Changes in femur neck bone density in US adults between 1988–1994 and 2005–2008: demographic patterns and possible determinants
- First Online:
- Cite this article as:
- Looker, A.C., Melton, L.J., Borrud, L.G. et al. Osteoporos Int (2012) 23: 771. doi:10.1007/s00198-011-1623-0
- 291 Views
This analysis compares femur neck bone mineral density (FNBMD) and bone determinants in adults between National Health and Nutrition Examination Survey (NHANES) III (1988–1994) and NHANES 2005–2008. FNBMD was higher in NHANES 2005–2008 than in NHANES III, but between-survey differences varied by age, sex, and race/ethnicity. The likelihood that FNBMD has improved appears strongest for older white women.
Recent data on hip fracture incidence and femur neck osteoporosis suggest that the skeletal status of older US adults has improved since the 1990s, but the explanation for these changes remains uncertain.
The present study compares mean FNBMD of adults ages 20 years and older between the third (NHANES III, 1988–1994) and NHANES 2005–2008. Dual-energy X-ray absorptiometry systems (pencil beam in NHANES III, fan beam in NHANES 2005–2008) were used to measure hip BMD, and several bone determinants are compared between surveys to assess their potential role in explaining observed FNBMD differences.
FNBMD was higher overall in NHANES 2005–2008 than in NHANES III, but between-survey differences varied by age, sex, and race/ethnicity. Although FNBMD differences in several groups were small enough (≤3%) to be attributable to use of different dual-energy X-ray absorptiometry (DXA) systems in the two surveys, variability in size and direction of the differences does not support artifactual differences in DXA methodology as the sole explanation. Several FNBMD determinants (body size, smoking, selected bone-active medications, self-reported health status, calcium intake, and caffeine consumption) changed in a bone-improving direction in older adults, but FNBMD in older non-Hispanic white women remained significantly higher in 2005–2008 even after adjusting for DXA methodology or for the selected bone determinants.
The likelihood that FNBMD has improved appears strongest for older white women, but the reason for the improvement in this group remains unclear.
KeywordsBone densityFemoral neckGenderRace/ethnicitySecular trends
We recently reported a significant decline in the prevalence of femur neck osteoporosis in older US adults between 1988–1994 and 2005–2006 . This improvement in femoral neck bone mineral density (FNBMD) is consistent with the decline in hip fracture incidence observed in both national and community-based studies during the same time period [2–5]. Although our previous study provided valuable public health information by documenting a decline in osteoporosis prevalence in older US adults, it suffered from some important constraints. First, the focus on osteoporosis prevalence, as defined by the World Health Organization approach , limited the examination of FNBMD changes to older adults only. Second, the availability of FNBMD data for only 2 years (e.g., 2005–2006) at the time of the previous report limited our statistical power and the number of possible explanatory factors that could be examined. Specifically, the previous study examined the impact of changes in body mass index (BMI) and use of osteoporosis medications on the observed difference in osteoporosis prevalence. Both were positively associated with higher FNBMD, and both increased during the relevant time period [7–11], but neither factor appeared to explain the observed FNBMD difference observed between 1988–1994 and 2005–2006.
We undertook the present study to address these important data gaps. In this report, we examine FNBMD changes across the entire adult age range, not just ages 50 years and over, in order to provide a more complete picture of changes in FNBMD in the US population. We also evaluated changes by race/ethnicity, and we greatly increased the number of potential explanatory factors that were examined for their potential impact on the observed FNBMD differences in the population. In addition to BMI and osteoporosis drugs, changes in 12 other bone determinants (i.e., smoking, health status, dietary calcium, sodium, caffeine, alcohol intake, milk intake, personal and maternal fracture history, weight history, height, and use of drugs that may decrease FNBMD) are explored to assess their potential role in explaining the observed FNBMD differences. This more detailed look at potential explanatory factors was made possible by the release of an additional two years of FNBMD data from the National Health and Nutrition Examination Survey (NHANES) for 2007–2008. Thus, the specific objectives of the present study are: (a) to describe the differences in FNBMD between NHANES III (1988–1994) and NHANES 2005–2008 by age, sex, and race/ethnicity for adults ages 20 years and older; and (b) to identify additional bone determinants that may have changed in a manner consistent with improved FNBMD and assess their potential impact on the observed FNBMD differences between surveys.
