We measured femoral neck and shaft dimensions and volumetric BMD with QCT. Relations of these measures to age were quantified in a cross-sectional study among 3358 men 65–100 years old. Relations of femoral neck dimensions and vBMD to age differed from those in the shaft, indicating that patterns of bone modeling and remodeling in the neck and shaft are distinct.
Introduction: Little is known about population variation in dimensions and volumetric BMD of the proximal femur or the relation of these measures to age among older men.
Materials and Methods: In a cross-sectional study, dimensions and volumetric BMD (vBMD) in the femoral neck and shaft were obtained from QCT scans among 3358 men 65–100 years of age in the Osteoporotic Fractures in Men cohort. Total bone size and size of the cortical and medullary compartments were measured with volumes in the femoral neck and with areas in the shaft. We quantified distributions of these measures and examined their relations to age with multivariable linear regression.
Results: Population variation in femoral neck and shaft dimensions and vBMD was substantial. In the femoral neck, total volume was minimally related to age, whereas cortical volume was 5% smaller and medullary volume was 10% larger (both p < 0.0001) in the oldest (85+ years) compared with the youngest (65–69 years) men. Across these ages, the percent of cortical bone declined from 46% to 42% (p < 0.0001). Integral and trabecular vBMD were 9% and 22% lower, whereas DXA femoral neck BMD was 4% lower, in the older men. Neck cortical vBMD was unrelated to age. In the shaft, cross-sectional area and medullary area were 9% and 22% larger, respectively, in the oldest men (both p < 0.0001), but cortical area was unchanged with age. The percent of cortical bone declined from 69% to 65% across these ages (p < 0.0001). Shaft cortical BMD was 4% lower in the older men (p < 0.0001).
Conclusions: There is substantial diversity of femoral morphology and vBMD among older U.S. men. Patterns indicative of modeling and remodeling in the femoral neck were distinct from those in the shaft. Notably, changes in periosteal and endosteal dimensions that underlie cortical thinning appear to differ in the neck and shaft.
The magnitude of periosteal apposition and endosteal resorption in the proximal femur in aging adults is an issue of long-standing interest.(1–5) Previous studies of this topic in human populations suggest that, over the course of adult life, both the femoral neck(1,6–13) and shaft(1,6–10) undergo relatively linear increases in total cross-sectional area (or width) and medullary area (or width), but a decrease in cortical area (or thickness). However, because most of these studies have evaluated femoral dimensions across a broad adult age range, it is unclear whether these patterns persist at the oldest ages. Moreover, there is some indication that the extent of periosteal apposition differs in the femoral neck and shaft. For example, reports of direct measurements on proximal femur cross-sections from cadavers suggest that periosteal apposition occurs to a lesser extent in the femoral neck than in the shaft and that there may be sex differences in these processes.(1,5,6) However, others using areal projections from absorptiometry report that the association of age with subperiosteal width, a surrogate measure of periosteal apposition, was the same in the femoral neck and shaft regardless of sex.(9) Others report no evidence for periosteal apposition in either the neck or shaft.(8) The relation of age with femoral neck volumetric BMD is also unclear. It has been reported that both total and trabecular volumetric BMD in the femoral neck decline with age.(12,13) However, cortical BMD was reduced with age in one study(13) but unchanged with age in two others.(12,14) Thus, despite decades of study in human populations, the extent of bone loss and structural drift in the femoral neck and shaft at older ages remains unclear.
Current knowledge about the relation of femoral neck and shaft dimensions and volumetric BMD (vBMD) to age is limited for several reasons. First, a variety of methods have been used to measure femoral morphology and BMD that may not produce comparable results.(1,6–14) Second, most of these studies have included few individuals ⩾65 years of age, so information on skeletal characteristics among the oldest adults, who experience the highest fracture rates, is scarce. Moreover, as others have noted,(9) the sample sizes of older adults may not have been sufficiently large to detect small changes in skeletal morphology. Third, most studies have been designed to compare men and women, so comprehensive sex-specific data are lacking. Finally, because few studies have reported on measures of bone density and dimensions simultaneously, little is known about the nature and extent of the modeling and remodeling processes occurring at advanced ages that result in periosteal apposition and endosteal resorption or how these influence bone density in the cortical and trabecular compartments.
