Endogenous Sex Hormones and Incident Fracture Risk in Older Men:  The Dubbo Osteoporosis Epidemiology Study FREE

One-third of all osteoporotic fractures occur in men. After the age of 60 years, the residual lifetime risk of hip or vertebral fracture in men is similar to that of prostate cancer.^1- 2 Moreover, in men, a prior osteoporotic fracture increases the risk of subsequent fracture by 3- to 4-fold to at least that of a woman of the same age with a fracture.³ Preventing the first such fracture may have major public health implications. Thus, understanding the determinants of fracture risk in men may reduce the burden of disease through facilitating better prevention strategies.

Male aging is associated with a gradual decrease in circulating testosterone,⁴ which may be detrimental to bone.⁵ However, the relationship between testosterone and incident fracture risk remains unclear. A recent study⁶ from Sweden reported that free testosterone within the normal range was independently associated with prevalent osteoporotic fractures in elderly men. In contrast, a subset analysis from the Rotterdam Study⁷ failed to confirm an association between testosterone and fracture risk. Data from the Framingham Study recently indicated a synergistic effect of sex hormones on fracture risk in that men with low serum testosterone and low estradiol (E₂) levels were at increased risk for incident hip fractures. Analyses restricted to either sex hormone alone, however, revealed that in elderly men, serum E₂ but not testosterone was associated with hip fracture risk.⁸ In these earlier studies,^7- 8 however, serum testosterone levels were measured using immunoassay-based methods that have been shown to be unreliable, particularly in the lower concentration range.⁹

The present study was aimed at determining whether endogenous sex hormones, measured by tandem mass spectrometry, are associated with incident vertebral and nonvertebral fractures over a median of 5.8 years in a large cohort of ambulatory community-dwelling men older than 60 years and determining how such effects relate to known fracture risk factors.

The present analysis is part of the ongoing Dubbo Osteoporosis Epidemiology Study.^1,3,10 Briefly, since 1989, all men and women 60 years or older living in Dubbo, a regional city of 32 000 predominantly white people in New South Wales, Australia, were invited to participate in the study. The age and sex distribution of the Dubbo population closely resembles that of the general Australian population.

By July 2004, 868 men were participating and followed up at approximately 2-year intervals and, of these men, 609 (70.2%) had a serum sample available. Baseline characteristics (age, weight, height, and bone mineral density [BMD]) were comparable between participants (men with serum samples available) and nonparticipants (men without samples [n = 259]). Study enrollment was defined as the time point at which a first blood sample was available. Men with incident symptomatic minimal trauma fractures during the prospective follow-up until December 2005 formed the fracture group (n = 113), whereas men without incident fractures during the same interval were defined as nonfracture controls (n = 496). The study was approved by St Vincent's Hospital Ethics Review Committee; all subjects gave written informed consent.

Participants were interviewed by a nurse coordinator who obtained anthropometric variables and administered a questionnaire to collect data on lifestyle factors and calcium intake.¹¹ Bone mineral density was measured at the lumbar spine (LSBMD) and the femoral neck (FNBMD) by dual-energy x-ray absorptiometry (LUNAR DPX-L; GE-LUNAR, Madison, Wisconsin). The same densitometer was used throughout the study; the coefficient of variation for the BMD measurements was 1.3% at the lumbar spine and 3.5% at the femoral neck.¹²

Dubbo has 3 radiological services to which the study has access. Therefore, all fractures that occurred within the city could be ascertained. Circumstances surrounding fracture were determined by telephone call after the fracture. All fractures included in the study were low-trauma fractures associated with a fall from standing height or less and were confirmed by radiograph. Because the first available blood sample may have been collected at the second or even third visit, some fractures had occurred before that sample. All fractures occurring before the first blood collection were defined as prevalent (ie, prior fracture history), whereas all fractures occurring thereafter were defined as incident. There was no systematic x-ray screening to identify asymptomatic vertebral fractures. All vertebral fractures in the present study were clinically symptomatic fractures, identified in individuals who underwent x-ray screening for back pain or symptoms that showed a vertebral deformity. Similar deformities identified in radiographs taken for other reasons were considered to be prevalent fractures.

Nonfasting blood samples were collected mostly in the morning. Serum samples were stored at −80°C until analysis. Samples were analyzed for serum testosterone and E₂ levels using a liquid chromatography tandem mass spectrometry method. The limit of quantitation for E₂ was 1.5 pg/mL (to convert to picomoles per liter, multiply by 3.671) and for testosterone it was 3.0 ng/dL (to convert to nanomoles per liter, multiply by 0.0347). Method imprecision was less than 10%.^13- 14 Serum levels of sex hormone–binding globulin (SHBG) were measured by a commercial immunoassay (DELFIA; Wallac Oy, Turku, Finland). Coefficients of variation were 10.2% at high (14.6 μg/mL), 5.3% at midrange (6.4 μg/mL), and 8.3% at low (2.2 μg/mL) concentrations (to convert to nanomoles per liter, multiply by 8.896).