The NHANES are conducted by the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention, to assess the health and nutritional status of large representative cross-sectional samples of the non-institutionalized, US civilian population. In NHANES III, a nationally representative sample was obtained in two 3-year cycles between 1988 and1994. Starting in 1999, data from a nationally representative sample have been collected each year, but data are released for 2-year periods to protect confidentiality and increase statistical reliability. The present study was based on data collected in NHANES III as well as NHANES 2005–2008. In each NHANES, data were collected via household interviews and standardized physical examinations conducted in specially equipped mobile examination centers [12, 13]. All procedures in both surveys were approved by the NCHS Research Ethics Review Board, and written informed consent was obtained from all subjects.
NHANES III and NHANES 2005–2008 were designed to provide reliable estimates for three race/ethnic groups (self-reported by the participants): non-Hispanic whites (NHW), non-Hispanic blacks (NHB), and Mexican Americans (MA). The analytic sample from NHANES III in the present study consists of 14,646 adults ages 20 years and older with valid FNBMD data, which represents 63% of the sample in this age range who were originally selected for the survey, 78% of the participants in this age range who were interviewed, and 88% of those who received physical examinations. The analytic sample from NHANES 2005–2008 consists of 8,220 adults ages 20 years and older with valid FNBMD data, which represents 56% of the subjects in this age range who were eligible to participate in the survey, 75% of the interviewed participants and 78% of those who were examined.
In the present study, BMD data at the femur neck were analyzed because it has been proposed as the reference skeletal site for defining osteoporosis in epidemiological studies . FNBMD was measured in NHANES 2005–2008 with Hologic QDR 4500A fan-beam densitometers (Hologic, Inc., Bedford, MA, USA) using Discovery version 12.4 software. Scanning was done in the fast mode . In NHANES III, FNBMD was measured with Hologic QDR 1000 pencil beam densitometers . Rigorous quality control (QC) programs were employed in both surveys, which included use of anthropomorphic phantoms and review of each QC and respondent scan at a central site (Department of Radiology of the University of California, San Francisco in NHANES 2005–2008, and Department of Diagnostic Radiology of the Mayo Clinic in NHANES III) [14, 15]. In both surveys, the left hip was scanned unless there was a history of previous fracture or surgery.
Potential explanatory bone determinants
Bone determinants which might potentially underlie observed trends in FNBMD between surveys were selected from risk factors that are identified in the National Osteoporosis Foundation (NOF) Guide to Prevention and Treatment of Osteoporosis  or used in the World Health Organization fracture risk assessment (FRAX®) model . Only bone determinants that were measured in a comparable manner in both surveys were used in the present study. They included the following:
Body mass index and weight at age 25 years
BMI was calculated as body weight (kilograms) divided by height (meters squared). Body weight was measured to the nearest 0.01 kg using an electronic load cell scale, and standing height was measured with a fixed stadiometer. Respondents self-reported their body weight in pounds at age 25 years.
Personal hip, wrist, or spine fracture history and maternal hip fracture history
History of hip, spine, or wrist fracture was based on a self-reported fracture at these sites that occurred after age 20 years. Maternal hip fracture history was based on self-report that the respondents’ biological mother had fractured her hip.
Cigarette smoking, high alcohol intake, and positive health status
Cigarette smokers were defined as respondents who self-reported that they currently or formerly smoked, while high alcohol users were defined as respondents who self-reported that they usually consumed three or more drinks per day when they drank alcohol. Respondents who self-reported that their health in general was excellent, very good, or good were considered to have positive health status.
Milk, calcium, sodium, and caffeine intake
Milk users were defined as respondents who reported drinking milk alone or on cereal one or more times per week during the past 30 days. Calcium, sodium, and caffeine intakes from foods and beverages were based on a single 24-h dietary recall. Respondents were also asked to report how often vitamin–mineral supplements and antacids were taken in the past month and how much was taken on each occasion. Nutrient composition of the supplements was based on the supplement’s label information. The average daily intake of calcium from supplements was calculated in the present study using the number of days the supplement was used, the amount taken per day and the serving size unit from the product label. The average daily calcium intake from supplements for each respondent was summed with their reported calcium intake from food to estimate total calcium intake.