To address these issues, we obtained QCT scans and measured BMD with DXA among U.S. men ⩾65 years of age participating in the Osteoporotic Fractures in Men (MrOS) study. At the enrollment visit, all 5995 participants received BMD scans with DXA, and a subset of 3785 were referred for 3D QCT scans. Using these cross-sectional data, we sought to quantify distributions of dimensions and vBMD of the femoral neck and shaft, determine the extent to which dimensions, volumetric, and areal BMD are correlated, and study the relation of dimensions and BMD to age among older men.
MATERIALS AND METHODS
The MrOS Study enrolled 5995 participants from March 2000 through April 2002 as described elsewhere.(15,16) Briefly, recruitment occurred at six U.S. academic medical centers in Birmingham AL, Minneapolis MN, Palo Alto CA, Pittsburgh PA, Portland OR, and San Diego CA. Eligible participants were at least 65 years of age, able to walk without assistance from another person, and had not had bilateral hip replacement surgery. All men enrolled in the MrOS cohort provided written informed consent, completed the baseline self-administered questionnaire, and attended the baseline visit at their local site at which skeletal, anthropometric, and other measures were obtained.
Information about demographic factors, physical activity, self-reported health status, smoking history, alcohol use, and physician-diagnosed medical conditions was obtained by self-reported questionnaire. Age was categorized into 5-year groups (65–69, 70–74, 75–79, 80–84, and ≥85 years). Participants were classified into mutually exclusive categories of race/ethnicity as white, black, Asian, Hispanic, or other (multirace persons) based on self-report. The physical activity scale for the elderly (PASE)(17) provided an overall physical activity score based on exercise, leisure, and occupational activities. Cigarette smoking was classified as current, past, or never, and pack-years were computed for those who ever smoked. Current alcohol consumption was computed as average drinks per week. Height (cm) was measured using a Harpenden stadiometer. Participants were weighed (kg) while wearing indoor clothing except shoes on balance beam or digital scales. Body mass index (BMI) was calculated as weight divided by the square of height in meters (kg/m2). Height and weight at age 25 were self-reported.
Areal BMD measures were obtained for the total hip and its subregions with fan-beam DXA (QDR 4500 W; Hologic). Participants were positioned and scanned by centrally certified densitometry technicians according to standardized procedures.(18) Scanners were calibrated at baseline, and quality control scans obtained daily showed no shift in scanner performance at any site during the enrollment period.
The first 650 participants enrolled at each site, and all men from minority backgrounds were referred for QCT scans of the hip and lumbar spine. Men with a history of hip replacement were ineligible for hip scans. All scans were obtained using a standardized protocol that specified scanning the pelvic region from the femoral head to 3.5 cm below the lesser trochanter at settings of 80 kVp, 280 mA, 3-mm slice thickness, and 512 × 512 matrix in spiral reconstruction mode. Scanner models used at the sites were GE Prospeed (Birmingham), GE Hispeed Advantage (Minneapolis), Phillips MX-8000 (Palo Alto), Seimans Somatom +4 (Pittsburgh), Phillips CT-Twin (April-July, 2000, 190 participants) and Toshiba Acquilion (December 2000-March 2002, 467 participants) (Portland), and Picker PQ-5000 (San Diego). Calibration standards with known hydroxyapatite concentrations (150, 75, and 0 mg/cm3; Image Analysis) were included with the participant in each scan. All scans were transferred to the University of California, San Francisco (UCSF) for central quality review. In total, 3785 participants (63.1% of the cohort) were referred for QCT scans. Of these, 101 had spine but not hip scans because of hip replacement. Of the 3684 hip scans, 112 (3.0%) were not available for processing, because they were lost or corrupted during transfer to UCSF, leaving 3572 hip scans available for processing.
