African-American women have a lower risk of fracture than white women, and this difference is only partially explained by differences in dual-energy X-ray absorptiometry (DXA) areal bone mineral density (aBMD). Little is known about racial differences in skeletal microarchitecture and the consequences for bone strength. To evaluate potential factors underlying this racial difference in fracture rates, we used high-resolution peripheral quantitative computed tomography (HR-pQCT) to assess cortical and trabecular bone microarchitecture and estimate bone strength using micro–finite element analysis (µFEA) in African-American (n = 100) and white (n = 173) women participating in the Study of Women's Health Across the Nation (SWAN). African-American women had larger and denser bones than whites, with greater total area, aBMD, and total volumetric BMD (vBMD) at the radius and tibia metaphysis (p < 0.05 for all). African-Americans had greater trabecular vBMD at the radius, but higher cortical vBMD at the tibia. Cortical microarchitecture tended to show the most pronounced racial differences, with higher cortical area, thickness, and volumes in African-Americans at both skeletal sites (p < 0.05 for all), and lower cortical porosity in African-Americans at the tibia (p < 0.05). African-American women also had greater estimated bone stiffness and failure load at both the radius and tibia. Differences in skeletal microarchitecture and estimated stiffness and failure load persisted even after adjustment for DXA aBMD. The densitometric and microarchitectural predictors of failure load at the radius and tibia were the same in African-American and white women. In conclusion, differences in bone microarchitecture and density contribute to greater estimated bone strength in African-Americans and probably explain, at least in part, the lower fracture risk of African-American women. © 2013 American Society for Bone and Mineral Research.
Cross-sectional and longitudinal cohort studies have consistently demonstrated that fracture risk is about 50% lower in African-American women than in white women.[1-3] The reasons for these differences are incompletely understood. Although areal bone mineral densities (aBMDs) of the lumbar spine and proximal femur, as measured by dual-energy X-ray absorptiometry (DXA), are higher in African-American women, differences in BMD account for less than one-half of the variation in fracture risk.[4-9] In theory, differences in bone shape, cortical and trabecular microarchitecture, properties of the bone tissue, and nonskeletal factors such as the risk of falling could contribute to this ethnic difference in fracture risk. Although bone strength is directly proportional to bone size, lumbar vertebral area and femoral neck area are both lower in African-American women than in white women.
Until recently, microarchitectural characteristics of bone could only be assessed by histomorphometric analysis of bone biopsy specimens, which because of its invasive nature is not suitable for widespread clinical assessment of fracture risk. Recently, however, noninvasive methods, such as high-resolution peripheral quantitative computed tomography (HR-pQCT), have become available for assessing cortical (Ct) and trabecular (Tb) bone density and microarchitecture and their potential role in skeletal integrity. Because HR-pQCT measures volumetric BMD (vBMD) rather than aBMD, it avoids projection artifacts resulting from differences in bone size that are inherent to DXA.[10-12] HR-pQCT can also be used to perform micro–finite element analysis (µFEA), a technique that incorporates geometric and material properties of bone into biomechanical measures that reflect whole-bone strength.[13, 14] Cadaveric studies suggest that µFEA predicts femoral and vertebral strength better than aBMD,[15-17] whereas prospective case-cohort studies suggest that QCT-based FEA predicts fracture even after adjustment for aBMD.[18, 19] In this study, we measured areal BMD of the spine, hip, and total body by DXA and assessed bone microarchitecture and estimated strength of the distal radius and tibia using HR-pQCT in 273 African-American or white women in order to (1) characterize racial differences in bone microarchitecture, volumetric density, and µFEA-derived measures of bone strength; (2) identify the microarchitectural predictors of estimated bone strength; and (3) determine whether these HR-pQCT derived parameters provide information independent of aBMD and other clinical covariates that may help to explain the lower fracture rates in African-Americans.
