Testosterone, Dihydrotestosterone, Bone Density, and Hip Fracture Risk among Older Men: The Cardiovascular Health Study

Abstract

Background:

Little is known about the relationships of dihydrotestosterone (DHT), a more potent androgen than testosterone (T), with bone mineral density (BMD) and fracture risk. Our objectives were to evaluate the relationships of T, DHT and sex hormone binding globulin (SHBG) with BMD, fracture risk, and lean body mass (LBM).

Methods:

We evaluated 1128 older men free of cardiovascular disease in a prospective cohort study using data from the Cardiovascular Health Study. T and DHT were measured by liquid chromatography–tandem mass spectrometry and SHBG by fluoroimmunoassay. Our outcomes included incident hip fracture (n=106) over a median of 10.2 years and BMD and LBM by dual-energy x-ray absorptiometry (n=439).

Results:

In Cox regression models mutually adjusted for T, SHBG, and covariates, each standard deviation increment in DHT (0.23 ng/ml) was associated with a 26% lower risk of hip fracture (adjusted hazard ratio [aHR] 0.74, 95% confidence interval (CI) 0.55-1.00, p=0.049). Similarly, SHBG was associated with fracture in mutually adjusted models (aHR HR 1.26, 95% CI, 1.01- 1.58, p=0.045). In contrast, T (aHR, 1.16, 95% CI, 0.86-1.56, p=0.324) was not significantly associated with fracture in mutually adjusted models. T, DHT and SHBG were not associated with BMD. T and DHT were both positively associated with LBM in individual models.

Conclusions:

In older men, DHT was inversely associated with hip fracture risk and SHBG was positively associated with hip fracture risk, while T was not. Future studies should elucidate the mechanisms by which DHT affects bone health.

1. Introduction

Osteoporotic fractures are one of the most common causes of disability and comprise a substantial portion of medical costs worldwide [1]. The lifetime risk of hip fracture after 50 years of age is approximately 17% for Caucasian females and 6% for Caucasian males in the United States [2]. Hip fractures in particular are strongly associated with bone mineral density (BMD) and cause more disability than other types of fracture [1]. Mortality increases after both hip and non-hip fractures, and this mortality risk is even higher in men than women [3–4].

Although the pathogenesis of hip fractures is multifactorial, sex steroids are imperative for growth and maintenance of both female and male skeletons [5]. Male aging is associated with a decrease in serum sex hormones, and this decline has been shown to influence bone health [6–7], although the links between androgen levels in men and BMD and fracture risk remain an ongoing source of debate [6–16]. There is evidence to suggest that estradiol affects BMD and fracture risk, but the contribution of testosterone (T) remains unclear [8, 17, 18].

An important limitation in previous literature has been restriction of androgens tested to total or sometimes bioavailable T. Also, few studies have evaluated the effect of dihydrotestosterone (DHT), the product of 5α-reductase activity, on bone health. Accurate assessment of circulating DHT has only been possible with the advent of mass spectrometry-based DHT assays that can measure the small concentrations of DHT in blood. DHT differs from T in its higher potency and affinity and slower dissociation from the androgen receptor [19]. Although a few studies have suggested little incremental benefit to measurement of DHT in assessing fracture risk [9], we have found DHT to be more strongly associated than T in its associations with several forms of chronic disease [20–22].

To elucidate the relationships of total and free T and DHT, and their transport protein sex-hormone binding globulin (SHBG), with BMD and risk of hip fracture we studied men enrolled in the Cardiovascular Health Study. We hypothesized that both high SHBG and low DHT levels would be associated with increased fracture risk. We hypothesized that testosterone would not have a relationship with hip fracture risk.

