Sex Steroids and All-Cause and Cause-Specific Mortality in Men

Background  Sex steroid levels are related to metabolic outcomes that could convey higher risk of premature death.

Methods  We examined whether total or free testosterone, dihydrotestosterone, and sex hormone–binding globulin levels are related to all-cause or cause-specific mortality in men. Data were obtained from the Massachusetts Male Aging Study, a population-based cohort study of 1709 men aged 40 to 70 years. Men were followed up for all-cause and cause-specific mortality.

Results  Complete data were available for 1686 men, with 395 deaths occurring during 15.3 years of follow-up. With age adjustment, dihydrotestosterone and sex hormone–binding globulin levels were associated with ischemic heart disease mortality, and free testosterone level was associated with respiratory mortality. In multivariate-adjusted models, higher free testosterone (P = .02) and lower dihydrotestosterone (P = .04) levels were significantly associated with ischemic heart disease mortality, although the latter association was not robust to differences in model selection. The relative risk of death from ischemic heart disease per 1-SD lower free testosterone level was 0.80 (95% confidence interval, 0.64-0.99). Free testosterone level was significantly associated with respiratory mortality (P = .002), with a multivariate-adjusted relative risk per 1-SD lower free testosterone level of 1.90 (95% confidence interval, 1.24-2.92). Total testosterone level was unrelated to mortality, and sex hormone–binding globulin was not significantly associated with mortality after multivariate adjustment.

Conclusions  In men, endogenous sex steroid levels seem to have relatively weak associations with mortality. These data provide little support for the hypothesis that endogenous sex steroid levels are associated with risk of premature death but suggest that further investigation of the relationship between sex steroids and mortality from ischemic heart disease and respiratory disease may be warranted.

Serum testosterone levels exhibit a gradual decline as men age, with longitudinal studies^1,2 showing declines of 0.4% to 2.6% per year. At the same time, sex hormone–binding globulin (SHBG) levels rise, resulting in a greater decline (≥2% per year) in free testosterone levels.¹ Approximately 6% to 12% of men older than 40 years have clinical androgen deficiency as defined by recent clinical guidelines (ie, low testosterone levels plus signs and symptoms).³

Although there are concerns about the risks of testosterone replacement (eg, increased risk of cardiovascular disease [CVD] and prostate cancer),⁴ its use has burgeoned in the past decade despite the absence of changes in approved indications for testosterone treatment.⁵ Whereas in the United States more than 1.75 million prescriptions for testosterone products were written in 2002, reflecting a 170% increase since 1999,⁶ in other countries the modest trend is no higher than that attributable to population aging.⁵

Low testosterone level and androgen deficiency have been associated with abdominal obesity, diabetes mellitus and prediabetic states (insulin resistance, impaired glucose tolerance, and metabolic syndrome), and dyslipidemia, among other outcomes.⁶ In addition, low testosterone level has been associated with increased mortality risk in male veterans.⁷ Sex hormone–binding globulin, which sequesters approximately 60% of total testosterone, and dihydrotestosterone, an active metabolite of testosterone, have been associated with CVD risk factors (eg, low high-density lipoprotein [HDL] cholesterol levels,⁸ fat mass,⁹ metabolic syndrome,¹⁰ and type 2 diabetes mellitus¹¹) that may increase risk of premature death. To our knowledge, no single large, long-term study has examined serum testosterone, dihydrotestosterone, and SHBG levels as predictors of mortality from multiple causes in men. We sought to address this issue with data from the Massachusetts Male Aging Study (MMAS).

The MMAS is an observational cohort study of health and endocrine function in a population-based random sample of men aged 40 to 70 years.¹² A total of 1709 respondents (52.5% of the 3258 eligible men) completed the baseline (1987-1989) protocol. The response rate reflected the requirements for early-morning phlebotomy and an extensive in-person interview. Participants received no financial incentive at baseline.

A trained field technician–phlebotomist visited each subject at home, administered a health questionnaire, and obtained 2 nonfasting blood samples.¹² Anthropometrics (height, weight, and hip and waist circumference) and blood pressure were directly measured according to standard protocols developed for large-scale fieldwork.¹³ The following information was collected via interviewer-administered questionnaire: demographics, psychosocial factors, history of chronic disease, self-assessed general health status, tobacco and alcohol use, nutritional intake, and physical activity and energy expenditure during the past 7 days. The MMAS received approval from the New England Research Institutes institutional review board, and all the participants gave written informed consent.