Subjects and Methods
Subjects and eligibility criteria
The Study of Women's Health Across the Nation (SWAN) is a seven-site, longitudinal cohort study in community-based samples of women. Women were initially recruited between 1996 and 1997 and were required to be 42 to 52 years old, have menstruated within the last 3 months, and belong to one of the site's predesigned race/ethnic groups. The Boston SWAN cohort initially used data from the annual census to provide a random sample of African-American and white women, the Boston site's predesignated race/ethnic groups. Ethnicity was determined by subject self-identification. Initial eligibility criteria, cohort recruitment, and determination of menopause stage have been described in detail. Subjects have been followed prospectively for 15 years, with follow-up visits every 1 to 2 years. The current study was conducted at the Boston SWAN site during the 11th and 12th follow-up visits, at which time the women were 56 to 66 years old and 93% were postmenopausal. For the current study, women were excluded if they had a contraindication to DXA and/or HR-pQCT scanning, a history of solid organ transplant, or weight greater than 330 lbs (as a result of weight limits of the HR-pQCT equipment). The protocol was approved by the Partners Healthcare Institutional Review Board, and all women provided written informed consent.
Assessment of clinical covariates
Clinical factors, including age (years), cigarette smoking, alcohol intake (drinks per day), medical diagnoses, medication use, reproductive history, menopause stage, and physical activity were assessed at the concurrent visit using standardized interviewer-administered and self-administered questionnaires. History of fractures occurring after age 20 years was obtained by self-report at the baseline SWAN visit, and subjects subsequently reported new fractures at each follow-up visit. For the present analysis, fractures of the hand, foot, and face were excluded. At the Boston SWAN site, all reported fractures occurring during the 15-year follow-up period were confirmed by X-ray or physician reports. Subjects also reported any use of osteoporosis medications (including all oral and intravenous bisphosphonates, selective estrogen receptor modulators, teriparatide, and calcitonin) at the baseline and all follow-up study visits.
Assessment of aBMD
aBMD of the lumbar spine in the posterior-anterior (PA) and lateral projections, total hip, femoral neck, and total body were measured by DXA (QDR4500A; Hologic Inc., Bedford, MA, USA). The head was excluded from total body DXA measurements to avoid artifacts from metal jewelry and dental work. For lateral spine scans, the DXA scanners utilized a C-arm to image L2–L4 vertebrae, and L4 was excluded if the pelvis overlaid the vertebral body. A standard quality control program was employed that included daily measurement of a Hologic DXA anthropomorphic spine phantom, visual review of every scan image by a local site investigator experienced in bone densitometry, and central review of a randomly-selected 5% of scans plus all problem scans by Synarc, Inc. (Newark, CA, USA).
Assessment of vBMD, bone microarchitecture, and bone strength
On the same day as their aBMD measurement, volumetric bone density (vBMD) and microarchitecture of the distal radius and tibia were assessed using HR-pQCT (Xtreme CT; Scanco Medical AG, Basserdorf, Switzerland) as described.[10-12] Quality control was maintained with daily scanning of the manufacturer's phantom. All HR-pQCT scans were reviewed for motion artifact and were repeated if significant motion artifact was noted. Two radius scans and one tibia scan were excluded as a result of motion artifact on both the initial and the repeat scan. In addition, seven tibia scans could not be obtained as a result of size limitations or difficulty with limb positioning.
Using Scanco analysis software version V6.0, total bone area (mm2), total and trabecular vBMD (total vBMD, Tb vBMD, mg HA/cm3), and trabecular number (Tb N, mm−1) were measured directly. Trabecular separation (Tb Sp, mm), trabecular thickness (Tb Th, mm), and trabecular distribution (Tb Sp SD, mm) were then calculated.
To characterize cortical microarchitecture, HR-pQCT images were processed by a semiautomated technique implemented in Scanco software.[21-23] After image segmentation of cortical bone, the following measures were obtained: cortical bone volume (Ct BV, mm3), cortical vBMD (Ct vBMD, mg HA/cm3), cortical thickness (Ct Th, mm), cortical area (Ct Ar, mm2), trabecular area (Tb Ar, mm2), cortical porosity (Ct Po, %), and endocortical perimeter (mm).
Linear µFEA of the entire bone was used to estimate radius and tibia metaphyseal biomechanical properties under uniaxial compression as described.[24, 25] Outcomes included stiffness (kN/m), failure load (kN), load fraction carried by the cortical and trabecular compartments at the proximal and distal regions of interest (%), and apparent modulus (E app, MPa).