2. Methods

2.1. Participants

The Cardiovascular Health Study (CHS) is a multicenter, prospective study of men and women 65 years and older recruited from population-based, Medicare-eligibility lists from clinic sites in Pittsburgh, PA, Sacramento, CA, Hagerstown, MD and Forsyth County, NC. Participants were not institutionalized or wheelchair-dependent, were not receiving cancer treatment at the time of enrollment, were expected to remain in the region from which they were recruited for at least three years and did not require a proxy for consent. In 1989-1990, 5201 predominantly Caucasian and African American participants were recruited and examined, and in 1992-1993, 687 predominantly African American participants were recruited into the study. The baseline examination included standardized medical history questionnaires, physical examination, and laboratory examination; these procedures were repeated in the original cohort in 1992-1993 when the African American cohort was added. Follow-up contact occurred every six months, alternating between telephone calls and clinic visits through 1999 and telephone calls every six months thereafter. Informed consent was obtained from all study participants. The study sample consisted of men participating in the CHS 1994 clinical exam who had no history of cardiovascular disease (myocardial infarction, coronary artery bypass grafting, percutaneous coronary intervention, heart failure, or stroke). Individuals without measured hormones were excluded from the study.

2.2. Sex steroid hormone levels

Levels of total T, total DHT, and SHBG were measured at the Seattle VA Puget Sound Health Care System, as described previously [10], for an ancillary study of sex steroid hormones and incident cardiovascular disease. As a result, hormone levels were measured on stored serum specimens collected in 1994-1995 among men free of myocardial infarction, coronary artery bypass grafting, percutaneous coronary intervention, heart failure, or stroke as of the 1994-1995 visit. The time of collection was not consistently recorded, but most samples were collected before noon. Samples were stored at −70 degrees Celsius at the CHS Central Laboratory in Burlington, VT. In 2010, the frozen samples were shipped on dry ice for hormone assays. Previous studies suggest that T and DHT are stable in prolonged frozen storage [23].

All assays were conducted in duplicate, and average values were used in analyses. Total T and DHT were measured simultaneously with a liquid chromatography–tandem mass spectrometry assay. The lower limit of detection for total T was 1.0 ng/dl, with an intra-assay coefficient of variation (CV) of 4.9% and an inter-assay CV of 5.1%. The T assay was certified for accuracy by the CDC Hormone Standardization Program. The lower limit of detection for total DHT was 0.02 ng/ml, with an intra-assay CV of 5.9% and an inter-assay CV of 6.2%. For total T, CVs were determined at 5.5 ng/ml; for DHT, CVs were determined at 0.30 ng/ml. We measured SHBG by a time-resolved fluoroimmunoassay (Delfia, Perkin Elmer, Norton, OH). The lower limit of detection for SHBG was 0.5 nmol/L, with an intra-assay CV of 1.4% and an inter-assay CV of 6.6% at 31 nmol/L [10].

We calculated free T and free DHT by the Mazer method [24]. We used the Mazer formula because it allowed calculation of the free fractions of both hormones.

2.3. Outcomes

Our primary outcome was incident hip fracture, while our secondary outcomes were total hip and femoral neck BMD and lean body mass (LBM) measured by dual-energy x-ray absorptiometry (DXA).

Incident hip fracture was identified using International Classification of Diseases, Ninth Revision (ICD-9) codes from hospitalization records from the time of the 1994/1995 visit through June 30^th, 2013, a method widely validated in previous studies [25]. Hospitalization data were collected every six months and checked against Medicare claims data to find hospitalizations that may not have been reported by a participant. Hip fracture was identified as ICD-9 code 820.xx in any position [26].

A subset of participants at the Sacramento and Pittsburgh field centers underwent DXA scanning in 1994-1995. Participants were offered scanning in the order in which they came to the study centers for their annual visits until funding was exhausted. Total hip and femoral neck BMD and LBM were measured using Hologic QDR-2000 densitometers (Hologic, Inc., Waltham, MA). Scans were read in blinded fashion using Hologic software, version 7.10 at the University of California, San Francisco. Additional details about the DXA protocol are detailed elsewhere [26, 27].

2.4. Covariates

Detailed descriptions of the data collection methods have been reported previously [22, 28]. Covariates were chosen based on their association with bone health. Most covariates, including clinic site, age, sex, alcohol consumption, smoking status, and weight were measured at the 1994 visit. Some covariates, including height, were measured at the 1992 visit. Medications were collected annually using a validated inventory [29].