Medication users were defined as those who were currently taking the prescription medications described below. In both surveys, respondents showed the containers for all current prescription medications to the interviewer, who recorded the name of the product. The December 2007 Multum Lexicon Drug Database (Cerner Multum Inc, Denver, CO, USA; http://www.multum.com/Lexicon.htm) was used to assign generic drug names and codes in both surveys. Medications that increase BMD were based on those included in a recent systematic review of the comparative effectiveness of treatments to prevent osteoporotic fractures . These medications were categorized into two groups: sex hormones (estrogen and testosterone) and other drugs such as bisphosphonates, selective estrogen receptor modifiers (SERMs), and other non-estrogen drugs. Medications that decrease BMD were defined as those identified in the NOF guide  and included glucocorticoids, antineoplastic drugs, anticonvulsants, barbiturates, anticoagulants, IV nutrition products, lithium, cyclosporine, and tacrolimus. The specific drugs in the subgroups of medications that increase or decrease BMD are shown in Appendix 1. Respondents who reported taking drugs that increase BMD concurrently with drugs that decrease BMD (n = 168) were defined as users of drugs that increase BMD.
Analyses were conducted with PC-SAS (Version 10.0, SAS Institute, Cary, NC, USA) and SUDAAN (Version 9.03, Research Triangle Institute, NC, USA). All analyses used sample weights and took into account the complex design of the surveys .
Unadjusted mean FNBMD by age, sex, race/ethnicity and survey were plotted. Differences in the FNBMD change between surveys by age, sex, and race/ethnicity were tested for statistical significance using first-order interaction terms (e.g., survey × sex, survey × age, survey × race/ethnicity) in linear regression models. All interactions terms were statistically significant (p < 0.05) so subsequent analyses were done separately by age, sex, and race/ethnicity.
Analyses were performed to assess the potential impact of changes in dual-energy X-ray absorptiometry (DXA) methodology between surveys. These analyses were based on results from a small number of studies which have compared BMD values measured on the same DXA systems used in NHANES III and NHANES 2005–2008 [19–23]. These studies produced varying results, so consensus on the best method to adjust BMD for potential DXA differences is lacking. As a result, two different approaches to explore the possible effect of the DXA method change were used in the present study. In the first analysis, the FNBMD value of each respondent from NHANES III was increased by 3%, since the largest discrepancy between the two scanner systems reported in published studies [19–23] was a difference of ±3% at either the femur neck or total hip. Estimates of mean DXA method-adjusted FNBMD values from NHANES III were then calculated and compared with mean FNBMD from NHANES 2005–2008 within race/ethnic and sex groups after also adjusting age. In the second analysis, we applied a linear regression equation developed previously by Cummings et al.  to compare BMD measurements collected with Hologic QDR 1000 and 4500 instruments in the Study of Osteoporotic Fracture and Osteoporotic Fractures in Men studies. This equation was used to adjust FNBMD values from NHANES III which were then compared with mean FNBMD from NHANES 2005–2008 within sex and race/ethnic group after also adjusting for age.
To identify possible determinants of observed differences in FNBMD between surveys, linear and logistic regression were used to find bone determinants that had changed significantly between surveys in a direction consistent with the positive difference in FNBMD observed between surveys in adults age 50 years and older. These analyses were limited to older adults because between-survey FNBMD differences were larger and more consistent in direction in this age range. Linear regression was used to compare FNBMD differences in older adults between surveys before and after adjusting for the risk factors identified in the previous step. A reduction in between-survey FNBMD differences after adjusting for these bone determinants would suggest that these factors may have played a role in the FNBMD differences. Co-linearity between inter-related risk factors, such as height, BMI, and weight at age 25, was addressed by ensuring that bivariate correlations were <0.60 or by comparing model results with and without including inter-related risk factors.
Additional analyses were conducted to further examine the effect of bone determinants on observed FNBMD difference by sex between surveys. These analyses were limited to NHW respondents age 50 years and older because FNBMD differences were relatively large in this race/ethnic group and estimates for some of the bone determinants in the two nonwhite groups were not statistically reliable. These analyses assessed whether between-survey differences in the bone determinants varied in the same manner as seen for between-survey FNBMD differences in the two sexes. Differences in between-survey changes in bone determinants by sex were evaluated using linear regression models that included a sex × survey interaction term. To illustrate these differences graphically, the difference in the age-adjusted mean or prevalence of the bone determinant between NHANES III and NHANES 2005–2008 was calculated and expressed as a percent of the mean or prevalence of that factor in NHANES III.
