Abstract
Objective
To assess the relationship between sex hormone-binding globulin (SHBG) and 18 other hormonal and metabolic parameters in well characterized, normally cycling premenopausal women.
Design
Cross-sectional study.
Setting
University general clinical research center.
Subjects
319 young, healthy women with ovulatory menstrual cycles.
Interventions
None.
Main Outcome Measure
Midfollicular serum SHBG concentrations.
Results
In our final linear regression model, SHBG was negatively associated with bioavailable T and positively associated with adiponectin, associations that were independent of other parameters. SHBG was also positively associated with estrone sulfate, but only when taking into account confounding variables. Unexpectedly there was no straightforward relationship between SHBG and insulin resistance as assessed by the homeostatic model assessment.
Conclusion(s)
Our results highlight the link between androgen action as reflected by bioavailable T and circulating SHBG concentrations in all premenopausal women, and speak to the importance of the relationship between SHBG and adiponectin, which is at least in part independent of androgen action.
Keywords: Sex hormone-binding globulin, premenopause, testosterone, adiponectin, insulin resistance
INTRODUCTION
Sex hormone-binding globulin (SHBG) is a homodimeric plasma glycoprotein that is produced by hepatocytes and serves as the principal transport protein for T and E2 (1, 2). According to the free hormone hypothesis, the biological activity of sex steroids is affected by their free or unbound concentrations (3), providing one means by which SHBG can regulate sex steroid action. Other means that involve a cell membrane receptor for SHBG have also been proposed (4), including receptor-mediated endocytosis of steroids (5). The role of SHBG in regulating sex steroid action has long been of interest (6), and in particular low SHBG concentrations in women have been linked to conditions of androgen excess (2). As well, more recent work has emphasized the potential value of SHBG measurements in assessing risks for several other conditions, including the metabolic syndrome, type 2 diabetes, cardiovascular disease, and breast cancer (2, 7, 8).
Many studies have assessed the relationships between SHBG and assorted hormonal and metabolic parameters in different populations. Nonetheless the precise mechanisms involved are poorly understood, complicated by many of the parameters having unidirectional or bidirectional cause and effect relationships with one another, and relationships explained by confounding factors. For example, estrogens increase hepatic SHBG production at least in part through a direct effect involving estrogen receptor α (9), but any resulting increase in SHBG might be expected to decrease bioavailable E2. An example of confounding is that the effect of oral contraceptives to decrease bioavailable T, often attributed to increased SHBG, could simply be due to decreased ovarian production of androgens because of negative feedback on gonadotropins. Still another example of the complexity of the relationships is that for women type 2 diabetes risk is associated with high T concentrations, the opposite of the case for men (10).
In any event, because few of the many studies of SHBG and physiologically related parameters have focused on well characterized normally cycling women, we performed a cross-sectional study of such women. We specifically used baseline data only from subjects who completed the Women in Steady Exercise Research (WISER) study, the primary purpose of which was to assess the effect of an exercise intervention on parameters associated with breast cancer risk (11, 12). As previously reported, the exercise intervention had no significant effect on SHBG or sex hormone concentrations despite significant increases in aerobic fitness and lean body mass, and a significant decrease in percent body fat (13, 14), possibly because overall the body composition changes that did occur were modest in degree.
MATERIALS AND METHODS
The WISER study was a randomized, controlled, parallel-arm study that investigated the effects of a 16-week aerobic exercise intervention on breast cancer biomarkers in young, healthy, sedentary, eumenorrheic women. The study was approved by the Human Subjects Review Committee at the University of Minnesota (Institutional Review Board; IRB ID#0505M69867). Written informed consent was obtained from each participant prior to participation.
Details of the study design and methods have been described previously (11). Briefly, non-smoking women aged 18-30 years with a body mass index (BMI) of 18-40 kg/m2 (inclusive), having self-reported menstrual cycle length of 24-35 days, exercising < 2 times a week, and residing in the Minneapolis-St. Paul metropolitan area, were randomized into exercise and control groups. Exclusion criteria included use of any hormonal contraception in the past 3 months or depot-medroxyprogesterone acetate in the past 12 months, gynecological disorders, metabolic or endocrine-related diseases, non-melanoma cancer in the past 5 years, alcohol consumption of > 7 servings per week, current or recent (past 6 months) pregnancy, and body weight changes greater than 10% over the past year. The primary study was completed by 319 of the 391 women who started the study by undergoing baseline measurements during the luteal phase of the menstrual cycle and subsequent follicular phase, and the findings presented here are based only on the baseline measurements of those 319 subjects (both exercisers and controls) who completed the study.