2.5. Statistical analyses

We contrasted the subset of individuals with available hormone levels and those without them. Baseline characteristics were computed overall and by quartiles of DHT. We used χ² tests to compare categorical variables and a linear trend test across quartiles of DHT for continuous variables. We also examined the unadjusted Pearson correlation between T and DHT.

Cox hazard models were used to estimate the hazard ratio (HR) of incident hip fracture associated with hormones. Several nested models were used. First, we adjusted for age, race, clinic site, height, and weight. We opted to adjust for weight and height as opposed to BMI as CHS has previously shown that weight is a better predictor than BMI for bone-related outcomes and that adjustment for individual terms is preferred to adjustment for their ratio [30– 31]. Second, we added adjustment for model 1 covariates along with current smoking, smoking pack-years, cystatin-based estimated glomerular filtration rate, daily hours spent sitting or lying down, alcoholic drinks per week, diabetes, and hypertension. All hormones were included in the second model as well to estimate their independent contributions. We carried forward covariate values from previous examinations for missing data at baseline (<2%).

In sensitivity analyses, we added adjustment for thyroid medication, thiazides, and loop diuretics, walking pace and number of reported falls over the past year, and excluded men using 5-α-reductase inhibitors or oral corticosteroids, and those with prostate cancer. We also conducted analyses that additionally adjusted for a quadratic term for age and for a potential interaction of age with weight. For cross-interpretability, we present estimates per standard deviation in each hormone. We also present results for free (rather than total) levels of hormones.

We performed linear regression with identical models as described above to determine the relationship between hormones and BMD and LBM.

We used splines in generalized additive models to address the functional form of hormones in the models; we found no meaningful departures from linearity.

Analyses were conducted using R [32].

3. Results

3.1. Baseline characteristics

There were 4,842 individuals who attended the 1994/1995 CHS visit: we excluded women and men without hormones measured, which included men with a history of cardiovascular disease (defined as myocardial infarction, coronary artery bypass surgery, percutaneous coronary intervention, heart failure or stroke), leaving 1128 men in the sample. Table 1 displays baseline characteristics for included participants based on quartile of DHT. Average DHT among all participants was 0.44 ng/ml (SD 0.23 ng/ml). Those with high DHT tended to be thinner than those with low DHT and those with lower DHT were more likely to have diabetes and hypertension. None of the participants reported taking bisphosphonate therapy or testosterone therapy. Those with lower DHT were more likely to be on steroids or alpha reductase inhibitors. The unadjusted correlation between T and DHT was 0.696 (95% confidence interval [CI], 0.665-0.725).

Table 1.

Baseline Characteristics According to Quartile of Total DHT

	All participants	Dihydrotestosterone (DHT) by quartile				P value
Characteristic	N=1128	Q1 N=302 (0.01-0.29)*	Q2 N=269 (0.30-0.41)*	Q3 N=277 (0.42-0.55)*	Q4 N=280 (0.56-1.87)*
Age, years	76.5 (5.1)	76.9 ± 5.5	76.5 ± 5.0	76.1 ± 4.9	76.6 ± 5.1	0.31
Weight, kg	80.0 (12.4)	84.4 ± 13.6	81.2 ± 11.3	78.9 ± 11.8	75.1 ± 10.5	<0.01
Height, cm	172.9 (6.5)	172.9 ± 7.0	173.1 ± 5.8	173.1 ± 6.5	172.5 ± 6.6	0.48
Body Mass Index, kg/m2	26.6 (3.6)	28.1 ± 3.9	27.0 ±3.4	26.2 ± 3.4	25.1 ± 3.1	<0.01
Body Surface Area, m2	1.95 (0.17)	2.01 (0.18)	1.97 (0.15)	1.94 (0.17)	1.89 (0.15)	<0.01
eGFRcys, mL/min/1.73m2	73.1 (17.1)	71.1 ± 17.2	71.8 ± 17.9	74.2 ± 15.5	75.2 ±17.6	<0.01
Sitting**, hours/day	14.6 (3.6)	14.8 ±3.6	14.5 ± 3.6	14.5 ± 3.7	14.6 ± 3.5	0.62
Black race	14.7%	15.9%	13.8%	13.0%	16.1%	0.66
Education (≥ 12th grade)	51.8%	48.3%	52.6%	52.7%	53.8%	0.56
Current Smoker	11.8%	10.8%	12.3%	11.6%	12.7%	0.91
Diabetes	9.9%	16.9%	8.6%	6.9%	6.8%	<0.01
Hypertension	52.6%	63.6%	55.0%	45.8%	45.0%	<0.01
Levothyroxine use	4.9%	4.6%	5.6%	2.9%	6.5%	0.24
Oral steroid use	2.8%	6.0%	3.7%	1.1%	0.0%	<0.01
Finasteride use	2.4%	8.6%	0.4%	0.0%	0.0%	<0.01
Testosterone (ng/dl)	381.7 (181.1)	225.6 (140.6)	344.3 (94.9)	421 (100.7)	546.5 (189.4)	<0.01
Dihydrotestosterone (ng/ml)	0.33 (0.23)	0.19 (0.09)	0.36 (0.03)	0.48 (0.04)	0.75 (0.19)	--

