Sex Steroid Hormone Levels and Body Composition in Men

Abstract

Background:

Previous studies indicate that testosterone (T) is positively correlated with lean mass and inversely correlated with fat mass in men; however, the directionality of these associations, as well as the association with other hormones including estradiol (E2) and SHBG, is unclear.

Methods:

We examined cross-sectional and longitudinal associations of E2, T, SHBG, and E2/T ratio with body composition among men ages 30 to 79 in the Boston Area Community Health/Bone Survey. Total, trunk, and appendicular lean and fat mass were measured by dual-energy x-ray absorptiometry at baseline, and weight and waist/hip circumference were measured at baseline and follow-up. Partial Pearson correlation coefficients were used to estimate the linear relationship between each body composition measure and log-transformed hormone variable.

Results:

In cross-sectional analyses of 821 men, T, calculated free T, and SHBG were inversely correlated with fat mass, weight, body mass index, waist/hip circumference, and waist-to-hip ratio, with multivariable-adjusted correlations ranging from −0.13 to −0.37. Calculated free E2 was positively correlated with percentage total (r = .13) and trunk (r = .15) fat mass, and E2/T was positively correlated with all measures examined (r = .13–.40). There were no significant multivariable-adjusted longitudinal associations between baseline hormone levels and change in weight, body mass index, waist/hip circumference, or waist-to-hip ratio after an average follow-up of 4.8 years.

Conclusions:

We observed significant cross-sectional associations between hormone levels, including E2, T, and E2/T, and body composition measures in men. Longitudinal analyses showing no influence of baseline hormone levels on change in anthropometric measures imply that body composition affects hormone levels and not the reverse.

Aging in men is associated with changes in body composition, including a decrease in skeletal muscle mass and an increase in total and abdominal fat mass (1). Concurrently, total testosterone (TT) and free testosterone (FT) levels decrease, whereas SHBG levels increase with age (2). Previous cross-sectional studies have reported significant correlations between lower testosterone (T) levels and increased adiposity in men (3–11); however, the directionality of this association is uncertain (2), as is the association with other hormones such as estradiol (E2) (4–7, 12–14).

In men, the major source of E2 is through its conversion from T via the enzyme aromatase (2). Aromatase is abundantly expressed in adipose tissue, suggesting that greater fat mass may favor synthesis of aromatase and consequently E2 (15). Although this suggests a possible association between adiposity and E2 in men, existing data on this relationship are equivocal, with some studies reporting a positive association (6, 7, 12) and others reporting no association (4, 5, 13).

Given these uncertainties, we examined cross-sectional and longitudinal relationships between sex steroid hormone levels, body composition, and anthropometric measures (weight, body mass index [BMI], waist and hip circumference, and waist-to-hip ratio [WHR]) in a racially diverse sample of men enrolled in the Boston Area Community Health/Bone (BACH/Bone) Survey. We first examined cross-sectional associations between baseline hormone levels and measures of body composition and anthropometric variables. We hypothesized that higher levels of E2 and free E2 (FE2) and lower levels of TT, FT, and SHBG would be associated with unfavorable body composition, including greater total and trunk fat mass. In addition, we used longitudinal data on anthropometric variables to examine associations between baseline hormone levels and change in anthropometric measures.

Materials and Methods

Study design

The BACH/Bone Survey is a population-based study of skeletal health in 1219 randomly selected Black, Hispanic, and white male residents of Boston aged 30 to 79 years. BACH/Bone enrolled a subset of the 2301 men in the larger BACH Survey (16). Details of the BACH Survey have been published previously (17); briefly, BACH used a multistage stratified cluster sampling design to recruit approximately equal numbers of participants in prespecified groups defined by gender, race/ethnicity, and age.

At the conclusion of the baseline BACH data collection visits in 2002 to 2005, male participants were asked if they would be willing to participate in BACH/Bone. Eligibility criteria included weight ≤300 lbs, ability to travel to Boston University School of Medicine, and consent for a dual-energy x-ray absorptiometry (DXA) scan and ability to get on the scan table. Between November 2002 and July 2005, 1219 (65%) of 1877 eligible BACH participants were enrolled in BACH/Bone, and 1209 completed baseline DXA scans. Additional study details are available elsewhere (18). Follow-up data on anthropometric measures were collected for 80.5% of participants as part of the BACH Follow-Up Survey in 2006–2010, an average of 4.8 years after the baseline data collection.

