Osteoporosis International

, Volume 22, Issue 5, pp 1377–1388

Ethnic differences in femur geometry in the women's health initiative observational study


    • Wayne State University School of Medicine
  • T. J. Beck
    • Johns Hopkins University School of Medicine
  • G. Wu
    • Mel and Enid Zuckerman College of Public HealthUniversity of Arizona
  • C. E. Lewis
    • University of Alabama Birmingham
  • T. Bassford
    • Mel and Enid Zuckerman College of Public HealthUniversity of Arizona
  • J. A. Cauley
    • Brigham and Womens Hospital, Harvard Medical School
  • M. S. LeBoff
    • University of Pittsburgh Graduate School of Public Health
  • S. B. Going
    • Mel and Enid Zuckerman College of Public HealthUniversity of Arizona
  • Z. Chen
    • Mel and Enid Zuckerman College of Public HealthUniversity of Arizona
Original Article

DOI: 10.1007/s00198-010-1349-4

Cite this article as:
Nelson, D.A., Beck, T.J., Wu, G. et al. Osteoporos Int (2011) 22: 1377. doi:10.1007/s00198-010-1349-4



Participants in the observational study of the Women's Health Initiative (WHI) were studied to determine if ethnic differences in femur geometry can help to explain differences in hip fracture rates. Structural differences in femurs of African and Mexican-American women appear to be consistent with lower rates of hip fractures vs. whites.


Ethnic origin has a major influence on hip fractures, but the underlying etiology is unknown. We evaluated ethnic differences in hip fracture rates among 159,579 postmenopausal participants in the WHI then compared femur bone mineral density (BMD) and geometry among a subset with dual X-ray absorptiometry (DXA) scans of the hip and total body.


The subset included 8,206 non-Hispanic whites, 1,476 African-American (AA), 704 Mexican-American (MA), and 130 Native Americans (NA). Femur geometry derived from hip DXA using hip-structure analysis (HSA) in whites was compared to minority groups after adjustment for age, height, weight, percent lean mass, neck-shaft angle and neck length, hormone use, chronic disease (e.g., diabetes, rheumatoid arthritis, cancer), bone active medications (e.g., corticosteroids, osteoporosis therapies), and clinical center.


Both AA and MA women suffered hip fractures at half the rate of whites while NA appeared to be similar to whites. The structural advantage among AA appears to be due to a slightly narrower femur that requires more bone tissue to achieve similar or lower section moduli (SM) vs. whites. This also underlies their higher BMD (reduces region area) and lower buckling ratios (buckling susceptibility). Both MA and NA women had similar advantages vs. whites at the intertrochanter region where cross-sectional area and SM were higher but with no differences at the neck. NA and MA had smaller bending moments vs. whites acting in a fall on the hip (not significant in small NA sample). Buckling ratios of MA did not differ from whites at any region although NA had 4% lower values at the IT region.


Differences in the geometry at the proximal femur are consistent with the lower hip fracture rates among AA and MA women compared to whites.


African AmericanCross-sectional geometryEthnic differencesHip structure analysisLocal bucklingMexican AmericanNative AmericanProximal femur


Ethnic origin is one of the most important factors governing osteoporotic fracture risk. A recent study by Cauley et al. examined ethnic differences in fracture rates and in risk factors among 159,579 participants in the multi-ethnic Women's Health Initiative (WHI) study [1]. In relative terms, overall fracture rates were similar in non-Hispanic white and Native American (NA) women but occurred at roughly half of those rates among African-American (AA) and Mexican-American (MA) women, respectively. While that paper did not look specifically at hip fractures, it is conceivable that femurs of women in the latter two groups may have a structural advantage that makes them less likely to suffer fragility fractures. However, a confounding observation is that obesity rates tend to be higher in ethnic minorities; a meta-analysis by De Laet showed that higher body mass index (BMI) in kilogram per square meter moderates hip fracture rates [2]. Recently, we showed that controlling for body size can permit one to distinguish the effects of obesity on the femur from those that might confer a size-independent mechanical advantage [3].

Although other factors play roles in fracture incidence, the fundamental clinical problem is that osteoporosis reduces the mechanical strength of the whole bone under trauma conditions. Bones fracture when applied forces generate internal stresses exceeding the capacity of bone tissue. Conceivably an ethnic group may have a strength advantage if their bone geometry generates smaller stresses under similar loads or if their bone tissue has a higher stress capacity i.e., greater tissue material strength. Evidence for an ethnic difference in bone tissue strength is lacking, suggesting that any strength advantage must be geometric. Ethnic differences in BMD suggest strength advantages but provide no information on what they may be since a given BMD can occur in bones of very different geometries. Methods like hip structure analysis (HSA) that derive geometry from dual X-ray absorptiometry (DXA) scan images can provide some insights into mechanical advantages that may underlie a BMD difference.

