Dr Cauley receives research support from: Merck & Co., Eli Lilly & Co., Pfizer, and Novartis. She also receives honoraria from Merck & Co. and Novartis in addition to serving on the Speaker's Bureau for Merck & Co. Dr Cummings receives research support from Eli Lilly & Co., Amgen, Pfizer, and Novartis. He also serves as a consultant to Pfizer, Novartis, Aventis, Procter & Gamble, and Zelos and received honoraria from Eli Lilly & Co., Novartis, and Merck & Co. Dr Orwoll receives funding from Aventis, Pfizer, Eli Lilly & Co., and Novartis. He also receives honoraria from Merck & Co. and serves as a consultant to Procter & Gamble, GlaxoSmithKline, Aventis, and TAP Pharmaceuticals. All other authors have no conflicts of interest.
Published online on July 17, 2006;
In a prospective study of 5384 older men, hip BMD was a very strong predictor of hip fracture, much stronger than spine BMD. The relationship between hip BMD and hip fracture risk seemed to be stronger than observed in a large prospective study of women. Hip BMD is an excellent test for predicting fracture risk in men.
Introduction: There have been few studies of the relationship between BMD and risk of fractures in men and none for the association between lumbar spine BMD and risk of hip and nonvertebral fractures. There is also controversy about whether the associations between BMD and risk of fracture are the same in men and women.
Materials and Methods: We measured proximal femur and lumbar spine BMD in 5384 men, ≥65 years of age. We compared the results to the very similar cohort of 7871 women ≥65 of age. During 4.4 years of 99% complete follow-up, we validated 317 nonvertebral (59 hip) fractures in men and 1169 nonvertebral (208 hip) fractures in women.
Results: Total hip BMD was very strongly associated with risk hip fracture in men (3.2-fold increased risk per sex-specific SD decrease in BMD; 95% CI, 2.4–4.1). The association was stronger than observed in SOF (2.1; 95% CI, 1.8, 2.4; p < 0.001 for interaction). Among the men, lumbar spine BMD was weakly associated with risk of hip fracture (relative risk [RR] per sex-specific SD decrease in BMD: 1.5; 95% CI, 1.2, 2.0). The association between total hip BMD and risk of nonvertebral fractures was somewhat stronger for men (RR = 1.6; 95% CI, 1.5, 1.8) than found for women (p = 0.01 for interaction). The risk of nonvertebral fracture was substantially higher in women than in men for all T scores of hip BMD, regardless of whether sex-specific or female reference values were used.
Conclusions: Hip BMD is strongly associated with risk of nonvertebral, and especially hip fracture, in older men. These associations are at least as strong as in women. As in women, lumbar spine BMD in men is only weakly associated with risk of hip fracture. Regardless of whether sex-specific or female reference values were used, T scores indicated different risks of fractures in men than in women.
Many prospective studies have shown that BMD predicts fractures in women. A few prospective studies suggest that the association is similar for men.(1–3) Schuitt et al.(3) reported that men and women ≥55 years of age participating in the Rotterdam Study had similar associations between BMD and risk of hip and other nonvertebral fractures. This cohort was included with several smaller studies in a meta-analysis by Johnell et al.(1); that analysis also reported that BMD was associated with risk of hip, “osteoporotic,” and nonvertebral fractures in men and the strengths of the BMD/fracture associations were similar in men and women. In a smaller prospective study not included in the meta-analysis, Nguyen et al.(2) also found similar patterns of BMD and fracture risk in men and women. These studies were based on hip BMD; the value of lumbar spine BMD for prediction of hip and nonvertebral fractures in men has not been studied.
There is growing clinical interest in measuring BMD in men. The results are typically reported as T scores, and clinicians may assume that a certain T score (such as “osteopenia” with a T score of −1 to −2.5) implies a similar risk of fractures in men and women. However, the same T scores may mean different risks of fracture in a man and a woman of the same age. Thus, guidelines recommending pharmacological treatments at a certain T score may indicate different potential benefits from treatment in women and men. There is debate about whether using female, instead of sex-specific, reference values for men is the best approach.