All participants provided written, informed consent separately for participation in each study, and all study protocols were approved by the Institutional Review Boards of Boston University School of Medicine (for BACH/Bone) and New England Research Institutes (for BACH and BACH/Bone).

Hormone assays

A venous blood sample (20 mL) was collected throughout the day during the baseline BACH interview as close to waking as possible (median time since waking was 3 h 38 min; 50% of samples were obtained before 11:00 am), and samples were stored at −80°C. Details of the measurement of E2, TT, and SHBG are described elsewhere (19). Briefly, E2 was measured at the Mayo Clinic Core Laboratory (Rochester, Minnesota) using liquid chromatography-tandem mass spectrometry (ThermoFisher Scientific, Waltham, Massachusetts and Applied Biosystems-MDS Sciex, Foster City, California). The lower limit of detection was 2.5 pg/mL (9.2 pmol/L). To measure E2 levels in the low range reliably, values <12.5 pg/mL (45.9 pmol/L) were calculated using manual integration of chromatograms. The interassay coefficients of variation for E2 concentrations of 25 to 373 pg/mL (91.8–1369 pmol/L) ranged from 9.1% to 16.6%. TT and SHBG were measured at the Children's Hospital Medical Center Research Laboratories (Boston, Massachusetts) using competitive electrochemiluminescence immunoassays on the 2010 Elecsys system (Roche Diagnostics, Indianapolis, Indiana). The lower limits of detection for TT and SHBG were 2 ng/dL (0.069 nmol/L) and 3 nmol/L, respectively. The interassay coefficients of variation were 1.7% to 7.4% for TT concentrations of 24 to 700 ng/dL (0.83–24.3 nmol/L) and 2.4% to 2.7% for SHBG concentrations of 25 to 95 nmol/L. FT and FE2 concentrations were calculated from TT, E2, and SHBG concentrations using mass action equations (20).

Anthropometric measures

During the baseline data collection visits in 2002–2005, BACH Survey interviewers measured participant height to the nearest 0.1 cm using a stadiometer and weight to the nearest 0.1 kg using a digital scale. Waist and hip circumference were measured twice using a standardized protocol, and the average of the 2 measures was used. BMI and WHR were calculated from height, weight, and waist and hip circumference, respectively. Follow-up data were collected in 2006–2010 using the same protocols among 1610 men who participated in the baseline BACH Survey and provided informed consent for the BACH Follow-Up Survey.

Body composition measures

Detailed body composition measures including total body and regional fat and nonfat mass were obtained at baseline using a Hologic QDR 4500W densitometer (Hologic, Inc, Waltham, Massachusetts). All mass quantities exclude the head. Total and trunk fat mass were measured, and appendicular fat mass was calculated by subtracting trunk from total fat mass. Total lean mass was calculated by subtracting bone mineral mass from nonfat mass. Appendicular lean mass was calculated by adding the nonfat mass measures for the arms and legs and subtracting the appendicular bone mineral content. Trunk lean mass was calculated by subtracting appendicular lean mass from total lean mass. Relative body composition measures were also obtained by dividing each total, trunk, or appendicular lean or fat mass measure by the sum of lean and fat mass for the relevant region and multiplying by 100.

Covariates

During the baseline BACH interview, participants provided data on covariates of interest including demographics, physical activity, alcohol intake, self-reported health, smoking, and medication use. Physical activity was assessed using the Physical Activity Scale for the Elderly (21). Pack-years of smoking were calculated by multiplying the number of packs of cigarettes smoked per day by the number of years the participant reported smoking. Interviewers assessed use of prescription and over-the-counter medications during the previous 4 weeks by direct observation/recording of medication labels, as well as participant self-report with prompts of indications and common brand names.

Statistical analysis

We excluded men with missing hormone data (n = 331). Most of these (N = 330) were missing data on E2, which was measured on stored serum samples, and much of the loss was explained by lack of informed consent to store serum samples. In addition, we excluded men with medical conditions or use of medications that may influence hormone levels, including HIV (n = 14), medications such as androgens and finasteride (n = 20), and current cancer treatment (n = 12). In addition, we excluded 17 men with outlying hormone values and 4 with missing data for covariates of interest, leaving 821 men in our analysis. These men comprised the sample for our cross-sectional analyses of baseline hormone levels, body composition, and anthropometric measures. Longitudinal information on one or more anthropometric variables was available for 650 participants. These men comprised the sample for our longitudinal analyses of baseline hormone levels and change in anthropometric measures over an average follow-up period of 4.8 years.

