JOURNAL OF BONE AND MINERAL RESEARCH Volume 24, Number 8, 2009 Published online on March 16, 2009; doi:

10.1359/JBMR.090307 2009 American Society for Bone and Mineral Research

Does Obesity Really Make the Femur Stronger? BMD, Geometry, and Fracture Incidence in the Womens Health Initiative-Observational Study
Thomas J. Beck,1 Moira A. Petit,2 Guanglin Wu,3 Meryl S. LeBoff,4 Jane A. Cauley,5 and Zhao Chen3

ABSTRACT: Heavier individuals have higher hip BMD and more robust femur geometry, but it is unclear whether values vary in proportion with body weight in obesity. We studied the variation of hip BMD and geometry across categories of body mass index (BMI) in a subset of postmenopausal non-Hispanic whites (NHWs) from the Womens Health Initiative Observational Cohort (WHI-OS). The implications on fracture incidence were studied among NHWs in the entire WHI-OS. Baseline DXA scans of hip and total body from 4642 NHW women were divided into BMI (kg/m2) categories: underweight (<18.5), healthy weight (18.5 24.9), overweight (2529.9), and mild (3034.9), moderate (3539.9), and extreme obesity (>40). Femur BMD and indices of bone axial (cross-sectional area [CSA]) and bending strength (section modulus [SM]) were extracted from DXA scans using the hip structure analysis (HSA) method and compared among BMI categories after adjustment for height, age, hormone use, diabetes, activity level, femur neck-shaft angle, and neck length. The association between BMI and incident fracture was studied in 78,013 NHWs from the entire WHI-OS over 8.5 2.6 (SD) yr of follow-up. Fracture incidence (cases/1000 person-years) was compared among BMI categories for hip alone, central body (hip, pelvis, spine, ribs, and shoulder girdle), upper extremity (humerus and distal), and lower extremity (femur shaft and distal but not hip). Femur BMD, CSA, and SM were larger in women with higher BMI, but values scaled in proportion to lean and not to fat or total body mass. Women with highest BMI reported more falls in the 12 mo before enrollment, more prevalent fractures, and had lower measures of physical activity and function. Incidence of hip fractures and all central body fractures declined with BMI. Lower extremity fractures distal to the hip trended upward, and upper extremity incidence was independent of BMI. BMD, CSA, and SM vary in proportion to total body lean mass, supporting the view that bones adapt to prevalent muscle loads. Because lean mass is a progressively smaller fraction of total mass in obesity, femur BMD, CSA, and SM decline relative to body weight in higher BMI categories. Traumatic forces increase with body weight, but fracture rates at the hip and central body were less frequent with increasing BMI, possibly because of greater soft tissue padding. There was no evident protective effect in fracture rates at less padded distal extremity sites. Upper extremity fractures showed no variation with BMI, and lower extremity fracture rates were higher only in the overweight (BMI = 2529.9 kg/m2). J Bone Miner Res 2009;24:13691379. Published online on March 16, 2009; doi: 10.1359/JBMR.090307 Key words: obesity, hip structural geometry, femur strength, fractures, postmenopausal women, Womens Health Initiative
Address correspondence to: Thomas J. Beck, ScD, The Johns Hopkins Outpatient Center, 601 North Caroline Street, Baltimore, MD 21287 USA, E-mail: tjbeck@jhmi.edu obesity seems to moderate the effects of osteoporosis by increasing BMD(4) and by making fractures less likely. DeLaet et al.,(3) in a meta-analysis of 12 multinational cohorts comprising nearly 60,000 adult subjects, showed that independent of sex, those with body mass indices (BMIs) >25 kg/m2 had signicantly lower rates of hip, osteoporotic, and all fractures. However, increasing body weight is mainly caused by greater fat mass, and whereas biological mechanisms have been postulated to the contrary, there is little evidence to suggest that fat directly inuences the skeleton in ways that are evident at the

INTRODUCTION
HROUGHOUT MUCH OF the developed and developing world, obesity and osteoporosis are important public health concerns. For example, >50% of U.S. adults are overweight or obese according to recent surveys.(1) The associations between obesity and osteoporosis have been studied in a recent review(2) and in a meta-analysis.(3) Although details remain unclear, the general consensus is that

The authors state that they have no conicts of interest.

1 Johns Hopkins University School of Medicine, Baltimore, Maryland, USA; 2University of Minnesota School of Kinesiology, Minneapolis, Minnesota, USA; 3Mel and Enid Zuckerman College of Public Health, University of Arizona, Tucson, Arizona, USA; 4 Brigham and Womens Hospital, Harvard Medical School, Boston, Massachusetts, USA; 5University of Pittsburgh Graduate School of Public Health, Pittsburgh, Pennsylvania, USA.

1369

1370 macroscopic (whole bone) level.(2) Moreover, from a biomechanical perspective, the fracture categories used by De Laet et al. may not tell the whole story of weight and fracture risk because of the differing effects of soft tissue padding over skeletal sites. Whereas BMD does increase with body weight, it is not clear that either BMD or bone strength remains in proportion to weight in the obese. Evidence in children suggests that femur geometric strength is reduced relative to body weight in the overweight,(57) but whether this is the case in elderly postmenopausal women has not been explored. The probability of fracture depends not only on the strength of the bone but on trauma severity and the likelihood that it will occur. Without question, bones adapt to prevalent loading conditions,(8) and skeletal loads at weight-bearing sites stimulate adaptive increases in bone density(4) and geometry.(7) However, heavier patients are generally less active and may have fewer trauma opportunities; when traumas do occur, their greater soft tissue padding may moderate impacts.(9,10) On the other hand, the magnitudes of traumatic impact forces on the skin surface (if not the bone) are a direct function of body weight (e.g., all else equal, a doubling of body weight doubles the impact force in a fall). Bones fracture when external forces (loads) cause internal stresses to exceed the strength of the tissue. Although we are unaware that the issue has been studied, there is no obvious reason why bone tissue strength in otherwise normal individuals would be inuenced by body weight. It is therefore likely that weight effects are mainly geometric (i.e., in the dimensional characteristics that determine stress magnitudes under a given load). Because femur stresses generated from physical activities and common traumas are mainly in axial compression and bending, we concentrated on measurements of bone crosssectional area (CSA) and section modulus (SM) that govern these stresses, respectively. CSA and SM can be computed from DXA scans of the hip with special software. In this study, we evaluated how femur BMD and geometry vary in proportion to body weight and body composition across categories of BMI (kg/m2) among non-Hispanic white (NHW) women 50 yr of age in the bone density cohort of the Womens Health Initiative Observational Study (WHI-OS). Because geometric strength should inuence fracture rates, we sought to evaluate fracture incidence by BMI category, but because few fractures occurred in those with DXA scans, the entire WHI-OS cohort of NHW women was used. To separate out effects of soft tissue padding, fracture incidence versus BMI was evaluated at both well-padded central body sites and at less padded extremity sites.

