Introduction

Osteoporosis is a significant public health problem and contributes to disability and premature death among older men [1]. The lifetime risk of shoulder, forearm, hip, and spine fracture for men at age 45 years was 4.4, 5.2, 11.2, and 8.6%, respectively [2]. The risk for any of these fractures in men was 23.8% compared to 47.3% in women [2]. An estimated 5% of men over the age of 50 have a hip BMD more than 2 SD below the young normal mean [3], and at least one-third of all hip fractures occur in men [4]. Longer life expectancy has resulted in large increases in the numbers of older men and women at risk of suffering hip fractures. However, gains in life expectancy have been greater in men than women [5], suggesting that the proportionate increase in the number of hip fractures may be greater in men.

The etiology of osteoporotic fractures is complex and multi-factorial. Nevertheless, bone mass is a major determinant. There have been few comprehensive studies of the correlates of BMD in a large sample of older men [6, 7]. Most previous studies involved small select samples of men and a limited number of potential determinants. The relative importance of many proposed risk factors has been difficult to assess. The objective of the current analysis was to determine the cross-sectional demographic, anthropometric, historical (medical and family), lifestyle, and neuromuscular factors associated with BMD of the lumbar spine and proximal femur in a large population-based sample of older men enrolled in The Osteoporotic Fractures in Men Study, or “MrOS.”

Subjects and methods

Participants

From March 2000 through April 2002, 5,995 men who were at least 65 years of age or older were recruited for participation in MrOS [8]. Men were recruited using population-based listings at six US clinical settings in Birmingham, Ala.; Minneapolis, Minn.; the Monongahela Valley near Pittsburgh, Pa.; Palo Alto, Calif.; Portland, Ore.; San Diego, Calif. [9]. We excluded men who had had a bilateral hip replacement or who were unable to walk without the assistance of another person. The Institutional Review Board (IRB) at each center approved the study protocol, and written informed consent was obtained from all participants.

Measurement of bone mass

Bone mineral density (BMD) (g/cm2) of the lumbar spine (L1 to L4) and the total hip and its sub-regions (femoral neck; trochanter) was measured using fan-beam dual-energy X-ray absorptiometry (QDR 4500 W, Hologic Inc., Bedford, Mass.). Standardized procedures for participant positioning and scan analysis were executed for all scans. All DXA operators were centrally certified on the basis of an evaluation of scanning and analysis techniques. Densitometry technicians at the Coordinating Center (University of California, San Francisco) reviewed a random sample of all the scans, scans with exceptionally high or low BMD, and potentially problematic scans flagged at the clinic to assure adherence to standardized techniques. Cross-calibration studies performed prior to the baseline MrOS visit found no linear differences across scanners, and the maximum percent difference in mean total spine BMD between scanners was 1.4%. Longitudinal quality control using daily scan data for standardized phantoms indicated no shifts or drifts in scanner performance.

Other measurements

Information on demographics, medical and family history and lifestyle was obtained by questionnaire and interview by trained clinical staff. Race/ethnicity was based on self declaration and included the following categories: Caucasian, Black/African-American, Asian, Hispanic or Latino, Native Hawaiian/Pacific Islander, American Indian/Alaskan Native, and multiracial. Weight was measured in indoor clothing without shoes using a calibrated balance beam scale. Height was measured using a Harpenden stadiometer (DyFed, UK). Body mass index was calculated using weight (kg)/height (m2). Height and weight loss since age 25 was defined as the self reported height or weight at age 25 minus the current height or weight.

Grip strength was measured twice by a hand held dynamometer (Jamar) in both right and left arms. The average grip strength in kg was used in the analysis. Gait speed was determined as usual time to complete a 6-m course and expressed in meters per second. Time to complete five chair stands (seconds) and ability to stand from a chair without using arms (yes/no) was also recorded.

Lifestyle factors included smoking history. Subjects were classified into current, past, or never smokers. Information on alcohol consumption was obtained by interview and expressed as drinks per week. Dietary information about the usual eating habits over the past year was obtained using the modified Block Food Frequency Questionnaire including information on calcium (mg/day), Vitamin D (mg/day), and caffeine intake (mg/day) [10]. Information on calcium and Vitamin D supplement use was also obtained and summed with the dietary intake to estimate total calcium and total Vitamin D intake. Subjects were asked if they walk for exercise (yes/no). Physical activity was also measured by computing the physical activity summary scale for the elderly (PASE) [11].