Sampling weights were used to produce estimates that are representative of the Black, Hispanic, and white male population in Boston between the ages of 30 and 79. All analyses were conducted using SAS 9.2 (SAS Institute, Inc, Cary, North Carolina) and SUDAAN 10.0.1 (RTI International, Research Triangle Park, North Carolina).

All hormone values were log-transformed to improve normality. We estimated weighted means, standard deviations, and percentages for covariates of interest across our entire study sample. We examined univariate associations between each covariate and exposure or outcome of interest to assess the potential for confounding, and retained in our models covariates that were significantly associated with at least one body composition measure and one or more hormone measures. All analyses were adjusted for continuous age, height, and physical activity and categorical race/ethnicity, alcohol intake, self-reported health, pack-years of smoking, and current/recent use of cholesterol-lowering or cardiovascular/antihypertensive medications. Previously published data from this cohort indicate no difference in BMI by race/ethnicity but higher lean mass in Black vs Hispanic/white men and higher fat mass in white vs Black/Hispanic men (22). Additional variables considered as potential confounders but excluded from the final models based on evidence that they did not confound the associations of interest included marital status, current smoking status, time between waking and blood draw, and current/recent use of H2 blockers, opiates, diuretics, anticonvulsants, oral glucocorticoids, or vasodilators. We additionally adjusted analyses of the absolute measures of total, trunk, or appendicular lean or fat mass for the corresponding lean or fat mass measure; for example, analyses of trunk lean mass were adjusted for trunk fat mass.

We used nonparametric locally weighted scatterplot smoothing to examine the functional form of the relationship between each hormone and body composition measure (23). We estimated the linear association between each exposure and outcome of interest using age- and multivariable-adjusted partial Pearson correlations and calculated P values for each association using corresponding linear regression models incorporating the sampling weights. We calculated the absolute difference in weight, BMI, waist circumference, hip circumference, and WHR between baseline and follow-up and examined these measures in relation to each hormone using locally weighted scatterplot smoothing, partial Pearson correlations, and linear regression models. In additional analyses, we examined associations among men with morning blood samples (before 10 am) and evaluated the impact of adjusting the analyses of FE2 for FT and vice versa, and adjusting the analyses of E2, TT, and SHBG for each of the other 2 hormones. We also examined variation in the results by covariates including age, race/ethnicity, smoking status, BMI, and physical activity.

Results

Characteristics of the 821 men included in our analyses are displayed in Table 1. On average, participants were 46.4 years of age and moderately overweight. In general, simple graphical displays suggested linear associations of increasing E2, FE2, and E2/T with less favorable body composition, and of increasing TT, FT, and SHBG with more favorable body composition.

Table 1.

Baseline Characteristics of Study Sample (n = 821)

Covariate	Mean ± SD or Percentagea
Age, y	46.4 ± 12.5
Race/ethnicity
Black	21.5
Hispanic	11.9
White	66.6
Height, cm	176 ± 7
Weight, kg	88.1 ± 15.1
Body mass index, kg/m2	28.4 ± 4.6
Waist circumference, cm	97.2 ± 11.9
Total body lean mass, kg	55.4 ± 7.6
Total body fat mass, kg	22.0 ± 8.3
Percent lean mass	0.72 ± 0.07
Trunk lean mass, kg	28.6 ± 3.9
Trunk fat mass, kg	12.7 ± 5.2
Percentage trunk lean mass	0.70 ± 0.08
Physical Activity Scale for the Elderly score	194 ± 112
Fair/poor self-reported health	11.4
Current smoker	25.1
Smoking pack-years
0 pack-years	46.3
1–5 pack-years	21.5
6–15 pack-years	13.1
16–30 pack-years	12.0
>30 pack-years	7.1
Alcohol intake
None	23.1
<1 drink per day	38.7
1–2 drinks per day	28.8
≥3 drinks per day	9.4
Hormone levels
Estradiol, pg/mL (pmol/L)	23.5 ± 9.1 (86.3 ± 33.4)
Free estradiol, pg/mL (pmol/L)	0.67 ± 0.26 (2.5 ± 0.95)
Total testosterone, ng/dL (nmol/L)	435 ± 176 (15.1 ± 6.1)
Free testosterone, ng/dL (nmol/L)	9.0 ± 3.5 (0.31 ± 0.12)
SHBG, nmol/L	33.5 ± 16.4

^aAll estimates are weighted according to sampling design (see Materials and Methods).