Nonfasting blood samples were collected within 4 hours of the subject's awakening to control for diurnal variation in hormone levels.¹⁴ Two samples were collected 30 minutes apart and pooled for analysis in equal aliquots to smooth episodic secretion.¹⁵ Blood was kept in an ice-cooled container for transport and was centrifuged within 6 hours. Serum samples were initially stored at –20°C, shipped to the laboratory within 1 week by same-day courier, and stored at –70°C until the time of assay.

Sex steroid measurements were performed at the Endocrine Laboratory at the University of Massachusetts Medical School (Worcester). Total testosterone level was measured in 1994 in stored baseline serum samples using a radioimmunoassay kit (Diagnostic Products Corp, Los Angeles, Calif). Previous analyses showed negligible assay drift due to storage.³ The interassay coefficient of variation was 8.0%. The level of SHBG was measured at baseline using a radioimmunoassay kit (Farmos Diagnostica, Farmos Group Ltd, Oulunsalo, Finland), with an interassay coefficient of variation of 10.9%. The free testosterone level was calculated from testosterone and SHBG using the mass action equations described by Södergård et al,¹⁶ with association constants for testosterone obtained from Vermeulen et al.¹⁷ The dihydrotestosterone level was measured at baseline using radioimmunoassay as previously described,¹⁸ with an interassay coefficient of variation of 12.2%. The HDL cholesterol level was measured at baseline using a standard technique at a Centers for Disease Control and Prevention–certified lipid laboratory (Miriam Hospital, Providence, RI).

Baseline variables were investigated for their associations with mortality. A common set of variables was used to control for confounding, including the following input as continuous variables: age, body mass index (calculated as weight in kilograms divided by height in meters squared), waist circumference, HDL cholesterol level, and systolic blood pressure. In addition, we adjusted for race (white vs other), alcohol consumption (<1, 1, and ≥2 drinks daily), calories expended in physical activity (0, <200, and ≥200 kcal/d), ever smoking, self-assessed health (excellent, very good, good, and fair/poor), and self-reported chronic disease (heart disease, hypertension, and diabetes mellitus).

The vital status of MMAS respondents was ascertained through 2004 by linking the MMAS database with the National Death Index.¹⁹ Cause of death was ascertained via the National Death Index Plus service, which provides causes of death according to the International Classification of Diseases (ICD). Before 1999, deaths were coded according to the ninth revision of the ICD (ICD-9)²⁰ and subsequently, according to the tenth revision (ICD-10).²¹ We categorized dead respondents according to underlying cause of death. We considered deaths from all causes and the following 5 specific causes: ischemic heart disease (IHD) (ICD-9/ICD-10 codes 410-414/I20-I25), other (non-IHD) diseases of the circulatory system (ICD-9/ICD-10 codes 390-459/I00-I99), malignant neoplasms (ICD-9/ICD-10 codes 140-208/D00-D09), diseases of the respiratory system (ICD-9/ICD-10 codes 460-519/J00-J99), and other causes. The National Death Index Plus matches nosologist coding of cause of death within organ systems in 97% of cases.²²

Sex steroids were divided into quintiles according to their distributions in the baseline sample. Person-years were accumulated from each subject's baseline visit to date of death or December 31, 2004. We computed mortality rates (number of deaths divided by number of person-years) and relative risks (RRs) of death in each quintile. We chose men in the highest sex steroid quintile as the reference category for computing RRs. Multivariate analysis was performed using Poisson regression.²³ Tests for linear trend across sex steroid quintiles were performed by fitting models with these as ordinal variables. In a complementary analysis, we present RRs of death per 1-SD decrease in sex steroid. Significance was determined as P<.05.

Of 1306 living men, 1291 had complete sex steroid data. There were 403 deaths in the cohort during follow-up, 395 of which had complete sex steroid data. In total, we included 1686 men who contributed 25 704 person-years of follow-up to the analysis. The largest number of deaths (n = 157, 39.7%) were due to diseases of the circulatory system: 101 (25.6% of all deaths) were due to IHD, and 56 (14.2% of all deaths) were due to other diseases of the circulatory system. There were 127 cancer deaths (32.2%) and 73 deaths with other causes (18.5%). Finally, 38 men (9.6%) died of diseases of the respiratory system. The 8 deaths in men excluded from the analysis because of missing data were due to IHD (n = 1), cancer (n = 3), respiratory disease (n = 1), and other causes (n = 3).