A geometric advantage that reduces hip fracture incidence should reduce stresses under a typical fall condition. As depicted in Fig. 1, a fall to the side that impacts on the greater trochanter generates a combination of bending and axial compression. Stresses are only depicted at the femoral neck although those at the intertrochanter would be similar (with a greater bending component). Stresses should generally be greatest in compression due to the additive effects of axial and bending loads, and should be highest on the lateral surfaces of cross-sections. This is especially true in the osteoporotic femur [4]. Better resistance of axial compressive stresses would result from greater bone surface [cross-sectional area (CSA)] in cross-sections. A shorter femoral neck or a smaller neck-shaft angle would reduce the bending moment on proximal cross-sections while a greater more peripherally distributed surface [larger section moduli (SM)] in the cross-section would better resist bending stresses. These geometric parameters are clearly greater in stronger bones [5, 6] and also in elderly subjects who do not suffer hip fractures [7, 8]. Interestingly however, CSA and SM do not discriminate fracture cases from controls as well as conventional BMD. The key to this conundrum is that the lower BMD in proximal femurs of fracture cases is due to, not only, less bone mass, but also to a larger outer diameter [7, 8]. Mechanically, a larger diameter tube should be stronger in bending (improves SM) unless the tube walls are locally so thin that they buckle under compressive loads. Susceptibility to local buckling in a hollow tube increases with the ratio of outer radius to wall thickness (buckling ratio). In several studies the only geometric parameter that appears to predict hip fracture as well as BMD is an estimate of buckling ratio (see Methods) [79]. If osteoporotic femurs do fail during falls in local buckling, fracture should begin on the superior-lateral compressive surface rather than on the inferior medial tensile surface where failure should otherwise initiate. Recently, de Bakker and colleagues used a high-speed video to show that nine of 12 elderly cadaver femurs begin to fail on the superior lateral surface of the femoral neck when loaded in an experimental simulation of a fall. [10] The relevance to the present study is that that men and women of African ancestry appear to have narrower femurs and may be less susceptible to local buckling. Mechanically, a narrower femur would require more bone tissue to achieve equivalent strength in bending thus would have higher BMD and a lower buckling ratio. An earlier small study conducted at an ancillary clinical center of the WHI, compared femur geometries between AA and white postmenopausal women from the Detroit area [11]. That study and a follow-up comparison with black and white women from Johannesburg, South Africa [12] showed smaller diameters underlying the higher BMD in groups of African origin compared to whites of both countries. Studies of femur geometry in other ethnic groups is currently lacking in women, but a recent cross-sectional comparison by Travison et al.[13] evaluated men of white, black and Hispanic origin in the Boston area. These investigators showed that black men had narrower femurs than whites of similar body size and that aging patters in femur geometry may be generally less advantageous in Hispanic men than in other groups. To date, a study of ethnic differences in femur geometry that is similar in size to the Boston study has not been conducted in women.
Fig. 1

Incident hip fracture rates in cases per 1,000 person-years among a total of 159,579 non-Hispanic white, African-American, Mexican-American, and Native American women. Error bars are 95% confidence intervals. Comparisons were adjusted for age, BMI, hormone use and a diagnosis of diabetes

The present study draws from WHI centers that had acquired DXA data to compare femur geometry of postmenopausal non-Hispanic white women to relatively large groups of AA and Mexican-American (MA) women as well as to a modest sample of NA. Our goal in the present study was to evaluate structural differences in the proximal femur that might help to explain ethnic differences in hip fracture rates. Because the analysis of ethnic differences in WHI fracture rates by Cauley and colleagues did not isolate hip fractures or account for differences in obesity rates [1] we re-analyzed that previously published data to provide context for the subset of women who had DXA scans of the hip.

Subjects and methods

WHI, the largest study of women's health in the United States, comprises an observational study, two hormone trials (estrogen alone or estrogen plus progestin), a trial of calcium and vitamin D supplements, and a dietary modification trial. Details regarding the inclusion and exclusion criteria, participants' characteristics, clinical trial design, randomization, blinding, and follow-up can be found in previously published papers [14, 15]. In brief, postmenopausal women who were between the ages of 50 and 79 years, and not likely to change residence or die within 3 years at the time of enrollment, were recruited from 40 clinical centers in the U.S. Three of the centers (University of Arizona at Tucson and its satellite center in Phoenix; University of Pittsburgh; and University of Alabama at Birmingham) were also designated for the acquisition of bone mass and body composition data using DXA. The study protocol and consent forms were approved by the institutional review boards for all participating institutions.