Cross-sectional analyses of hormone levels and body composition

In multivariable-adjusted analyses, we observed statistically significant positive correlations between E2 and weight (r = .14), BMI (r = .13), and hip circumference (r = .12) and between FE2 and total, trunk, and appendicular lean mass (r = .11–.12), percentage total (r = .13) and trunk (r = .15) fat mass, and anthropometric measures including weight, BMI, and waist and hip circumference (r = .15–.20) (Table 2). The correlation coefficients for relative measures of lean mass (percentage total, trunk, and appendicular lean mass) were identical in magnitude but opposite in direction to the correlations for relative fat mass (results not shown).

Table 2.

Age- and Multivariable-Adjusted Partial Pearson Correlation Coefficients for Cross-sectional Associations Between Hormone Levels (Log-transformed) and Absolute Body Composition Measures in Men

Body Composition Measure	Age-adjusted						Multivariable-adjusteda
	E2	FE2	TT	FT	SHBG	E2/T	E2	FE2	TT	FT	SHBG	E2/T
Total lean mass, kgb	0.12	0.16	−0.19	−0.10	−0.23	0.26	0.09	0.12	−0.08	−0.01	−0.14	0.15
Trunk lean mass, kgb	0.12	0.16	−0.21	−0.12	−0.24	0.28	0.09	0.11	−0.04	−0.01	−0.08	0.13
Appendicular lean mass, kgb	0.11	0.15	−0.16	−0.08	−0.20	0.21	0.08	0.12	−0.11	−0.02	−0.19	0.16
Total fat mass, kgb	0.08	0.14	−0.34	−0.26	−0.30	0.36	0.04	0.06	−0.24	−0.19	−0.17	0.23
Trunk fat mass, kgb	0.09	0.15	−0.36	−0.25	−0.35	0.38	0.03	0.07	−0.25	−0.18	−0.23	0.23
Appendicular fat mass, kgb	0.07	0.10	−0.28	−0.24	−0.19	0.29	0.04	0.04	−0.19	−0.20	−0.07	0.19
Percentage fat massc	0.05	0.10	−0.33	−0.24	−0.28	0.32	0.08	0.13	−0.31	−0.22	−0.28	0.31
Percentage trunk fat massc	0.06	0.13	−0.35	−0.23	−0.35	0.34	0.09	0.15	−0.34	−0.21	−0.35	0.34
Percentage appendicular fat massc	0.03	0.05	−0.25	−0.23	−0.15	0.23	0.06	0.08	−0.23	−0.20	−0.13	0.23
Weight, kg	0.11	0.17	−0.30	−0.21	−0.30	0.35	0.14	0.20	−0.34	−0.23	−0.33	0.40
Body mass index, kg/m2	0.13	0.19	−0.35	−0.25	−0.32	0.41	0.13	0.19	−0.34	−0.23	−0.32	0.39
Waist circumference, cm	0.07	0.13	−0.37	−0.27	−0.33	0.38	0.09	0.15	−0.37	−0.26	−0.34	0.38
Hip circumference, cm	0.10	0.14	−0.30	−0.24	−0.24	0.35	0.12	0.16	−0.31	−0.24	−0.24	0.37
Waist-to-hip ratio	0.00	0.06	−0.28	−0.18	−0.29	0.23	0.01	0.07	−0.27	−0.16	−0.30	0.22

All estimates are weighted according to sampling design; bold indicates P < .05.

^aAdjusted for continuous age, height, and physical activity and categorical race/ethnicity (Black, Hispanic, white), alcohol intake (0, <1, 1–2, or ≥3 drinks per day), self-reported health (fair/poor or good/very good/excellent), pack-years of smoking (0, 1–5, 6–15, 16–30, or >30 pack-years), and use of cholesterol-lowering or cardiovascular/antihypertensive medications.

