fracture or control), and gender. It was postulated that modeling in the frequency domain would make it possible to mathematically predict cortical widths at any location in the femoral neck. A model with fewer independent variates (i.e., fewer than 128 separate width measurements) has equal or greater power to achieve between-group separation at specific locations within the femoral neck cortex.
The circumferential distribution of cortical widths in both the control (male or female) and fracture (complete or incomplete) groups were modeled using the JMP statistical package. To normalize the distribution of residuals, the square root of the cortical width (dependent variable) was adopted. Least squares regression models used (1) a simple Fourier series, periodic functions of the measurement angle (sin angle + cos angle, sin 2× angle + cos 2× angle, sin 3× angle + cos 3 × angle, sin 4× angle + cos 4× angle, sin 6× angle + cos 6× angle, sin 8× angle + cos 8× angle, sin 9× angle + cos 9× angle), or (2) the 128 width measurements as independent categorical variables. For both models, subject and gender or disease were used as independent categorical variables, and disease or gender were modeled as an interaction with the angular location of width. Comparison of these two models demonstrated that the model with 128 independent widths measurements improved the goodness of fit, but because it increased the numbers of degrees of freedom assigned to the model this improvement was not statistically significant in the comparison for the effect of gender (p = 0.071) or for that of disease (p = 0.94). From the first (frequency domain) model the circumferential distribution of cortical widths was predicted (with 95% confidence intervals [CIs]) for each subject group.
For the control subjects, there were no significant differences between the males and females in their ages (p > 0.75). Time since death was also unrelated to the amounts of cortical or cancellous bone (p > 0.21). There was no significant difference in the ages of the female fracture and female control groups (p > 0.21). In the female fracture group, there was no association between time since fracture and the amounts of cortical (p > 0.44) or cancellous bone (p > 0.93).
Total bone area
In the control samples, male subjects had a significantly greater maximum (male 34.32 ± 0.79 mm, female 30.76 ± 0.55; p = 0.0016) and minimum (male 29.66 ± 1.13, female 24.52 ± 0.61; p = 0.0012) cross-section diameter. However, there were no differences in these dimensions between the female control and the female fracture group (fracture: maximum 31.67 ± 1.08, p = 0.503; minimum 25.64 ± 0.72, p = 0.271). The ratio of maximum to minimum diameters was not different between fractures and controls (fractures: 1.25 ± 0.04; female controls 1.26 ± 0.03, male controls 1.16 ± 0.04; p > 0.05 Tukey–Kramer HSD test).
Six of the 13 biopsies from the fracture cases had substantial proportions of the posterior and inferoposterior regions missing, with the other regions being intact. There were no differences in the maximum and minimum diameters between the complete and incomplete biopsies nor in the ratio of these diameters (p > 0.635). The proportion of subcapital (n = 5) and transcervical (n = 1) fractures was not significantly different from that in the group of complete biopsies (p = 0.31, Chi-square analysis).
The total bone area of the whole cross-section in the control group showed that males had significantly more bone than females (males 227.55 ± 14.9 mm2; females 184.8 ± 6.83 mm2; p = 0.015). Although total bone area was lower in samples from femoral neck fractures than female controls, this was not significant (fractures 165. 85 ± 10.08; p = 0.159). However, when the total amount of bone was expressed as a proportion of bone + marrow, it was significantly reduced in the fracture group (Tt.Ar: female fracture 27.83 ± 1.18%, female control 33.62 ± 1.47%; p = 0.0054; male control 30.81 ± 2.23). In the control group, there were no differences in the Tt.Ar (%) between females and males (p = 0.298). There were no differences in either the total bone area or the Tt.Ar (%) between the complete and incomplete biopsies from the fracture group.
Cancellous bone distribution
In the control group, there were no significant differences in the amount of cancellous bone between the females and the males in any of the eight regions (p > 0.241, Fig. 2A). There were no significant differences (p > 0.178) in the amount of cancellous bone in the fracture group compared with the female control group (Fig. 2A). There were no differences between the complete and incomplete biopsies from the fracture group.