All biological, anthropometric, and body composition measures were taken at the General Clinical Research Center at the University of Minnesota. Height was measured by a stadiometer without shoes to the nearest 0.1 cm (Scale Tronix, White Plains, NY). Body mass was measured to the nearest 0.1 kg, using an electronic scale (Scale Tronix, White Plains, NY). Body mass index was calculated by dividing weight in kg by height in meters squared (kg/m2). Body fat percentage and android/gynoid fat ratio were assessed by dual energy x-ray absorptiometry (DXA) using a Lunar Prodigy DXA apparatus (Lunar Radiation Corp., Madison, WI). A sub-maximal treadmill test was used to provide a measure of aerobic fitness, metabolic equivalents (METs) at 85% maximal heart rate, with workload converted into METS using a standard conversion formula (15).
Biological Sample Analyses
Blood samples for serum and plasma concentrations were drawn between 6:45 and 11:00 am after an overnight fast, with specimens stored frozen at −70°C. Blood draws took place during specific days of the menstrual cycle. Occurrence and timing of ovulation were assessed using a commercial 9-day Assure LH™ ovulation kit (Conception Technologies, San Diego, CA). Day of ovulation was considered to be the day following a positive LH surge result. A midluteal phase blood draw was scheduled 6-9 days after ovulation for analysis of progesterone, followed by a single midfollicular phase blood draw between cycle days 7 and 10 of the following cycle for all other serum or plasma concentrations. During this same cycle urine was collected for three consecutive 24-hour periods on cycle days 7-9, with the 3 refrigerated 24-hour collections pooled for aliquots, which were then stored at −20 °C until analysis.
Serum concentrations of T, E2, estrone sulfate, progesterone, and SHBG were measured using commercially available RIA kits for sex steroids and an ELISA for SHBG as previously described (13). Bioavailable fractions of T and E2 were calculated using the equations of Vermeulen et al. (16) and association constants estimated by Mazer (17). Plasma concentrations of insulin were measured by chemiluminescent immunoassay and combined with glucose concentrations to determine the homeostatic model assessment (HOMA) index, a measure of insulin resistance that correlates well with insulin sensitivity as determined by the gold standard euglycemic clamp technique (18). Plasma concentrations of insulin-like growth factor 1 (IGF-1), IGF binding protein 3 (IGFBP-3), and adiponectin were determined by ELISA, those of leptin and C-reactive protein by multiplex bead array assays, and those of F2-isoprostanes by a gas chromatography-mass spectrometry method, all as previously described (11, 19, 20), with IGF-1/IGFBP-3 ratios calculated on a mass basis. The urinary estrogen metabolites 2-hydroxyestradiol (2-OHE2) and 16α-hydroxyestrone (16α-OHE1) were measured by liquid chromatography/tandem mass spectrometry as previously described (11, 12), with 2-OHE2/16α-OHE1 ratios calculated on a molar basis.
Statistical Analyses
In keeping with the focus on SHBG, in our analyses SHBG concentration was used as the dependent variable, with all other parameters considered as independent variables, even though many of these are related to one another, such as body fat % and BMI. Furthermore, values for two parameters, namely serum bioavailable T and serum bioavailable E2, were derived using SHBG values.
For 14 of the 19 parameters analyzed, including SHBG and 13 others considered as independent variables, results were available for 319 women, as shown in Table 1. For the 5 remaining independent variables (estrone sulfate, adiponectin, leptin, C-reactive protein, and urinary 2-OHE2/16α-OHE1 ratio), there were missing observations, with all results available for 203 women. Values of all serum and plasma concentrations aside from insulin were log-transformed to normalize distribution of the data, as were values of HOMA and urinary 2-OHE2/16α-OHE1 ratios. We performed linear regression to study the association between serum SHBG concentrations and the 18 independent variables. Model selection was performed using Mallow's Cp statistic as criterion for adding variables (21) and the final model included the 7 variables shown in Table 2, which provides data for the 203 women for whom all results were available.
TABLE 1.