^*DHT quartile ranges, units ng/ml

^**Sitting=Number of hours spent sitting or lying down each day

Continuous variables are shown as mean (± standard deviation)

Categorical variables are shown as percent

On average, participants were followed for 10.2 years (interquartile range 5.9-15.5 years). We documented 106 cases of hip fracture among the 1128 men, for an incidence rate of 0.89 per 100 person-years (95% CI, 0.66-1.21).

In general, the participants included in our study were older, had better renal function, were less likely to be smokers and were less likely to have diabetes than those not included in the study.

3.2. Androgens and fracture risk

Table 2 provides hazard ratios for risk of hip fracture per standard deviation increase in T, DHT, and SHBG. In mutually adjusted models (Model 2), each standard deviation increment in DHT was associated with a 26% lower risk of hip fracture (adjusted HR per SD increment 0.74, 95% CI, 0.55-1.00 p=0.049). When further adjusting for other covariates such as the use of thyroid medication, thiazide, loop diuretics, walking pace and number of falls over the past year, DHT remained significantly associated with lower risk of fracture (adjusted HR 0.69, 95% CI, 0.50-0.97, p = 0.033). In contrast, total T (mutually adjusted [Model 2] HR 1.16, 95% CI, 0.86-1.56, p=0.324) was not associated with hip fracture. SHBG was also significantly related to hip fracture in mutually adjusted models (mutually adjusted [Model 2] HR 1.26, 95% CI, 1.01- 1.58, p=0.045). Additional analyses excluding participants using 5-α-reductase inhibitors, oral corticosteroids, and those with prostate cancer yielded results that were virtually unchanged (data not shown). Results were also similar when a quadratic term for age and an interaction of age with weight were included in our models.

Table 2.

Associations Between Total Hormone Levels and Hip Fracture Risk

	Hazard Ratio (95% Confidence Interval) for risk of hip fracture per standard deviation increment
	Model 1	Model 2
Total T (SD 181.08 ng/dl)	0.99 (0.79-1.22)	1.16 (0.86-1.56)
Total DHT (SD 0.23 ng/ml)	0.85 (0.68-1.06)	0.74 (0.55-1.00)
SHBG (SD 29.57 nmol/l)	1.12 (0.92-1.37)	1.26 (1.01-1.58)

Model 1 adjusts for age, race, clinic site, height and weight

Model 2 adjusts for model 1 covariates and smoking status and pack years, renal function, number of hours spent sitting or lying per day, alcohol use, diagnosis of diabetes, diagnosis of hypertension along with other hormone (s) and SHBG

Although primary analyses focused on estimated levels of total hormones, we repeated our analyses using free hormones, which showed largely similar results (Supplemental Table 1). In contrast to the analyses using total hormones, SHBG did not have a significant relationship with hip fracture risk, supporting a physiological role for the active form of the hormone or free level.