For the present study, we analyzed the baseline data from those clinical trial (CT) and observational study (OS) participants who had hip and total body DXA scans at one of the three WHI DXA locations. A total of 10,563 women had usable results from the HSA at baseline, including: white (8,156 or 77.2%); AA (1,466 or 13.9%); Mexican-American (MA, 702 or 6.6%); NA (124 or 1.2%); and Asian/Pacific Islander (36 or 0.3%). There are also 76 participants with unknown ethnic affiliations (0.7%). We did not use data for the present study from the Asian/Pacific Islander or “unknown” ethnic groups because of the small sample sizes leaving 10,448 participants.

Data collection for other covariates

Demographic, anthropometric, and medical history data were collected from WHI participants using methods described previously [1618]. Questionnaires were used at baseline to collect information on age, years since menopause, race/ethnicity, smoking, recreational physical activities, hormone therapy use, physical function, disease history, and alcohol use. A medication inventory was used to assess current medication intake, including dietary supplements. A modified block food frequency questionnaire [19] was used to determine intake of certain nutrients (calcium, Vitamin D) as well as Kcal intake (energy). Calcium and vitamin levels were the total of dietary plus supplemental intakes. Recreational physical activity was assessed by questions on the frequency and duration of several types of recreational activity, and metabolic equivalent task (MET) scores (defined as the ratio of work metabolic rate to a standard resting metabolic rate, with one MET roughly equivalent to the resting metabolism while sitting quietly) were computed as the product of days per week, minutes per day, and MET value for each activity group. The methodology was based on the method of Ainsworth et al.[20] as modified by McTiernan and colleagues[21]. Physical function was measured using the 10-item medical outcomes study scale [22] for which a higher score indicates better physical function. Weight was measured to the nearest 0.1 kg on a balance beam scale with the participant dressed in indoor clothing without shoes. Height was measured to the nearest 0.1 cm using a wall-mounted stadiometer. BMI was calculated as weight in kilogram per height in square meter. Reported body composition values were measured by DXA in the form of total body lean and fat masses which by this method exclude bone mineral mass.

Fracture data were collected prospectively over the period of study and previously reported for all fractures in the entire WHI cohort [1]. The hip fracture data are presented here for each ethnic group from the whole study as an estimate of fracture risk for the participants in the bone sub-study. We have added additional adjustments to the crude fracture incidence for covariates that might identify factors that would not be bone related.

DXA measurements

For the WHI, measurements were obtained from DXA scans of the total body and proximal femur using Hologic DXA instruments (QDR2000, QDR2000 + or QDR4500W, Hologic, Inc. Waltham, MA). Bone densitometry technicians at each site were trained and certified by Hologic, Inc. and by the WHI DXA coordination center at the University of California at San Francisco to use standard protocols for positioning and analysis. The quality assurance program utilized spine and hip phantom scans, and involved the review of a random sample of all scans. Hardware and software changes were tracked with in vitro and in vivo cross-calibrations and by scans of calibration phantoms across instruments and clinical sites. A separate cross-calibration was conducted on all the WHI DXA instruments using a special phantom designed for the HSA method. For the present study, conventional DXA data for both the proximal femur and total body from baseline study visits at the four DXA scanner locations were used. Of the conventional measurements, we report BMD, BMC and region of interest area in the femoral neck (FN) from hip scans, and from total body scans we include total body lean and fat masses in absolute terms and as a percentage of total body mass.

Hip structure analysis

Scan files were analyzed by HSA software developed at the Johns Hopkins School of Medicine [16]. All subject identifiers were automatically removed at the coordination site at the University of Arizona prior to analysis by HSA and were only restored after all analysis was complete. The HSA method uses the principle that a line of pixels across the bone axis is a projection of the mineral in a cross-section from which certain geometric properties can be measured [6]. Geometry is calculated from five profiles spaced one pixel apart and then averaged at each region. The narrow-neck region traverses the neck at its narrowest point and the intertrochanter is centered across the bisector of the angle between the neck and shaft axes. The shaft region was not used in the present study as it is a rare site of hip fractures. Average pixel value in region profiles is reported as BMD, and outer diameter is measured from the outer profile margins corrected for image blur. To determine the bone surface in the cross-section, pixel values are divided by the average mineral density of normal adult cortical bone (1.053 g/cm3), yielding a linear thickness; the profile integral is thus the area (CSA) in square centimeter. The center of mass (COM) of the profile is determined then the cross-sectional moment of inertia (CSMI) is measured as the integral weighted by the square of the distance of each pixel from the COM. Maximum bending stress in a cross-section is a function of SM, computed as CSMI divided by the maximum distance to the medial or lateral profile margin (dmax). Research has suggested that homeostatic mechanisms tend to preserve SM in aging bone cross-sections [23, 24] but in a progressively thinner cortical shell. If cortices become thin enough to buckle (fold) under compressive loads the SM will overestimate the actual strength. This complex phenomenon cannot be fully characterized by the limited information in a DXA scan but an estimate of the buckling ratio (BR) can suggest that a cross-section may be susceptible. The BR is used in engineering designs incorporating hollow tubes; ideally the ratio of outer radius to wall thickness should be kept below ~10 to avoid strength loss due to local buckling [25]. BR is estimated in HSA by modeling the cross-section as a hollow circular (NN) or elliptical (IT) annulus with a fixed proportion (60%, 100% and 70%, respectively) of the CSA in the cortical shell. Though undeniably crude, this estimate appears to provide the mechanical explanation for why conventional BMD predicts hip fractures [8].

