Hormonal and Biochemical Parameters and Osteoporotic Fractures in Elderly Men


  • Dr. Jacqueline R. Center,

    Corresponding author
    1. Bone and Mineral Research Division, Garvan Institute of Medical Research, Sydney, Australia
    • Bone and Mineral Research Program, Garvan Institute of Medical Research, St. Vincent's Hospital, 384 Victoria Street, Sydney, NSW 2010 Australia
    Search for more papers by this author
  • Tuan V. Nguyen,

    1. Bone and Mineral Research Division, Garvan Institute of Medical Research, Sydney, Australia
    Current affiliation:
    1. Liverpool Hospital, Liverpool, Sydney, Australia
    Search for more papers by this author
  • Philip N. Sambrook,

    1. University of Sydney, Royal North Shore Hospital, St. Leonards, Sydney, Australia
    Search for more papers by this author
  • John A. Eisman

    1. Bone and Mineral Research Division, Garvan Institute of Medical Research, Sydney, Australia
    Search for more papers by this author


Low testosterone has been associated with hip fracture in men in some studies. However, data on other hormonal parameters and fracture outcome in men is minimal. This study examined the association between free testosterone (free T) estradiol (E2), sex hormone-binding globulin (SHBG), 25-hydroxyvitamin D [25(OH)D], parathyroid hormone (PTH), insulin-like growth factor I (IGF-I), and fracture in 437 elderly community-dwelling men. Age, height, weight, quadriceps strength, femoral neck bone mineral density (FN BMD), and fracture data (1989–1997) also were obtained. Fractures were classified as major (hip, pelvis, proximal tibia, multiple rib, vertebral, and proximal humerus) or minor (remaining distal upper and lower limb fractures). Fifty-four subjects had a fracture (24 major and 30 minor). There was no association between minor fractures and any hormonal parameter. Risk of major fracture was increased 2-fold for each SD increase in age, decrease in weight and height, and increase in SHBG, and risk of major fracture was increased 3-fold for each SD decrease in quadriceps strength, FN BMD, and 25(OH)D (univariate logistic regression). Independent predictors of major fracture were FN BMD, 2.7 (1.5–4.7; odds ratio [OR]) and 95% confidence interval [CI]); 25(OH)D, 2.8 (1.5–5.3); and SHBG, 1.7 (1.2–2.4). An abnormal value for three factors resulted in a 30-fold increase in risk but only affected 2% of the population. It is not immediately apparent how 25(OH)D and SHBG, largely independently of BMD, may contribute to fracture risk. They may be markers for biological age or health status not measured by methods that are more traditional and as such may be useful in identifying those at high risk of fracture.


ALTHOUGH COMMONLY seen as a disease of postmenopausal women, recently osteoporosis and its clinical endpoint of fracture has been recognized as an important disease of elderly men. It has been estimated that up to one-third of all fractures occur in men.(1) The social and economic burden of osteoporotic fractures is an increasing public health concern as an aging population compounds the age-specific increase in fractures observed for both men and women. A deeper understanding of the pathogenesis of osteoporosis and fracture will aid in rational prevention and treatment of this disease, especially in men, who have not been studied as much as women.

Bone mineral density (BMD) is one of the best predictors of osteoporotic fracture. A 1 SD decrease in BMD results in a 1.5- to 2.5-fold increase in fracture risk.(2–4) However, bone density incompletely separates those who will fracture from those who will not fracture. Other risk factors such as the quality of bone as well as the propensity to fall, also are determinants of fracture risk, although they are harder to quantify. There has been more recent evidence that hormonal and biochemical parameters may also affect both bone density and fracture risk. However, the data concerning fracture risk, particularly in men, is scarce.

Hypogonadism is an established risk factor for fractures in women and estrogen therapy has been shown to decrease the rate of fracture in postmenopausal women.(5) In men, several case-control studies also have shown a relationship between hypogonadism and hip fractures,(6, 7) although the benefits of replacement therapy in elderly men is not yet clear.