Same-day reproducibility (with repositioning) for HR-pQCT measurements at the radius and tibia in our laboratory ranged from 0.2% to 1.4% for vBMD parameters; 0.3% to 8.6% for trabecular microarchitecture parameters; 0.6% to 2.4% for cortical microarchitecture parameters; 7.3% to 20.2% for cortical porosity measurements; and 2.1% to 3.0% for µFEA measures. These ranges are similar to published reports.[10, 22]
Statistical analysis was performed using SAS 9.2 software (SAS Institute Inc., Cary, NC, USA). Clinical characteristics of African-American and white women were compared using independent samples two-sided t tests and/or chi-square tests. Primary outcomes were HR-pQCT-derived microarchitectural and volumetric density measures and µFEA results at the radius and tibia. Unadjusted differences in means of aBMD and HR-pQCT parameters between African-American and white women were examined by using independent samples two-sided t tests. The group comparisons were then repeated using a multivariate linear regression model (PROC REG in SAS) while adjusting for clinical covariates known to affect bone health and those that were significantly different between groups by univariate analysis, including age, weight, current tobacco and alcohol use, current physical activity score, diabetes, and history of systemic use of hormone replacement therapy (HRT), osteoporosis medications (oral or intravenous bisphosphonates or raloxifene), and significant glucocorticoids (as defined by self report of glucocorticoid use >3 months at the baseline visit or report of use at three or more subsequent follow-up visits). Additional analyses were performed to adjust for aBMD at the total hip in addition to all the aforementioned covariates with the purpose of determining whether HR-pQCT differences between racial groups persisted after adjusting for DXA aBMD. Next, to determine microarchitectural predictors associated with failure load at the radius and tibia, we utilized a general linear model with multiple predictors. To account for colinearity in the predictor model, microarchitectural variables were grouped using oblique component variable cluster analysis, and, based on biomechanical relevance, one variable was chosen from each cluster to enter into the model. Race was also included to test whether the microarchitectural predictors were sufficient to explain racial differences in failure load. Last, sensitivity analysis was performed by repeating all unadjusted and multivariate-adjusted analyses after exclusion of 142 women with a history of one or more of the following conditions known to affect bone: previous systemic HRT use (n = 94), previous bisphosphonate or selective estrogen receptor modulator use (n = 26), current tamoxifen or aromatase inhibitor use (n = 8), prior significant glucocorticoid use (n = 31), hypercalcemia (n = 2), hyperthyroidism (n = 7), and anorexia nervosa (n = 4); some subjects had more than one of the above conditions. Data are reported as mean ± SD unless otherwise noted, and p values ≤0.05 are considered statistically significant.
One hundred African-American women and 173 white women underwent HR-pQCT scanning of distal radius and tibia (Table 1). On average, they were 59.9 ± 2.7 years old, and 93% were postmenopausal at this time. Time since the final menstrual period, and history of glucocorticoid, osteoporosis medication, and hormone replacement therapy use were similar between racial groups. White women weighed less (76.4 ± 16.8 kg versus 84.6 ± 19.1 kg, p < 0.01) and were more likely to drink alcohol than African-American women (p < 0.01). African-American women had a lower physical activity score (p < 0.01) and were more likely to smoke (p = 0.03) than white women. Medical comorbidities were generally similar between races, although diabetes was more prevalent in the African-American women (25% versus 5.8%, p < 0.01).