The HSA program also measures the femoral neck length (NL) as the distance from the center of the femoral head to the intersection of the neck and shaft axes as well as the angle between them. These parameters influence the moment arm of forces causing bending of the proximal femur.

Statistical analyses

Descriptive analyses were used either to calculate mean, SD or 95% confidence interval (continuous variables), or frequencies for each key variable and the covariates. In order to test for significant differences in cross-sectional geometric variables among the ethnic groups, we utilized multiple linear regression analysis and tested for differences in mean levels of the variables, adjusting for covariates overall. We also compared each minority ethnic group to whites by calculating the percentage difference from the mean for whites for each geometric variable. Ethnic group was included in regression models as the independent variable of primary interest. Potential confounding factors in the multivariable linear regression models included age, weight, height, percent lean mass (from whole body DXA), neck-shaft angle and neck length (from HSA); history of hormone use, history of chronic disease (e.g., diabetes, rheumatoid arthritis, cancer survivors), medications that affect bone and mineral metabolism (e.g., corticosteroids, osteoporosis therapies), and clinical center were examined and controlled for in the regression analyses. Neck-shaft angles are not body-size dependent and were not adjusted. Neck-length is body (bone) size dependent but should not be influenced by other (modifiable) parameters, thus was adjusted for height and weight alone.

Differences in fracture incidence rates between ethnic groups in the whole WHI-OS cohort were tested using a proportional hazards model and adjusted for age, BMI, diagnosis of diabetes, and hormone therapy use.


Demographic characteristics of the four ethnic groups in the total population are listed in Table 1. Minorities were somewhat younger, AA women were taller, while MA and NA women were shorter on average than white women. Body weight has a major influence on the variability in femur geometry and to a lesser but still significant effect on BMD. Both AA and NA women had larger body masses, while all three minorities had significantly less lean mass and greater body fat as a percentage of total mass compared to whites. Consistent with their body composition, minority women tended to be less active than white counterparts with lower MET scores, less time spent walking for exercise, and fewer had physical function scores above 90 (not significant in NA). Total vitamin D intakes were higher in whites than in minorities in total population. Total calcium intake was significantly smaller than whites in AA women and in NA women, but not different in MA women. Minority women also were less likely than whites to have used hormone therapy and more likely to be diagnosed as diabetic.
Table 1

Mean ± SD or N (%) characteristics of total study population by ethnicity


Total Population

Non-Hispanic white



Native American







63.9 ± 7.3

61.9 ± 7.4*

60.7 ± 7.3*

60.1 ± 7.5*

Height (cm)

161.8 ± 6.3

162.7 ± 6.0*

157.9 ± 5.8*

159.8 ± 7.7*

Weight (kg)

72.3 ± 15.4

83.5 ± 18.5*

72.6 ± 15.4

79.1 ± 17.9*

BMI (kg/m2)

27.6 ± 5.5

31.4 ± 6.5*

29.0 ± 5.8*

30.6 ± 6.6*

TB lean mass (kg)

37.3 ± 5

41.4 ± 6.1*

36.3 ± 5.1*

37.5 ± 5.0

TB fat mass (kg)

31.6 ± 11

38.2 ± 12.8*

32.6 ± 10.4*

36.7 ± 11.9*

Percent total body lean

53.6 ± 7

51.6 ± 6.7*

52.1 ± 6.4*

50.2 ± 6.7*

Percent total body fat

43.5 ± 7.3

45.6 ± 6.9*

45.0 ± 6.6*

47.1 ± 6.9*

Weekly energy exp. in physical activity (METS)

12.4 ± 14.2

7.6 ± 11.4*

9.6 ± 12.4*

10.5 ± 16.8

Minutes per week spent walking






3,278 (44.9%)

877 (63.2%)

369 (59.8%)

62 (50.8%)

One to 150

3,021 (41.4%)

353 (25.4%)

179 (29%)

39 (32%)


1,004 (13.8%)

158 (11.4%)

69 (11.2%)

21 (17.2%)

Physical functioning construct > 90

2,713 (34%)

337 (23%)*

207 (33%)

22 (18%)*

Total Ca intake (mg)

1,136 ± 693

684 ± 488*

1,093 ± 690

952 ± 631**

Total vitamin D intake (mcg)

9.7 ± 6.9

6.7 ± 5.6*

7.1 ± 5.7*

6.9 ± 7.7*

Hormone use





Never used

3699 (45.1%)