BECK ET AL. centers across the United States. There were two components in the WHI: an observational study (OS) and a set of three clinical trials. This study included a cross-sectional analysis of BMD and other variables at baseline and used incident fracture determined on prospective follow-up of the participants in the WHI-OS.

The bone density cohort

At WHI enrollment, subjects were scanned at the total body and hip using Hologic QDR2000 (Hologic, Bedford, MA, USA) machines at 3 of the 40 WHI clinic centers in the United States: Tucson and Phoenix, Arizona; Pittsburgh, Pennsylvania; and Birmingham, Alabama. The hip structure analysis (HSA) program was subsequently used on available hip DXA scans acquired at baseline on 6032 WHI-OS participants. The complex effects of race and obesity on femur geometry are the subject of a separate analysis; hence, this study was restricted to 4642 women self-identied as NHW. For comparison purposes, women were divided into six standard BMI (kg/m2) categories including underweight (<18.5), healthy weight (18.524.9), overweight (2529.9), mild obesity (3034.9), moderate obesity (3539.9), and extreme obesity (>40).

The fracture cohort

Fractures were assessed among NHW women enrolled in the entire WHI-OS cohort (including the bone density subset). These included 78,013 subjects who were followed an average of 8.5 2.6 (SD) yr for a total of 5180.4 personyears. Within the WHI-OS, fracture assessment for all fracture types was by annual self-report, with the exception that hip fractures were adjudicated by review of medical records. The overall validity of self-reported fractures in WHI was evaluated and, although there was some variation by skeletal site, the investigators concluded that incidence rates were acceptably accurate.(11)

Data collection for other covariates

Questionnaires were used at baseline to collect information on age, race/ethnicity, recreational physical activities, energy intake, and factors such as hormone use and the presence of chronic disease. Physical activity was assessed by questions on the frequency and duration of a range of different types of activities using the modied coding scheme of Ainsworth et al.(12) This scheme attempts to assign intensity in terms of metabolic equivalent task (MET) scores (dened as the ratio of work metabolic rate to a standard resting metabolic rate). Scores are dened relative to 1 MET roughly equivalent to resting metabolism while sitting quietly as the product of days per week, minutes per day, and MET value for each activity.(13) Physical function was measured using the 10-item Medical Outcomes Study Scale,(14) where 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. Body composition was measured by DXA in the

MATERIALS AND METHODS Overview of the WHI

The WHI is one of the largest prospective health studies among postmenopausal women. Participants between 50 and 79 yr of age at baseline and lacking a condition likely to cause death within 3 mo were recruited from 40 clinical

OBESITY AND GEOMETRY OF THE PROXIMAL FEMUR IN WHI

1371

FIG. 1. A typical DXA scan of the adult hip showing locations of the ve mineral mass proles that are averaged in each region for geometry measurements with the HSA software.

form of total body lean and total body fat mass in kilograms, which by denition exclude bone mineral mass.

HSA
The HSA method uses the principle that a line of pixels traversing the bone axis in a bone mineral mass image is a projection of the mineral in the corresponding crosssection.(15) Analysis sites include the following: narrow neck (NN) across the femur neck at its narrowest point; the shaft across the diaphysis at 1.5 times minimum neck width distal to the intersection of the neck and shaft axes, and intertrochanter (IT) along the bisector of the angle between neck and shaft axes. As shown in Fig. 1, ve parallel proles are generated one pixel (;1 mm) apart at each region, and measurements are averaged over the ve proles. Average pixel value in the prole is reported as BMD, and outer diameter (OD) is measured from outer 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 prole integral is thus bone CSA (cm2). The center of mass (COM) of the prole is determined, and the cross-sectional moment of inertia (CSMI) is measured as the integral weighted by the square of distance of each pixel from the COM. Maximum bending stress in a cross-section is a function of its SM, computed as the CSMI divided by the maximum distance of COM to the medial or lateral prole margin (dmax).

Research has suggest that homeostatic mechanisms tend to preserve the SM in aging bone cross-sections,(16,17) but the SM can overestimate strength of tubular objects if walls become so thin that they buckle (fold) under compressive loads. This complex phenomenon cannot be fully characterized by the limited information in a DXA scan, but a hint that a cross-section may be susceptible can be obtained from the buckling ratio (BR). BR is used in engineering designs incorporating hollow tubes where the ratio of outer radius to wall thickness should be kept below ;10 to avoid local buckling.(18) BR is estimated in HSA by modeling the cross-section as a hollow circular (NN) or elliptical (IT) annulus with a xed proportion (60%, 100%, and 70%, respectively) of the CSA in the cortical shell for the narrowneck, intertrochanter, and shaft regions, respectively. Cross-calibration of DXA scanners for HSA was achieved by circulating a specically designed calibration phantom across all scanners in the project. Precision for the HSA method was not specically evaluated in the WHI study, although it should be comparable to results on Hologic scanners as reported by Khoo et al.(19)

Statistics
For the purposes of general exploration, a stepwise multiple linear regression was used to identify independent variables with a signicant inuence on conventional femoral neck DXA variables, as well as those from the three HSA regions. Independent variables included age,

1372

BECK ET AL.