We prospectively studied the association between BMD of the proximal femur and lumbar spine and risk of nonvertebral fracture in a community-based cohort of 5995 men, ≥65 years of age: the Osteoporotic Fractures in Men Study (MrOS). The methods of MrOS intentionally resembled those of the Study of Osteoporotic Fractures (SOF), a large cohort of older women.(4,5) This allowed us to compare the associations between BMD and fractures after the same length of follow-up for both men and women.
MATERIALS AND METHODS
The design of MrOS was derived from SOF.(4) Both are prospective studies of large cohorts recruited from several communities in the United States. In MrOS, as in SOF, BMD of the hip and lumbar spine were measured, and participants were periodically followed-up to ascertain and validate new fractures.
In MrOS, we recruited 5995 men ≥65 years old by local advertisements and mass mailings in several U.S. urban areas: Birmingham, AL; Minneapolis, MN; Palo Alto, CA; the Monongahela Valley near Pittsburgh, PA; Portland, OR; and San Diego, CA. (Participants in SOF had been recruited from Baltimore, MD; Minneapolis, MN; the Monongahela Valley near Pittsburgh, PA; and Portland, OR, USA.) Greater details about the methods of MrOS and SOF have been published elsewhere.(4–6) Six hundred ten (10.2%) of the men recruited to MrOS were nonwhite or Hispanic. Because of their low rate of hip fracture, nonwhite women were excluded from SOF. Therefore, to allow comparisons between MrOS and SOF, we limited analysis to the 5384 white men.
In MrOS, BMD was measured in the proximal femur and lumbar spine using DXA measured by Hologic QDR 4500 densitometers during the baseline visit in 2000–2002. In SOF, the proximal femur and lumbar spine BMD were assessed by DXA using Hologic QDR 1000 densitometers at the second clinic visit in 1989–1990. An in vivo cross-calibration for spine and hip was performed on the Hologic QDR1000 and QDR4500 scanners with 15 MrOS and 15 SOF participants at the Pittsburgh and Minneapolis sites. Correction factors derived from linear regression models were applied to the QDR1000 data for comparison with QDR4500 results.
Follow-up and ascertainment of fractures
In the MrOS and SOF cohorts, we contacted participants every 4 months through mailed questionnaires to ask about recent fractures. Follow-up for fractures and vital status was 99% complete during 4.4 years of follow-up. All reported fractures were validated by physician review of radiology reports or X-rays if no radiology report was available. The physician rated fractures as associated with “severe” trauma if they occurred during a motor vehicle accident or trauma equivalent to falling from more than one stair step above standing height. Main analyses included all fractures, and subanalyses were completed that excluded fractures associated with severe trauma.
To create comparable periods of follow-up after measurement of BMD, follow-up time in SOF was truncated to produce an average follow-up time of 4.44 years (range: 30 days to 5.6 years) to be similar to the 4.41 years (range: 7 days to 5.7 years) of follow-up in MrOS.
In both studies, height was measured on a Harpenden stadiometer (DyFed, UK) and weight was measured on balance beam scales (except the MrOS Portland site, which used a digital scale) that were calibrated with standard weights. Body mass index (BMI) was calculated as weight (kg)/height (m)2. Smoking status and alcohol use were by self-report in both studies.
To compare baseline characteristics between men and women, χ2 tests were used for categorical variables, and t-tests were used for continuous variables. Age-adjusted fracture rates were calculated for sex-specific quartiles of total hip, femoral neck, and lumbar spine BMD and are expressed as rate per 1000 person-years. The association between BMD and risk of fracture was analyzed by proportional hazards models with 95% CIs. We express the risks as change in risk per SD in that measurement within that sex. The values of SDs differ for men and women, and therefore, for comparisons of women and men, we also expressed the BMD–fracture relationship as change in relative risk per 0.1 g/cm2 decrease in BMD.