Cortical bone distribution
There was a significant difference in the proportion of cortical bone between the female and male control groups in the inferoanterior region (p < 0.05) but not in any other region (Fig. 2B). However, the proportion of cortical bone was significantly reduced in the female fractures compared with the female control group in the inferior, inferoanterior, and anterior regions (p < 0.05; Fig. 2B). When compared with the complete biopsies, the incomplete biopsies in the fracture group had a significantly lower proportion of cortical bone in the inferior and inferoanterior regions (p < 0.05).
Mean cortical width
There were no significant differences (p > 0.05) in mean cortical widths between the male and female control group (Fig. 3). Mean cortical width (Fig. 3) was significantly reduced (p < 0.05) in certain regions of the femoral neck in the female fracture group compared with the female control group (inferoanterior (31.8% reduction): control 2.566 ± 0.16 mm, fracture 1.751 ± 0.143; anterior (30.5%): control 1.327 ± 0.078, fracture 0.923 ± 0.091; Superoposterior (24.9%): control 0.822 ± 0.039, fracture 0.617 ± 0.049). Cortical width was proportionally reduced by generally smaller amounts in the other regions (inferior 21.4%; posterior 18.1%; inferoposterior 21.2%; superior 21.7%; superoanterior 25.5%) but these did not reach significance (p > 0.05). The incomplete biopsies had a significantly lower mean cortical width than the complete biopsies in the inferior (complete 3.167 ± 0.194 mm, incomplete 2.288 ± 0.209; p = 0.010) and inferoanterior (complete 2.006 ± 0.152, incomplete 1.395 ± 0.18; p = 0.027) regions but were no different in the other regions.
Minimum cortical width
This was investigated because of its potential for being associated with stress risers, points of likely failure during mechanical overload. There were no significant differences (p > 0.05) in minimum cortical widths between the male and female control group (Fig. 3). Although the minimum cortical width was lower in all regions of the female osteoporotic group (Fig. 3), this was only significant in the inferoanterior (45.2% lower; control 1.725 ± 0.126 mm, fracture 0.944 ± 0.013; p = 0.0004), anterior (35% lower; control 0.735 ± 0.09, fracture 0.478 ± 0.06; p = 0.012 [Wilcoxon rank sum test]), and the superoposterior (39.7% lower; control 0.469 ± 0.049, fracture 0.283 ± 0.044; p = 0.0103) regions. When complete and incomplete biopsies were compared, the minimum cortical width was lower in the inferior (complete 2.43 ± 0.21, incomplete 1.61 ± 0.21; p = 0.022) and inferoanterior (complete 1.15 ± 0.14, incomplete 0.64 ± 0.17; p = 0.049) regions but was not significantly (p > 0.05) different in the other regions.
Modeling of the cortical widths in the frequency domain
The modeled circumferential distribution of cortical widths was calculated for each biopsy from a model (R2adj = 0.79), which used periodic functions of the measurement angle (sin angle + cos angle, sin 2× angle + cos 2× angle, sin 3× angle + cos 3× angle, sin 4× angle + cos 4× angle, sin 6× angle + cos 6× angle, sin 8× angle + cos 8× angle, sin 9× angle + cos 9 × angle) as independent continuous variables and subject, gender, and disease as independent categorical variables. In the controls, the male subjects had greater cortical widths than the female subjects in the superior region but had generally thinner cortices in the other regions (Fig. 4). Comparison of the female controls to the fracture cases demonstrated that the cases had a similar distribution of cortical width (Fig. 5). However, the fracture cases had generally thinner cortices and this was most marked in the inferoanterior zone. The cortical widths of the incomplete biopsies from the fracture cases showed more exaggerated thinning than the complete biopsies in the inferior and anterior regions. When comparing the proportional loss of cortical bone in the fracture cases, this was most marked (up to 30%) in the inferoanterior and superoposterior regions (Fig. 6). Cortical bone was preserved in the inferior and superior regions. In the incomplete biopsies, the loss in the inferoanterior region was more marked (up to 50%).