Characteristics of study subjects and univariate analysis results for SHBG and independent variables
| n | Median (10th - 90th percentile) | Estimate (SE)a | P value | |
|---|---|---|---|---|
| Age (years) | 319 | 25.3 (20.0 - 30.0) | 0.01 (0.01) | 0.34 |
| Height (cm) | 319 | 164.5 (155.5 - 174.5) | 0.00 (0.00) | 0.55 |
| BMI | 319 | 23.3 (19.8 - 30.8) | −0.23 (0.01) | 0.004 |
| Body fat % | 319 | 35.2 (26.4 - 48.9) | −0.01 (0.00) | 0.0001 |
| Android/gynoid fat ratio | 319 | 0.34 (0.22 - 0.49) | −1.82 (0.32) | < 0.0001 |
| Assessed fitness (METS at 85% maximal heart rate) | 319 | 7.06 (4.65 - 8.49) | 0.01 (0.02) | 0.64 |
| Serum SHBG (nmol/L) | 319 | 24.4 (13.7 - 47.9) | -- | -- |
| Serum bioavailable T (ng/mL) | 319 | 0.22 (0.09 - 0.43) | −0.65 (0.05) | < 0.0001 |
| Serum bioavailable E2 (pg/mL) | 319 | 41.2 (22.8 - 62.2) | −0.34 (0.09) | 0.0003 |
| Serum estrone sulfate (ng/mL) | 319 | 2.15 (1.30 - 3.38) | −0.10 (0.10) | 0.31 |
| Serum progesterone (ng/mL) | 312 | 17.8 (2.3 - 32.0) | 0.06 (0.04) | 0.12 |
| Plasma insulin (mU/L) | 319 | 6.2 (2.0 - 13.0) | −0.02 (0.01) | 0.01 |
| HOMA | 319 | 1.21 (0.43 - 2.70) | 0.26 (0.07) | 0.0003 |
| Plasma IGF-1/IGFBP-3 ratio | 319 | 0.07 (0.06 - 0.10) | −7.23 (2.12) | 0.001 |
| Plasma adiponectin (μg/mL) | 228 | 8.3 (4.3 - 15.6) | 0.33 (0.07) | < 0.0001 |
| Plasma leptin (μg/mL) | 226 | 10.8 (3.8 - 25.0) | −0.12 (0.05) | 0.01 |
| Plasma C-reactive protein (μg/mL) | 217 | 3.3 (0.7 - 25.5) | −0.07 (0.03) | 0.01 |
| Plasma F2-isoprostanes (pg/mL) | 319 | 48.6 (32.1 - 78.1) | −0.10 (0.10) | 0.35 |
| Urinary 2-OHE2/16α-OHE1 ratio | 304 | 12.3 (2.6 - 88.3) | 0.06 (0.03) | 0.028 |
All variables log-transformed with the exceptions of age, height, BMI, body fat %, android/gynoid ratio, assessed fitness, plasma insulin, and IGF-I/IGFBP-3 ratio.
TABLE 2.
Full and final model mulivariate linear regression results for SHBG and independent variables (n = 203)
| Full model | Final model | |||
|---|---|---|---|---|
| Estimate (SE)a | P value | Estimate (SE)a | P value | |
| Age (years) | −0.0050 (0.0091) | 0.58 | -- | -- |
| Height (cm) | −0.0033 (0.0037) | 0.37 | -- | -- |
| BMI | −0.0002 (0.0092) | 0.95 | -- | -- |
| Body fat % | −0.0017 (0.0068) | 0.81 | -- | -- |
| Android/gynoid fat ratio | −0.54 (0.42) | 0.14 | −0.48 (0.28) | 0.09 |
| Assessed fitness (METS at 85% maximal heart rate) | 0.018 (0.023) | 0.39 | -- | -- |
| Serum bioavailable T (ng/mL) | −0.60 (0.07) | < 0.0001 | −0.59 (0.06) | < 0.0001 |
| Serum bioavailable E2 (pg/mL) | −0.20 (0.09) | 0.02 | −0.22 (0.07) | 0.004 |
| Serum estrone sulfate (ng/mL) | 0.31 (0.10) | 0.001 | 0.36 (0.08) | < 0.0001 |
| Serum progesterone (ng/mL) | 0.022 (0.030) | 0.47 | -- | -- |
| Plasma insulin (mU/L) | −0.0026 (0.0072) | 0.72 | -- | -- |
| HOMA | 0.20 (0.07) | 0.004 | 0.24 (0.06) | < 0.0001 |
| Plasma IGF-1/IGFBP-3 ratio | −0.38 (2.09) | 0.87 | -- | -- |
| Plasma adiponectin (μg/mL) | 0.10 (0.06) | 0.10 | 0.12 (0.06) | 0.04 |
| Plasma leptin (μg/mL) | 0.08 (0.06) | 0.18 | -- | -- |
| Plasma C-reactive protein (μg/mL) | −0.0002 (0.26) | 0.95 | -- | -- |
| Plasma F2-isoprostanes (pg/mL) | −0.024 (0.092) | 0.77 | -- | -- |
| Urinary 2-OHE2/16α-OHE1 ratio | 0.030 (0.023) | 0.19 | 0.002 (0.005) | 0.60 |
All variables log-transformed with the exceptions of age, height, BMI, body fat %, android/gynoid ratio, assessed fitness, plasma insulin, and IGF-I/IGFBP-3 ratio.