Low levels of vitamin D (25-hydroxyvitamin D [25(OH)D]) and secondary hyperparathyroidism also have been associated with hip fractures in the elderly,(8–10) but the populations under examination have consisted predominantly of women. One Dutch study found lack of sunshine exposure to be the primary determinant of the low vitamin D status in the hip fracture subjects even though the majority of the subjects were living independently.(9) Supplementation of vitamin D and calcium for elderly nursing home women was reported to reduce hip fracture rate(11, 12) but in a study of elderly men and women who were either living independently or in homes for the aged, vitamin D supplementation had no effect on hip fracture rate.(13)

The aim of the present study was to determine if there were any association between occurrence of fracture and sex hormone levels, calcitropic hormones, and growth factors in elderly ambulatory men and whether any such effect was dependent or independent of bone density.



Subjects in this study were participants of the Dubbo Osteoporosis Epidemiology Study, a prospective community-based study, details of which have been previously described.(1, 14, 15) The study was carried out in Dubbo, a semiurban city 400 km northwest of Sydney, Australia, with a population of approximately 32,000 of whom 98.6% are white. The community is relatively stable with its own centralized health services, making it suitable for epidemiological study. Although open to the whole population, all participants were of white background. The study commenced in 1989 and recruited men and women over the age of 60 years. From 1993, blood was collected from subjects at initial visits or follow-up visits. This study reports on the first 437 men who had blood samples available for analysis.

Clinical data collection

Subjects were interviewed by a nurse coordinator at initial and subsequent visits at approximately 2-year intervals. A structured questionnaire was used to collect data including lifestyle factors such as cigarette smoking (pack-years of use) and alcohol use. Anthropometric variables including weight and current height were measured at each visit. Quadricep strength (maximum isometric contraction) was measured in the dominant leg of the seated subject with a horizontal spring gauge calibrated to a maximum 50 kg force.

BMD (g/cm2) was measured at the femoral neck (FN) and lumbar spine (LS) by dual-energy X-ray absorptiometry with a LUNAR DPX-L densitometer (LUNAR Corp., Madison, WI, U.S.A.). The measurement was performed at the same visit or within 1 month of the blood sample collection in 99% of cases. In those cases in which the blood was taken at a time other than at the scheduled visit, the BMD used was that closest to the time of blood collection. The CV for the BMD measurement in normal subjects at our institution is 1.3% at the LS and 3.5% at the FN.(16)

Fractures were recorded from the two radiology services supplying the Dubbo area and the circumstances surrounding the fracture were determined by personal interview after the fracture. All fractures included in the study were low-trauma fractures caused by a fall from a standing height or less. Vertebral fractures collected were those coming to clinical attention. There was no systematic X-ray screening before the study to identify all prevalent or asymptomatic vertebral fractures. Incidentally found, that is, asymptomatic, vertebral fractures were included, provided there was no known malignancy or metabolic bone disease. One fracture subject found to have a malignancy was excluded. No other subject had known cancer or metabolic bone disease. Because of the design of the study and because the blood was collected at the time of the second or even third visit, fractures occurred both before and after the data collection. Information on fractures was gathered from the onset of the study in June 1989 until March 1997. Fractures were divided into major and minor according to type of fracture: major fractures included hip, pelvic, distal femur, proximal tibia, multiple rib, vertebral, and proximal humerus; minor fractures included all remaining peripheral fractures.

Measurement of biochemical parameters

Nonfasting venous samples were separated and stored at −80°C. Assays were performed during 1996 and 1997 for the following parameters: free testosterone (free T), estradiol (E2), sex hormone-binding globulin (SHBG), parathyroid hormone (PTH), 25(OH)D, and insulin-like growth factor 1 (IGF-1). Free T was measured as a solid-phase unextracted analogue radioimmunoassay (RIA) using125I-testosterone analogue as the labeled analyte with interassay CV of 9% (Coat-A-CountR Free Testosterone; Diagnostics Product Corp., Los Angeles, CA, U.S.A.). PTH was measured by a solid-phase two site immunoenzymometric assay also from Diagnostics Products Corp., with interassay CV of 6%. E2 was measured by commercial RIA with a polyethylene glycol (PEG)-enhanced separation step with interassay CV of 10% (Sorin Biomedica Diagnostics S.P.A., Saluggia, Italy). The interassay limit of detection for E2 was 15 pmol/liter. SHBG was measured by two-site immunoradiometric assay from Orion Diagnostica (Espoo, Finland) with interassay CV of 6%. 25(OH)D was measured using an in-house pre-extraction competitive binding protein assay with charcoal separation with interassay CV of 12%. IGF-I was measured using a commercial pre-extraction RIA with a PEG-enhanced second antibody separation step (Bioclone Australia Pty, Ltd., NSW, Australia) with interassay CV of 10%.

