Our findings regarding differences in mean BMD for the total body and selected subregions in men versus women or in NHB versus NHW agree with previous studies [2
], and are also consistent with fracture differences observed for these groups [21
]. However, BMD values among MA in NHANES 1999–2004 were lower than in NHW of the same age and sex at the different skeletal sites, which contrasts with findings for the proximal femur from NHANES III [19
]. A comparison of selected relevant demographic, acculturation, and body size variables of Mexican Americans in NHANES III versus NHANES 1999–2004 did not identify a clear explanation for this discrepancy. BMD differences between Hispanics and whites (when not adjusted for body size) have also varied in direction in community-based studies [2
] and, in some cases, have varied by skeletal site within studies. For example, Araujo et al [2
] reported lower femur BMD in older white men compared to Hispanic men, but found no difference in BMD between these groups at the total body, lumbar spine, or forearm. In contrast, Morton et al [4
] reported similar unadjusted BMD values in Hispanic and white women at the hip but lower values in Hispanic women at the lumbar spine and total body. Finally, Taaffe et al [5
] reported similar BMD in whites and Hispanics at the hip, spine, or total body prior to adjusting for height.
The differences in BMD between age decades in older adults at the different skeletal sites were greater in women than men, which is consistent with longitudinal studies that have documented greater BMD loss with age in women than in men [22
]. Within sex, we did not find strong evidence for different age patterns in BMD between race/ethnic groups. This agrees with findings reported previously for the proximal femur from NHANES III and from other cross-sectional studies [2
]. There are few prospective data on ethnic differences in BMD change with age, and results have varied somewhat. For example, two longitudinal studies in elderly men and women showed lower rates of BMD loss from the hip in blacks when compared to Caucasians [24
]. A third longitudinal study suggested that ethnic differences in the rate of loss varied by age and skeletal site, with white women having faster loss at the forearm than blacks at menopause but not in older age groups, and no difference in spine BMD loss between races regardless of age [26
]. To date, longitudinal data by race for the skeletal subregions examined in this study have not been published, so it is not known whether these sites show differential BMD loss by age between whites and blacks. We are also unaware of comparable longitudinal data comparing rates of BMD loss in Hispanics with other groups for any skeletal site. Each of the three longitudinal studies comparing BMD loss rates of blacks versus whites had some limitations. Depending on the study, the limitations included a relatively short time period (1.5–4 years) between measurements, a small sample size, use of different densitometers for measurements at different time points, or use of different time intervals between measures in the different race/ethnic groups. On the other hand, the cross-sectional age patterns presented here can suffer from cohort effects, and they have been shown to overestimate loss from some skeletal sites and underestimate loss from other sites when compared to longitudinal rates of BMD loss at those sites [22
]. Thus, the cross-sectional age patterns by race/ethnicity seen in the present study should be interpreted with caution and need to be confirmed with longitudinal data.
The strength of the cross-sectional relationship between age and BMD in older adults differed among the skeletal subregions examined in the present study. Of particular interest is that the relationship strength tended to follow degree of weight-bearing. Specifically, the strongest relationships were observed for the pelvis, which differed from that seen at the leg as well as the arm in both sexes. The relationship at the right leg also differed from that seen at the arm in women, although not in men. Longitudinal data describing BMD loss at the pelvis and leg could not be located for comparison, but other cross-sectional studies have also reported greater decreases in BMD with age at the pelvis relative to other skeletal subregions in both sexes [3
Finding a stronger negative relationship at these weight-bearing sites is consistent with Frost’s suggestion that osteoporotic fragility is a disuse phenomenon [11
]. Melton et al [10
] found declines in habitual skeletal loading were generally accompanied by reductions in bone strength indices at several skeletal sites. Loss of loading from gravitational, weight-bearing forces could contribute to lower skeletal loads with age. Both body weight and body mass index tend to decline with age in older adults [27
], as do physical activity levels [28
]. The importance of weight-bearing in maintaining BMD at the pelvis and leg is also consistent with results from a study of BMD loss during spaceflight, as most of the total loss that occurred during the space missions examined came from these two sites [29
Changes in lean mass with age could also play a role in differences in the age-BMD relationships, since muscle forces also contribute to skeletal loads. Of particular interest is the consistency in the relative ranking of age patterns in lean mass (trunk>legs>arms) with the relative ranking of age patterns in BMD (pelvis>legs>arms) observed in the present study. These findings suggest that the loss of muscle forces on bone may be greatest in the pelvis, followed by the legs and then the arms. The stronger negative relationship between lean mass and age in the leg than arm observed in our study is consistent with results from several longitudinal studies that have reported greater loss of lean mass from the leg than arm [30
]. One of these longitudinal studies also found larger changes in trunk lean soft tissue than in the legs (−1.2% vs. −0.8% in men and −0.6% vs −0.3% in women after two years) [30
]. The differences between the trunk and legs in this longitudinal study were not tested statistically, but they appear consistent with the cross-sectional age patterns in lean mass observed in the present study.