|White (n = 173)||African-American (n = 100)||p|
|Age (years)||60.0 ± 2.8||59.6 ± 2.6||0.21|
|Weight (kg)||76.4 ± 16.8||84.6 ± 19.1||<0.01|
|Height (cm)||164.5 ± 5.9||163.7 ± 6.7||0.31|
|BMI (kg/m2)||28.2 ± 5.8||31.5 ± 6.5||<0.01|
|Physical Activity Score||8.5 ± 1.8||7.7 ± 1.9||<0.01|
|Tobacco, n (%)||16 (9.3)||18 (18)||0.03|
|EtOH, n (%)|
|None||19 (11.0)||36 (36)||<0.01|
|<2/day||140 (75.1)||63 (63)|
|≥2/day||24 (13.9)||1 (1)|
|Current calcium supplement use, n (%)||67 (39.6)||28 (28.9)||0.08|
|Current vitamin D supplement use, n (%)||74 (43.8)||31 (32.0)||0.06|
|Current multivitamin use, n (%)||71 (42.0)||38 (39.1)||0.65|
|Significant glucocorticoid use, n (%)||18 (10.4)||13 (13)||0.51|
|Osteoporosis medication use, n (%)||17 (9.8)||9 (9)||0.82|
|Current aromatase inhibitor/tamoxifen use, n (%)||6 (3.6)||2 (2.1)||0.51|
|Diabetes, n (%)||10 (5.8)||25 (25)||<0.01|
|Osteoporosis, n (%)||19 (11.0)||8 (8)||0.43|
|Breast cancer, n (%)||10 (5.8)||6 (6)||0.94|
|Hyperthyroidism, n (%)||2 (1.2)||5 (5)||0.05|
|Confirmed fracture, n (%)||22 (12.7)||9 (9)||0.35|
|Age at menarche (years)||12.6 ± 1.4||12.8 ± 2.1||0.30|
|Menopause duration (months)||92.0 ± 41.7||96.0 ± 43.2||0.46|
|Systemic HRT use, n (%)||62 (35.8)||32 (32)||0.52|
aBMD of the PA spine, lateral spine, total hip, femoral neck, and total body was significantly higher in African-American women than in whites (p < 0.01 for all, Table 2). After adjustment for all clinical covariates (see Subjects and Methods), African-American women had higher total hip, femoral neck, and total body aBMD, but differences in PA and lateral spine aBMD were no longer statistically significant.
|DXA aBMD||Unadjusted||Clinical covariate adjusted|
|Total body (g/cm2)||1.077 ± 0.008||1.158 ± 0.013||<0.0001||1.088 ± 0.009||1.138 ± 0.012||0.002|
|Total hip (g/cm2)||0.911 ± 0.010||1.006 ± 0.015||<0.0001||0.928 ± 0.010||0.976 ± 0.013||0.005|
|Femoral neck (g/cm2)||0.761 ± 0.009||0.879 ± 0.014||<0.0001||0.775 ± 0.009||0.854 ± 0.012||<0.0001|
|PA spine (g/cm2)||0.986 ± 0.013||1.065 ± 0.017||0.0002||1.004 ± 0.012||1.033 ± 0.016||0.184|
|Lateral spine (g/cm2)||0.705 ± 0.009||0.747 ± 0.015||0.014||0.714 ± 0.010||0.728 ± 0.015||0.447|
vBMD and microarchitecture
HR-pQCT images of the radius and tibia of a representative African-American subject and white subject are presented in Fig. 1. In unadjusted analyses, all differences between African-American and white women favored improved skeletal characteristics in African-Americans (Table 3). Total vBMD and total cross-sectional area of the radius and tibia were significantly higher in African-American women than white women. Trabecular vBMD at the radius and cortical vBMD of the tibia were also significantly higher in African-Americans. Trabecular thickness, cortical area, cortical bone volume and cortical thickness were significantly greater in African-Americans at both the radius and the tibia, and cortical porosity was significantly lower at the tibia in African-American women.