Statistical analysis

The major and minor fracture groups were compared with their non-fracture counterparts using two-sided t-tests. A significant difference was only considered for p < 0.05. Fracture risk was determined by logistic regression analysis. None of the biochemical parameters related to minor fractures, so risk for major fracture only was considered in logistic regression models and subjects with minor fractures were excluded from the modeling. Forward and backward models, after adjustment for FN BMD were used to obtain the most empirical model in multiple logistic regression analysis. Separate analyses also were performed according to whether fractures occurred before or after the blood sample collection.

Population attributable risk percentages were calculated for each of the independent risk factors determined in multiple regression analysis for major fracture. In this calculation, the risk factors were analyzed as dichotomous variables. Attributable risk percent was calculated as p(RR − 1)/[p(RR − 1) + 1], where p is the proportion of subjects with the specified risk factor in the population considered and RR is the relative risk of major fracture for subjects with the specified risk factor, estimated from the logistic regression model. In this formulation, attributable risk percent is the proportion by which the outcome (major fracture) in the population would be reduced if the risk of a particular risk factor (measured by p) were eliminated.


Thirty subjects sustained a minor fracture and 24 sustained a major fracture, leaving 383 participants who did not suffer a fracture. Major fractures included 7 hip fractures, 2 pelvic fractures, 12 vertebral fractures, 2 humeral fractures, and 1 fracture of the proximal tibia. The majority of fractures (77%) occurred before the blood sample was collected with the time interval (mean ± SD) between fracture event and collection of serum being 1.7 ± 1.7 years. Ten fractures (five major and five minor) occurred after the blood was collected.

Those men who had suffered a minor fracture had lower LS BMD (p = 0.001; Table 1) than those who did not fracture. However, there was no other difference between those who had suffered a minor fracture and those who did not sustain a fracture. Thus, this group was not examined further in relation to the biochemical parameters in subsequent regression models.

Table Table 1. Characteristics of the Men According to Fracture Class8
original image

Those who sustained a major fracture were older (3 years; p = 0.0.02), shorter (4 cm; p = 0.04), weighed less (8 kg; p = 0.002), had lower quadriceps strength (9 kg; p < 0.0001), and had lower LS (0.16 g/cm2 [13%]; p = 0.001) and FN BMD (0.15 g/cm2 [16%]; p < 0.0001) than subjects who had not fractured (Table 1).

Major fracture subjects also had higher SHBG (14 nmol/liter; p = 0.01) and lower 25(OH)D (14 nmol/liter; p = 0.0001) levels. PTH was marginally higher in those with major fracture (0.9 pmol/liter; p = 0.08) but there was no difference in sex hormone or IGF-I levels between those with and without major fracture (Table 1).

Major fracture subjects also were analyzed depending on whether the fracture occurred before (retrospective group) or after (prospective group) the blood sample was collected. Both groups still had a lower 25(OH)D and higher SHBG than the nonfracture group, although this did not reach statistical significance in the prospective group of fractures, because of the small numbers (n = 5). In addition, fractures occurring within 6 months before the blood collection were analyzed separately, to determine whether there was any effect of vitamin D on fracture status. There was no difference in the vitamin D levels between the two groups (p = 0.55).

Univariate regression analysis

There was an approximate 2-fold increase in risk of major fracture for each 10 year age increase, 10 kg weight reduction and 5 cm reduction in height (univariate logistic regression, Table 2). Moreover, there was an approximate 3-fold increase in major fracture risk for each 10 kg decrease in quadriceps strength and SD (0.15 g/cm2) decrease in FN BMD. In relation to the biochemical parameters, there was a 1.9-fold increase in risk of major fracture (confidence intervals (CI) 1.3–2.6) for each SD (16 nmol/l) increase in SHBG and a 3.0-fold increase in major fracture risk (confidence intervals (CI) 1.7–5.3) for each SD (22 nmol/l) decrease in 25(OH)D.