Some caveats apply, however. For example, the trunk includes more skeletal regions than just the pelvis. In addition, trunk lean mass is not solely skeletal muscle. Magnetic resonance imaging data from a small sample of young adults suggest skeletal muscle makes up a greater proportion of trunk tissue volume than visceral organs [33
], but it is not known whether this applies to older adults. It is also important to note that the age patterns in lean mass and BMD were not completely consistent in men: the age pattern in lean mass differed significantly between the arm and leg in men, whereas the age pattern in BMD did not. Age patterns in the present study are cross-sectional, and may not accurately reflect longitudinal lean mass loss. Finally, although Melton et al [10
] found some correspondence in age-related declines in bone strength and habitual loading, none of the loading variables tested explained >41% of the bone variation in bone strength in their cross-sectional study. Thus, skeletal loading may only partially explain the loss of bone with age.
The relationship between age and BMD at the lumbar spine also differed significantly from that seen at the other three subregions examined. In men, the association was positive in direction. In women, the association was negative but weaker than at the other skeletal sites. Other cross-sectional studies have also reported either no BMD loss or apparent BMD gain in men when based on anterior-posterior spine scans [3
]. Interpretation of age patterns in lumbar spine BMD is likely confounded by changes in the spine with age that mask BMD loss [34
]. Indeed, Melton et al [22
] found substantial BMD loss at the lumbar spine with age when lateral spine scans were used to measure BMD and the same observation has been made in studies based on quantitative computed tomography (QCT) of the lumbar spine [35
Limitations of the present study include use of cross-sectional data to describe changes with age and possible confounding in the observed age patterns by bone size. BMD is only partially corrected for bone size, so we also adjusted for leg length to reduce bone size confounding. However, it is not clear that this adjustment can effectively remove bone size effects for the different skeletal subregions considered. Other limitations include the potential for nonresponse bias in the sample, and the exclusion of institutionalized people, who may have lower BMD [38
], from the survey sampling frame. Nonresponse bias due to refusal to participate in the physical examinations in NHANES is reduced by a nonresponse adjustment factor included in the calculation of the sample weights for use with examinee data. However, about 8% of the respondents age 20 years and older who came to the exam centers lacked usable total body DXA data, and this nonresponse is not addressed by the sample weight adjustments. These were mostly pregnant women; a smaller number either had highly variable imputed DXA values or were amputees. Imputed values for those with missing DXA data were used in the present study (unless highly variable) to help reduce nonresponse bias by other factors, such as increasing age or body mass index.
In summary, total body DXA data from NHANES 1999–2004 offer the opportunity to examine bone density of the total body and skeletal subregions for a wide age range of adult men and women from several race/ethnic groups. Mean BMD of the total body and selected subregions varied in expected ways for some demographic characteristics (men>women and NHB>NHW) but not in others (NHW>MA). Differences in age-related patterns in BMD also emerged for some characteristics (sex) but not others (race/ethnicity). Differences in age patterns in BMD by degree of weight-bearing in older adults were detected for the pelvis, leg and arm. In women, the relationship differed between all three sites, (e.g., pelvis > right leg > left arm). In men, the relationship differed for the pelvis when compared to the leg or arm, but not between the arm and leg. Similar patterns seen for lean mass in these three skeletal subregions suggest muscle forces may play a role in BMD differences by age. Age patterns for the lumbar spine seen in the present study are likely confounded by degenerative changes in the spine with age that mask BMD loss when standard anterior-posterior DXA scans are used. Longitudinal data are needed to confirm the age patterns seen in this study.