845 (57.3%)

390 (55.6%)

85 (65.4%)

Past user

1,338 (16.3%)

212 (14.4%)

82 (11.7%)

16 (12.3%)

Current user

3,167 (38.6%)

418 (28.3%)

230 (32.8%)

29 (22.3%)

Diabetes (%)

431 (5%)

246 (17%)*

79 (11%)*

30 (24%)*

*p < 0.001, **p < 0.05, vs. Non-Hispanic white

Femur BMD and geometry for the four ethnic groups (Mean, SE) are listed in Table 2 after adjustment for age, height, weight, percent lean mass, femoral neck length (NL), neck shaft angle (NSA), hormone use and a diagnosis of diabetes.
Table 2

Adjusted mean (standard error) of BMD and hip geometry measurements by ethnicitya


Non-Hispanic white



Native American






Conventional femoral neck

 BMD (g/cm2)

0.712 (0.001)

0.792 (0.003)*

0.722 (0.004)**

0.742 (0.009)***

 BMC (gram)

3.53 (0.006)

3.82 (0.014)*

3.56 (0.019)

3.66 (0.046)***

 Region area (cm)

4.96 (0.003)

4.83 (0.008)*

4.93 (0.012)**

4.93 (0.028)

Hip structure Analysis

 Neck-shaft angle

130.7 (0.055)

130.1 (0.13)*

131.1 (0.188)

131.1 (0.437)

 Neck length (cm)

4.69 (0.01)

4.74 (0.01)*

4.64 (0.02)**

4.62 (0.05)

Narrow neck


 BMD (g/cm2)

0.711 (0.001)


0.711 (0.004)

0.730 (0.01)

 Cross-section area (cm2)

2.04 (0.003)

2.14 (0.008)*

2.04 (0.011)

2.08 (0.027)

 Outer diameter (cm)

3.02 (0.002)

2.98 (0.005)*

3.02 (0.008)

3.01 (0.018)

 Section modulus (cm3)

0.92 (0.002)

0.94 (0.004)*

0.91 (0.006)

0.92 (0.015)

 Buckling ratio

12.6 (0.03)

11.7 (0.17)*

12.5 (0.1)

12.2 (0.2)



 BMD (g/cm2)

0.715 (0.001)


0.726 (0.004)**

0.747 (0.01)***

 Cross sect. area (cm2)

3.47 (0.006)

3.53 (0.014)*

3.60 (0.02)*

3.64 (0.048)***

 Outer diameter (cm)

5.12 (0.003)

4.97 (0.008)*

5.22 (0.011)*

5.14 (0.027)

 Section modulus (cm3)

2.90 (0.006)

2.78 (0.013)*

3.00 (0.019)*

3.05 (0.045)***

 Buckling ratio

10.4 (0.03)

9.80 (0.1)*

10.5 (0.1)

10.0 (0.2)**

*p < 0.001, **p < 0.05, ***p < 0.01, vs. Non-Hispanic white.

aResults are adjusted for age, height, weight, percent total body lean mass, neck-shaft angle, neck length, hormone use, and diabetes. Neck-shaft angles are unadjusted; neck lengths are adjusted only for height and weight

Femur differences: Whites vs. African Americans

Conventional femoral neck BMD was 13% higher in AA women vs. whites and was mostly but not completely due to a greater BMC; 3% of the BMD difference was due to a smaller femoral neck region. Since region length is normally fixed by DXA software, the smaller region reflects a narrower femoral neck within the region. The narrower bone effect is more directly evident in the femur outer diameter (OD) by the HSA method. Regions were 1% narrower at the NN and 3% narrower at the IT region in AA women. AA differences in HSA BMD in the NN were due to both less bone (CSA) and to a narrower region, but at the IT region, the larger BMD was mostly due to a narrower bone. From a mechanical standpoint AA women had higher CSA than whites at both the NN and IT regions. SM was significantly larger at the NN region but significantly smaller than whites at the IT region. Buckling ratios were significantly lower than whites at both regions. AA women showed slightly longer NL and slightly smaller NSA. These latter differences have opposing effects on the bending moment in a fall and may yield no net effect on strength.

Femur differences: whites vs. Mexican-Americans

Conventional femoral neck BMD was significantly (1%) higher than whites; but this was due to a smaller region area not a larger BMC. No differences were apparent at the HSA NN region; however at the IT region, MA had 2% higher BMD due to a 4% greater CSA opposed by a 2% wider OD (These parameters have opposing effects on BMD, as do BMC and region area respectively on conventional BMD). Section moduli at the IT region were significantly higher than whites but MA women showed no significant BR differences from whites in either region. The shorter femoral neck length in MA women should improve strength by reducing the neck bending moment in a fall on the hip.