Table Table 2. Univariate and Age-Adjusted ORs for Major Fracture Risk Factors8
original image

Multivariate regression analysis

Serum levels of SHBG and 25(OH)D were independent predictors of major fracture, even after adjusting for FN BMD, albeit with slightly lower odds ratios (ORs). For 1 SD decrease in 25(OH)D the OR was 2.8 (95% confidence interval [CI] 1.5–5.3) and for a 1 SD increase in SHBG the OR was 1.7 (95% CI 1.2–2.4). Age was not a significant independent predictor of major fracture risk in this model and adjustment for age did not alter these ORs significantly (Table 3).

Table Table 3. Risk Factors for Major Fracture (Multivariate Analysis)8
original image

Because of the correlation between FN BMD and quadricep strength, its addition only marginally improved the model. There also was no significant change in the point estimates for either SHBG or 25(OH)D after quadricep strength was included. Addition of E2 to models including SHBG did not improve the model fit or alter the point estimate of SHBG.

Attributable risk

The three independent risk factors [FN BMD, SHBG, and 25(OH)D] for major fracture were analyzed as dichotomous variables in order to calculate population attributable risks (Table 4). Osteoporosis was defined as an FN BMD 2.5 SD below the value for young normal males (<0.74 g/cm2); quartile cut-offs were used for both SHBG and 25(OH)D. The lowest quartile of 25(OH)D (<58 nmol/liter) was compared with the upper three quartiles and the highest quartile of SHBG (>194 nmol/liter) was compared with the lowest three quartiles. The attributable risks of major fracture for 25(OH)D and SHBG were 47% and 37%, respectively, comparable with that attributable to the presence of osteoporosis (39%). Because of the low prevalence of subjects having osteoporosis and also being in the high-risk quartile of these biochemical parameters, addition of one of these variables to the presence of osteoporosis or visa versa did not add to and in some cases lessened the individual attributable risks. Indeed, the combination of these three risk factors affected less than 2% of the population and had an attributable risk of 32%, which was less than that of BMD alone (39%) or low 25(OH)D and low BMD (42%). However, this rare combination was associated with a 30-fold increased risk of major fracture.

Table Table 4. Prevalence, Relative, and Attributable Risks of Major Fracture for Groupings of Biochemical Risk Factors8
original image


This study has shown an association between low 25(OH)D and high SHBG levels and the presence of a major fracture, independent of BMD, in elderly ambulatory men. Consistent with previous observations, BMD also was an independent risk factor for major fracture in these men. Recently, high SHBG as well as low 25(OH)D levels have been associated with hip and vertebral fracture in case-control studies, predominantly or only in women.(8–10, 17) However, to our knowledge there are no published data on these parameters and fracture in men.

Of considerable interest, associations between both 25(OH)D and SHBG and risk of major fracture were not significantly altered by adjusting for BMD in the multivariate analysis. This suggests that the major part of their association with fracture is distinct from any effect on BMD, which is one, albeit a major, risk factor for fracture.

The calcium-regulating hormones and vitamin D metabolites [e.g., 25(OH)D] are associated intimately with bone metabolism. Thus, it is quite plausible that perturbations in these variables may affect bone strength. However, it seems unlikely that SHBG per se would have a direct role on bone structure. More likely, SHBG and even possibly 25(OH)D may be markers for some other risk for fracture such as biological age or degree of frailty, which, although recognizable risks, are hard to measure. Both SHBG and 25(OH)D were weakly correlated with quadricep strength in a previous analysis of this study group.(18) However, addition of quadriceps strength in multivariate analysis of major fracture risk did not significantly change the point estimates of either of these parameters, suggesting that in this population any association between SHBG or 25(OH)D and fracture risk is not via strength alone.

There is some evidence suggesting that increasing SHBG may be a marker of increasing biological age or frailty. It has been shown to increase with age in men,(19) consistent with previous findings from the current study population.(18) In women this increase is less certain.(19, 20) SHBG is related to other general metabolic factors as well as age, including nutritional status and growth.(20) There is an inverse relationship between SHBG and both weight(21) and IGF-I levels,(22) consistent with evidence that IGF-I may be an important regulator of SHBG.(19, 20) Thus, in a state of good nutrition or weight gain, high levels of IGF-I and low levels of SHBG are observed.(23) Therefore high levels of SHBG (associated with greater chronological age and low weight) may be a marker for increased biological age or even increased catabolism and as such represent a risk factor for fracture. Obviously, E2 is largely bound to SHBG such that higher SHBG may produce lower free E2 levels. However, SHBG was independently associated with fracture after adjusting for total E2 levels. In addition, initial estimates of free E2 and non-SHBG-bound E2(24, 25) were not associated with fracture in multivariate and stepwise analyses.