The purpose of this study was to examine the whole femoral neck cross-section from intracapsular fracture cases to identify any structural differences in the bone distribution between fracture cases and age-matched controls. Although it is well documented that bone loss occurs in the femoral neck,(8-16) it was not clear whether this is due to generalized bone loss affecting cortical and cancellous bone equally or due to more localized reductions in bone mass. By dividing the femoral neck cross-section into eight regions (as used by Crofts et al.(19) and previously by Jowsey et al.(20) we were able to identify specific areas where cortical bone was deficient in comparison with controls.
Initially, the effects of the modifications to the image analysis package on reproducibility were determined. Evaluation of these changes has established that this program will provide reliable estimates of the regional amounts of cortical and cancellous bone within a femoral neck cross-section.
A potential confounding factor in this study is possible variations in the location along the femoral neck at which the biopsies are taken. We, therefore, measured the minimum (anterior-posterior) and maximum (inferior-superior) diameters of the whole femoral neck cross-section. These and their ratios (to take account of genetic differences) were close to those reported by Backman(21) for the middle of the femoral neck as compared with lateral (trochanter) or medial (neck) sites (middle: maximum 36.9 mm, minimun 32 mm, ratio 1.15; lateral: maximum 45 mm, minimum 27.2 mm, ratio 1.65; medial: maximum 49.1 mm, minimum 48.6 mm, ratio 1.01). Furthermore, they were not significantly different between the fractures and the female controls although the male controls were of course larger. Similarly, the mean or minimum cortical widths for each octant were not related to the ratio of maximum/minimum diameters of the whole femoral neck cross-section (p > 0.078 and p > 0.118, respectively).
As expected, the cross-sections of the femoral neck from male control subjects had significantly greater minimum and maximum diameters compared with the female subjects reflecting bone size differences between the genders. Although the total bone area was higher in males than females when this was expressed as a proportion of bone + marrow area, there were no differences. Similarly, the proportion of the total area that was occupied by cancellous or cortical bone and the mean regional cortical width showed no significant gender difference. Modeling of the cortical widths suggested that males had a generally thinner cortex except in the superior region where the cortices were thicker. Structural integrity is dependent not only on bone mass but also on its circumferential distribution: the same amount of bone distributed at a greater distance from the center of area will increase stiffness.(22) Thus, the tendency of males to have a larger neck diameter, which might increase with age at a greater rate than in females,(23) may confer some additional protection against hip fracture despite the thinner cortical shell.
The major differences between the fracture group and control material were the overall and localized reductions in cortical bone. In the fracture cases, the most striking abnormality was the thinning of the inferoanterior and superoposterior cortices. The reason for this selective thinning of these cortices is unclear. It may be due to changes in patterns of exercise leading to an underloading of these regions. Certain types of exercise involving hip extension and adduction (e.g., resistance training, weight lifting, rowing) in postmenopausal women have been shown to increase bone mineral density in the Ward's triangle region(24,25) without overall increases in femoral neck bone mineral density. The Ward's region consists predominantly of anterior and posterior cortical bone. Loading of the anterior cortex is suggested, by anatomical considerations, to be dependent on a major extent on the psoas major and iliacus muscles which insert into the lower trochanter and flex the hip. Loading will also occur in more rapid forms of locomotion due to torsion as the hip is flexed and extended.