Medical history

Subjects were asked whether a doctor or health care provider had ever told them they had certain medical conditions including diabetes, osteoarthritis (OA), rheumatoid arthritis (RA), thyroid disease, prostate cancer, hypertension, chronic obstructive pulmonary disease (COPD), myocardial infarction (MI), stroke, osteoporosis, or kidney stones. Information was obtained on whether or not they had ever had a gastrectomy. Fracture history since age 50 was obtained including specific information on hip, wrist, and spine fractures. Maternal and paternal history of any fracture and specifically hip fracture was also obtained. “Don’t know” responses to these family history questions were coded as “No.” Subjects were asked about their general health status compared to others their own age (good/excellent vs. fair, poor, very poor).

Participants were asked to bring all prescription and non-prescription medications to the clinic for verification. Current use was defined as use within the preceding 30 days. A study-specific medication dictionary was used to categorize the type of medication from product brand and generic names obtained from the medication containers. Dose or duration of use or specific indication was not queried.

Statistical analyses

To detect potential associations among predictor variables and BMD, we first analyzed the data using bivariate age-adjusted linear regression models. These associations were further adjusted for body weight. Variables with bivariate associations in which the P values were less than 0.10 at any one site were included for consideration in stepwise multiple regression models. We used forward stepwise regression procedures with variables significant at the P =0.15 level for entry into the model to determine the final multivariable models. Age, weight, and race were forced into the stepwise regression models. Backward and best fit selection procedures produced similar results. The variables entered into the multivariable stepwise models included age, clinic site, race, weight, height, height loss since age 25, current smoking, daily caffeine intake, average number of alcohol drinks per week, PASE score, daily calcium intake (dietary), daily vitamin D intake (dietary), non-trauma fracture after age 50, maternal and paternal history of fracture, diabetes, gastrectomy, hypertension, COPD, Parkinson’s disease, stroke, OA, prostate cancer, kidney stones, select medications (use of COX-II inhibitor, thiazide-diuretic, betablocker, statin, selective serotonin reuptake inhibitors (SSRI), or steroid use [12]), self reported health status, ability to stand without arm use, and grip strength.

Because of the cumulative effect of missing values, the sample size in the multivariable regression was 5,591 for the total hip and femoral neck and 5,590 for the lumbar spine. To avoid bias from specific therapies for osteoporosis, or a history of osteoporosis, we excluded men from the multivariable results who reported ever taking an osteoporosis medication ( n =143) and men who self-reported osteoporosis ( n =212), two categories that were not mutually exclusive. To express the strength of the associations, units of percent difference in BMD for continuous variables were chosen to approximate 1 SD in the distribution of a given variable. For age and body weight, we used standard units, 5 years for age and 10 kg for weight. Percent differences were calculated from the regression coefficients using the formula (beta * unit/mean BMD).

The data collected included sets of variables related to broad areas of interest such as measures of obesity and muscle strength and estimates of physical activity. Within each set, several variables were associated with BMD in bivariate analyses. We chose to model the variable within each set that had the lowest amount of missing data and showed the strongest relationship to BMD in the bivariate analyses. We evaluated several interactions, specifically age × race, weight × race, and age × weight. There was no systematic pattern of interaction, and results are not reported. To compare the determinants of BMD in men and a similar population of older women [13], we ran our multivariable models limiting this analysis to Caucasian men.

Results

The study population included 5,362 (89.4%) Caucasian men, 244 (4.1%) African-American men, 191 (3.2%) Asian men, 127 (2.1%) Hispanic/Latino men, 5 (0.08%) American Indian men, 7 (0.12%) Native Hawaiian/Pacific Islander men, and 59 (0.98%) multiracial men. For these analyses, Pacific Islander, American Indian, and multiracial men were grouped as “other.”

The average age of the men was 73.7±5.9 years. The mean total hip BMD (g/cm2) was 0.96±0.14 (range 0.31 to 2.1); femoral neck, 0.78±0.13 (range 0.27 to 1.87), and lumbar spine 1.07±0.19 (range 0.47 to 2.10). The age- and age and body weight-adjusted bivariate analyses of femoral neck BMD and lumbar spine BMD are summarized in Tables 1 and 2, respectively. All references to “hip” BMD refer to the femoral neck unless specified (Table 1).