Femur differences: whites vs. Native Americans

NA women had significantly higher BMD and BMC than their white counterparts. There were no differences in BMD or geometry at the NN region but at the IT region, BMD, CSA, and SM were significantly greater and the BR was significantly smaller than in whites. Like in MA women, NL was shorter than in whites but differences did not reach significance.

Hip fracture incidence by ethnicity

The annualized rate of hip fracture per 1,000 person-years by ethnic group is plotted in Fig. 1 and is based on data from all women participating in WHI. This data was previously published in reference one although not restricted to the hip or adjusted for BMI. The general pattern is, however, quite consistent with the previous report showing significantly lower fracture rates among AA and MA women vs. whites. It is also consistent with the fracture pattern in the subsample with DXA measurements used in the present study although MA fracture rates did not reach significance vs. whites (not shown). Even in the entire WHI study, the numbers of NA subjects were relatively small and only eight incident hip fractures were recorded. While far from conclusive, it would appear that hip fracture rates among NAs are similar to whites. There were no significant differences in hip fracture location (cervical vs. trochanter) across ethnic groups (not shown).


Among participants in the WHI observational study, adjusted hip fracture rates among white women were nearly three times that of AA or MA women, while NA women appear to fracture at similar rates. The fracture incidence pattern among the subset of women with DXA measurements used in the present study was entirely consistent with that in the larger study. The incidence data did not show ethnic differences in distribution of cervical vs. trochanteric hip fractures, although we looked for structural differences at both corresponding femur regions. These observations would lead one to expect mechanical advantages at both neck and intertrochanter regions among both AA and MA women but no such advantage vs. whites among NA women. Our results are generally consistent with those expectations except that NA women appeared to have a mechanical advantage similar to that of MA women. The significant geometric differences from whites are summarized for the three minority groups in Fig. 2.
Fig. 2

Graphic of proximal femur showing definitions for neck-length, neck-shaft angle and region locations for conventional femoral neck BMD and the narrow-neck and intertrochanter BMD and geometry analyzed by HSA. Tables show adjusted mean percent differences from whites for the three ethnic groups; only statistically significant (P < 0.05) differences are shown

African American vs. white women

At present there is little doubt that AA women suffer fewer osteoporotic fractures than white women and while they are more likely to be obese, the underlying explanations for their fracture advantage are not as clear as one might think. For example, AA women in this study were significantly less active, had lower physical function status and lower intakes of Ca and vitamin D (Table 1). They were also less likely to use hormone therapy and more likely to be diagnosed with diabetes than their white counterparts. These factors should point toward increased rather than reduced fracture rates as would their slightly taller stature [26]. Consistent with prior reports [27, 28], AA women had significantly higher BMD in the conventional FN region and in the NN and IT regions analyzed by HSA, independent of body size and other confounders, but as previously noted, BMD is not in itself a strength property and a closer look at the components of BMD and the underlying femur structure suggests that their main advantage over whites is a proximal femur that is geometrically more stabile and less susceptible to local buckling.

In both femur regions, the greater BMD in AA women actually overestimates their advantage in terms of the amount of bone (BMC or CSA) because of the inverse relationship between BMD and region area (i.e., BMD = BMC/region area). Region sizes are effectively smaller among AA women because their femoral neck and intertrochanter regions are significantly narrower. This was previously observed at the narrow neck by Nelson et al. [11, 12] in US and South African women of African descent and in AA men by Travison and colleagues at both the narrow-neck and intertrochanter regions [5]. At the narrow neck region, SM and CSA were significantly greater than whites at the narrow neck by 5% and 3% respectively but their longer femoral neck counters these positive effects on strength in a fall. At the intertrochanter, the SM were actually smaller than in whites although CSA was 1% larger. Again this is generally consistent with findings by Travison in AA men over 60, where SM were similar to whites at the narrow neck and non-significantly smaller at the intertrochanter [29].

Taken together, these differences do not seem adequate to explain a nearly threefold reduction in hip fracture rates. The key to this conundrum may be that the relatively narrower diameter of the proximal end of the AA femur retains its structural stability with age better than that of whites. The argument is as follows: Mayhew [24] and colleagues evaluated the apparent structural changes accompanying the aging process in an Australian sample of Caucasian cadaveric femurs spanning adulthood in both males and females. They concluded that age-related periosteal apposition and changes in the distribution of bone tissue within the femoral neck tend to preserve its bending strength (SM) especially with respect to mechanical loads produced by normal ambulation, though not necessarily for abnormal loads produced by a fall on the hip. In upright stance, forces generated from the body weight and muscle action on the infero-medial neck are in the same (compressive) direction and their sum is typically several times the body weight [30], but on the opposite (superior-lateral) cortex these forces are in opposition and that cortex remains relatively unloaded. Not surprisingly Mayhew observed preservation of the highly stressed inferior medial cortex with aging in both genders but considerable thinning of the minimally stressed superior-lateral cortices [24]. The thinned cortices would bear high-compressive stresses [4] in a fall and may not be able withstand local buckling.