QCT-derived femoral outcome measures
Image processing occurred centrally at UCSF using software specifically developed for this purpose. Image processing was performed by one of us (TFL) using published methods.(19,20) Images were calibrated from the native scanner Hounsfield units (HU) to equivalent concentration (g/cm3) of calcium hydroxyapatite contained in the calibration standard, which was contained in the field of view in all scans. This step, which has been an accepted practice for some time,(21) reduces variability in image attenuation across scanners, because all HU are scaled to materials of known density. We also evaluated the utility of applying additional adjustment factors derived from scans of a commercial torso phantom (Image Analysis), which were obtained on all machines used in our study. However, because these adjustments did little to further reduce between-scanner variability in the measures, we did not use them in our image processing protocol.
Regions of interest (ROIs) in the left proximal femur were identified in the QCT images reformatted along the neutral axis of the femoral neck. The periosteal boundary of the femur was determined with a region growing algorithm with thresholds to distinguish bone from adjacent soft tissue.(19) Using this boundary, the cross-sectional area along the neutral axis in each slice was calculated and plotted, the minimum and maximum areas were recorded, and the neck axis length between the two cross-sections measured. The femoral neck ROI was defined as the portion along the femoral neck axis bounded medially at minimum cross-sectional area and laterally to a distance of 25% of the neck axis length toward the maximal cross-sectional area.
Within the femoral neck ROI, the following measures were obtained. The cross-sectional area (cm2) was computed as the area within the periosteal boundary at the minimum cross-section. Integral volume (cm3) was computed as the total volume within the periosteal boundary. A trabecular volume was obtained by applying an erosion process to the integral volume to retain the same shape in a region fully contained within the medullary space. The cortical volume was then defined by applying a threshold of 0.35 g/cm3 to all voxels between the periosteal boundary and the outer boundary of the trabecular volume. Medullary volume was computed by subtracting the cortical volume from the integral volume. The percent cortical volume was computed as cortical volume divided by integral volume × 100%. vBMD was computed as the concentration of calcium hydroxyapatite averaged over all voxels in the integral, trabecular, and cortical volumes. Trabecular vBMD values in the negative range can occur because the voxel density is more consistent with marrow fat, which has negative HU.
The femoral shaft ROI was centered 3 cm beneath the inferior aspect of the lesser trochanter. The following measures were obtained in a 10-mm transverse section reconstructed perpendicular to the neutral axis of the shaft. The cross-sectional area was defined as the area within the periosteal boundary of the shaft. The cortical area was defined by applying a threshold of 0.35 g/cm3 to all voxels within the shaft cross-section and computing the area delineated by those voxels. Medullary area was computed by subtracting the cortical area from the cross-sectional area. The percent cortical area was computed as cortical area divided by cross-sectional area times 100%. Cortical BMD was computed as the concentration of calcium hydroxyapatite averaged from all voxels in the cortical area.
Of the initial 3572 hip scans available for processing, 133 (3.7%) failed image processing. Reasons for failure were insufficient number of images, interference from metal implants, calibration standard not visible, or unrecorded cause. Shaft measures could not be obtained in an additional 355 (9.9%) scans because the image did not extend below the lesser trochanter. Thus, analyses were based on 3358 men (94%) with complete data for the femoral neck outcome measures (cross-sectional area; integral, cortical, and medullary volumes; percent cortical volume; integral, cortical, and trabecular vBMD) and on 2986 men (83%) with complete data for shaft outcome measures (cross-sectional area, cortical area, medullary area; percent cortical area; cortical BMD). Measures of total bone size (femoral neck cross-sectional area and integral volume, and shaft cross-sectional area) served as surrogate measures of periosteal apposition. Measures of medullary size (femoral neck medullary volume and shaft medullary area) served as surrogate measures of endosteal resorption. Measures of cortical thinning were femoral neck cortical volume, percent cortical volume, shaft cortical area, and percent cortical area.