Selected bone determinants by NHANES survey period among adults age 50+ years
Mean or percent
NHANES III 1988–1994
Body weight (kg)
Weight at age 25 (lbs)
Ever smoked (%)
Positive health status (%)
Drink 3+ units alcohol (%)
Use medications that decrease BMD (%)c
Use medications that increase BMD (%)
Bisphosphonates, SERMS, other non-estrogen drugs
Mother had a hip fracture (%)
Self-reported previous hip, wrist or spine fracture (%)
Drink milk 1+ times/week (%)
Calcium from food and supplements (mg/day)
Sodium from food (mg/day)
Mean femur neck BMD by NHANES survey period among adults age 50+ years
Adjusted for age
Adjusted for age and other selected bone determinantsa
Adjusted for age and potential DXA method differenceb
Two approaches were used to adjust for the potential impact of DXA methodology differences in older adults, but because conclusions were similar in all subgroups, only results for the approach in which FNBMD values from NHANES III were increased by +3% are shown in Table 2. Although reduced in size (from 6% to 4%), the FNBMD difference in NHW women remained statistically significant after applying this adjustment. However in NHW men and MA women, FNBMD differences were no longer significant after adjusting for DXA methodology differences. Among MA men, the adjustment increased the size of the FNBMD difference between surveys (from −2% to −6%), so that mean FNBMD became significantly lower in NHANES 2005–2008 compared to NHANES III. FNBMD differences in NHB men and women did not differ significantly either before or after adjusting for DXA method differences.
The present study examined FNBMD changes over time across the entire adult age range, and thus extends our previous study of osteoporosis prevalence in older adults by providing a more complete picture of changes in the skeletal status of all US adults. FNBMD in US adults was higher overall in NHANES 2005–2008 than in NHANES III (1988–1994), consistent with the decline in hip fracture incidence reported during the same time period [2–5]. However, the extent of the difference in FNBMD between these two surveys varied noticeably by age, sex and race/ethnicity. The inconsistency was more evident in those <50 years of age, in whom between-survey FNBMD differences tended to be small and variable in direction. Furthermore, these differences in young adults all fell within the range that could be attributable to differences in DXA methodology between the surveys (e.g., ≤3%). In older adults, the mean age-adjusted FNBMD point estimates were 2–6% higher in NHANES 2005–2008 than in NHANES III in all but one group (MA men), and the increase in mean FNBMD between these two time periods was statistically significant in three of the six sex–race/ethnic groups examined (NHW of both sexes and MA women). These BMD differences, while small if considered in an individual, represent relatively large shifts in the population distribution of FNBMD.
The lack of consistency in magnitude of the observed BMD differences across all population subgroups suggests that the explanation for the observed BMD differences may be multi-factorial. Because FNBMD differences between surveys were larger and more consistent in adults age 50 years and older, we focused on this age range in subsequent analyses. Two possibilities were explored: (a) the observed increase reflected true changes in skeletal status of the population or (b) they were an artifact of DXA methodology differences between surveys.
The pattern of FNBMD differences in the various population subgroups was less uniform than might be expected if DXA methodology differences were the sole explanation, but we lack comparison data for the specific DXA systems used in the two surveys. Only a small number of studies have directly compared FNBMD measured on the DXA systems used, and they have produced varying results [19–23]. Thus, uncertainty exists regarding the need to adjust measurements for the DXA system used and, if so, the best approach to do so. In the present study, we used two approaches to explore the effect of potential DXA measurement changes on FNBMD differences. Results suggested that DXA method changes could potentially account for FNBMD differences observed among the subjects under 50 years of age and in two of the three groups of older adults in whom FNBMD increased significantly between surveys (NHW men and MA women) but not in the third group (NHW women).
The alternative possibility that the observed FNBMD increase reflects a real change in the skeletal status of the older adult population is supported by finding that several BMD determinants have changed in a direction that could result in higher FNBMD in 2005–2008: increased body size (body weight, height, BMI, and self-reported body weight at age 25 years), greater use of bisphosphonates, SERMs and other non-estrogen drugs that increase BMD, higher calcium consumption, less smoking, and lower caffeine consumption. Similar trends for several of these factors in the adult US population have been reported elsewhere [2, 24–26].