^bMultivariable-adjusted models additionally adjusted for corresponding measure of total lean or fat mass (eg, analysis of trunk lean mass adjusted for trunk fat mass).

^cCorrelation coefficients for percentage total, trunk, and appendicular lean mass are identical in magnitude but opposite in direction to those for percentage total, trunk, and appendicular fat mass.

In multivariable-adjusted models, higher levels of TT and SHBG were associated with lower levels of absolute fat mass, percentage fat mass, and anthropometric measures, with correlation coefficients ranging from −0.19 to −0.37 for TT and −0.13 to −0.35 for SHBG. Similarly, FT was significantly inversely associated with all fat mass and anthropometric measures (r = −.16 to −.26). SHBG was also associated with lower levels of total (r = −.14) and appendicular (r = −.19) lean mass, while TT and FT were unassociated with lean mass after adjustment for fat mass.

The ratio of E2/T was significantly positively correlated with all body composition and anthropometric measures in both age- and multivariable-adjusted models. After adjustment for confounding, the correlation coefficients ranged from 0.13 to 0.40, with the strongest associations for increasing E2/T and higher weight (r = .40), BMI (r = .39), waist circumference (r = .38), hip circumference (r = .37), and percentage trunk fat mass (r = .34).

The results for men with a morning blood sample were similar to those for all men combined; although some correlation coefficients changed somewhat due to the smaller sample size (n = 260), the overall interpretation of the results did not change (results not shown). In analyses adjusted for levels of other hormones (FE2 adjusted for FT and vice versa, and analyses of E2, TT, and SHBG adjusted for each other), we observed stronger positive correlations with E2 and FE2 and weaker correlations with SHBG, when compared to the multivariable-adjusted results shown in Table 2. After adjustment for other hormones, E2 and FE2 were significantly positively correlated with all body composition and anthropometric measures examined. In contrast, the correlations with TT and FT were similar to those shown in Table 2, although the inverse correlations with TT were slightly attenuated whereas the inverse correlations with FT were slightly stronger after adjustment for other hormones (results not shown).

Longitudinal analyses of baseline hormone levels and change in anthropometric measures

There were no statistically significant multivariable-adjusted associations between baseline hormone levels and change in anthropometric measures between baseline and follow-up (Table 3). The results were similar after adjustment for other hormones and in analyses restricted to men with a morning blood draw.

Table 3.

Age- and Multivariable-adjusted Partial Pearson Correlation Coefficients for Longitudinal Associations Between Baseline Hormone Levels (Log-transformed) and Absolute Change in Anthropometric Measures in Men

Anthropometric Variableb	Age-adjusted						Multivariable-adjusteda
	E2	FE2	TT	FT	SHBG	E2/T	E2	FE2	TT	FT	SHBG	E2/T
Change in weight, kg	−0.05	−0.06	0.01	−0.01	0.03	−0.04	−0.07	−0.07	0.01	0.00	0.03	−0.05
Change in body mass index, kg/m2	−0.06	−0.06	−0.01	−0.01	0.02	−0.04	−0.08	−0.08	−0.01	−0.02	0.02	−0.04
Change in waist circumference, cm	−0.07	−0.09	0.03	−0.01	0.08	−0.08	−0.09	−0.10	0.03	−0.01	0.08	−0.09
Change in hip circumference, cm	−0.03	−0.06	0.09	0.04	0.10	−0.10	−0.06	−0.08	0.06	0.02	0.08	−0.10
Change in waist-to-hip ratio	−0.05	−0.05	−0.01	−0.03	0.03	0.00	−0.04	−0.05	0.01	−0.01	0.04	−0.03

All estimates are weighted according to sampling design; bold indicates P < .05.

^aAdjusted for continuous age, height, and physical activity and categorical race/ethnicity (Black, Hispanic, white), alcohol intake (0, <1, 1–2, or ≥3 drinks per day), self-reported health (fair/poor or good/very good/excellent), pack-years of smoking (0, 1–5, 6–15, 16–30, or >30 pack-years), and use of cholesterol-lowering or cardiovascular/antihypertensive medications

^bCalculated as the value at follow-up (collected in 2006–2010) minus the value at baseline (collected in 2002–2005).