Distributions of baseline characteristics among men who were and were not referred for QCT scanning were compared with χ2 tests for contingency tables and t-tests for continuous variables. Simple correlations among the skeletal variables were estimated with Pearson correlation coefficients. We used multivariable general linear models (GLM) to estimate least square (LS) means of each femoral outcome variable in 5-year age categories. We examined several variables that could potentially confound the association of the skeletal measures with age, including height, weight, BMI, race/ethnicity, smoking history, current alcohol consumption, physical activity (expressed as the PASE score(17)), and history of selected medical conditions (stroke, Parkinson's disease, osteoporosis, rheumatoid arthritis, osteoarthritis, cancer, heart disease, and nontraumatic fracture before age 50). Our final regression model included adjustments for enrollment site, race/ethnicity, current height, and BMI. Volume measures were also additionally adjusted for the neck axis length. None of the other variables examined confounded the associations of age with the femoral outcomes and therefore were not included as covariates. Adjustments for current height and BMI were made because we observed that younger men (65–69 years) compared with the oldest men (85+ years) had greater mean height (176 ± 6.8 versus 171 ± 6.7 cm), weight (86.7 ± 14.3 versus 74.7 ± 10.4 kg), and BMI (28.1 ± 4.1 versus 25.5 ± 3.3 kg/m2) (all p < 0.0001). Adjustment for BMI is reported because the age-specific LS means of the femoral measures were similar when we replaced the term for BMI with a term for weight in the final models. Mean reported height and weight at age 25 were also significantly greater among the youngest compared with the oldest men, but adjustment for these produced results similar to those adjusted for current body size. Finally, men with self-reported history of osteoporosis (n = 111) were retained, because the LS means estimated with and without these men were virtually identical.
We examined the possibility that extreme observations biased the LS means by refitting the final regression models after trimming distributions of each outcome measure at 2 and 3 SD from the mean. In these analyses, the estimated LS means were attenuated when trimmed at either cutpoint, but the percent difference in the estimated means between the oldest and the youngest men was not substantially altered for any variable. Therefore, to provide conservative LS mean estimates, we report means estimated from distributions trimmed at 3 SD. Trimming reduced the sample sizes by 140–161 persons for the femoral neck analyses and by 125–158 persons for the shaft analyses. All analyses were conducted with SAS statistical software (SAS Institute).
Characteristics of the MrOS participants who did and did not receive QCT scans at enrollment were similar (Table 1). Compared with those without QCT scans, those with were slightly younger on average (0.3 years), more likely to be from a minority population, and less likely to have a previous diagnosis of hypertension. Because the study design specified that all nonwhite men be recruited for QCT scans, greater racial diversity in the QCT cohort is expected.
Table Table 1.. Characteristics of Participants Who Did Not and Did Undergo QCT Scans at Enrollment, the MrOS Study
We observed substantial variation in all dimensions and vBMD measures (Figs. 1–3). All femoral neck dimensions (Fig. 1) were approximately normally distributed, with some tendency toward skew to higher values. The femoral neck vBMD measures also exhibited substantial variation and were approximately normally distributed (Fig. 2). The distributions of shaft dimensions (Fig. 3) and cortical vBMD (Fig. 2) were approximately normal, showed substantial variation, and tended to be skewed toward higher values. Shaft data were missing for 17% of participants, but they were missing at random with regard to age. The mean age among those without shaft data (73.4 ± 5.7 years) and among those with (73.5 ± 5.9 years) was not significantly different (p = 0.42). For all femoral neck and shaft measures, the site-specific means and SDs were similar to those for the entire study population (data not shown).
Correlation analyses (Table 2) suggest that the QCT-derived measures provide unique information about dimensions and vBMD in the femoral neck and in the shaft. Measures of total bone size in the femoral neck (cross-sectional area and integral volume) were modestly correlated (r = 0.49). Correlations between the size of the cortical and medullary compartments were modest (r = 0.47) in the femoral neck and weak (r = −0.11) in the shaft. Nearly all the vBMD measures in the neck and shaft were inversely correlated with measures of total bone size and medullary size but were positively correlated with measures of cortical size. However, femoral neck cortical vBMD was uncorrelated with cortical volume (r = 0.08), and shaft cortical vBMD was only modestly correlated with shaft cortical area (r = 0.19). The correlation between femoral neck cortical and trabecular vBMD was weak (r = 0.10). Femoral neck BMD (from DXA) was uncorrelated with measures of femoral neck size, positively correlated with cortical volume, inversely correlated with medullary volume, and strongly correlated with the integral and trabecular vBMD measures.