Adjusting for changes in these bone determinants between surveys had varying effects in the different subgroups of older adults, however. For example, the between-survey increase in FNBMD in older NHW men was no longer significant after adjusting for the bone determinants, which suggests that these factors played a role in improving FNBMD in this group. In contrast, controlling for the bone determinants had very little effect on the FNBMD increase between surveys in older NHW and MA women. The disparate impact of adjusting for risk factor changes by sex in NHW may have occurred because the magnitude of changes in some of these risk factors varied in these two groups. For example, mean BMI, weight at age 25 years, and smoking differed more between the two time-periods in NHW men than in NHW women.
In contrast to the other sex–race/ethnic groups, FNBMD was significantly lower in 2005–2008 among older MA men both before and after adjusting for the selected bone determinants. Several body size parameters, such as height, weight, and BMI did not increase between surveys in these men, like they did in the other groups, which may account for this paradoxical finding. However, after adjusting for the DXA method change in addition to changes in the bone determinants, FNBMD did not differ between surveys in these older MA men, so it is not clear whether BMD has truly declined or merely stayed stable in this group.
Not all relevant BMD determinants could be examined in the present study. For example, bone size was not directly measured in either survey. O’Neill et al.  and Reid et al.  reported that secular trends in hip axis length (HAL) had occurred in British or Australian women prior to the 1990s, but data on bone size-related parameters like HAL are currently not available from NHANES 2005–2008, so that possibility could not be explored. The bone area measurement from the DXA scan could not be used to examine bone size trends because bone area measurements from pencilbeam and fanbeam systems cannot be compared directly due to geometric projection effects. In addition, bone area measurements from a DXA scan are not true volumetric measures of bone. Thus, we used standing height as proxy for bone size. Standing height has increased slightly in the US population since the 1990s , which was also reflected in our study sample, but standing height may not fully capture variability in bone size. For example, Guglielmi et al.  found height was unrelated to 3-D quantitative computed tomography estimates of proximal femur bone size, and Lochmuller et al.  reported that adjusting for body height, weight, and projected femoral area did not fully correct for bone volume of excised femurs. It is also possible that cohort effects in diet or other health behaviors prior to adulthood played a role in the observed FNBMD differences, but these could not be evaluated in the present study. Yet other factors, such as physical activity, were not measured comparably in the two surveys and so could not be validly compared. However, the proportion of US adults who engaged regularly in moderate or vigorous leisure time physical activity did not change significantly between 1997 and 2006 , which suggests that physical activity is unlikely to play a major role in the FNBMD differences observed over a similar time period. Finally, measurement error in the bone determinants that could be included in the present study might also affect their ability to account for observed FNBMD differences.
In addition to the lack of a robust method to account for differences in DXA methodology, limitations of the present study include possible nonresponse bias in the estimates from both surveys. Use of sample weights in the analysis addresses this bias to some extent, since a nonresponse adjustment factor is included in their calculation. However, 12% and 22%, respectively, of the NHANES III and NHANES 2005–2008 respondents ages 20 years and older who came to the examination centers lacked valid hip DXA data, and this is not addressed by the sample weight adjustments. Results of a detailed non-response bias analysis conducted by NCHS prior to the public release of the NHANES 2005–2006 femur data suggested that additional adjustments for non-response were not necessary, however. A final limitation is the exclusion of institutionalized people, an important at-risk group for osteoporosis  from the NHANES sampling frame by design.
In summary, FNBMD was higher overall in NHANES 2005–2008 than in NHANES III (1988–1994), which suggests that hip BMD has improved in US adults. This improvement is consistent in direction with the decrease in hip fracture incidence reported among older adults in the US [2–5]. However, FNBMD differences between surveys were generally greater in older than in younger adults, and they also varied considerably by sex and race/ethnicity. The variability in the size of the BMD differences in the different population groups suggests the causes may be multifactorial. The FNBMD differences in several groups were within the magnitude that could be attributed to use of different DXA systems in the two surveys. However, the lack of uniformity in the size and direction of the FNBMD differences across the various age, sex, and race/ethnic groups, and the identification of several BMD determinants that changed in a manner consistent with improved BMD over the same time period, support the possibility that a real change in FNBMD may have occurred. This likelihood appears strongest for older NHW women, since the FNBMD difference remained statistically significant after adjusting for DXA methodology. More work is needed to identify the underlying basis for the increase in FNBMD in these older NHW women, since adjusting for selected bone determinants that had improved between surveys had little impact on the magnitude of the FNBMD difference observed in this group.
Conflicts of interest