Figure 1 summarizes the cross-sectional and longitudinal relationships between levels of E2, TT, E2/T, and SHBG and BMI or change in BMI over time. Consistent with Tables 2 and 3, the cross-sectional figures indicate positive associations with E2 and E2/T and inverse associations with TT and SHBG, whereas the longitudinal figures indicate no relationship between baseline hormone levels and change in BMI.

Figure 1.

Associations between baseline hormone levels (log-transformed) and baseline BMI or absolute change in BMI over time in men.

Variation in the results by covariates of interest

There was no clear evidence of variation in the results by level of any covariate examined (results not shown). However, there was a suggestion of variation in the associations of E2 and FE2 with total, trunk, and appendicular lean mass by smoking status, with stronger positive associations among never and past smokers and nonsignificant inverse associations among current smokers. Further, there was evidence of variation in the associations of TT and FT with anthropometric measures by physical activity, with stronger inverse correlations among individuals with lower levels of physical activity. Future studies should examine these and other interactions of interest for confirmation.

Discussion

The results of these cross-sectional analyses suggest that higher levels of E2 and the ratio of E2/T in men are associated with greater fat mass, whereas higher levels of T and to some extent SHBG are associated with lower fat mass. In multivariable-adjusted models, we observed significant positive associations of FE2 and E2/T with percentage total and trunk fat mass, as well as weight, BMI, and waist and hip circumference; in contrast, higher levels of TT and FT were associated with lower absolute and relative measures of total, trunk, and appendicular fat mass and lower weight, BMI, waist and hip circumference, and WHR. In longitudinal analyses, we did not observe significant associations between hormone levels at baseline and change in anthropometric measures over time. These results suggest that the association between hormone levels and body composition is driven by the influence of adiposity on hormone levels, rather than an effect of hormone levels on adiposity, because baseline hormone levels were not correlated with change in BMI, waist circumference, and other anthropometric measures over an average follow-up of 4.8 years. However, additional analyses, including analyses of changes in hormone levels and DXA over time, are needed to evaluate the longitudinal relationship between hormones and body composition further.

Previous studies have consistently reported significant associations between low T levels and poorer body composition in men (3–11). The mechanisms linking sex hormones with body composition are not completely understood, but possible biological mechanisms include effects of adiposity on SHBG levels (2), or effects of T on regulation of mesenchymal stem cell differentiation (24) and muscle protein synthesis (25) through androgen receptor-mediated pathways (24), activation of inflammatory pathways (26, 27), or increases in cortisol (28). In our study, TT and FT were significantly inversely associated with measures of adiposity but were not associated with absolute measures of lean mass after adjustment for corresponding measures of fat mass. In analyses of the Osteoporotic Fractures in Men Study US (29) and Sweden (12) cohorts, TT and FT also were unassociated with lean mass in elderly men, which Vandenput and colleagues speculated could be due to the use of measures of nonbone lean body mass, which includes organs and connective tissue, rather than measures of actual muscle mass, or the unavailability of data on the effect of T on androgen receptors in the muscle tissue (12). Other studies have reported significant positive associations between T levels and muscle mass or strength (30–32), suggesting that muscle mass may be a more relevant measure than lean mass.

In longitudinal analyses, baseline T levels were not associated with change in anthropometric measures over an average of 4.8 years. These data suggest that correlations between T and body composition are driven by effects of fat mass on T levels, and that T at physiological concentrations may not influence changes in body composition in men. A recent article by Sartorius et al reported that obesity but not aging was associated with a decrease in T over time in healthy men ages 40 and over (33), supporting an effect of fat mass on changes in T levels in men. Further, in a study by Haring et al, obesity was associated with risk of T deficiency but men with BMI < 30 kg/m² had no change in levels of T over time (34).

After adjustment for SHBG and/or T, our cross-sectional analyses showed significant positive associations between both E2 and FE2 and measures of lean mass, fat mass, and anthropometric variables. The positive association between E2 and lean mass in our study also was observed in a recent analysis of the Osteoporotic Fractures in Men Study (Sweden cohort), in which E2 and FE2 were positively associated with both fat mass and lean mass, whereas FT was unassociated with lean mass (12). However, other studies have reported no association between E2 levels and lean mass (5), muscle mass (31), or muscle strength (5) in men. In addition, clinical trials of aromatase inhibition in elderly men have failed to show an improvement in body composition (35, 36), suggesting that E2 does not contribute importantly to body composition in men. However, relatively few studies are available and additional research is needed on associations between estrogens and lean mass.