|Unadjusted||Clinical covariate adjusted||aBMD and clinical covariate adjusted|
|Total area (mm2)||260.6 ± 3.6||272.7 ± 5.1||0.049||262.1 ± 3.8||270.1 ± 5.2||0.246||260.9 ± 3.8||272.3 ± 5.2||0.097|
|Total vBMD (mg HA/cm3)||300.1 ± 5.3||321.8 ± 7.1||0.014||303.2 ± 5.6||316.4 ± 7.6||0.185||308.5 ± 4.7||306.7 ± 6.4||0.830|
|Tb vBMD (mg HA/cm3)||148.5 ± 2.9||159.5 ± 3.8||0.023||150.6 ± 3.0||155.9 ± 4.1||0.324||154.0 ± 2.4||149.9 ± 3.3||0.345|
|Ct vBMD (mg HA/cm3)||945.2 ± 4.6||952.3 ± 4.9||0.317||945.8 ± 4.5||951.1 ± 6.1||0.505||948.2 ± 4.3||946.5 ± 5.7||0.825|
|Tb number (mm−1)||1.85 ± 0.02||1.87 ± 0.03||0.624||1.88 ± 0.02||1.83 ± 0.03||0.311||1.90 ± 0.02||1.80 ± 0.03||0.012|
|Tb thickness (mm)||0.066 ± 0.001||0.071 ± 0.001||0.002||0.066 ± 0.001||0.071 ± 0.001||0.009||0.067 ± 0.001||0.070 ± 0.001||0.146|
|Tb separation (mm)||0.489 ± 0.008||0.485 ± 0.014||0.811||0.482 ± 0.009||0.497 ± 0.013||0.368||0.475 ± 0.008||0.511 ± 0.011||0.017|
|Tb distribution (µm)||0.226 ± 0.007||0.216 ± 0.010||0.428||0.221 ± 0.007||0.224 ± 0.010||0.828||0.215 ± 0.007||0.234 ± 0.009||0.112|
|Tb area (mm2)||209.5 ± 3.7||216.2 ± 5.2||0.291||210.4 ± 4.0||214.8 ± 5.4||0.539||208.7 ± 3.9||217.8 ± 5.3||0.201|
|Ct area (mm2)||53.79 ± 0.67||59.45 ± 1.01||<0.0001||54.53 ± 0.69||58.18 ± 0.94||0.003||54.93 ± 0.65||57.39 ± 0.89||0.037|
|Ct bone volume (mm3)||447.5 ± 6.2||498.6 ± 9.0||<0.0001||453.8 ± 6.3||487.7 ± 8.6||0.003||457.8 ± 5.9||479.9 ± 8.1||0.039|
|Ct thickness (mm)||0.851 ± 0.014||0.9261 ± 0.019||0.001||0.858 ± 0.014||0.914 ± 0.020||0.032||0.867 ± 0.014||0.898 ± 0.019||0.208|
|Ct porosity (%)||2.69 ± 0.13||2.42 ± 0.12||0.159||2.65 ± 0.12||2.49 ± 0.16||0.456||2.61 ± 0.12||2.57 ± 0.16||0.854|
|Endocortical perimeter (mm)||66.47 ± 0.62||67.13 ± 0.87||0.525||66.61 ± 0.66||66.88 ± 0.89||0.815||66.33 ± 0.64||67.40 ± 0.88||0.354|
|Total area (mm2)||677.1 ± 8.5||713.6 ± 11.9||0.011||679.7 ± 8.7||709.0 ± 11.7||0.059||676.6 ± 8.6||713.8 ± 11.7||0.017|
|Total vBMD (mg HA/cm3)||274.0 ± 4.2||288.9 ± 5.8||0.036||278.6 ± 4.4||280.8 ± 6.0||0.778||282.4 ± 3.6||273.7 ± 4.9||0.181|
|Tb vBMD mg HA/cm3)||160.7 ± 2.8||165.4 ± 3.6||0.297||163.5 ± 2.8||160.5 ± 3.8||0.537||165.9 ± 2.4||156.2 ± 3.3||0.027|
|Ct vBMD (mg HA/cm3)||857.9 ± 5.4||891.2 ± 6.6||0.0002||863.4 ± 5.2||881.8 ± 7.1||0.049||866.5 ± 4.6||875.1 ± 6.3||0.306|
|Tb number (mm−1)||1.87 ± 0.03||1.85 ± 0.04||0.751||1.90 ± 0.03||1.80 ± 0.04||0.035||1.91 ± 0.02||1.78 ± 0.03||0.002|
|Tb thickness (mm)||0.072 ± 0.001||0.075 ± 0.001||0.047||0.072 ± 0.001||0.075 ± 0.001||0.100||0.073 ± 0.001||0.074 ± 0.001||0.376|
|Tb separation (mm)||0.481 ± 0.008||0.486 ± 0.011||0.787||0.474 ± 0.008||0.500 ± 0.011||0.075||0.469 ± 0.008||0.510 ± 0.010||0.004|
|Tb distribution (µm)||0.237 ± 0.009||0.229 ± 0.009||0.557||0.233 ± 0.009||0.235 ± 0.012||0.894||0.228 ± 0.008||0.245 ± 0.011||0.250|
|Tb area (mm2)||570.1 ± 8.7||597.7 ± 12.3||0.063||570.7 ± 9.1||596.7 ± 12.4||0.112||566.8 ± 8.9||602.9 ± 12.2||0.026|
|Ct area (mm2)||112.1 ± 1.2||121.0 ± 2.3||0.0002||114.3 ± 1.3||117.3 ± 1.8||0.210||115.0 ± 1.3||115.8 ± 1.7||0.702|
|Ct bone volume (mm3)||890.6 ± 11.2||978.4 ± 20.5||<0.0001||909.6 ± 12.0||945.4 ± 16.4||0.097||916.5 ± 11.1||931.1 ± 15.1||0.466|
|Ct thickness (mm)||1.152 ± 0.016||1.234 ± 0.027||0.005||1.170 ± 0.018||1.203 ± 0.024||0.303||1.180 ± 0.017||1.183 ± 0.023||0.923|
|Ct porosity (%)||7.58 ± 0.23||6.47 ± 0.27||0.003||7.42 ± 0.22||6.74 ± 0.31||0.094||7.33 ± 0.21||6.95 ± 0.29||0.319|