Table Table 2.. Correlations Among Dimensions and BMD in the Femoral Neck and in the Femoral Shaft. The MrOS Study
Variation in femoral neck dimensions and BMD according to age is shown in Table 3. Crude age-specific means for cross-sectional area and integral volume were smaller in the oldest men compared with the youngest men. However, the adjusted means differed by ∼2% among the oldest compared with the youngest men, but the difference was statistically significant only for integral volume (p = 0.02). In contrast, adjusted mean cortical volume was 5.4% lower, and adjusted mean medullary volume 10% greater, among the oldest men compared with the youngest. Consistent with the foregoing observations, the percent cortical volume was significantly lower in the oldest compared with the youngest men. Adjusted mean integral and trabecular vBMD were, respectively, 9% and 22% lower in the oldest compared with the youngest men, whereas cortical vBMD did not vary significantly by age. Mean areal femoral neck BMD also was significantly lower in the oldest compared with the youngest men, but the difference in means at 4% was smaller than the difference observed for integral vBMD. Although the observed associations of the femoral neck measures with age tended to be highly statistically significant, age explained a small amount of the variance in these measures, with the partial r2 for age ranging from 0.1% to 2.0%.
Table Table 3.. Variation in Femoral Neck Dimensions and Bmd according to Age. The MrOS Study
Variation in femoral shaft dimensions and cortical vBMD according to age is shown in Table 4. Adjusted mean cross-sectional area was 9% greater among the oldest compared with the youngest men. Mean cortical area did not vary significantly by age. However, adjusted mean medullary area was nearly 22% larger among the oldest compared with the youngest men. The percent of cortical bone was significantly lower in the older compared with the younger men. Adjusted mean cortical vBMD declined significantly with age. In contrast to observations for the femoral neck, age explained about 2% of variation in most of the shaft measures.
Table Table 4.. Variation in Femoral Shaft Dimensions and BMD according to Age. The MrOS Study
To evaluate the potential that the different scanners affected our results, we conducted site-specific linear regression analyses. Within each site, the estimated means for each femoral measure in the 5-year age categories were all of similar magnitude and in the same direction as those reported for the entire cohort (data not shown). We noted no systematic deviations from the overall age association in any site, indicating that potential scanner differences produced no obvious bias in the results reported for the cohort overall.
In this cross-sectional study among U.S. men ⩾65 years of age, we observed substantial population variation in dimensions and vBMD of the proximal femur. Several of these measures were associated with age, independent of body size and race/ethnicity. Our data suggest that, with advancing age, femoral neck structure is affected by minimal periosteal apposition and considerable endosteal resorption that results in marked cortical thinning and reduced cortical volume. Integral and trabecular vBMD were substantially lower in older compared with younger men, whereas cortical vBMD was unchanged. Femoral shaft structure likewise was affected by periosteal apposition and endosteal resorption, which resulted in cortical thinning without a reduction in cortical area. Shaft cortical vBMD also declined. Thus, age-related bone loss and structural drift seem to occur through periosteal apposition and endosteal resorption in the proximal femur, but to differing degrees depending on anatomic location. These data provide fresh insight into features of femoral morphology at advanced ages that may influence biomechanical strength and fracture risk.
The primary strength of the MrOS study is that it was specifically designed to investigate skeletal characteristics in a diverse cohort of older men. The large size of the MrOS QCT cohort permitted us to characterize femoral dimensions and vBMD among men specifically in their seventh through ninth decades, the ages when fracture rates are highest. The extensive array of skeletal measures in the cohort permitted us to compare femoral characteristics from QCT and DXA. Men in the MrOS cohort are geographically and racially diverse, generally healthy, and well educated. Distributions of total hip and femoral neck BMD measured with DXA in the MrOS cohort and among similarly aged men in the Third National Health and Nutritional Examination Survey (NHANES III) are comparable, although the MrOS participants are slightly heavier and have mean BMD that is 2–8% higher depending on age.(15) Thus, results from this study are likely to be broadly applicable to similarly aged, generally healthy U.S. men.