The Cox proportional hazards model was used to characterize the association, and estimate the magnitude of association, between serum testosterone and/or serum E₂ and fracture risk. The association was further adjusted for known risk factors, such as advancing age, baseline weight, FNBMD, prior fracture, dietary calcium intake, and smoking. Femoral neck BMD, rather than LSBMD, was considered because in elderly men FNBMD measurement is less likely to be affected by degenerative changes than LSBMD measurement. However, if FNBMD was replaced by LSBMD, hazard ratios remained essentially unchanged. Because weight and BMD were significantly interrelated (r = 0.41, P < .001) and the effect of sex hormones on fracture risk might be independently mediated through body mass and bone mass, the second model was fitted to the data by replacing weight in the first model with FNBMD. Because the distributions of testosterone, SHBG, E₂, and dietary calcium intake were skewed, these variables were subjected to a log (x + c) transformation, making them normally distributed and stabilizing their variance (x indicates an original variable and c indicates a constant). Because analyses of subcategories of fractions were considered secondary, we did not adjust the respective P values for multiple comparisons. To quantify the impact of serum testosterone on fracture risk, the partial attributable risk fraction was estimated from the multivariate variables.¹⁵ All analyses were done using the R statistical environment on a Windows platform.¹⁶

In total, 609 men (mean [SD] age, 72.6 [5.7] years) had been observed for a median of 5.8 years (range, 0-12.6 years). During the follow-up, 113 men sustained at least 1 low-trauma fracture. Twenty-five men experienced multiple incident fractures. A total of 149 incident fractures were reported, including 55 vertebral, 27 hip, 28 rib, 6 wrist, and 16 upper and 17 lower extremity fractures. The incidence of all new fractures was 3.4 per 100 person-years (95% confidence interval, 3.4-3.5 per 100 person-years); 79.3% of all fractures occurred in men aged 70 years or older, in whom the incidence was 4.7% (95% confidence interval, 4.6%-4.8%), representing a 2.7-fold higher incidence than that seen in those younger than 70 years.

Serum testosterone levels were inversely related to age (r = − 0.13, P = .001) and weight (r = − 0.13, P = .002) and positively related to serum E₂ (r = 0.45, P < .001) and SHBG (r = 0.35, P < .001) concentrations. There were no statistically significant associations between testosterone levels and BMD (LSBMD: P  = .43; FNBMD: P  = .49). Serum E₂ levels were positively related to BMD (LSBMD: r = 0.16, P < .001; FNBMD: r = 0.11, P = .01). Serum SHBG levels were positively correlated with age (r = 0.15, P < .001) and inversely correlated with height (r = − 0.11, P = .007), weight (r  = − 0.30, P < .001), LSBMD (r = − 0.20, P < .001), and FNBMD (r = − 0.23, P < .001).

At baseline and compared with men without fractures, men with incident fractures were older (P < .01), shorter (P = .02), and had lower body weight (P < .01) and had lower dietary calcium intake (P = .002) and a lower LSBMD and FNBMD (P < .01). Serum testosterone levels were significantly lower (median [interquartile range], 383 [282-519] ng/dL vs 403 [297-562] ng/dL; P = .03) and serum SHBG levels were significantly higher (6.2 [4.8-7.6] μg/mL vs 5.4 [4.1-7.1] μg/mL; P = .003) in men with incident fractures compared with men without fractures during follow-up. In contrast, baseline E₂ concentrations were comparable between both groups (P = .20).

In univariate analysis (Table 1), apart from established risk factors (eg, age, weight, prior fracture, and BMD), the hazard ratio of any osteoporotic fracture for each standard deviation decrease in serum testosterone was 1.33 and for each standard deviation increase in serum SHBG was 1.36. Serum E₂ levels were not significantly associated with fracture risk in bivariate analysis.

Low levels of serum testosterone were significantly associated with an increased risk of fracture after adjustment for SHBG, age, and weight; however, when weight was replaced by FNBMD, the association was similar but did not remain statistically significant (Table 2). On the other hand, low levels of serum E₂ were only significantly associated with fracture risk after adjustment for SHBG and not when age and weight or FNBMD were included in the model (Table 2).

In a multivariate analysis with all known risk factors of fracture being considered simultaneously, baseline serum testosterone, SHBG, age, FNBMD, weight, and calcium intake were significantly and independently associated with any low-trauma fracture (Table 3). Each standard deviation decrease in serum testosterone was associated with an increased hazard of fracture by 1.37-fold with weight in the model and 1.28-fold after adjusting for FNBMD and covariates. Depending on whether BMD was taken into account, the fraction of fracture cases attributable to the variation in total testosterone was about either 3% or 6.7%.