TABLE 1. Standardized b Coefcients From Stepwise Multiple Linear Regressions of Conventional and HSA- Derived DXA Parameters on Independent Variables Including Age, Height, Total Body Lean and Fat Masses, Femoral Neck-Shaft Angle and Neck Length, Use of Hormone Therapy, and a Diagnosis of Diabetes in NHW Subjects From the BMD Cohort of WHI-OS (N = 4642) Age Conventional femoral neck BMD BMC Area BMD CSA Outer diameter SM BMD CSA Outer diameter SM BMD CSA Outer diameter SM 20.207 20.114 0.205 20.239 20.147 0.248 20.025 20.173 20.067 0.295 0.062 20.153 20.06 0.227 0.097 TB lean mass 0.275 0.357 0.216 0.275 0.371 0.211 0.409 0.256 0.400 0.389 0.466 0.338 0.472 0.318 0.542 TB fat mass 0.244 0.188 20.098 0.177 0.159 20.054 0.156 0.206 0.179 20.077 0.100 0.190 0.184 0.124 Height 0.139 0.346 0.103 0.313 0.157 20.089 0.284 0.127 20.114 0.280 0.136 Neck-shaft angle 20.149 20.132 0.026 20.188 20.198 20.028 20.223 20.241 20.246 20.177 20.156 20.164 20.132 Neck length 20.033 0.121 0.214 0.189 20.07 0.167 0.134 0.164 0.072 0.136 HRT use 0.144 0.137 0.110 0.117 0.092 0.101 0.102 0.091 0.104 0.079 20.055 0.033 Diabetes 20.028 20.028 20.173 20.067 0.295 0.062 20.153 20.060 0.227 0.097

HSA narrow-neck

HSA intertrochanter

HSA shaft

, nonsignicant contribution to the regression (p > 0.05).

height, lean and fat masses, neck-shaft angle and neck length, hormone therapy use, and a diagnosis of diabetes. Differences in descriptive characteristics between BMI groups were tested using ANOVA or the x2 test for proportions as appropriate. Differences in bone geometry between BMI groups were evaluated by analysis of covariance (ANCOVA) using height, age, weekly energy expended in physical activity, use of hormone therapy, diabetes, neck-shaft angle, and femur neck length as covariates. The latter two variables inuence the length of the bending moment arm in the proximal femur and may have an independent effect on stress magnitudes. To evaluate how BMD and geometry vary in proportion to body mass, they were evaluated again with the same covariates after normalizing parameters to total body mass and then to lean body mass. A signicance level of p = 0.05 was used in all tests. Fracture incidence by BMI group was calculated as numbers of incident fractures divided by the amount of at-risk experience and expressed as fractures/1000 person-years of risk. Differences in fracture incidence rates between BMI groups were tested using a proportional hazards model and adjusted for group differences in age, hormone use, and a diagnosis of diabetes.

greatest effects on SM. Fat mass was a signicant contributor to all models except shaft OD, and contributions to SM were generally small. Fat mass had larger b coefcients on BMD than on BMC or its HSA counterpart CSA. It also had a weak negative effect on FN region area and on OD at the NN and IT regions. Height effects reect general size scaling, and the effects of NSA and NL are mainly caused by the inuence of height on the size of bending moments at the proximal femur in ambulation. Negative coefcients on age are evident for BMD, BMC, and CSA at all regions, but age effects on region area and OD are all positive. Age effects on SM were weakly negative at the NN but weakly positive at the IT and shaft regions. Hormone therapy use had generally positive effects on BMD, BMC, CSA, and SM and a weak negative effect on shaft OD. A diagnosis of diabetes had negative effects on region area and OD, some negative effects on BMD and CSA, and a small positive effect on SM at the IT and shaft regions.

Descriptive characteristics
General characteristics of NHW participants by BMI category at baseline are listed in Table 2. There is a trend toward younger age with increasing BMI, and heaviest groups are slightly shorter in stature (p < 0.05 for both versus healthy weight). In absolute terms, total body lean and fat masses increase with BMI, and the proportion of body fat increases from 28% in the underweight to 53% in the extremely obese. There was a signicant trend toward lower physical activity and reduced physical function with increasing BMI; fewer women walked at all for exercise, and of those that did, few walked >150 min/wk. Heavier women expended less energy in exercise, and fewer had physical function scores >90 in the 10-item Medical Outcomes Study Scale.(14) Women with BMIs greater than healthy weight were signicantly more likely to be diagnosed with diabetes. Healthy weight women were signicantly more likely to be current users of hormone therapy.

RESULTS Exploration of body composition and other factors on BMD and geometry
Standardized b coefcients for signicant contributors to the stepwise multiple linear regression of age, height, total body lean and fat mass, neck-shaft angle and neck length, use of hormone therapy, and a diagnosis of diabetes on conventional femoral neck DXA and HSA geometry parameters are listed in Table 1. A complete exposition is beyond the scope of this paper, but note that lean body mass produced the largest contribution to all models with

OBESITY AND GEOMETRY OF THE PROXIMAL FEMUR IN WHI

TABLE 2. Average Characteristics of Subjects by BMI Category Underweight Category BMI limits (kg/m2) Subjects in category (%) Age (yr) Height (cm) Weight (kg) Total body lean mass (kg) Total body fat mass (kg) Percent lean mass Percent fat mass Minutes/week spent walking for exercise None (%) 0150 min (%) 150 min (%) Weekly energy expenditure from physical activity (METS) Physical function score >90 (%) One or more falls in last 12 mo (%) Diagnosis of diabetes Current user of hormone therapy History of prior fracture of any bone (%) History of prior fracture on/after age 55 (%) Incident fracture (any bone) (%) Incident hip fracture (incl. pathological) (%) 18.5 205 (4.4%) 65.4 6.9 162.6 6.6 46.3 4.7 30.4 3.6 10.6 2.7 72.0 7.2 25.3 7.1 * 31% 43% 26% 19.8 52% 30% 3% 43% 47% 16% 15% 2.9% Healthy weight 18.524.9 1744 (37.6%) 64.3 7.3 162.4 6.2 59.5 6.0 31.7 3.7 21.9 4.7 58.7 6.0 38.7 6.1 Overweight 25.029.9 1601 (34.5%) 64.2 7.3 161.7 6.1* 71.4 6.2* 33.6 3.9* 31.4 4.5* 51.2 4.4* 46.5 4.5* * 43% 44% 13% 15.0* 32%* 33% 5%* 44% 42% 21% 16% 1.9% Mild obesity 3034.9 688 (14.8%) 64.48 7.2 161.1 6.0* 83.4 7.2* 36.5 4.2* 40.2 5.4* 47.2 4.2* 50.71 4.2* * 51% 39% 10% 13.1* 19%* 34% 9% 35% 48% 19% 14% 1.6% Moderate obesity 35.039.9 243 (5.2%) 62.3 7.1* 161.39 6.73 97.2 9.1* 39.7 4.5* 50.4 6.7* 43.7 3.7* 54.27 3.80* * 66% 27% 7% 9.9* 13%* 34% 8% 36% 44% 18% 13% 0.8%

1373

Extreme obesity 40 161 (3.5%) 61.2 7.1* 159.7 9.3* 111.1 11.7* 42.9 5.9 59.4 9.3* 42.5 4.9* 55.6 5.0* * 76 % 20 % 4% 9.0* 6%* 41% 17% 25% 54% 17% 17% 1.2%

34% 46% 20% 18.3 44% 32% 2% 49% 39% 21% 15% 2.1%

* p < 0.01, p < 0.05, and p < 0.0001 vs. healthy weight.