These cross-sectional data cannot be interpreted confidently to represent actual change with age. Like others,(7,13) we observed evidence of birth cohort effects, such that the oldest men were shorter and weighed less than the youngest men in both early and later adult life. Cohort effects may have influenced peak femoral neck size, because in crude analyses, femoral neck cross-sectional area and integral volume were on average the smallest among the oldest men. Thus, it is possible that secular differences in peak bone size could have obscured our ability to detect evidence for periosteal apposition in the femoral neck. This would occur if among the oldest men the femoral neck did undergo periosteal apposition greater than that observed in our study, but it resulted in femoral neck size that on average was similar to that in the youngest men at the same time point. This explanation seems unlikely, however, because the crude mean shaft area increased, rather than declined, with age.
Limitations specific to the use of QCT scans also pertain to this study. We took several steps to minimize sources of variability in scan acquisition and image processing that are possible in a multisite study with different scanners. First, we obtained scans with a standardized protocol that specified all machine settings to limit variation in scan acquisition across the enrollment sites. Nonetheless, some scans were not obtained according to protocol and could not be processed. Such errors resulted in substantial missing data for the femoral shaft. However, the shaft data were missing at random with regard to age, which reduces the precision of the estimated means and may underestimate the observed associations with age. Second, to reduce variability in image processing, all scans were read at a single location with a specialized software developed specifically for this purpose.(19,20) Third, all images were calibrated to standards in the field of view containing known calcium hydroxyapatite equivalents. We observed that the site-specific means and SDs were similar to the overall means and SDs for each of the femoral measures, indicating that the use of different scanners did not result in any obvious systematic bias. Nonetheless, we controlled for enrollment site as a proxy for scanner in our analyses. Despite these attempts to reduce the impact of scanner variability, measurement error is still possible but should be random with respect to age, because age distributions at each site were similar. Random error caused by scanner variation would underestimate the associations of age with the femoral measures.
An important technical limitation in our study is the effect of partial volume averaging in the cortical bone ROI.(14) Partial volume averaging occurs when the thickness of the imaged structure is smaller than the spatial resolution of the imaging system.(14,22) Our image processing approach used an edge detection algorithm to delineate the outer periosteal boundary. Then, we defined the cortical ROI using a threshold of 0.35 g/cm3 calcium hydroxyapatite equivalent applied to all voxels within the periosteal boundary but outside of the trabecular ROI.(19) Thus, it is possible that use of the threshold excluded true cortical bone from the ROI because of partial volume averaging. This would underestimate femoral neck cortical BMD, either because of increased cortical porosity(23) or extreme thinness of the structure.(14,22) However, our method produced femoral neck cortical vBMD values that are similar to those reported by Riggs et al.,(13) who applied the same threshold for cortical bone.(24) Nonetheless, the potential extent of partial volume averaging is reflected in the discrepancy between means for femoral neck and shaft cortical vBMD, which in each age group was twice that reported in the femoral neck.
Despite the limitations of our study, the MrOS cohort adds to knowledge about variability in femoral dimensions and vBMD in men 65–100 years of age. Our data show a remarkable diversity in the size and density of the femoral neck and shaft that previously has not been described in detail for men at advanced ages. Most dimensions were only moderately to weakly correlated, indicating that they represent unique structural features. This claim is supported by the observation that associations with age differed even among measures that were more highly correlated such as integral and medullary volume. Although femoral neck BMD from DXA was correlated with certain dimensions and vBMD, our data show that a single areal BMD measurement does not capture the extent of diversity of femoral features. Moreover, the relation of age with DXA BMD was relatively weaker than the relations of age with either femoral neck integral or trabecular vBMD, suggesting that the areal BMD measure may underestimate age-related bone loss.