Mexican American vs. white women

Figure 1 shows that like AA women, MA women suffer hip fractures at nearly one third the rate of white women, but the underlying femur geometry varied from whites in different ways. In the present study MA women showed 1% significantly greater FN BMD, but this was due to a slightly smaller region, not to a greater BMC. This effect was not evident at the HSA narrow-neck cross-section nor were there any other geometric differences. However, MA femoral neck lengths were significantly shorter than those of white women and this would improve strength by reducing the bending stress on the neck in a fall on the hip even if the CSA and SM were similar. The shorter NL would also reduce the bending moment at the IT region, where CSA and SM values were 4% and 3% greater, respectively than in whites. The fact that MA women were significantly shorter in stature should also reduce fracture risk [26] presumably by diminishing impact forces in a fall from height. Interestingly, and unlike AA women, there were no apparent advantages of MA women over whites with regard to BR. Similar studies of femur geometry in MA women could not be found in the literature but the study by Travison et al. included a cohort of 394 Hispanic men. The latter group was significantly lighter in weight, shorter in stature, and younger than whites. After correcting for these factors and physical activity, the authors found that BMD at the NN and IT regions was 6% and 8% higher, respectively in Hispanic men. The higher BMD at the NN region was due to a combination of a 3% narrower OD containing 3% greater CSA, but since OD was similar at the IT region the BMD difference was due to a 7% greater CSA. Section modulus did not differ from whites at either region although unlike our study BR values were 9–11% smaller. In Hispanic men, it appears that any mechanical advantage over whites is due to higher CSA and lower BR values, whereas in MA women the differences are largely at the IT region and in the form of greater SM and CSA and smaller stature and fall-mode bending moments (not reported by Travison et al.)[13].

Native American vs. white women

Figure 1, based on a small number of fracture cases in an NA population drawn only from the southwest that is unlikely to represent that of the greater US, suggests that hip fractures may occur at similar rates as in whites. Reports of conventional BMD or geometry among NA women are also scarce, hence the results of this sample may be of some value even if numbers are few (N = 130). In this sample, there was a small BMD advantage at the femoral neck in NA women due to a greater BMC compared to whites. Like MA women, there were no significant differences from whites in the narrow-neck region, but there were differences at the intertrochanter. At that region NA women had 4–5% larger BMD, CSA and SM than whites, while buckling ratios were 4% smaller. Like in MA women, femoral neck lengths were shorter but differences did not reach significance. Slightly smaller stature should confer some mechanical advantage in reducing impact stress in falls on the hip. Note that femurs of NA women differed from whites in a generally similar manner as in MA women. Little is known about osteoporotic fragility and its factors among NA women, but these data should provide some initial insights.


The cohort of women who had DXA scans among the WHI is not considered to be representative of the WHI as a whole or of the US population. Caution is especially warranted with regard to the small sample of NAs drawn mainly from the US South-West in a highly heterogeneous subgroup. These sampling issues and their implications have been discussed previously by Cauley et al. [1]. With respect to measurement issues, the DXA-based methods evaluate geometric differences that influence the stresses generated by loads but not the capacity of the tissue to resist them. Although unlikely, it is possible that bone tissue strength differences may play some role in ethnic differences in fracture rates, but this has not been studied. The HSA method used here for measuring femur geometry from DXA scans has a number of other limitations that have been previously explored [31]. For one, measurement precision is somewhat worse than that of BMD on the same data because geometry is less tolerant of positioning error [32]. The net effect is that larger sample sizes are required to show an effect with HSA geometry than with conventional BMD. The present study shows that there are important geometric differences in the femur associated with ethnicity, but the method measures dimensions from the two-dimensional projected image of the femur as it is positioned for hip scans. Local buckling is a complex phenomenon and is technically difficult to model. The buckling ratio (BR) computed from the DXA data while useful in research studies, is at best a crude index of buckling susceptibility.

Finally, ethnic differences in body size and shape are likely to influence the mechanics of the hip in complex ways and that the simplistic height and weight adjustments used here are less than ideal for ensuring that groups are compared in mechanically equivalent ways. There is an undoubtedly considerable variation in the way that people fall on the hip, and it is not known whether different ethnic groups fall differently in ways that may independently influence their risk of hip fracture.

Summary and conclusions

It is well established that bones dynamically adapt their geometry to prevalent loading conditions [33] so that at steady state, the strengths of bones in an individual should be in equilibrium with the normally encountered loads. One subtle question for the present study is whether a geometric advantage of an ethnic group suggests that their bones are intrinsically stronger, i.e., that under equivalent load stimuli their bodies produce a relatively stronger femur. For a definitive answer one would require complex three-dimensional models that incorporate the entire pelvic geometry, but given the limited geometric data in this study we suspect that observed differences in CSA and SM between groups are more likely due to body shape heterogeneity that influences the way that physiologic forces are applied to the hip joint rather than a difference in biological response to load stimuli. Suggestive evidence for a difference in biological response might be a consistent ethnic effect on SM across both NN and IT regions. We did not observe such an effect.