Mean ± SD follow-up in the entire cohort was 15.3 ± 4.0 years, 17.0 ± 0.5 years in surviving subjects, and 9.4 ± 4.9 years in subjects who subsequently died. Table 1 gives baseline characteristics of the analysis sample by vital status. Men who subsequently died were older, were less likely to be employed, had lower educational levels and household incomes, considered themselves to be in worse health, and overall had a worse risk profile than men who survived.

The Figure shows age-adjusted mortality rates by sex steroid quintile. Total testosterone level was not significantly associated with age-adjusted mortality. Low free testosterone level was associated with decreased IHD mortality (P = .08 for trend) and increased respiratory disease mortality (P = .01 for trend; RR, 1.62; 95% confidence interval [CI], 1.10-2.40, per 1-SD [4.35-ng/dL (0.15-nmol/L)] decrease in free testosterone). The SHBG concentration displayed an inverse association with IHD mortality (P = .003 for trend; RR, 1.41; 95% CI, 1.10-1.83, per 1-SD [16.2-nmol/L] decrease in SHBG), and dihydrotestosterone was inversely associated with IHD mortality (P = .005 for trend; RR, 1.36; 95% CI, 1.05-1.78, per 1-SD [0.17-ng/mL (0.58-nmol/L)] decrease in dihydrotestosterone).

Tables 2, 3, 4, and 5 show multivariate-adjusted associations of mortality with total testosterone, free testosterone, dihydrotestosterone, and SHBG levels, respectively. As in age-adjusted models, total testosterone level was not associated with any mortality outcome in multivariate models. Results of a model predicting all-cause mortality with an alternative categorization of total testosterone into 4 categories yielded similar results, with the following RRs (reference group: total testosterone ≥400 ng/dL [≥13.9 nmol/L]): <200 ng/dL (<6.9 nmol/L): 1.23 (95% CI, 0.71-2.12); 200 to less than 300 ng/dL (6.9 to <10.4 nmol/L): 1.29 (95% CI, 0.90-1.84); and 300 to less than 400 ng/dL (10.4 to <13.9 nmol/L): 1.13 (95% CI, 0.87-1.47) (P = .13 for trend). Because there were only 51 deaths in men with total testosterone levels less than 300 ng/dL (<10.4 nmol/L), examination of specific causes is not possible. Low free testosterone level was significantly associated with decreased IHD mortality (P = .02 for trend) and increased respiratory disease mortality (P = .002 for trend). The multivariate-adjusted RR of IHD death per 1-SD decrease in free testosterone level was 0.80 (95% CI, 0.64-0.99), whereas for respiratory disease death it was 1.90 (95% CI, 1.24-2.92). In addition, men in the lowest free testosterone quintile were more than 2 times less likely to die of IHD (RR, 0.45; 95% CI, 0.23-0.89) and 5 times more likely to die of respiratory disease (RR, 5.02; 95% CI, 1.09-23.09) compared with men in the highest free testosterone quintile. Dihydrotestosterone level was significantly associated with IHD mortality in the trend across quintiles (P = .04 for trend), but elevated risk associated with lower dihydrotestosterone level across the different models was small, and neither the continuous dihydrotestosterone variable nor the comparison between the lowest and highest quintiles was significant in predicting IHD mortality.

In this prospective study of 40- to 70-year-old men followed up for 15.3 years, free testosterone level was positively associated with IHD mortality, independent of confounding effects. The increase in risk associated with free testosterone, although robust, was relatively small, with a 20% decrease in risk per 1-SD decrease in free testosterone level. We also observed a strong negative association of free testosterone level with respiratory disease death. The relationship between dihydrotestosterone level and IHD mortality, although also significant in a test for trend across quintiles, was relatively weak and, unlike the free testosterone results, was not consistent across different models. Moreover, dihydrotestosterone level and IHD mortality were inversely related, in contrast to free testosterone level and IHD mortality; thus, these results must be viewed with caution. Total testosterone and SHBG levels were not associated with mortality from any cause in multivariate models.