Studies of femoral morphology among adults across a wide age range have yielded important insights into patterns of modeling and remodeling in the proximal femur.(1,6–13) The sum of the evidence is consistent with both periosteal apposition and endosteal resorption occurring in the femoral neck and shaft into older ages. Whether the extent of these processes, especially periosteal apposition, is the same in the neck and shaft, and whether there are sex-specific patterns is less clear. Observations from this study support previous reports that periosteal apposition occurs at advanced ages, but that the magnitude varies by location in the proximal femur.(1,6,7) Our data indicate that the balance of periosteal apposition and endosteal resorption results in cortical thinning in both the femoral neck and shaft, as shown by the inverse association of age with the percent cortical bone measures. However, we observed an inverse association of age and cortical volume in the femoral neck but no association of age with cortical area in the shaft, suggesting that net bone formation on the periosteal surface of the shaft but not the neck offsets net bone loss on the endosteal surface. Patterns similar to that which we observed in the shaft have also been reported for the distal tibia(1,7,13,25) and distal radius(13) among men. Histologic evidence may provide a possible explanation for our observation of more minimal periosteal expansion at the femoral neck. Allen and Burr(26) showed that the femoral neck surface had significantly less cellular periosteum than the midfemoral shaft. Thus, the reduced number of cells in the femoral neck periosteum may limit apposition.
Information about the relation of vBMD to age is limited. We noted that in the femoral neck, integral and trabecular vBMD declined by 9–22% between the seventh and ninth decades, but cortical vBMD was relatively unchanged. Inverse associations of integral and trabecular vBMD with age have been reported by others,(12,13) whereas cortical vBMD has been reported to be inversely associated with age in one study(13) but not associated in others.(12,14) However, like others,(23) we did observe that shaft cortical vBMD declined with age. It is unclear why aging would affect femoral neck and shaft cortical vBMD differently. Some have argued that observed inverse associations of age with femoral neck cortical vBMD are spurious owing to partial volume averaging in thin cortices.(14) However, if partial volume averaging affected our results in the same manner, we would have expected to observe an inverse relation of age to femoral neck cortical vBMD, which we did not. Alternatively, there may be biologic differences in the way aging affects femoral neck and shaft cortical vBMD. Recently, Mayhew et al.(27) showed that the thickness of the inferior cortex in the femoral neck increases in older relative to younger men and women, whereas the superior cortex thins substantially. Because cortical vBMD is averaged over all voxels contained in our femoral neck cortical ROI, the thicker inferior cortex contributes primarily to this estimate. Thus, increasing thickness in the inferior region of the cortex offers a plausible explanation for the lack of association we observed between age and cortical vBMD in the femoral neck. To clarify the extent of cortical vBMD changes in the femoral neck or shaft with age, prospective studies are necessary.
Although most of the observed associations between age and the femoral measures were highly statistically significant, we noted that a small amount of variation in the femoral measures was explained by age. This is most likely caused by the broad variability observed in each of the measures and the narrow age range of the cohort. In this regard, our results are consistent with those reported in a cohort with a wider adult age range in which the variability explained by age was also generally modest.(13) The majority of men in the MrOS cohort report themselves to be in good or excellent health. However, it seems unlikely that health status moderated the biologic processes underlying bone loss and structural drift to such an extent that we observed only a highly select group of men with minimal bone loss or structural change since young adulthood. As we described above, the mean total hip and femoral neck BMD from DXA in the MrOS cohort is similar to that reported in the NHANES III study, which is a random sample of the U.S. population. Therefore, it conceivable that other factors influence morphology and BMD at advanced ages.
Knowledge about the magnitude and extent of periosteal apposition and endosteal resorption has implications for fracture etiology, because both the amount and location of cortical bone as well as BMD affect the biomechanical strength.(1–3) We observed evidence for endosteal resorption and cortical thinning in both the femoral neck and shaft. However, the degree to which cortical bone was displaced outward by periosteal apposition seems to be substantially more in the shaft than in the neck. In addition, there was also substantial reduction of trabecular BMD in the femoral neck and cortical BMD in the shaft. These observations underscore the need for further study of fracture risk with sex-specific measures to determine whether BMD, dimensions, or a combination explain femoral fragility.
The Osteoporotic Fractures in Men (MrOS) Study is supported by the National Institutes of Health. The following Institutes provide support: the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), the National Institute on Aging (NIA), and the National Cancer Institute (NCI) through Grants U01 AR45580, U01 AR45614, U01 AR45632, U01 AR45647, U01 AR45654, U01 AR45583, U01 AG18197, and M01 RR000334. The authors thank all the participants in the Osteoporotic Fractures in Men study for continued participation, Benjamin KS Chan, MS, for assistance with graphical methods, and Michael Bliziotes, MD, for thoughtful editorial comments.