3.3. Androgens and BMD

A total of 439 men underwent measurement of BMD of the hip. The average total hip BMD was 0.94 grams/meters² (SD 0.16), and the average femoral neck BMD was 0.78 grams/meters² (SD 0.14).

In adjusted models, we observed no significant associations of T, DHT or SHBG with total hip or femoral neck BMD. In mutually adjusted models, a one standard deviation change in testosterone was associated with 0.016 (95% CI, −0.005-0.038) g/m² higher total hip BMD (p = 0.14). The corresponding coefficients were −0.008 g/m² (95% CI, −0.028-0.011, p=0.40) for DHT and −0.02 g/m² (95% CI, −0.037-0.0025, p=0.09) for SHBG. Associations were similarly not significant at the femoral neck.

3.4. Androgens and LBM

Given the association of DHT with lower risk of fracture, but not with higher BMD, we tested whether these hormones were associated with greater LBM assessed by DXA, with the hypothesis that DHT might specifically increase LBM and thereby protect against injurious falls. The average LBM as measured during BMD scans was 18.3 kg/m² (SD 1.7).

In adjusted models, both T and DHT were comparably associated with higher LBM with increments of 0.17 kg/m² per SD (95% CI, 0.1-0.29, p = 0.005) and 0.17 kg/m² per SD (95% CI, 0.1-0.29, p = 0.005) in T and DHT, respectively. In mutually adjusted models, the associations with LBM remained similar to each other (regression coefficients 0.14, 0.11) but neither was statistically significant.

4. Discussion

Among older men without cardiovascular disease, DHT was significantly associated with lower risk of incident hip fracture, while T was not: SHBG was significantly associated with higher risk of incident hip fracture. T, DHT and SHBG were not significantly associated with BMD, but T and DHT had similar positive associations with LBM.

Findings from cohort and case-control studies in elderly men report inconsistent findings related to T, estradiol and SHBG and their associations with hip fracture. Some studies report no association, while others report increased risk of fracture with low T, low estradiol and high SHBG [6– 7]. Like several other studies [8–11], we did not find a significant association between T and hip fracture, although we did find a significant relationship between SHBG and fracture risk. The studies evaluating the association between sex steroids and fracture risk are challenging to compare, as they frequently vary in hormonal assay used, circulating fractions of testosterone measured, and specific fracture outcome assessed. Of note, after accounting for DHT in our study, the relationship between T and hip fracture was non-significant but numerically positive, arguing strongly against an association of T itself with fracture independent of its link as a precursor to DHT.

Studies evaluating the relationship between T and BMD have shown varied results with some studies showing a positive relationship [6, 12–14], and others finding no such association [7–10, 15, 16]. Our data coincide with the latter group of studies, which did not demonstrate relationships of either T or DHT with BMD.

Few studies have evaluated the effects of total DHT or free DHT on bone health. To our knowledge, ours is the first study to demonstrate that DHT is linked to lower risk of hip fracture. In contrast, the Concord Health and Ageing in Men Project (CHAMP) observed no significant relationship between DHT and either risk of hip fracture or BMD [9]. CHAMP used total fracture as an outcome, as opposed to hip fracture, and documented only a handful of hip fractures, making it difficult to compare studies. However, in another study, men with osteoporosis had significantly lower DHT than those without osteoporosis [33].

DHT binds more tightly to the androgen receptor than testosterone, and the DHT- androgen receptor complex is more readily transferred to the DNA-binding site and transactivates genes more efficiently than testosterone [33]. In vitro studies have shown that DHT interacts with androgen receptors on osteoclast cells to inhibit bone resorption [34]. Given the inverse relationship we observed between DHT and hip fracture, it is perhaps surprising that we documented no comparable relationship between DHT and BMD. Although this may reflect the play of chance, as the number of men who underwent BMD was limited, it suggests that other factors may mediate the relationship between DHT and fracture risk. Beyond BMD per se, it is possible that other determinants of bone strength, such as bone size and shape, degree of mineralization, microarchitecture and bone turnover, are responsible [35]. Our results suggest that DHT may increase LBM, potentially leading to greater stability and protection during falls, but this did not appear unique to DHT. As a result, further exploration of the effects of DHT on bone in clinical populations is needed.