To test the statistical significance of differences in risk of fracture and BMD level in men and women, we pooled the data for MrOS and SOF and tested for the presence of interaction using cross-product terms for sex and age, sex and BMD, BMD and sex, and the three-way interaction between age, sex, and BMD. We tested for interactions for all three BMD sites (total hip, femoral neck, and lumbar spine). In the pooled data, there was significant (p < 0.05) interaction between sex and age, indicating that the ratio of rates of fracture in men and women changes as they age. Furthermore, there was a significant interaction between sex and total hip BMD for both hip and nonvertebral fractures, indicating that the association between total hip BMD and risk of hip and nonvertebral fracture is significantly different for men than women. The interaction between femoral neck BMD and sex for was not significant for hip fractures (p = 0.11) or nonvertebral fractures (p = 0.37). Because we found significant interaction between sex and BMD and sex and age, results are stratified by age or by sex.
We used means and SDs for total hip and femoral neck BMD from the National Health and Nutrition Examination Survey III to calculate T scores.(7) We used sex-specific T scores: male reference values for men in MrOS and female reference values for women in SOF. We also explored using reference values for women to calculate T scores for men. Estimates of 3-year risk of fracture by T score were based on logistic regression models. Risk of fracture was modeled separately for men and women and adjusted for age.
The men in MrOS had a mean age of 73.8 years, had an average BMI of 27.4 kg/m2, and only 3% currently smoked cigarettes. Compared with the women in SOF, MrOS men were taller and drank more alcohol (p < 0.001; Table 1). During 4.4 years of 99% complete follow-up, we validated that 317 (5.9%) of the 5384 men had at least one nonvertebral fracture. Of these, 36 (11.4%) were attributed to severe trauma. In contrast, 1169 (14.9%) of 7871 women in SOF had at least one nonvertebral fracture during the same interval, and 57 (4.9%) were classified as traumatic. In older men and women, about 18% of nonvertebral fractures were hip fractures (Table 2).
Table Table 1.. Characteristics of the MrOS Men and SOF Women
Table Table 2.. Number of Men and Women With at Least One Nonvertebral Fracture* and Location of First Nonvertebral Fracture
Hip fractures in men
Among older men, the relationship between total hip BMD and risk of hip fracture was very strong: each sex-specific SD decrease in total hip BMD was associated with a 3.2-fold increased risk of hip fracture (95% CI, 2.4, 4.1; Table 3). The age-adjusted annual rate of hip fracture was particularly high (Fig. 1, 7.1 per 1000 person-years, 95% CI, 4.9, 9.3%) among men in the lowest quartile of total hip BMD (Fig. 1). The association between lumbar spine BMD and risk of hip fracture (relative risk [RR]: 1.5 per sex-specific SD decrease; 95% CI, 1.2, 2.0) was substantially weaker than the association between hip BMD and risk of hip fracture. Results for femoral neck BMD are similar, and exclusion of fractures because of severe trauma did not significantly alter the results (data not shown).
Table Table 3.. Age-Adjusted Relative Hazard (RH) of Fracture per 0.1 g/cm2 or Sex-Specific SD Decrease in BMD at the Total Hip, Femoral Neck, and Lumbar Spine in Men and Women
Nonvertebral fractures in men
In this cohort of older men, each sex-specific SD decrease in total hip BMD was associated with a 1.6-fold (95% CI, 1.5, 1.8) increased risk of nonvertebral fracture. Excluding hip fractures from the nonvertebral fractures attenuated results a little (RR, 1.4 per sex-specific SD decrease; 95% CI, 1.2, 1.6). Results for femoral neck BMD were very similar. Lumbar spine, femoral neck, and total hip BMD had similar associations with risk of nonvertebral fracture: each sex-specific SD decrease in BMD increase the risk of fracture ∼1.5-fold. As with hip fractures, exclusion of fractures because of severe trauma did not significantly alter the results (data not shown).
Comparisons between men and women
Between 65 and 70 years of age, SOF women were 2.2-fold (95% CI, 1.6, 3.1) more likely to experience a nonvertebral fracture than MrOS men, even after controlling for total hip BMD and weight (Table 4). This sex difference in risk of fracture decreased with age, so that the risk of fracture for men and women became similar and not significantly different among participants ≥80 years of age after adjustment for total hip BMD (RR, 1.2; 95% CI, 0.9, 1.7).