Table 1 Correlates of bone mineral density of the femoral neck in age- and age and weight-adjusted bivariate models
Full size table
Table 2 Correlates of bone mineral density of the lumbar spine in age- and age and weight-adjusted bivariate models
Full size table

Demographics

Each 5-year increase in age was associated with 2.6% lower femoral neck BMD, but 7% higher spine BMD. Additional adjustment for body weight attenuated the effect of age at the hip, but increased the magnitude of the effect at the spine. African-American men had 6–11% higher hip and spine BMD than Caucasian men, even after adjusting for age and body weight. Asian men had 3 to 4% lower hip and spine BMD than Caucasian men, but this difference was not apparent after adjusting for body weight. Hispanic men had higher hip but lower spine BMD than Caucasians, but these differences were not statistically significant.

Anthropometrics

A 10-kg higher body weight and a 4-unit increase in BMI were associated with 4% greater femoral neck BMD and 2 to 3% greater lumbar spine BMD in age-adjusted models. A weight increase of 11.5 kg since age 25 was associated with 3% greater femoral neck BMD in age-adjusted models. Weight change was not related to spine BMD. Taller men had 2 to 3% higher femoral neck and lumbar spine BMD for each 10-cm increase in height in age-adjusted analysis. After additional adjustment for body weight, a 1 SD increase in height was associated with a 1% lower hip BMD. After adjusting for age and body weight, each 3 cm of height loss since age 25 was associated with a 1.4% lower hip, but not spine BMD.

Lifestyle

Current smoking was associated with 2% lower hip and spine BMD, but this effect was not significant and was attenuated even further after adjustment for body weight. Caffeine intake was not significantly associated with either femoral neck or lumbar spine BMD. A 1 SD (approximately seven drinks per week) increase in alcohol consumption was associated with about a 1% higher hip and spine BMD. Greater physical activity as measured by the PASE scale was associated with about a 1% higher hip, but not spine BMD. Men who reported walking for exercise had higher BMD in age and weight-adjusted analyses, but the difference was not statistically significant.

Diet

Men who reported higher dietary calcium intakes had modestly higher hip BMD, but there was no association with lumbar spine BMD. In contrast, calcium intake from supplements was associated with lower hip and spine BMD. Vitamin D was not related to either hip or spine BMD.

Fracture history

A positive history of any fracture since age 50 was a strong predictor of lower hip and spine BMD. Men who had a history of a hip fracture had 7 to 9% lower hip and spine BMD than men without a history of hip fracture. A history of a wrist or spine fracture was also associated with 5 to 6% lower BMD. Controlling for body weight had little effect on these associations.

Family history

Both maternal and paternal histories of fracture, including a hip fracture, were each associated with 2 to 4% lower BMD at both anatomic sites. Adjustment for weight had no effect on these results.

Medical history

Self reports of diabetes and hypertension were associated with significantly higher hip and spine BMD, even after adjusting for age and weight. OA and a history of an MI were associated with higher spine, but not hip BMD. Histories of gastrectomy, COPD, Parkinson’s disease, prostate cancer, stroke, osteoporosis, and kidney stones were associated with lower BMD at both sites. In general, these associations were slightly attenuated after adjusting for age and body weight and in certain instances were borderline significant. There was no significant relationship with a history of RA or thyroid disease.

We examined a number of medications and their relationship with BMD. Use of an SSRI was associated with 4 to 5% lower hip and spine BMD, even after adjusting for age and body weight. Use of corticosteroids, either inhaled or oral, was associated with lower BMD at both sites, but the relationship was stronger for spine BMD. Current use of a beta-blocker was associated with about a 1% higher hip and spine BMD after adjusting for age and body weight. Thiazide diuretics were associated with higher spine, but not hip BMD in age and weight-adjusted analyses. Use of NSAIDS was associated with higher spine BMD. We found no associations between the use of androgens, testosterone injections, anti-androgens, non-thiazide diuretics, anti-convulsants, tricyclic anti-depressants, nitrates, or thyroid supplements and BMD at the spine or hip. In age and weight-adjusted models, statin use was not significantly associated with BMD. Men who reported treatment with a bisphosphonate, calcitonin, or any osteoporosis medication had 12 to 19% lower BMD than non-users.

General health

Men who reported good to excellent health had significantly higher hip BMD.

Neuromuscular function

Longer time to complete five chair stands was associated with about 1% lower BMD. Men who could stand from a chair without using arms had 2 to 4% higher BMD in age and weight-adjusted models. One standard deviation higher grip strength (8 kg) was associated with greater BMD, but the effect was modest (<1%). Gait speed was not related to BMD.