Variation of BMD and bone geometry with BMI

Mean (SE) values of both conventional and HSA BMD are listed by BMI category together with the HSA geometry at the three femur regions in Table 3. With the exception of OD among the underweight, all parameters in all regions differ signicantly from healthy weight. To show the proportional variation of BMD and geometric properties with BMI category, gures were plotted after dividing by the mean of the healthy weight BMI category, thus setting healthy weight values to 1.0 and lower and higher values to <1 and >1, respectively. Most differences from healthy weight were highly signicant, and except for Figs. 3C and 5, only nonsignicant (p > 0.05) comparisons were marked on Figs. 24. Figure 2 shows normalized, adjusted mean BMD values by BMI category for the three HSA and two conventional femur regions, illustrating similar scaling with BMI despite large differences in absolute value. All covariates were signicant contributors to most adjustment models, but some variables did not reach signicance (p < 0.1) in some models. Bone CSA and SM showed similar variation with BMI at the three HSA regions; hence, we show only the NN region in Fig. 3. As in Fig. 2, the adjusted BMD and geometric strength indices are larger with greater BMI in absolute terms (Fig. 3A), but when divided by body weight (Fig. 3B), the trend with BMI is reversed for all three parameters. Relative to their body weight, heavier individuals

have lower BMD, CSA, and SM. However, when plotted relative to total body lean mass (Fig. 3C), variation with BMI is much smaller and is nearly eliminated on SM. Despite signicant differences in underweight (p < 0.002) and overweight (p < 0.04), SM normalized to lean mass was within 1.4% of healthy weight in all heavier categories and was only 3% lower in the underweight. Although BMI differences were greatly reduced, BMD and CSA per unit lean mass remained signicantly different from health weight with larger differences at BMI extremes. BMD and CSA values in extremely obese were 10% and 8%, respectively, below health weight, and like SM, were 3% lower in the underweight.

Femur OD and estimated BR

Underlying the association of weight with BMD and SM is a subtle but signicantly greater OD with higher BMI that differed across HSA regions (Fig. 4A). OD trended more steeply upward with BMI at the IT and shaft than at the NN region. Whereas OD in all three regions was signicantly wider than healthy weight, no differences were apparent in the underweight. Larger OD produced a positive effect on SM, but in a thin-cortex bone, the strength conferred by a given SM may be offset by greater susceptibility to local buckling,(20) as estimated by a higher BR. In Fig. 4B the relative pattern of BMI on BR (higher values

1374

BECK ET AL.
TABLE 3. Adjusted Baseline Mean (SE) Measurements of BMD and Proximal Femur Geometry by BMI Category Underweight Healthy weight 18.524.9 0.660 (0.002) 0.782 (0.002) 0.663 (0.002) 1.89 (0.007) 3.00 (0.004) 0.838 (0.003) 13.5 (0.06) 0.663 (0.002) 3.17 (0.013) 5.04 (0.007) 2.65 (0.012) 11.1 (0.05) 1.07 (0.004) 2.87 (0.010) 2.81 (0.004) 1.49 (0.005) 3.95 (0.02) Overweight 25.029.9 0.706 (0.002)* 0.841 (0.002)* 0.708 (0.002)* 2.03 (0.007)* 3.02 (0.005) 0.908 (0.004)* 12.6 (0.07)* 0.711 (0.003)* 3.45 (0.014)* 5.10 (0.008)* 2.88 (0.013)* 10.4 (0.06)* 1.15 (0.004)* 3.10 (0.011)* 2.84 (0.004)* 1.62 (0.006)* 3.66 (0.02)* Mild obesity 3034.9 0.749 (0.004)* 0.882 (0.004)* 0.739 (0.004)* 2.14 (0.012)* 3.04 (0.008)* 0.964 (0.006)* 12.1 (0.11)* 0.748 (0.004)* 3.68 (0.022)* 5.18 (0.013)* 3.09 (0.021)* 9.99 (0.09)* 1.20 (0.007)* 3.29 (0.017)* 2.88 (0.007)* 1.73 (0.009)* 3.51 (0.03)* Moderate obesity 35.039.9 0.794 (0.008)* 0.938 (0.007)* 0.786 (0.008)* 2.26 (0.021)* 3.03 (0.014) 1.04 (0.011)* 11.2 (0.19)* 0.794 (0.008)* 3.95 (0.038)* 5.23 (0.022)* 3.31 (0.036)* 9.52 (0.16)* 1.28 (0.012)* 3.54 (0.029)* 2.92 (0.012)* 1.87 (0.016)* 3.30 (0.06)* Extreme obesity 40 0.823 (0.010)* 0.954 (0.009)* 0.821 (0.010)* 2.39 (0.027)* 3.07 (0.018)* 1.15 (0.014)* 10.9 (0.25)* 0.824 (0.010)* 4.11 (0.049)* 5.27 (0.028)* 3.50 (0.046)* 9.23 (0.21)* 1.33 (0.015)* 3.72 (0.037)* 2.95 (0.015)* 2.01 (0.021)* 3.20 (0.08)*

Category BMI limits (kg/m2) Conventional DXA Femoral neck BMD (g/cm2) Total hip BMD (g/cm2) HSA Narrow neck BMD (g/cm2) CSA (cm2) Outer diameter SM Buckling ratio (dimensionless) Intertrochanter BMD (g/cm2) CSA (cm2) Outer diameter (cm) SM (cm3) Buckling ratio (dimensionless) Shaft BMD (g/cm2) CSA (cm2) Outer diameter (cm) SM (cm3) Buckling ratio (dimensionless)

18.5 0.615 (0.008)* 0.720 (0.007)* 0.616 (0.008)* 1.74 (0.021)* 2.98 (0.014) 0.768 (0.011)* 14.5 (0.20)* 0.611 (0.008)* 2.91 (0.038)* 5.02 (0.022) 2.42 (0.036)* 12.2 (0.16)* 0.982 (0.012)* 2.61 (0.030)* 2.80 (0.012) 1.35 (0.017)* 4.44 (0.06)*

* p < 0.0001, p < 0.01, and p < 0.05 vs. healthy weight.