Table Table 4.. Ratio of Rates of Non-Spine Fractures Observed in Women to the Rates of Men With Increasing Age
BMD has different SDs in men and women. For example, the SD for femoral neck BMD is 0.13 for men and 0.10 for women. Therefore, for comparisons between men and women, we expressed the associations as change in relative risk per 0.1 g/cm2 decrease in BMD instead of sex-specific SD. The association between total hip BMD and risk of hip fracture was stronger for men than women (p < 0.001 for interaction): 2.3-fold (1.9, 2.8) per 0.1 g/cm2 for men, and 1.7-fold (1.6, 1.9) for women (Table 3). For femoral neck BMD, men had a nonsignificantly higher risk of hip fracture than women per 0.1 g/cm2 decrease. Men had a somewhat higher relative risk of nonvertebral fracture per 0.1 g/cm2 decrease in total hip BMD than women (p = 0.01 for interaction). Again, for femoral neck BMD, men had a nonsignificantly higher risk of nonvertebral fracture than women per 0.1 g/cm2 decrease in BMD.
Lumbar spine BMD was a similarly weak predictor of nonvertebral fractures in men and women: each 0.1 g/cm2 decrease in lumbar spine BMD increased the risk of nonvertebral fractures 1.3-fold (95% CI, 1.2, 1.3) in men and 1.2-fold (95% CI, 1.1, 1.2) in women. Nonvertebral, nonhip fractures were less strongly associated with decreasing BMD levels than nonvertebral fractures. As above, exclusion of fractures because of severe trauma did not significantly alter the results (data not shown).
T scores and risk of fracture in men and women
Figure 2 provides an estimate of the approximate 3-year risk of nonvertebral fractures based on sex, age, and total hip T score. For example, a 65- to 69-year-old white man with a T score of −2.5 has a 7.6% (95% CI, 5.6, 10.2) cumulative 3-year risk of fracture, whereas a woman of the same age- and sex-specific T score has a 16.8% (95% CI, 14.2, 19.7) cumulative 3-year risk of nonvertebral fracture. Men had a lower risk of fracture at every T score of hip BMD. As age increased, the association between T score and risk of fracture became more similar.
Figure 2 also displays the association between BMD and rates of nonvertebral fracture using female reference values. Very few men had very low female-specific T scores (such as T score < −2.5), and confidence limits around the risk estimates for men with low female T scores were very wide. Using these reference values, men continued to have lower risk of fracture than women at every T score. Using absolute values of total hip BMD in place of female T scores produces the same relationships because female T scores simply correspond to the absolute values of total hip BMD.
We found that BMD in the total hip or femoral neck (hip BMD) is an excellent measurement for assessing the risk of fracture in men. In this study, the association between total hip (or femoral neck) BMD and risk of hip fracture was very strong in men, due in part to a very high rate of hip fracture observed among men with the lowest levels of BMD. Men in MrOS had a statistically significantly higher risk of hip fracture per 0.1 g/cm2 decrease in total hip BMD than women in SOF. In older men (as in older women) lumbar spine BMD was substantially weaker than hip BMD for prediction of hip fracture. The association between BMD and risk of nonvertebral fracture was at least as strong for men as for women.
There have been very few prospective studies of the association between BMD and risk of fracture in men. Schuitt et al.(3) reported that hip BMD was significantly related to risk of hip and other types of fracture among men ≥55 years old who were recruited from one community in the Netherlands; the associations were similar to those observed in women in the same population. A meta-analysis of several cohort studies that included the results from the Netherland cohort came to similar conclusions.(1) A smaller study by Nguyen et al.(2) also reported that, among men, hip BMD was significantly associated with risk of hip fracture. In that study, the association was similar for men and women.
Our study is the first to show that spine BMD predicts fractures in men. Our results indicate that, as has been found in women, spine BMD is a much weaker predictor of hip fracture than is BMD of the hip. When hip fractures are removed from the nonvertebral fractures, hip and spine BMD have very similar, but modest, associations with risk of fracture.