Serum levels of 25(OH)D decrease and PTH levels increase with aging although there does not appear to be a difference in 1,25-dihydroxyvitamin D [1,25(OH)2D] levels in either elderly patients(8) or in hip fracture subjects.(8, 9) However, even after adjustment for age and low BMD, low 25(OH)D was still a predictor of major fracture. Hence, although 25(OH)D or PTH may affect bone strength independent of BMD, effects on muscle strength and possibly falls(3) may influence fracture risk indirectly. The difficulty in analyzing the direction of these relationships may relate to more sunlight exposure and thus higher vitamin D levels in ambulatory men versus less sunlight exposure and even poorer diet leading to lower vitamin D levels in less healthy non-ambulatory men. There have been several case-control studies documenting the association between low levels of 25(OH)D and hip fracture in women.(8–10, 26) A benefit of vitamin D therapy on hip fracture was observed in institutionalized(11, 12) but not community-dwelling(13) individuals. However, BMD was either not measured or only measured in a small proportion of subjects. Thus, any effect of vitamin D status on fracture risk, independent of BMD, has not been examined closely.

Surprisingly, there was no association in this study between major fractures and levels of the sex hormones. Hypogonadism has been previously demonstrated in hip fracture subjects.(6, 7) Although the mechanism relating testosterone levels to fracture risk has been postulated to be via its effects on bone density, the present study did not show even a univariate association between either E2 or free T levels and major fracture. Part of the explanation for this finding may be the small numbers involved, although this seems unlikely to be the only explanation, given that the numbers of fracture subjects were lower in several of the reported “positive” studies. Alternatively, the grouping together of major fractures may not have the power to detect what might be observable with a similar number of hip fractures. The population studied here was more healthy than the hip fracture subjects reported in case-control studies, thus decreasing any observed differences in testosterone levels between nonfracture and relatively “healthy” fracture subjects. Even allowing for these factors, the sex hormones did not have as big an impact on fracture risk as SHBG or 25(OH)D, which were risk factors even in this healthy, heterogeneous fracture group. In addition, as previously mentioned, total rather than free E2, the latter arguably a more sensitive measure, was analyzed, which also may have decreased the power of the study to detect a relationship between E2 and major fracture. However, an association between both free testosterone and total estradiol with bone density was previously demonstrated in this population.(18) Thus it is still possible that there is an effect of sex hormones on fracture through bone density, but which was not strong enough to be detected in the present study.

The results of this study should be interpreted within the context of its limitations. Although the subjects studied were part of a prospective cohort study, the fracture data were cross-sectional and thus causal inferences cannot be made. However, when analyzed according to whether the fracture occurred before (retrospective fracture group) or after (prospective fracture group) the blood sample was collected, both the retrospective and the prospective group of major fractures still had a lower 25(OH)D and higher SHBG than the nonfracture group. In addition, consistent with a European study,(26) timing of the fracture did not have any effect on the vitamin D levels. The small number of fractures necessitated grouping into major and minor fracture groups, which would have tended to decrease the strength of any association with any special type of fracture, for example, hip fractures. Thus, subtle relationships with some of the biochemical parameters may not have been observed but the positive results are unlikely to have been exaggerated. These findings need to be confirmed in prospective studies of both men and women.

In conclusion, this study has shown a relationship between levels of 25(OH)D and SHBG and major fracture risk in elderly men independent of BMD. Although the mechanism by which these parameters may affect fracture risk is not immediately obvious, it is possible that they may be markers for biological age or relative health not measured by other traditional methods. This novel hypothesis has the potential to influence future determination of fracture risk and suggests avenues for further exploration of the complex and multifactorial nature of fracture risk.


We acknowledge the help of Dr. R. Slack-Smith and Mr. M. Russell in radiological procedures and Janet Watters and Angela Ferguson in the measurement of bone densitometry. We also acknowledge the support of the staff of Dubbo Base Hospital (particularly Mr. B. Luton and Mr. B. Ayrton). This work was supported by the Australian Institute of Health and National Health and Medical Research Council of Australia. J.R. Center is the recipient of a medical postgraduate scholarship from the National Health and Medical Research Council of Australia.