The association between serum testosterone and fracture risk was further examined by fracture type (Table 4). After adjusting for covariates and weight, serum testosterone was significantly associated with an increased risk of nonvertebral fracture and hip fracture, but not with symptomatic vertebral fracture. However, in a model with adjustment for covariates and FNBMD, serum testosterone was only associated with hip fracture risk.

The association between serum E₂ and fracture risk was noted in the model with covariates and body weight: a decrease in E₂ by 1 SD was associated with an increased hazard of fracture. However, when weight was replaced by FNBMD, the relationship of serum E₂ with fracture risk was no longer significant (Table 3). Analysis stratified by fracture site did not reveal any significant association between serum testosterone and hip fracture, nonvertebral fracture, and symptomatic vertebral fracture after adjusting for covariates and weight or FNBMD.

When the entire sample was analyzed by quartiles of baseline serum testosterone levels, adjusted for major independent covariables, the risk for any fracture tended to increase with decreasing concentrations of circulating testosterone. The hazard ratio was higher in men with testosterone levels in the lowest quartile (< 294 ng/dL) compared with men with testosterone levels in the highest quartile (Figure 1A). When the entire sample was analyzed by quartiles of baseline serum SHBG levels (adjusted for covariates), a nonsignificant increase in fracture risk with increasing baseline SHBG levels was observed (Figure 1B). Quartile analyses for serum E₂, adjusted for covariates, showed no significant association with fracture risk (Figure 1C).

Based on their baseline serum testosterone levels, subjects were then grouped into 3 categories: low testosterone group (quartile 1, < 294 ng/dL), intermediate testosterone group (quartiles 2 and 3, 294-559 ng/dL), and high testosterone group (quartile 4, > 559 ng/dL). Time-to-event (time-to–first fracture) analyses demonstrated that men in the lowest testosterone group had a greater risk of low-trauma fracture during the follow-up than did men in the highest testosterone group. Fracture risk did not differ between the groups with intermediate and high serum testosterone levels (Figure 2 and Table 5). After adjustment for age, weight, fracture history, smoking status, calcium intake, and serum SHBG levels, the risk of fracture was more than doubled in men with serum testosterone levels lower than 294 ng/dL compared with men with high serum testosterone levels. Similar time-to-event analyses for serum unadjusted or adjusted E₂ levels showed no difference in fracture risk between lower and higher E₂ groups (Table 5).

The present prospective study shows that men with lower serum testosterone levels had increased risk of osteoporotic fracture, and this effect was independent of established risk factors, such as age and BMD. In contrast, there was no significant association between serum E₂ levels and fracture in the presence of BMD and age.

Observational studies^4,17- 18 suggest that male aging is associated with a gradual decrease in circulating testosterone levels. However, the clinical significance of this change remains unclear, as does any justification for testosterone treatment.¹⁹ Data relating serum testosterone levels to the prospective risk of osteoporotic fractures in community-dwelling older men are scarce and limited. A recent population-based study⁷ of 178 elderly men from the Rotterdam Study detected no association between sex hormone levels and fracture risk in men; however, the statistical power of this study was inadequate for any firm conclusions. In the Framingham Study,⁸ men with low serum testosterone levels tended to have a higher risk of hip fracture; however, the risk was only observed among a subset of men with low serum testosterone and low serum E₂ levels.

In the present prospective study, baseline serum testosterone was a risk factor for osteoporotic fractures independent of age and BMD. After adjusting for established risk factors, the risk of any fracture increased between 30% and 40% for each 1-SD lowering in serum testosterone levels. Circulating testosterone was associated with the risk of incident nonvertebral fractures, including hip fractures, and similarly, but not significantly, with the risk of vertebral fractures. This difference in the strength of statistical association might reflect the fact that vertebral fractures were clinically diagnosed without systematic x-ray screening to identify asymptomatic vertebral deformities. Therefore, the number of nontraumatic vertebral fractures is likely to be underestimated. Clinical^6,20 and histomorphometric²¹ data suggest that testosterone may have a different influence on BMD at cortical than trabecular bone sites.

The association between circulating testosterone and fracture risk is only partially understood. Several community-based observational studies^20,22- 25 suggest that the association between blood E₂ levels and BMD or bone turnover in men is stronger than the association with serum testosterone concentrations, perhaps reflecting the role of aromatization in the effects of testosterone on bone. Earlier longitudinal studies suggested that low serum E₂ levels are associated with an increased vertebral²⁶ and hip⁸ fracture risk in elderly men. In these studies, measurements of sex hormones were performed using immunoassay-based methods, which are known to be particularly unreliable in the lower concentration range.⁹ Based on a highly specific and sensitive tandem mass spectrometry technique, our results show only a weak relationship between serum E₂ levels and fracture risk in older men.²⁷ From these results, a role, albeit modest, of local aromatization of testosterone to E₂ in bone cannot be excluded.