Multivariable models

Multivariable models explained 19 and 10% of the overall variance in the BMD at the femoral neck (Table 3) and lumbar spine (Table 4), respectively. The corresponding model R2 for the total hip was 23% (data not shown). Race, age, and weight alone explained 15% of the variance in the femoral neck and 5%, in the lumbar spine. At both anatomic sites, African-American race was the strongest predictor of BMD. A 5-year increase in age was associated with lower hip, but greater spine BMD. In addition, body weight and a history of diabetes were strongly related to higher BMD at each site, while a history of non-traumatic fracture and SSRI use were related to lower hip and spine BMD. A history of OA was associated with higher BMD, but the association was much stronger for the spine than hip BMD. Corticosteroid use was associated with lower spine BMD. Ability to stand from a chair without using arms was related to higher hip BMD. In addition, maternal and paternal history of fracture, history of COPD, and kidney stones were associated with lower BMD, while alcohol consumption was associated with higher BMD at both anatomic sites. Statin use was not significant in the bivariate models, but in the multivariable models, statin use was associated with about 1% higher hip BMD. Other variables were weakly (<1%) associated with greater hip BMD, including grip strength, calcium intake, and physical activity. Greater height and height loss were associated with lower hip, but greater spine BMD.

Table 3 Multivariable correlates of femoral neck BMD in older men
Full size table
Table 4 Multivariable correlates of lumbar spine BMD
Full size table

Results of multivariable models restricted to Caucasian men were generally similar to the models based on the whole cohort with the following exceptions. Among Caucasian men, a history of Parkinson’s disease was a significant predictor of lower femoral neck BMD (−4.98%; 95% CI, −9.5 to −0.6). A positive history of prostate cancer was associated with −1.3% (−2.6 to −0.01) lower hip BMD. Finally, use of COX-II inhibitors was associated with greater femoral neck BMD (1.82%; 0.01 to 3.6). There was no difference in the predictors of spine BMD in models restricted to Caucasian men.

Discussion

Low BMD is an important risk factor for fractures in older men [14, 15, 16, 17]. It is estimated that the age-related decline in BMD accounted for a 60% increase in the risk of hip fracture in older men [18]. To better understand the factors that influence BMD, we characterized the demographic, historical, anthropometric, and lifestyle factors that are correlated with hip and spine BMD in the largest population-based study of BMD in older US men. Several features of the study are unique, including its large size and careful ascertainment and standardized measurement protocols for the large number of variables of potential interest.

Our study demonstrated that cross-sectional correlates of BMD were similar for both the proximal femur and lumbar spine (Table 5) and were in general similar to those reported for women [13]. Race/ethnicity, body weight, neuromuscular function, personal and family history of fracture, diabetes, and the use of select medications (SSRIs, corticosteroids, COX-II inhibitors) were the strongest independent determinants of BMD in older men. Somewhat weaker although statistically significant associations were found for age, alcohol use, height, calcium intake, and physical activity.

Table 5 Summary of independent correlates of BMD from multivariable models
Full size table

The strongest factor determining BMD was race/ethnicity. A genetic contribution to osteoporosis is well documented [19]. Comparisons of mono- and dizygotic twins demonstrate that up to 80% of the variance in BMD is under genetic control [20]. The importance of race/ethnicity for BMD likely reflects the important role of genes in determining BMD. African-American men had 12% higher hip and 6% higher spine BMD than Caucasian men, a finding consistent with previous findings in both men and women [3]. The higher bone mass among African-American men is sufficiently large to account for their lower risk of fracture. The higher BMD among African-American men could not be explained by weight or other historical or lifestyle factors.

Most previous studies have been confined to comparisons between two ethnic groups. In MrOS, we enrolled men of a number of different ethnicities, although we had limited power to detect differences between some groups. Hispanic men had higher hip but lower spine BMD than Caucasian men, but these differences were not statistically significant. In NHANES III, Mexican American men had a lower prevalence of hip osteoporosis, but the number of minority men was small and prevalence estimates less reliable [3]. In the present study, Asian men had lower BMD than Caucasian men in age-adjusted analysis, but this difference was entirely explained by body weight. In a study of women, after restricting the analysis to women weighing less than 70 kg, lumbar spine BMD was actually greater in Chinese and Japanese women compared to Caucasian women [20].