In this study we sought to determine whether geometric differences at two common proximal femur fracture sites might help to explain why older ethnic minority women suffer hip fractures at different rates than their white counterparts. After correction for body size we found significant differences in geometry that provide mechanical evidence for differences in hip fracture rates. Analysis of the conventional femoral neck BMD data revealed that the higher BMD in AA women is partially due to a smaller region of interest, due in turn to a narrower femur. A narrower femur requires more bone material within it to achieve a similar strength. But for osteoporotic fragility the more important aspect may be that the narrower thicker cortex configuration may improve cortical stability in cross-sections under compressive loads so that local buckling is less likely. The geometric advantages of MA women compared to whites appear to be quite different from AA women and are similar in some ways to that of NA women. Both groups showed higher FN BMD than whites but in MA women it was due to a smaller region not greater BMC while NA women had a similar-sized region with more BMC. There were no significant geometric differences from whites at the HSA narrow-neck region in either MA or NA women, but both groups had shorter femoral necks (not significant in NA), and higher SM and CSA at the intertrochanter. These parameters together with shorter stature should convey some advantage in bending stresses compared to whites in a fall on the hip. Buckling ratios did not differ from whites in MA women but were smaller at the intertrochanter among NA women perhaps increasing their structural advantage in resisting intertrochanter fractures. Although these results cannot be considered to represent the WHI observational population as a whole, these data should provide the groundwork for future hypothesis-driven research. Such work may help determine which skeletal determinants of hip fracture risk arise from acquired or inherited characteristics in various ethnic groups and could help to focus fracture prevention efforts in particular groups or individuals within groups.

Conflicts of interest

The hip structure analysis software method has been licensed to Hologic Inc. by the Johns Hopkins University (Dr. Beck's Institution). Dr. Beck has also co-founded the company Quantum Medical Metrics, to build DXA systems specialized to measure bone geometry. Other authors have no conflicts of interest.

Glossary of terms


Women self identified as African-American


Location of the center of mass of the hard tissue in the cross-section, measured by HSA from the mineral mass profile. Used in computing the CSMI and SM, but also useful in evaluating the symmetry of a cross-section since fragile NN and IT cross-sections are more asymmetric due to greater differences in cortical thickness on opposite margins

Cross-sectional area (CSA)

Total bone surface in a cross-section, exclusive of soft tissue spaces and pores. Forces directed along a long bone (axial forces) are distributed over the CSA, hence axial compressive stress varies inversely with CSA

Cross-sectional moment of inertia (CSMI)

The cross-sectional surface weighted by the square of distance from the center of mass of the cross-section. Bending stress within a cross-section depends inversely on the CSMI divided by the distance from the center of mass.

Geometry or structural geometry

The dimensions of the supporting material within an object, expressed in engineering terms especially within cross-sections

Geometric strength

That component of strength that governs stresses, not the ability of the material to withstand them, (currently cannot be measured by non-invasive methods). A method that evaluates only geometry (HSA and CT based methods) can only assess geometric strength.


Software program used to extract femur geometry from conventional DXA scan images of the hip at specific regions consisting of five parallel lines traversing the femur neck, separated by one pixel spacing. Measurements are computed for each line then averaged.


Intertrochanter region HSA program, traversing the femoral neck at its narrowest point.


Narrow-neck region used in HSA program, traversing the femoral neck at its narrowest point.


Shaft region located at a distance of 1.5× minimum neck outer diameter, distal to the intersection of the neck and shaft axes (ethnic differences in S region results were less remarkable and are not reported here).


Women self identified as Mexican-American


Women self-identified as Native American or American Indian

Local buckling

A failure mode where thin, relatively flat regions of a structure under compressive loads may bend (fold or crumple) locally leading to complete failure of a cross-section. Propensity to this failure mode is crudely estimated from DXA data with the buckling ratio (BR) where a higher value is associated with greater buckling susceptibility (see Methods).

Section modulus (SM)

Maximum bending stress in a cross-section is located at maximum distance from the center of mass (dmax) and is thus inversely related to the SM, where SM = CSMI/dmax.


force concentrations (per unit area) within an object from applied loads. Types of stress (i.e., tension, compression, torsion and shear) depend on how loads are applied, but magnitudes depend entirely on geometry of the object.


The loading force applied under a specific condition that causes internal stresses to exceed the material limits. Strength cannot actually be measured in vivo although it can be predicted using engineering information on the bone geometry, the loading conditions and the material strength.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2010