FIG. 2. The relative scaling of femur BMD with BMI for the HSA subset of NHW women from WHI-OS after adjustment for signicant covariates and normalization to mean of healthy weight (18.524.9 kg/m2). All means are signicant vs. healthy weight.

are bad) is similar in all three HSA regions, although the low absolute values in the shaft would make buckling irrelevant at that site. Signicantly higher BRs in underweight women would suggest that strength of their femurs are more likely to be compromised by local buckling, whereas lower BRs in heavier women would suggest that local buckling would be less likely than in healthy weight.

Fracture incidence
Using data from NHW women in the entire WHI-OS, fracture incidence is plotted by BMI category in Fig. 5.

Compared with healthy weight women, fracture incidence at central body sites including the hip was 54% more likely in underweight women; rates of these fractures decline signicantly with BMI and were 40% less likely in the extremely obese. Underweight women suffered hip fractures at twice the rate of healthy weight, but the rate in the extremely obese was less than one half of healthy weight. Upper extremity fracture rates did not vary with BMI. There was trend of increasing of lower extremity (distal to hip) with increasing BMI, but differences in fracture rates were signicantly greater than healthy weight only in overweight (BMI = 2529 kg/m2).

OBESITY AND GEOMETRY OF THE PROXIMAL FEMUR IN WHI

1375

FIG. 3. The relative scaling of narrow-neck BMD, bone CSA, and SM after adjustment for signicant covariates and normalization to mean of healthy weight in an HSA subset of NHW women from WHI-OS: (A) adjusted means, (B) adjusted means expressed relative to body weight, and (C) adjusted means expressed relative to total body bonefree lean mass (note expended vertical scale for clarity). All parameters are signicantly different from healthy weight in A and B. In C, *p < 0.05, **p < 0.01, and ***p < 0.001 vs. healthy weight.

FIG. 4. The relative scaling of (A) femur OD (note expended vertical scale for clarity) and (B) estimated BR at the narrow-neck, intertrochanter, and shaft regions of the proximal femur. BMI category mean values are normalized to mean value of healthy weight and adjusted for signicant covariates in the HSA subset of NHW women from WHI-OS. Except for OD in underweight, all parameters are signicant vs. healthy weight.

The general pattern of fracture incidence in the much smaller bone density cohort mirrored that of the whole WHI, but most differences between BMI groups did not reach signicance (data not shown). Compared with NHW in the entire WHI-OS, the bone density cohort was slightly older (64.2 versus 63.9 yr, p = 0.02), more likely to have suffered a fracture after age 50 (20% versus 18.2%, p = 0.007), more likely to suffer an incident fracture of any bone (2.2% versus 2.0%, p < 0.0001), and fewer had physical function scores >90 in the 10-item Medical

Outcomes Study Scale (33.7% versus 40.0%, p < 0.0001). There were no differences in BMI, history of osteoporosis, prior fracture, incident hip fractures, and number of reported falls in the year before enrollment (data not shown).

DISCUSSION
In this cross-sectional analysis of baseline data from 4642 NHW women between 50 and 79 yr of age from the bone density subset of the WHI-OS, we examined how femur

1376

BECK ET AL.

FIG. 5. Relative fracture incidence in cases per 1000 person-years in NHW women in WHI-OS by BMI category after adjustment for age, hormone use, and diabetes. Relative incidence (95% CI) is plotted for hip alone, central body including hip + spine, pelvis, ribs, sternum, clavicles and scapula, lower extremity (femur shaft and all distal sites), and upper extremity (humerus and all distal sites). *p < 0.05 vs. healthy weight.

BMD and geometry varied with BMI. Our purpose was to evaluate whether obesity strengthens the geometry of the proximal femur as commonly believed. We found that, in absolute terms, heavier women do have stronger femurs, but the increment in strength is not commensurate with their higher weight. This would suggest that heavier women should be more likely to suffer hip fractures, but rates of incident hip fractures among 78,013 NHW women in the entire WHI-OS during 8 2.6 (SD) yr of follow-up tended to decline with higher BMI.

Scaling of femur BMD and geometry with BMI

Femur BMD, CSA, and SM do get larger with greater BMI; the effect is proportionate to total body lean mass and not to total fat or to total body mass. For example, extremely obese women were 88% heavier on average than healthy weight women. Their bodies contained 38% more lean mass but 167% more fat. Their difference in NN region BMD, CSA, and SM of 25%, 26%, and 37% higher, respectively was more in line with the lean mass difference, especially for SM. Overall, the average ratio of SM to lean mass varied <1.4% from healthy weight through extreme obesity. It was 3% lower than healthy weight in underweight women (p < 0.007) but essentially the same relative difference as in BMD and CSA per unit lean mass. Lean mass is predominantly muscle and that femur geometry scales in proportion to lean mass is consistent with the Frost mechanostat(21) and with theories that bone adapts primarily to dynamic muscle loads. In physical activity, muscle forces generally exceed the action force because muscles act on bones at a mechanical disadvantage through a short-arm lever. For example, the hip abductor forces required to keep the pelvis level in a one-legged stance are several times body weight.(22) Although total body lean mass is a relatively crude index of the loading forces to which bones adapt, similar relationships to SM have been shown in children,(5) young adults,(23) middle-aged to older men,(24) and in another study of postmenopausal women.(7) There remains considerable controversy regarding whether fat mass has a direct and positive effects on bones, presumably mediated by fat-derived hormones. We do not attempt to resolve this controversy; the reader is directed

to the review by Zhao et al.(2) for an overview of the literature on this topic. Whereas lean mass dominated effects on femur geometry across the extremes of BMI in our observations, multiple linear regressions (Table 1) did show signicant, independent positive correlations of fat mass with BMD and geometry. It is worth noting that fat mass has a positive effect on BMC, but the effect on region area is negative. Because BMD is equivalent to BMC/area, both effects increase BMD, but the bone mass effect (BMC) is overestimated by looking only at BMD. Although still signicant, the mechanical implications of fat effects are much less evident in the SM. These observations may be worthy of further exploration in future research.(25)