Our findings have potential clinical implications, as many men take 5α-reductase inhibitors, such as finasteride or dutasteride, for benign prostatic hyperplasia or male pattern baldness. To date, studies have not demonstrated that men taking these medications have lower BMD or higher risk of falls or hip fracture [36–39]. The lack of observed harm from these medications may simply be due to lack of sufficient surveillance to date. Still, these medications lead to a relative increase in testosterone that can then be aromatized to produce estradiol, a potent agent for bone health. DHT is a non-aromatizable androgen, and our study suggests that non-aromatizable androgens may have an important impact on the risk of fracture. Clinical guidelines for men with osteoporosis currently recommend laboratory evaluation of testosterone in all men, but DHT may be of additional utility to assess the risk of fracture in these patients [40].

While this study is the first to demonstrate a relationship between DHT and fracture risk, it is not without important limitations. Our population was limited to largely Caucasian elderly men and therefore may not be generalizable to young men or those of other races, whose hormonal profile differs. Sensitivity analyses were performed to evaluate the effect of race on bone mineral density and fracture, but due to the modest number of African Americans in the cohort, and in particular the small number who sustained a hip fracture, we could not adequately compare outcomes between African American and non-African American men. DHT levels are limited by a single measurement and given diurnal variation [41], our data would have been strengthened by multiple measurements and consistent morning collection time. We relied on stored specimens for hormone measurements, although data suggest that hormone levels remain reliable despite extensive frozen storage [23]. Our laboratory has extensive expertise in measurement of sex steroids, and measurements demonstrated high reproducibility, but DHT measurements were not specifically subject to the CDC Hormone Standardization Program as T measurements were. We were also limited by the few clinic sites that performed BMD testing, so not all of our subjects with hormonal testing had BMD measurements. The number of hip fractures among men who had BMD measurements was quite low, which limited our ability to include both BMD and hip fracture in a model together. We were therefore unable to specifically examine the mediating effect of BMD in the full cohort who underwent follow up for hip fracture.

Estradiol is known to influence bone health in both men and women [42], and our study lacks data on estradiol levels, which would have added to the richness of our findings. The effects of testosterone on the skeleton may be exerted through action at the androgen receptor or indirectly through the aromatization of testosterone to estradiol [5, 16], but the correlation between estradiol and testosterone levels in men is modest, indicating that other factors also impact estradiol levels in men [43]. Sex hormone binding globulin, by binding testosterone, reduces the amount of bioavailable testosterone for aromatization, yet higher SHBG levels have been associated with lower estradiol levels independent of testosterone [43]. Estrogen is also known to increase SHBG, although this is most common in high SHBG states [44]. It is possible that the relationship we found between high SHBG and higher fracture risk is related to lower estradiol levels although other researchers have found a relationship between SHBG and fracture risk independent of estradiol [45]: Nonetheless, we have no reason to believe that the addition of estradiol would explain the difference we observed in the relationships of T and DHT with hip fracture.

Specific strengths warrant mention. Our study consisted of a large, well-characterized group of elderly men, a population at risk for hip fracture and its morbidity and mortality. Other strengths include a long follow-up period and sensitive assessment of hormone levels using liquid chromatography–tandem mass spectrometry assay for both T and DHT. We also evaluated both LBM and BMD, albeit on a smaller number of participants.

4.1. Conclusions

Osteoporosis is a common condition that affects both genders, and the risk of death and significant disability due to osteoporotic fractures are higher among men than women. In a group of older men without cardiovascular disease, DHT was significantly and substantially related to a lower risk of hip fracture, despite having no clear association with BMD and no unique relationship with LBM. More research is needed to determine the mechanism(s) by which DHT may affect bone health and whether interventions that regulate DHT might be used to reduce risk of hip fracture. While our results require confirmation, there may be a role for measurement of DHT along with T when the clinical scenario requires measurement of male hormone levels.

Availability of data:

All CHS data are publicly available through the NHLBI BioLincc (https://biolincc.nhlbi.nih.gov/studies/chs/).