|Endocortical perimeter (mm)||98.8 ± 0.8||100.6 ± 1.0||0.156||98.9 ± 0.8||100.4 ± 1.1||0.292||98.6 ± 0.8||101.0 ± 1.1||0.112|
After adjustment for clinical covariates, differences in vBMD were largely eliminated, except for higher cortical vBMD at the tibia in African-American women. Cortical area, cortical bone volume, and cortical thickness also remained statistically higher in African-American women after covariate adjustment at the radius, although these latter cortical differences did not persist at the tibia. After adjustment for clinical covariates, the trabecular thickness of the African-American women remained significantly greater than the trabecular thickness of the white women at the radius, but at the tibia the trabecular number of the African-American women unexpectedly was significantly lower than the trabecular number of the white women.
Further adjustment for femoral aBMD in addition to clinical covariates led to greater total cross-sectional area for African-American women at the tibia. Cortical area and cortical bone volume remained significantly higher at the radius for African-American women. However, adjustment for aBMD led to surprising improvements in trabecular parameters among white women, including statistically higher trabecular number and decreased trabecular spacing at the radius and tibia along with higher trabecular vBMD at the tibia as compared with African-American women.
Estimated bone stiffness and failure load, derived from µFEA, were significantly greater in African-American women than in white women at both the radius and the tibia, and these differences remained significant after adjustment for clinical covariates and for aBMD (Fig. 2). E App, an index of resistance to compressive forces independent of bone geometry, was also greater in African-American women in unadjusted analyses, although this difference did not persist after adjustment for clinical covariates and aBMD (data not shown). The percentage of load carried by cortical and trabecular bone at proximal and distal sites did not differ between races at either the radius or the tibia (data not shown).
Predictors of failure load
HR-pQCT–derived microarchitectural and BMD variables were grouped using cluster analysis into the following six categories: total cross-sectional area and perimeter; trabecular microarchitecture; cortical area and volume; cortical pore characteristics; trabecular volumetric density; and cortical volumetric density. Based on a priori assumed biomechanical relevance, one variable from each cluster was chosen to be incorporated into the multivariate model for predictors of the µFEA-estimated failure load at the radius and tibia. All women were pooled into this single analysis. The strongest predictors of estimated failure load at the radius and tibia were total cross-sectional area, trabecular vBMD, cortical thickness, and to a lesser extent trabecular number (Table 4). Cortical porosity was a modest predictor of failure load at the tibia but not at the radius. Cortical vBMD was not a predictor of failure load at either site. Altogether, these predictors explained 88% and 90% of the variation in µFEA-estimated failure load at the radius and tibia, respectively. Race was not a predictor of failure load independent of these variables (p = 0.21 and 0.12 at the radius and tibia, respectively).