Fractures and body weight

The observation that femur BMD and geometry get smaller in proportion to body weight in the obese would seem disadvantageous, because traumatic impact forces increase in proportion to body weight. Greater impact forces should increase incidence of hip fractures, but the opposite was observed here (Fig. 5) and in the meta-analysis of DeLaet et al.(3) Indeed hip and other central body fractures trended downward with increasing BMI, suggesting that the strength disadvantage versus weight is more than offset by thicker soft tissue padding that reduces the force transmitted to the bone within. A secondary factor may be the lower BR values in the obese, which should mean that buckling failure is less likely (i.e., femur strength is better characterized by SM and CSA than would be the case in lower BMI individuals where buckling is more likely). In the obese, there is less tissue padding at more distal weight-bearing sites, and torsional fractures where fat padding would have little effect are more common. If the geometry at those sites scales similarly with BMI as at the proximal femur, obesity should confer little or no advantage. This is consistent with our observation that fracture incidence distal to the hip did not decline with increased BMI and actually trended upward, although only reaching signicance in the overweight. Fracture incidence is a complex phenomenon. Reduced activity was associated with diminished physical function;

OBESITY AND GEOMETRY OF THE PROXIMAL FEMUR IN WHI proportions of women scoring high in physical function steadily declined with increasing BMI, and they may be less steady on their feet. Interestingly extremely obese women reported signicantly more falls in the year before study entry and were more likely to report a prevalent fracture of any bone. Even if fractures are not more common, the analysis of U.S. Medical Expenditure Panel Surveys by Finkelstein et al.(26) showed 15% higher musculoskeletal injuries requiring medical intervention in overweight and 48% higher in extremely obese individuals.

1377

underweight and that this is probably not the case in the overweight.

Method limitations
The HSA method has been used in a number of large epidemiologic studies, and its limitations are reasonably well understood. Our work shows that there are important geometric differences in the femur associated with obesity, but the method measures dimensions from a 2D projected image; thus, the main source of imprecision and systematic error is in how the femur is positioned for imaging. BMD is more tolerant of small inconsistencies in femur rotational position than geometry; exceptional care is needed to position the femur consistently in dissimilar patients.(19) There are systematic uncertainties in DXA data that are inuenced by tissue composition as shown by Bolotin et al.(34); these may inuence geometry measurements in obese subjects, although this has not been studied. DXA scan image quality tends to worsen with increasing body thickness so that geometry precision degrades on heavier patients. Future technical developments may improve precision of DXA derived geometry to levels adequate for clinical use on individual patients, but the current HSA method is best suited for the research setting where imprecision can be overcome with large numbers of subjects. The cross-sectional sample used in this study limits the conclusions one might draw from changes in body weight, although in an earlier longitudinal study, we previously showed that femur SM in postmenopausal women changed in an appropriate direction with increasing, decreasing, and static body weight, even though BMD declined in all groups.(7)

Signicant covariates
Our statistical models of statistical models of BMI associations with femur BMD, geometry, and fracture incidence included reported physical activity score, use of hormone therapy, and a diagnosis of diabetes. Hormone therapy has shown to reduce fragility fractures and has known effects on BMD and geometry,(27) and heavier women were less likely to have used hormones. Obesity is a known risk factor for diabetes; heavier women in this study were two to eight times more likely to be diagnosed with the disease, which carries a risk of increased fractures(28) and has effects on BMD.(29) Before adjusting fracture incidence for covariates, lower limb fracture rates were signicantly higher in heavier than healthy weight women except in moderate obesity (data not shown). We suspect that this may be associated with diabetes because other explanations were less likely; activity effects were small, and lower hormone use in heavier women should increase rather than decrease fracture incidence. Effects of diabetes on femur geometry in the WHI are being studied separately.

Femur BMD, geometry, and hip fracture rates in the underweight

Our primary emphasis was on the effects of obesity, but observations in the underweight are worthy of comment. Soft tissue padding should have little moderating effects on fall impact in underweight women, but this does not seem to explain why they suffered hip fractures at twice the rate of healthy weight women. In absolute terms, adjusted BMD, CSA, and SM are lower in the underweight but only by 78% (Fig. 2A). These parameters were also within 3% of healthy weight when scaled to lean mass, suggesting no major decit in adaption to muscle force. Whereas other nonbone factors may be important, it would still seem that fracture risk is inadequately explained by BMD, CSA, and SM. The observation that BR in the underweight was 8% and 10% higher (more susceptible) at the NN and IT regions than healthy weight may provide an explanation. Although BR can only be crudely estimated from DXA data, studies show elevated values in hip fracture cases.(3033) Effectively, high BR values suggest that strength is overestimated by SM, because bending may cause local folding of thin cortices on surfaces subjected to high compressive stresses. Unfortunately, there is insufcient information in a DXA scan to compute a strength estimate incorporating local buckling. These data can only suggest that geometry is compromised by buckling instability in the

Comment on generalizability of results

The WHI-OS is one of the largest observational studies of postmenopausal women in the world, although participants are believed to be healthier than the U.S. population as a whole; for example, hip fracture rates were one half the expected rate of similar age groups.(35) This study was restricted to women self-reported as NHW because of ethnic differences in both BMI and in fracture rates, especially among women of African heritage, which we are studying separately. The subset of NHW women in the bone density cohort was on average slightly older, fewer had physical function scores >90, and more had prevalent fractures than those in the whole WHI population. However, they did not differ in BMI distribution or in any other remarkable respect. We expect that the conclusions on patterns of obesity effects on femur geometry should be consistent with the WHI-OS.

Summary and conclusions

We conclude that, in absolute terms, femur BMD and geometric strength are greater with overweight in postmenopausal women, but they vary in proportion to lean (mostly muscle) mass and not to body weight or fat mass. In effect, femur strength is reduced relative to body weight in the obese but, although traumatic forces increase in

1378 proportion to body weight and that the most obese women reported more falls, they had fewer fractures at hip and other central body sites. The most logical explanation is that greater soft tissue padding at those sites more than compensates for the greater impact forces resulting from falls in the obese. Their thicker cortices should also reduce susceptibility to local buckling compared with lower weight women. The obesity advantage did not extend to fractures at extremity skeletal sites where padding is less important; fracture rates were not signicantly affected by BMI.