After race/ethnicity, anthropometric factors, in particular body weight, had a strong positive effect on BMD, independent of age, race, or any other variable examined. This finding is consistent with reports in women [13] and other studies in men [6, 7, 21, 22, 23]. Further analyses should test whether this relationship reflects fat and lean mass equally. The underlying mechanism for this association may reflect mechanical factors related to loading, genetic influences, and/or greater levels of endogenous sex steroid and growth hormones.

Older age was associated with lower hip BMD, but higher spine BMD. This finding is consistent with observations in women [24] and may reflect age-related changes in the spine, e.g., osteophyte formation.

Taller men and men who had lost more height since age 25 had lower hip, but greater spine BMD. The observation that taller men had greater spine BMD may reflect the larger vertebrae among taller men. The lower BMD at the hip among taller men is consistent with the observation that taller people and people who have experienced greater height loss have an increased risk of hip fractures [25, 26, 27].

History of previous fracture was associated with lower BMD, consistent with reports in women [13]. A positive family history of fracture was also associated with lower BMD, similar to reports from the Rancho Bernardo Cohort [28], but the magnitude of the association was smaller. It is interesting to note that a positive history of fracture in both the mother and father was equally and independently related to lower BMD in men, emphasizing the need to better understand the genetic determinants of bone mass in men.

Several prevalent medical conditions were significantly related to higher BMD, including hypertension, diabetes, and OA, while prostate cancer, COPD, and kidney stones were associated with lower BMD. The greater BMD among diabetic men is consistent with data in women [12, 29, 30] and men [30]. The greater BMD among men with diabetes was independent of body weight and could reflect other biologic mechanisms. Greater abdominal obesity, insulin and growth factors could contribute to higher BMD among the diabetics [31]. Hypertension has been associated with abnormalities in calcium metabolism, leading to increased calcium losses and secondary activation of the parathyroid gland [32], but these mechanisms would lead to lower BMD among hypertensive people. The positive association between hypertension and BMD may reflect higher insulin levels or may result from confounding by the use of medications that may increase BMD.

In general, previous studies have also reported higher bone mass in women with clinical or radiographic OA [33, 34]. This association may reflect shared risk factors, e.g., obesity, but may also reflect shared genetic factors and other systemic factors such as sex steroid hormones. Higher spine BMD in those with OA may also reflect osteophyte formation [7].

A history of COPD was associated with lower hip and spine BMD, independent of steroid use and smoking, consistent with the higher fracture risk observed for COPD [35]. It may reflect inactivity or immobility resulting from the progression of obstructive lung disease. Men who reported a history of prostate cancer had lower BMD. Only 3.5% of men reported current anti-androgen therapy, and among these men, hip and spine BMD was lower. Men with prostate cancer may have used anti-androgen therapy or had surgery that made them hypogonadal, both of which are associated with bone loss [36, 37]. A history of kidney stones was associated with lower BMD, consistent with observations that subjects who are calcium stone formers have a tendency towards low bone formation rates, increased bone resorption, lower BMD [38], and higher cytokine activity [39].

Current use of SSRIs was independently associated with lower BMD. Depression and anti-depressive treatment have been linked to an increased risk of hip fracture [40, 41, 42] and lower BMD in some [43, 44], but not all studies [41]. We did not assess depression at baseline, and it’s possible that the association between SSRI use and low BMD reflects weight loss and poor diet in persons with depression. We included a single question on how often the subject “felt blue” in the past year. Additional adjustment for this variable had no effect on our results. If the SSRI-BMD association reflected an indication bias, then one would have expected men on other types of anti-depressive therapy to have lower BMD, but there was no relationship between the use of tricyclic anti-depressants and BMD. In older women, SSRI use was not associated with lower hip BMD, but was associated with faster rates of bone loss [45]. In a large case-control study of hip fracture in older men, the use of psychotropic medications was associated with a two-fold increase risk of hip fracture [46]. The underlying mechanism may reflect both low BMD and/or an increased risk of falls. Future research will need to determine the underlying mechanisms whereby SSRIs influence BMD in older men. Serotonin is important in the pathophysiology of depression; SSRIs act by blocking the serotonin transporter that has been documented in bone [47]. Mice lacking the serotonin transporter gene have lower BMD than wild-type mice [48]. Given the high prevalence of the use of SSRIs in the general population, the finding of significantly lower BMD could have important public health consequences.