Our observation that total hip BMD was more strongly associated with the risk of hip and nonvertebral fracture in a cohort of men than in a similar cohort of women is different from the results reported from previous studies. The reason for the difference is not clear. The analysis of Johnell et al.(1) included more (149) hip fractures in men, but the study by Nguyen et al.(2) had fewer than did our study. MrOS started 12 years after SOF, but it seems unlikely that the association between BMD and fracture would change over time or differ in different birth cohorts of older adults. The studies by Schuitt et al.(3) and Nguyen et al.(2) were done in men and women from single communities, whereas MrOS and SOF recruited volunteers from several cities with three of the four or five sites common to both cohorts. To account for this potential effect, we adjusted our results for clinical sites. The differences we observed between men and women was caused, in part, by the very high risk of hip fractures that we observed among the men with very low BMD; it is not clear whether such subjects are overrepresented in MrOS or under-represented in other studies.
There has been debate about whether T scores in men should be based on comparisons with young adult means for men or women.(8) As observed in our studies, men and women have different T score/nonvertebral fracture risk relationships.(8) From a clinical perspective, it would simplify clinical decision-making useful if a certain T score or BMD value corresponded to the same risk of fracture in men and women. The risk of fractures in men and women with the same T score was quite different at younger ages, regardless of BMD. The ratio of these rates decreased with age, partly reflecting the fact that the rates of fractures in men (who serve as the denominator for the comparison) rise with advancing age.
Rather than searching for ways to express BMD in a fashion that produces similar absolute rates for men and women, we believe that the most useful clinical approach is to use BMD and age to directly estimate a man or woman's risk of hip and other types of fractures, as estimated for 3 years in Fig. 2 and has been previously published for hip, nonvertebral, and vertebral fractures for women.(9) These estimates may be improved by including risk factors for fractures.(10,11)
MrOS and SOF have many strengths. They are prospective studies of volunteers from several communities, and the rates of follow-up and X-ray validation of fractures have been very high. Nevertheless, this study has limitations. We did not analyze the relationship between BMD and risk of vertebral fractures in men; this study is underway in MrOS. We did not have enough fractures in MrOS to separately compare the associations between BMD and fracture risk for high- and low-trauma fractures. Both cohorts included only people ≥65 years old, and the results cannot be generalized to younger people. Participants in these studies are predominantly white, so we are unable to specifically study the value of BMD for prediction of fractures in nonwhite men or women.
The comparisons between men and women are also limited because the MrOS study began 12 years later than SOF. The higher prevalence of smoking in SOF than in MrOS may reflect that declining rates of smoking during that period. However, it is unlikely that changes in lifestyle or the environment would substantially alter the relationships between BMD and risk of fracture observed in these two prospective studies. BMD was measured on two different models of Hologic densitometers. However, we corrected the BMD based on in vivo cross-calibrations between the machines used in these studies. The models used in SOF used a pencil-beam mode, whereas those in MrOS used a fan-beam mode, but it is not clear how this could alter the comparisons of the association between BMD and risk of fracture in men and women.
We conclude that BMD is a very strong predictor of hip fracture in men and predicts a man's overall risk of nonvertebral fracture. This association between BMD and risk of hip and other nonvertebral fractures is at least as strong in men as in women. In older men, as in women, hip BMD is a much stronger predictor of hip fracture than is lumbar spine BMD. We found that T scores generally indicate different risks of nonvertebral fracture in men and women, and changing the reference from sex-specific to female values had little impact on these discrepancies. Thus, to have comparable benefit for men and women, decisions about treatment to prevent fracture should be based on estimates of risk rather than T score.
The authors thank Liezi Concepcion and Jamie Low for preparing and editing the paper for publication. The Osteoporotic Fractures in Men (MrOS) Study is supported by National Institutes of Health funding. 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), under the following grants: U01 AR45580, U01 AR45614, U01 AR45632, U01 AR45647, U01 AR45654, U01 AR45583, U01 AG18197, and M01 RR000334. The Study of Osteoporotic Fractures (SOF) is supported by Public Health Service research grants from the National Institutes of Health (AG05407, AR35582, AG05394, AR35584, and AR35583).