There are limited data on the relationship between testosterone level and all-cause mortality from studies of healthy men randomly selected from the population. Because concurrent illness is associated with lower sex steroid levels in men,^1,24 results of clinic-based studies may be biased by the strong possibility that low testosterone level is a mere epiphenomenon of concurrent, and possibly occult, illness. Smith et al²⁵ used data from the large (n = 2512) population-based Caerphilly study and found no association between total testosterone level and all-cause mortality. In contrast, the results of 3 small studies^26-28 suggest that low testosterone level is associated with increased short-term all-cause mortality. In addition, Shores et al⁷ found an RR of 1.88 for all-cause mortality in men with consistently low total or free testosterone levels in a study of 858 male veterans. A strength of the latter study is that men were classified as having low testosterone levels only if this was demonstrated on more than 1 occasion. However, this strength must be balanced against weaknesses of the study. Most important, and characteristic of other studies,^26-28 the patients (mean age, 61 years) studied were quite sick, with approximately 24% dying within 4.3 years. In contrast, in the Caerphilly study (mean age, 52 years) 19% of the subjects died during 16.5 years of follow-up, and in the MMAS (mean age, 55 years) 24% of the cohort died during 15 years.

The contribution of sex steroids, if any, to CVD in men is unclear.^29,30 With respect to CVD mortality, large cohort studies with long follow-up generally find no relationship between testosterone level and CVD mortality,^25,31,32 and low SHBG level has been shown to have no³³ or an inverse³⁴ relationship with CVD death. To our knowledge, risk of death associated with serum dihydrotestosterone level has not been examined previously.

Intervention trials have shown that testosterone therapy lowers the HDL cholesterol level,^30,35 raising concern that testosterone may contribute to cardiovascular risk.⁴ Such evidence is consistent with the present finding that higher free testosterone level is associated with increased IHD mortality. However, this finding seems to conflict with emerging data on the adverse metabolic consequences associated with low testosterone levels in men. Findings from a recent meta-analysis³⁶ indicate that lower testosterone level increases risk of type 2 diabetes mellitus and that testosterone withdrawal increases insulin resistance.³⁷ Furthermore, data from the MMAS indicate that low testosterone level increases risk of obesity,³⁸ metabolic syndrome,¹⁰ and type 2 diabetes mellitus.¹¹ The free testosterone and dihydrotestosterone findings with respect to IHD mortality seem to be, at first glance, in direct conflict with one another because in both cases their mechanism of action is presumably via the androgen receptor. Because dihydrotestosterone is almost completely derived from testosterone via conversion by 5α-reductase, these results could suggest that interference with 5α-reductase activity might increase the risk of CVD. However, data from intervention trials of 5α-reductase inhibitors, including the large, randomized, placebo-controlled Prostate Cancer Prevention Trial,³⁹ provide no indication that treatment with 5α-reductase inhibitors increases CVD events or mortality. Considering the free testosterone finding and the lack of association with total testosterone level, this provides further support for our conclusion that the association of dihydrotestosterone level with IHD mortality, if any, is not robust and is perhaps due to chance. That the association between dihydrotestosterone level and IHD mortality showed only marginal statistical significance and was not robust to differences in model selection in a study as large as the MMAS probably reflects some combination of the weak association and that the MMAS is a study of relatively healthy men. In this analysis, statistical power is driven mainly by the number of deaths in the sample. Extended follow-up on the existing cohort will allow us to examine this association with more precision. Further work is needed to clarify the cardiovascular risks and benefits of androgen replacement in men.

Low SHBG levels have been associated with CVD risk factors,^8,10,11 presumably because of the negative effect of insulin on SHBG production.⁴⁰ Consistent with this, we observed a higher, but nonsignificant, increased risk of IHD death in men with lower SHBG levels that was similar in magnitude to that in previous studies.^33,34 Furthermore, statistical control for HDL cholesterol, diabetes mellitus, central obesity, and heart disease factors, which are associated with insulin resistance, confounded the association between SHBG level and IHD mortality.

To our knowledge, the association of free testosterone level with respiratory disease death has not been reported previously. However, patients with Klinefelter syndrome (specifically, the 47,XXY karyotype) were shown to have a higher risk of death from respiratory disease than males in the general population,⁴¹ which could be due to a higher prevalence of smoking or the lower socioeconomic status of men with Klinefelter syndrome.⁴² Patients with chronic obstructive pulmonary disease, which places them at increased risk for death from respiratory illness, have been shown to have depressed testosterone levels.⁴³ The biologic mechanism linking low testosterone levels with respiratory disease morbidity or mortality is not known. Because smoking is associated with higher serum testosterone levels,⁴⁴ the possibility of a greater proportion of smokers in the lower testosterone groups cannot explain our observation. The efficacy of treating patients with chronic obstructive pulmonary disease with anabolic steroids is unclear, with short-term, small clinical trials^45,46 of anabolic steroids in patients with chronic obstructive pulmonary disease showing little effect on lung function. More comprehensive placebo-controlled trials examining the impact of testosterone replacement in men with lung disease may be warranted.