Statins were associated with modestly higher femoral neck BMD in multivariable models, a finding consistent with some observations in women. This association was independent of MI and other selection factors for statin use such as body weight [49]. Steroid use, both inhaled and oral, was associated with lower BMD, especially at the lumbar spine consistent with previous reports [50]. Prior and current exposure to corticosteroids was associated with an increased risk of fracture that was only partially explained by their lower BMD [51]. Use of COX-II inhibitors was associated with greater BMD. The combination of COX-II selective NSAIDS and aspirin was associated with greater areal and volumetric BMD in both men and women, and may reflect an effect on prostaglandin [52].

A number of lifestyle variables were associated with either higher hip or spine BMD or both, including alcohol consumption, dietary calcium intake, and physical activity. The positive association of dietary calcium with BMD is consistent with previous observations in women [53]. In contrast, calcium intake from supplements was associated with lower BMD. This latter observation may reflect an indication bias where men with lower BMD were advised to take calcium [54]. Previous studies have also shown that moderate alcohol consumption (fewer than two drinks per day) has a positive association with bone mass in men [6, 55]. The mechanism for this association is unclear, but it could reflect a “healthy user” effect, since moderate alcohol consumption is associated with a lower total mortality [56]. Alcohol may also increase BMD by raising sex steroid hormone levels [57, 58] or by direct effects on the aromatase gene [59].

Physical activity was positively associated with BMD in the femoral neck multivariable model, a finding consistent with observations in women [13]. Greater neuromuscular function as measured by grip strength and ability to stand from a chair without using arms was also independently related to higher hip BMD, as previously reported in women [13]. These observations support the hypothesis that the relationship between physical activity and bone mass may be mediated by several pathways, including direct effects of loading or indirectly through muscle tension.

The overall effect of these variables was modest, but remained statistically significant in the multivariable model. The public health implications of these findings highlight the need for older men to maintain a healthy lifestyle, including adequate calcium intake, maintenance of an active lifestyle, and modest alcohol consumption to preserve bone health.

Several medical conditions were related to BMD in age and weight-adjusted models and were of borderline significance in the multivariable regression models and deserve further exploration. Parkinson’s disease has been associated with an increased risk of hip fracture [60], possibly because of an increased risk of falls, but our data show that men with Parkinson’s disease also had lower BMD, especially Caucasian men. Men with a history of stroke had lower BMD, which may contribute to their higher risk of fracture [60]. Surgical removal of the stomach was associated with lower BMD, consistent with previous observations in men and women [6, 7, 13].

In multivariate models, we found no association with current or past cigarette smoking, caffeine intake, dietary and supplemental vitamin D intake, thyroid disease, and use of thiazide and non-thiazide diuretics, thyroid supplements, nitrates, anti-convulsants, and androgens. Some of these null associations could reflect the low statistical power to detect an association because of the low prevalence of use. For example, <4% of men reported current smoking.

Overall, our multivariable models explained 19% of the variance in BMD of the femoral neck. This compares well to other reports in men [6, 7] and women [13]. We could explain only 10% of the variance in lumbar spine BMD, compared to about 25% in women [13]. Osteophyte formation may have a greater influence on the measurement of spine BMD in men compared to women [7] and may have confounded associations with the various measures. These findings raise the possibility that there are many as of yet undetermined factors that contribute to BMD in older men.

The current report is the largest, most comprehensive study to describe the determinants of BMD in a population-based cohort of older men. A large number of variables were examined, and areal BMD was measured using state-of-the-art and well-standardized densitometry. Nevertheless, our study also had several limitations. We examined older, independently living volunteers whose characteristics may differ from those of other groups. Less than 4% of the men were current smokers, reinforcing the healthy group of men participating in MrOS. Although we attempted to recruit ethnic minorities, only 10% of our participants were minority. We could not examine different subsets of Asians or Hispanics. We used a cross-sectional design, and therefore we can establish associations, not causation. We examined areal BMD, which may be confounded by bone size. Future reports from MrOS will examine the correlates of volumetric BMD. We examined a number of variables and by chance alone, 5 of 100 could be significant. Associations that are marginally significant may need to be replicated in other studies.

In summary, a number of lifestyle and behavioral characteristics and medical conditions were associated with BMD in older men. We uncovered previously unrecognized correlates, especially the lower BMD among men reporting SSRIs. Our findings reinforce the need to understand these relationships to increase our understanding of etiologic mechanisms. Family history was a major independent predictor of low BMD, underscoring the need to explore the genetic factors that determine BMD in older men. From a clinical perspective, these results provide information useful for the identification of men who may be at risk for low bone density and fractures and who deserve further evaluation.