The mechanisms by which E2 may be associated with body composition and the directionality of this association are not fully understood. Estrogens in men are synthesized locally in many tissues by aromatization of T (37, 38). The positive association between E2 and fat mass may be explained in part by the fact that men with higher BMI tend to have more E2 due to increased aromatase activity. According to Cohen, this aromatization results in decreased T levels, which in turn favors visceral fat deposition (39). Hence, from a clinical perspective, E2 may be considered a marker of adiposity in men. This is consistent with our longitudinal results, where baseline E2 and FE2 levels were not associated with change in BMI or other anthropometric measures over time, suggesting that adiposity may influence levels of E2 rather than hormone levels influencing adiposity. Changes in peripheral aromatase activity with aging may also contribute to this association (40). As men age, their fat mass increases, lean mass decreases, and T levels also tend to decrease, although this decrease may be attributable to comorbidities rather than age itself (33); however, plasma E2 levels are scarcely affected because of increased aromatase activity (39). These changes result in a higher E2/T ratio, which could have a greater effect on body composition than each hormone alone. This increased ratio has been previously linked to the development of atherogenesis and subsequent vascular damage (39). This is further confirmed by our analysis of E2/T, which suggested that as this ratio increases, percentage fat mass and overall body size also increase.

Previous studies have reported inverse associations between SHBG and adiposity (3, 4, 7, 12), and SHBG levels have been shown to increase with age (2, 4, 5). In our analysis, SHBG was inversely correlated with several body composition and anthropometric measures including percentage total and trunk fat mass, BMI, waist circumference, and WHR, although these associations were attenuated after adjusting for TT and E2. The latter observation likely reflects the fact that SHBG is a determinant of TT and E2 levels or their action. Indeed, studies have shown that the receptor-mediated action of sex steroid-bound SHBG uses cAMP as a second messenger, which in turn modulates androgen receptor transcriptional activity (41–43). Our results showing that SHBG is associated cross-sectionally, but not longitudinally, with body composition are consistent with observations in younger men showing that SHBG levels, but not SHBG polymorphisms, are associated with measures of body composition, leading the authors to infer that body composition influences SHBG but not vice versa (44). Adipose tissue may influence SHBG concentrations through a number of mechanisms (14, 44), including the secretion of adipokines or inflammatory markers, as well as reduced production of SHBG in liver, the latter potentially mediated via obesity-related metabolic disturbances such as insulin resistance.

The analyses presented here have several strengths. The BACH/Bone cohort is a large, randomly selected population of men, diverse in age and race/ethnicity, and inferences should therefore be applicable to the broader community of aging men. Anthropometric variables were measured at baseline and follow-up by trained interviewers, rather than reported by participants. In addition, we have detailed information on body composition measured by DXA, several hormones of interest, including E2 measured by liquid chromatography-tandem mass spectrometry, and relevant covariates, which allowed for careful analysis of these associations and control for potential confounding. Further, we were able to examine both total and free hormone fractions and the strength of their associations with adiposity. Limitations of our analysis include the cross-sectional nature of the study design, which precluded a complete examination of the directionality of the associations. Although we were able to use longitudinal data on anthropometric measures to examine the influence of baseline hormone levels on change in body size, the unavailability of longitudinal data on DXA measures or hormone levels is a limitation. Blood was collected from study participants at different times of the day, which may have influenced the results due to diurnal variation in hormone levels. However, there were no clear differences in the results when the analysis was restricted to 260 men with a morning blood sample. In addition, T levels were measured using a platform-based immunoassay that is known to underestimate T levels across the range of concentrations in men and women (45), which may have introduced some error.

Conclusions

The results of these cross-sectional analyses of hormones and body composition suggest that TT and FT are associated with decreased adiposity, whereas E2 and FE2 are associated with increased adiposity in men. Our longitudinal analyses of hormone levels and change in anthropometric measures, although limited in scope, indicate that these associations are driven by the influence of adipose tissue on hormone levels. However, additional studies are needed to examine the directionality of these associations further. In particular, large prospective studies assessing change in hormone levels and change in body composition would help elucidate the clinical predictive role of endogenous sex hormones on adiposity in men.