Using finite element analysis, it has been reported that the reduction in bone density due to osteoporosis causes the peak stresses to be increased by almost 50% in the subcapital region.(26) However, this model was based on cadaveric (nonfractured) femoral necks and used a standardized but general bone loss to simulate the osteoporotic state. The stress patterns in the femoral neck are similar for both cortical and cancellous bone with peak stresses occurring in the inferior neck during gait.(26) Finite element analysis of the effects of standing, heel strike and toe-off showed that the femoral neck was deformed in a superior to inferior direction. A fall, however, resulted in a deformation in an inferoanterior to superoposterior direction. Clearly our data suggest that while the preferential preservation of the inferior region of the femoral neck cortex may protect against fracture during normal gait, the marked losses in the inferoanterior and the superoposterior regions will increase the propensity to fracture following a fall on to the greater trochanter. This weakening of the cortical shell coupled to the high impact stresses following a fall may enhance the likelihood of fracture initiation and propagation in this region.
There are a number of limitations to this study. First, the small number of cases and controls means that the power of the study to demonstrate statistically significant differences is relatively low. With a larger data set it may be possible to identify further changes in the cortical width or cancellous bone area. Similarly, the small numbers of transcervical fractures meant that it was not possible to determine differences between these and the subcapital fractures. Second, due to the nature of this material selection bias may have occurred. Collection of biopsies from the fracture cases was not sequential because not all suitable cases gave or were able to give informed consent. With regard to the control material, biopsies could only be obtained from those subjects who had died in hospital and whose relatives gave consent to a postmortem. Studies are in progress to address these limitations. Third, approximately half of the fracture cases were incomplete. This may be due to fracture propagation rather than clumsy handling of the biopsy during hemiarthroplasty. Comparison of the complete and incomplete biopsies suggests that the incomplete samples had thinner cortical shells and therefore increased fragility. There was no evidence from the shape of the biopsies and the ratio of maximum/minimum diameters to suggest that the incomplete samples were from a more medial position along the neck and the proportion of subcapital fractures in this group was not significantly different from that in the group of complete biopsies. Fourth, the age range of the subjects was limited so that any effects of age in this small data set on cortical and cancellous bone parameters could not be identified. Last, although previous studies have noted that osteopenia of the cancellous bone may be a contributory factor to fracture risk,(10,15) this was not apparent in our study. This is not surprising since Eventov et al.(10) commented that a third of the femoral necks examined had no evidence of osteopenia and our sample size was relatively small. Using computer tomography, Høiseth et al.(27) suggested that in the femoral neck there is substantially less age-dependent loss of cancellous than cortical bone and that loss of cortical bone mass is the dominating factor. A lack of any significant difference in the cancellous osteopenia between the fracture and nonfracture cases may not, however, negate the influence of the amount of cancellous bone on fracture risk. Loss of cancellous bone may have been dominant immediately following the menopause. Mori et al.(28) have also demonstrated that the trabecular bone area in the femoral head of fracture cases was not different from nonfracture cases but was significantly reduced compared with femoral heads from younger women.
In conclusion, this is the first study to investigate regional changes in both cortical and cancellous bone in cross-sections of the femoral neck cases of fracture in comparison with control material. In intracapsular fracture of the femoral neck, loss of cortical rather than cancellous bone is the predominant feature. Although there is generalized thinning of the cortex, this is lowest in those regions that have been predicted to bear the greatest stresses during normal gait.(26) Conversely, those regions of the cortex that are maximally loaded on impact from a fall showed the greatest proportional loss of cortical thickness. Understanding the physiological (e.g., the type of physical activity) and cellular causes (e.g., osteocyte death) of this loss will enable the development of preventative strategies.
This work was supported by an MRC Program Grant 9321536. We thank Prof. A. K. Dixon, Department of Radiology, University of Cambridge for the categorization of fracture types from the preoperative anteroposterior and lateral X-rays; Mr. N. Rushton, Orthopaedic Research Unit, and Mr. F. Norman-Taylor and Mr. M. Bowditch, Department of Orthopaedics, Addenbrooke's Hospital, for the supply of fracture biopsies; and Shaun Mosley, Rachel Graves, Nick Stevenson, and Dr. D.G.D. Wight, Department of Histopathology, Addenbrooke's Hospital for the supply of the postmortem samples.