Lower BMD and accelerated bone resorption have been shown to be independent determinants of fracture risk in elderly men. However, the 2 factors collectively accounted for only 20% of fracture cases in the general population.¹⁰ Serum testosterone levels and BMD were independently associated with fracture risk in the present study; therefore, the effects of testosterone are only partially explained by its changes on BMD. Low serum testosterone levels may result in decreased bone strength by affecting determinants of bone quality, including bone geometry. Also, in addition to their skeletal effects, androgen actions affect body composition; thus, androgen deficiency results in decreased muscle mass,^28- 29 thereby potentially contributing to fracture risk via impaired balance and decreased muscle strength, presumably through increased falls and less effective protective actions.^30- 31

Androgen deficiency in the aging male has become a topic of growing interest and has led to a marked increase in the prescribing of testosterone products in the United States, if not elsewhere.^32- 33 Three randomized placebo-controlled studies^34- 36 have evaluated the effects of testosterone on bone turnover and BMD in elderly nonosteoporotic men with low to normal testosterone concentrations. The effects varied according to testosterone dose and the degree of prior androgen insufficiency.³⁵ Our data suggest that if androgen therapy for the prevention of fragility fractures in healthy elderly men is justified on efficacy and safety grounds, it is most likely to be justifiable only in those with the most severe testosterone deficiency.

The present study's findings should be interpreted within the context of its strengths and limitations. Clearly, our findings should not be interpreted as an indication of any causal relationship between circulating testosterone and fracture risk. Although the homogeneity of ethnicity in the study is a strength, its results cannot be generalized to other populations. As is true for any long-term longitudinal study, serum samples were stored in the freezer for up to 13 years. While testosterone and E₂ are known to be stable over extended periods of storage at −80°C, SHBG levels may change slightly with time in storage. However, these latter changes are marginal and more likely because of changes in assay technique than in the analyte itself. Finally, serum was not collected consistently in the morning, which could introduce random measurement error. However, the circadian rhythmicity of testosterone synthesis is lost in older men³⁷; therefore, diurnal variability was unlikely to have had a significant effect on these results.

We conclude that, while low levels of E₂ and testosterone, determined by tandem mass spectrometry, were associated with an increased fracture risk in community-dwelling men older than 60 years, only the effect of testosterone was independent of FNBMD and other established risk factors of osteoporotic fracture. While testosterone may affect fracture risk via skeletal and nonskeletal mechanisms, the present findings suggest that measurement of serum testosterone provides additional clinical information for the assessment of fracture risk in elderly men.

Correspondence: Christian Meier, MD, Bone Research Program, ANZAC Research Institute, University of Sydney, Concord, New South Wales 2139, Australia (christian.meier@unibas.ch).

Accepted for Publication: August 20, 2007.

Author Contributions:Study concept and design: Meier, Nguyen, Handelsman, Center, Eisman, and Seibel. Acquisition of data: Nguyen, Handelsman, Kushnir, Center, Eisman, and Seibel. Analysis and interpretation of data: Meier, Nguyen, Handelsman, Schindler, Rockwood, Meikle, Eisman, and Seibel. Drafting of the manuscript: Meier, Nguyen, Handelsman, Eisman, and Seibel. Critical revision of the manuscript for important intellectual content: Meier, Nguyen, Handelsman, Schindler, Kushnir, Rockwood, Meikle, Center, Eisman, and Seibel. Statistical analysis: Meier, Nguyen, Handelsman, and Schindler. Obtained funding: Nguyen, Center, Eisman, and Seibel. Administrative, technical, and material support: Meier, Nguyen, Handelsman, Kushnir, Rockwood, Meikle, and Seibel. Study supervision: Nguyen, Eisman, and Seibel.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the National Health and Medical Research Council of Australia; and the ARUP Institute for Clinical and Experimental Pathology. Dr Meier is the recipient of medical research fellowship 81BS-67544 from the Swiss National Science Foundation and a research fellowship from the Margarete und Walter Lichtenstein Stiftung der Universität Basel.

Role of the Sponsor: The funding bodies had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.

Additional Contributions: David Haynes, Janet Watters, Jodie Rattey, and Shaye Field (Garvan Institute of Medical Research) James Modzelewski and Mark Jimenez (ANZAC Research Institute), and Marina Varshavsky (ARUP Institute for Clinical and Experimental Pathology) provided technical assistance.