BECK ET AL.
12. Ainsworth BE, Haskell WL, Leon AS, Jacobs DR Jr, Montoye HJ, Sallis JF, Paffenbarger RS Jr 1993 Compendium of physical activities: Classication of energy costs of human physical activities. Med Sci Sports Exerc 25:7180. 13. Ainsworth BE, Leon AS, Richardson MT, Jacobs DR, Paffenbarger RS Jr 1993 Accuracy of the College Alumnus Physical Activity Questionnaire. J Clin Epidemiol 46:1403 1411. 14. Ware JE Jr, Sherbourne CD 1992 The MOS 36-item shortform health survey (SF-36). I. Conceptual framework and item selection. Med Care 30:473483. 15. Martin RB, Burr DB 1984 Non-invasive measurement of long bone cross-sectional moment of inertia by photon absorptiometry. J Biomech 17:195201. 16. Beck TJ, Looker AC, Ruff CB, Sievanen H, Wahner HW 2000 Structural trends in the aging femoral neck and proximal shaft: Analysis of the Third National Health and Nutrition Examination Survey dual-energy X-ray absorptiometry data. J Bone Miner Res 15:22972304. 17. Mayhew P, Thomas C, Clement J, Loveridge N, Beck T, Boneld W, Burgoyne C, Reeve J 2005 Relation between age, femoral neck cortical stability, and hip fracture risk. Lancet 366:129135. 18. Young W 1989 Elastic stability formulas for stress and strain. In: H C, S T (eds.) Roarks Formulas for Stress and Strain, 6th ed. McGraw-Hill, New York, NY, USA, pp. 688. 19. Khoo BC, Beck TJ, Qiao QH, Parakh P, Semanick L, Prince RL, Singer KP, Price RI 2005 In vivo short-term precision of hip structure analysis variables in comparison with bone mineral density using paired dual-energy X-ray absorptiometry scans from multi-center clinical trials. Bone 37:112121. 20. Mayhew PM, Thomas CD, Clement JG, Loveridge N, Beck TJ, Boneld W, Burgoyne CJ, Reeve J 2005 Relation between age, femoral neck cortical stability, and hip fracture risk. Lancet 366:129135. 21. Lanyon LE, Rubin CT 1984 Static vs dynamic loads as an inuence on bone remodelling. J Biomech 17:897905. 22. Burr DB 1997 Muscle strength, bone mass, and age-related bone loss. J Bone Miner Res 12:15471551. 23. Petit MA, Beck TJ, Hughes JM, Lin HM, Bentley C, Lloyd T 2008 Proximal femur mechanical adaptation to weight gain in late adolescence: A six-year longitudinal study. J Bone Miner Res 23:180188. 24. Travison TG, Araujo AB, Esche GR, Beck TJ, McKinlay JB 2008 Lean and not fat mass is associated with male proximal femur strength. J Bone Miner Res 23:189198. 25. Ruff CB 2000 Body size, body shape, and long bone strength in modern humans. J Hum Evol 38:269290. 26. Finkelstein EA, Chen H, Prabhu M, Trogdon JG, Corso PS 2007 The relationship between obesity and injuries among U.S. adults. Am J Health Promot 21:460468. 27. Greenspan S, Beck T, Resnick N, Bhattacharya R, Parker R 2005 Effect of hormone replacement, alendronate, or combination therapy on hip structural geometry: A 3-year, doubleblind, placebo-controlled clinical trial. J Bone Miner Res 20:15251532. 28. Bonds DE, Larson JC, Schwartz AV, Strotmeyer ES, Robbins J, Rodriguez BL, Johnson KC, Margolis KL 2006 Risk of fracture in women with type 2 diabetes: The Womens Health Initiative Observational Study. J Clin Endocrinol Metab 91:34043410. 29. Rakel A, Sheehy O, Rahme E, Lelorier J 2008 Osteoporosis among patients with type 1 and type 2 diabetes. Diabetes Metab 34:193205. 30. Kaptoge S, Beck TJ, Reeve J, Stone KL, Hillier TA, Cauley J, Cummings SR 2008 Prediction of incident hip fracture risk by femur geometry variables measured by hip structural analysis in the Study of Osteoporotic Fractures. J Bone Miner Res 23:18921890. 31. Rivadeneira F, Zillikens MC, De Laet CE, Hofman A, Uitterlinden AG, Beck TJ, Pols HA 2007 Femoral neck BMD is a strong predictor of hip fracture susceptibility in elderly

ACKNOWLEDGMENTS
This work is the result of WHI Ancillary Study 153, which was supported by NIH Grant NIAMS/NIH: R01 AR049411. The WHI program is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, U.S. Department of Health and Human Services through Contracts N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-32119, 32122, 42107-26, 42129-32, and 44221.

REFERENCES
1. Wyatt HR 2003 The prevalence of obesity. Prim Care 30:267 279. 2. Zhao LJ, Jiang H, Papasian CJ, Maulik D, Drees B, Hamilton J, Deng HW 2008 Correlation of obesity and osteoporosis: Effect of fat mass on the determination of osteoporosis. J Bone Miner Res 23:1729. 3. De Laet C, Kanis JA, Oden A, Johanson H, Johnell O, Delmas P, Eisman JA, Kroger H, Fujiwara S, Garnero P, McCloskey EV, Mellstrom D, Melton LJ III, Meunier PJ, Pols HA, Reeve J, Silman A, Tenenhouse A 2005 Body mass index as a predictor of fracture risk: A meta-analysis. Osteoporos Int 16: 13301338. 4. Looker A, Flegal K, Melton LR III 2007 Impact of increased overweight on the projected prevalence of osteoporosis in older women. Osteoporos Int 18:307313. 5. Petit M, Beck T, Shults J, Zemel B, Foster B, Leonard M 2005 Proximal femur bone geometry is appropriately adapted to lean mass in overweight children and adolescents. Bone 36:568576. 6. Petit MA, Beck TJ, Lin HM, Bentley C, Legro RS, Lloyd T 2004 Femoral bone structural geometry adapts to mechanical loading and is inuenced by sex steroids: The Penn State Young Womens Health Study. Bone 35:750759. 7. Beck TJ, Oreskovic TL, Stone KL, Ruff CB, Ensrud K, Nevitt MC, Genant HK, Cummings SR 2001 Structural adaptation to changing skeletal load in the progression toward hip fragility: The study of osteoporotic fractures. J Bone Miner Res 16:11081119. 8. Ruff CB 2000 Body size, body shape, and long bone strength in modern humans. J Hum Evol 38:269290. 9. Bouxsein ML, Szulc P, Munoz F, Thrall E, Sornay-Rendu E, Delmas PD 2007 Contribution of trochanteric soft tissues to fall force estimates, the factor of risk, and prediction of hip fracture risk. J Bone Miner Res 22:825831. 10. Robinovitch SN, Hayes WC, McMahon TA 1991 Prediction of femoral impact forces in falls on the hip. J Biomech Eng 113:366374. 11. Chen Z, Kooperberg C, Pettinger MB, Bassford T, Cauley JA, LaCroix AZ, Lewis CE, Kipersztok S, Borne C, Jackson RD 2004 Validity of self-report for fractures among a multiethnic cohort of postmenopausal women: Results from the Womens Health Initiative observational study and clinical trials. Menopause 11:264274.