Limitations of the present study should be acknowledged. The MMAS included mostly white men of highersocioeconomic status, so these results may not be generalizable to more diverse populations. However, the MMAS was representative of the greater Boston, Mass, male population at the time of sampling,⁴⁷ and the distribution of the MMAS causes of death is consistent with the leading causes of death for the Massachusetts population at the midpoint of the study (1996).⁴⁸ The low response rate at baseline (52%) is cause for concern. In a telephone survey of 206 nonrespondents to the MMAS, we found that although nonrespondents were older, less likely to report cancer or heart disease, and more likely to report their health as fair or poor compared with the entire cohort, we found no differences in the prevalence of diabetes mellitus, high blood pressure, history of prostate surgery, or restriction in activity due to poor health. There were only 403 deaths in the cohort, and 395 with complete sex steroid data were analyzed in this study. This limited examination to the most common causes of death that occur in aging men and may have limited power to detect very weak associations. We had only 1 serum measure for defining exposure, which could have biased results toward the null hypothesis. Finally, the use of calculated free testosterone levels has limitations,⁴⁹ including overestimation of free testosterone level mainly in the low range (ie, <150 ng/dL [<5.2 nmol/L]), but it is doubtful that this would affect the results because less than 1% of the cohort had such low levels. These limitations must be considered in light of the strengths of this study, including a random, population-based sample of generally healthy, well-characterized men from a defined geographic area, the ability to statistically adjust for a variety of factors that could confound the association between sex steroids and mortality, the uniformity of the endocrine measurements, the length of follow-up, and the examination of all-cause and cause-specific mortality.

In summary, sex steroids overall seem to have a relatively weak or no association with mortality. Generally, statistically significant associations were confined to calculated free testosterone levels, with no robust associations observed for the sex steroids that were directly measured. Although previous research indicates that sex steroid levels may be associated with several adverse metabolic outcomes, the present data provide little support for the hypothesis that endogenous sex steroid levels are associated with risk of premature death but suggest that further investigation into the relationship between sex steroids and mortality from IHD and respiratory disease may be warranted.

Correspondence: Andre B. Araujo, PhD, New England Research Institutes, 9 Galen St, Watertown, MA 02472 (aaraujo@neriscience.com).

Reprints: John B. McKinlay, PhD, New England Research Institutes, 9 Galen St, Watertown, MA 02472 (mmas@neriscience.com).

Accepted for Publication: February 28, 2007.

Author Contributions: Dr Araujo had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Araujo, Handelsman, and McKinlay. Acquisition of data: Araujo, Page, and McKinlay. Analysis and interpretation of data: Araujo, Kupelian, Page, Handelsman, Bremner, and McKinlay. Drafting of the manuscript: Araujo, Page, and McKinlay. Critical revision of the manuscript for important intellectual content: Araujo, Kupelian, Handelsman, Bremner, and McKinlay. Statistical analysis: Araujo and Kupelian. Obtained funding: Araujo and McKinlay. Administrative, technical, and material support: Page. Study supervision: Bremner and McKinlay.

Financial Disclosure: Dr Page has received grant support from GlaxoSmithKline. Dr Bremner is an inventor on issued US patent No. 7 138 389; has been a consultant to Quatrx and Indevus; and was the principal investigator on a study supported by GlaxoSmithKline.

Funding/Support: This work was supported by grant AG 04673 from the National Institute on Aging; grants DK 44995 and DK 51345 from the National Institute of Diabetes and Digestive and Kidney Disorders; the Endocrine Society Solvay Clinical Research Award; and a VA Special Fellowship in Advanced Geriatrics.

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

Acknowledgment: We acknowledge the many contributions of Christopher Longcope, MD, who passed away in 2004. For nearly 20 years Dr Longcope was an indispensable colleague on the MMAS. His scientific expertise and collegiality are missed. We also thank Thomas G. Travison, PhD, and Amy B. O’Donnell, MPH, for their careful review of the manuscript.