|Standardized effect||p||Standardized effect||p|
|Total cross-sectional area||0.647||<0.0001||0.671||<0.0001|
All analyses were repeated after excluding subjects with a history of glucocorticoid use, osteoporosis medication use, systemic hormone replacement therapy use, current aromatase inhibitor or tamoxifen use, hypercalcemia, hyperthyroidism, and/or anorexia nervosa. Results of unadjusted and multiple covariate-adjusted comparisons of African-American women and white women were similar in this subset (n = 131) to results obtained from the entire cohort (data not shown).
In this study, we found that African-American women had more favorable values for a variety of cortical and trabecular bone densitometric and microarchitectural indices, resulting in higher estimated bone strength that persisted even after adjustment for clinical covariates and aBMD. Total vBMD was higher in African-American women than in white women at both the radius and the tibia in unadjusted analyses. In the radius, this difference was largely a result of higher trabecular vBMD, whereas in the tibia the difference was largely a result of higher cortical vBMD in the African-American women. In general, the differences in bone microarchitecture between African-American women and white women tended to be more pronounced in measures of cortical than trabecular bone. Adjustment for weight and other covariates attenuated racial differences to a greater extent in the tibia than in the radius, suggesting that the etiology of the racial differences in bone microarchitecture is different in weight-bearing versus non–weight bearing bone. Predictors of failure load were the same in white women and African-American women, suggesting that advantages in these parameters in African-American women lead to their greater estimated bone strength.
To our knowledge, this is the first study to report differences between postmenopausal white and African-American women in bone microarchitecture and predicted strength using HR-pQCT. In unadjusted analyses, African-American women had larger and denser bones, greater trabecular and cortical thickness, and greater cortical bone volume at both the radius and the tibia. Cortical porosity was also lower in the tibia of African-American women. These results are similar to several small studies that used bone histomorphometry in predominantly younger women to show that African-Americans have higher cortical and trabecular thickness than whites,[27-29] although not all histomorphometric studies have been consistent. Although DXA-measured lumbar vertebral area and femoral neck area values are lower in African-American women than in white women, studies using QCT have reported densitometric and geometric advantages in African-Americans compared to whites, primarily within cortical bone at the femur.[31-33] These skeletal advantages are also apparent in studies of African-American children using pQCT.[34, 35] However, all these QCT and pQCT studies were limited by larger voxel size, in which microarchitectural features of bone could not be distinguished.
HR-pQCT can detect variation in bone microarchitecture that may account for well-known racial/ethnic differences in fracture rates. For example, in comparison to whites, Chinese-American women have smaller bones but higher cortical vBMD and better trabecular bone microarchitecture,[36, 37] findings that are consistent with their lower fracture rate. In contrast, our findings in African-American and white women suggest that differences in bone size, as well as cortical density and cortical microarchitecture, likely explain the lower fracture risk of African-American women. Longitudinal observations are needed to substantiate this hypothesis.
The factors that account for differences in bone microarchitecture between African-Americans and whites are not well-characterized. In our cohort, physical activity levels were lower and tobacco use and rates of diabetes and hyperthyroidism were higher in African-American women. However, these characteristics are generally associated with greater skeletal fragility, and therefore cannot explain the improved microarchitectural and densitometric characteristics observed in African-American women. We performed multivariate adjustment to explore clinical factors (including weight) that might explain the skeletal differences observed between African-Americans and whites. This eliminated many of the racial differences in bone microarchitecture at the weight-bearing tibia, but not at the non–weight-bearing radius, consistent with the hypothesis that the higher average body mass of African-Americans may contribute to their improved skeletal characteristics at skeletal sites exposed to mechanical loading.