OBESITY AND GEOMETRY OF THE PROXIMAL FEMUR IN WHI

men and women because it detects cortical bone instability: The Rotterdam Study. J Bone Miner Res 22:17811790. Duan Y, Beck TJ, Wang XF, Seeman E 2003 Structural and biomechanical basis of sexual dimorphism in femoral neck fragility has its origins in growth and aging. J Bone Miner Res 18:17661774. Gnudi S, Sitta E, Fiumi N 2007 Bone density and geometry in assessing hip fracture risk in post-menopausal women. Br J Radiol 80:893897. Bolotin HH, Sevanen H, Grashuis JL 2003 Patientspecic DXA bone mineral density inaccuracies: Quantitative effects of nonuniform extraosseous fat distributions. J Bone Miner Res 18:10201027. Cauley JA, Robbins J, Chen Z, Cummings SR, Jackson RD, LaCroix AZ, LeBoff M, Lewis CE, McGowan J, Neuner J, Pettinger M, Stefanick ML, Wactawski-Wende J, Watts NB 2003 Effects of estrogen plus progestin on risk of fracture and bone mineral density: The Womens Health Initiative randomized trial. JAMA 290:17291738.

1379

32.

33. 34.

35.

Received in original form August 22, 2008; revised form December 10, 2008; accepted March 11, 2009.

APPENDIX 1: WHI INVESTIGATORS

Program Ofce (National Heart, Lung, and Blood Institute, Bethesda, MD, USA): Elizabeth Nabel, Jacques Rossouw, Shari Ludlam, Joan McGowan, Leslie Ford, and Nancy Geller. Clinical Coordinating Center: (Fred Hutchinson Cancer Research Center, Seattle, WA, USA) Ross Prentice, Garnet Anderson, Andrea LaCroix, Charles L. Kooperberg, Ruth E. Patterson, Anne McTiernan; (Medical Research Labs, Highland Heights, KY, USA) Evan Stein; (University of California at San Francisco, San Francisco, CA, USA) Steven Cummings. Clinical Centers: (Albert Einstein College of Medicine, Bronx, NY, USA) Sylvia Wassertheil-Smoller; (Baylor College of Medicine, Houston, TX, USA) Aleksandar Rajkovic; (Brigham and Womens Hospital, Harvard Medical School, Boston, MA, USA) JoAnn E. Manson; (Brown University, Providence, RI, USA) Charles B. Eaton; (Emory University, Atlanta, GA, USA) Lawrence Phillips; (Fred Hutchinson Cancer Research Center, Seattle,

WA, USA) Shirley Beresford; (George Washington University Medical Center, Washington, DC, USA) Lisa Martin; (Los Angeles Biomedical Research Institute at HarborUCLA Medical Center, Torrance, CA, USA) Rowan Chlebowski; (Kaiser Permanente Center for Health Research, Portland, OR, USA) Yvonne Michael; (Kaiser Permanente Division of Research, Oakland, CA, USA) Bette Caan; (Medical College of Wisconsin, Milwaukee, WI, USA) Jane Morley Kotchen; (MedStar Research Institute/Howard University, Washington, DC, USA) Barbara V. Howard; (Northwestern University, Chicago/Evanston, IL, USA) Linda Van Horn; (Rush Medical Center, Chicago, IL, USA) Henry Black; (Stanford Prevention Research Center, Stanford, CA, USA) Marcia L. Stefanick; (State University of New York at Stony Brook, Stony Brook, NY, USA) Dorothy Lane; (The Ohio State University, Columbus, OH, USA) Rebecca Jackson; (University of Alabama at Birmingham, Birmingham, AL, USA) Cora E. Lewis; (University of Arizona, Tucson/Phoenix, AZ, USA) Cynthia A Thomson; (University at Buffalo, Buffalo, NY, USA) Jean Wactawski-Wende; (University of California at Davis, Sacramento, CA, USA) John Robbins; (University of California at Irvine, Irvine, CA, USA) F. Allan Hubbell; (University of California at Los Angeles, Los Angeles, CA, USA) Lauren Nathan; (University of California at San Diego, LaJolla/Chula Vista, CA, USA) Robert D. Langer; (University of Cincinnati, Cincinnati, OH, USA) Margery Gass; (University of Florida, Gainesville/ Jacksonville, FL, USA) Marian Limacher; (University of Hawaii, Honolulu, HI, USA) J. David Curb; (University of Iowa, Iowa City/Davenport, IA, USA) Robert Wallace; (University of Massachusetts/Fallon Clinic, Worcester, MA, USA) Judith Ockene; (University of Medicine and Dentistry of New Jersey, Newark, NJ, USA) Norman Lasser; (University of Miami, Miami, FL, USA) Mary Jo OSullivan; (University of Minnesota, Minneapolis, MN, USA) Karen Margolis; (University of Nevada, Reno, NV, USA) Robert Brunner; (University of North Carolina, Chapel Hill, NC, USA) Gerardo Heiss; (University of Pittsburgh, Pittsburgh, PA, USA) Lewis Kuller; (University of Tennessee Health Science Center, Memphis, TN, USA) Karen C. Johnson; (University of Texas Health Science Center, San Antonio, TX, USA) Robert Brzyski; (University of Wisconsin, Madison, WI, USA) Gloria E. Sarto; (Wake Forest University School of Medicine, WinstonSalem, NC, USA) Mara Vitolins; (Wayne State University School of Medicine/Hutzel Hospital, Detroit, MI, USA) Michael Simon.