Many studies, including ours, have demonstrated that African-American women have higher DXA-measured aBMD than other ethnic groups.[4-9] However, higher aBMD does not entirely account for the reduction in fracture risk seen in African-American women. For example, Cauley and colleagues found that the relative risk of fracture was lower in African-Americans even after adjustment for aBMD and other risk factors. Therefore, to explore racial differences in bone microarchitecture and volumetric density independent of aBMD, we performed additional analyses while including aBMD in the multivariate adjustment model. We found that cortical area and cortical bone volume at the radius, as well as total cross-sectional area and trabecular area at the tibia remained significantly higher in African-American women even after adjustment for aBMD. In contrast to findings on the unadjusted analyses, adjustment for aBMD led to white women having more favorable trabecular microarchitecture, including increased trabecular number and decreased trabecular separation at the radius and tibia and higher trabecular vBMD at the tibia, as compared with African-American women. Despite this improved trabecular microarchitecture in white women, estimated bone stiffness and failure load remained significantly greater in African-American women after multivariate adjustments that included aBMD. The finding that estimated bone strength remains higher in African-American women despite inferior features of trabecular microarchitecture suggests that the favorable cortical bone characteristics of African-American women have a greater impact on whole-bone strength than the potential advantages of trabecular bone microarchitecture observed in white women.
Finally, to determine the contribution of individual bone microarchitectural characteristics to whole bone strength, we assessed the microarchitectural predictors of failure load, as estimated by µFEA, at the radius and tibia. We found that total cross-sectional area, trabecular vBMD, and cortical thickness were the strongest HR-pQCT-derived predictors of failure load at both the radius and tibia, with cortical porosity also associated with failure load at the tibia. These HR-pQCT-derived predictors of bone strength are similar to those identified in other populations including healthy postmenopausal women, young female athletes, and non-athletes.[38, 39] Importantly, the predictors of failure load at the radius and tibia were the same in African-American and white women. Because African-Americans tended to have superior values for each of these predictors compared to whites in unadjusted analyses, advantages in these bone microarchitectural and densitometric characteristics explain their greater predicted bone strength and likely contribute to their reduced fracture risk.
Our study has several important strengths, including the relatively large sample size, the detailed clinical information regarding factors affecting skeletal health, and the use of a high-resolution noninvasive imaging technique to assess bone microarchitecture and estimate bone strength via μFEA. In addition, interobserver variability was minimized by using a single operator for all HR-pQCT analyses. Our study also has several limitations. Although SWAN was designed to study women across the menopausal transition, nearly all women were already postmenopausal by the time we obtained measurements using HR-pQCT. Thus, our cross-sectional design cannot capture the dynamic changes that occur in the skeleton during the menopause transition, nor can we make inferences about skeletal status in premenopausal white and African-American women. Because women in our primary analysis were not excluded on the basis of medical conditions and medications known to affect bone, it is possible that these factors may affect the results despite our multivariate adjustments for them. However, results were similar in sensitivity analyses in which all women with bone-modifying disorders or taking bone-active medications were excluded. Finally, the HR-pQCT measurements were obtained at the peripheral skeleton only, and further studies are needed to determine whether the racial differences observed here are representative of other skeletal sites.
In summary, African-American women have structurally advantageous differences in bone density and microarchitecture compared with white women, some of which remained significant after multivariate adjustment including aBMD. Differences in bone microarchitecture and density explain most of the variation in µFEA-predicted failure load and likely account for the lower fracture risk observed in African-American women.
All authors state that they have no conflicts of interest.
The Study of Women's Health Across the Nation (SWAN) has grant support from the National Institutes of Health (NIH), DHHS, through the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR) and the NIH Office of Research on Women's Health (ORWH) (NR004061, AG012505, AG012535, AG012531, AG012539, AG012546, AG012553, AG012554, AG012495). The HR-pQCT measurements were made possible by an NCRR Shared Equipment Grant (1S10RR023405-01). The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the NIA, NINR, ORWH, or the NIH.
Authors' roles: Study conduct and data collection: APT, ES, ECT, JSF, and RN. Data analysis: EWY, HL, and MSP. Data interpretation: MSP, EWY, JSF, MLB, and RN. Drafting manuscript: MSP, EWY, and JSF. Revising manuscript content: JSF, MLB, and RN. Approving final manuscript: all authors. MSP and EWY take responsibility for the integrity of the data analysis.