Additionally, values of the 7 independent variables in the final model were categorized into tertiles and general linear models were used to test for differences in SHBG concentrations by tertiles of these variables. We report P values for comparisons between both unadjusted and multivariable-adjusted mean SHBG concentrations in the upper and middle tertiles versus the lowest tertile for all independent variables included in the final model.
All statistical analyses were performed using procedures reg and glm of SAS software (version 9.4; SAS Institute, Cary, NC).
RESULTS
Table 1 provides median values for all 19 parameters that were analyzed, and results of univariate analyses (ie, simple linear regression) to assess relationships between SHBG and the 18 independent variables. To better understand the nature of these relationships all 18 independent variables listed in Table 1 were included in the full model. The final model accounted for 51.8% of total variability of SHBG concentrations. Full and final model results are provided in Table 2, which shows significant positive associations between SHBG and three variables, namely estrone sulfate, HOMA, and adiponectin. In contrast there were significant negative associations between SHBG and both bioavailable T and bioavailable E2. There was no significant association between SHBG and either the urinary 2-OHE2/16α-OHE1 ratio or the android/gynoid fat ratio, although there was a trend towards a negative association for the latter. The final model did not include several independent variables that in the univariate analyses were significantly associated with SHBG, including notably the negative associations for BMI, body fat %, leptin, and IGF-1/IGFBP-3 ratio, suggesting that in large part these associations were on the basis of confounding.
Additional insight about the relationships studied may be obtained from Table 3, which shows results for differences in SHBG by tertiles of the variables included in the final linear regression model, and from Figure 1, which provides scatter plots for SHBG and 4 independent variables. Both univariate and multivariate linear regression showed a significant positive association between SHBG and HOMA. However, this was not a straightforward direct relationship, given that the multivariate-adjusted mean SHBG concentration in the first HOMA tertile was intermediate between that for the second and third HOMA tertiles, and as is evident by simple inspection of the pertinent scatter plot in Figure 1.
TABLE 3.
Unadjusted and multivariate-adjusted mean serum SHBG concentrations (nmol/L) by tertiles of variables
| Geometric means (95% CI) | |||
|---|---|---|---|
| Tertile 1 | Tertile 2 | Tertile 3 | |
| Android/gynoid fat ratio | |||
| Unadjusted (n = 319) | 33.4 (30.0 - 37.1) | 27.7 (24.9 - 30.8) | 21.1 (19.0 - 23.4) |
| P value (vs. tertile 1) | 0.028 | < 0.0001 | |
| Adjusted (n = 217) | 26.0 (23.6 - 28.6) | 26.4 (24.1 - 28.8) | 22.7 (20.7 - 24.9) |
| P value (vs. tertile 1) | 0.97 | 0.11 | |
| Serum bioavailable T (ng/mL) | |||
| Unadjusted (n = 319) | 41.7 (38.2 - 45.5) | 27.0 (24.8 - 29.5) | 17.2 (15.8 - 18.8) |
| P value (vs. tertile 1) | < 0.0001 | < 0.0001 | |
| Adjusted (n = 217) | 34.4 (31.1 - 37.9) | 24.9 (22.0 - 27.3) | 18.1 (16.4 - 20.0) |
| P value (vs. tertile 1) | < 0.0001 | < 0.0001 | |
| Serum bioavailable E2 (pg/mL) | |||
| Unadjusted (n = 319) | 35.0 (31.5 - 38.9) | 24.5 (22.0 - 27.2) | 22.8 (20.5 - 25.3) |
| P value (vs. tertile 1) | < 0.0001 | < 0.0001 | |
| Adjusted (n = 217) | 29.0 (26.5 - 31.8) | 24.1 (22.0 - 26.3) | 22.2 (20.3 - 24.4) |
| P value (vs. tertile 1) | 0.0081 | 0.0003 | |
| Serum estrone sulfate (ng/mL) | |||
| Unadjusted (n = 319) | 28.2 (25.3 - 31.6) | 27.6 (24.6 - 30.8) | 24.8 (22.2 - 27.8) |
| P value (vs. tertile 1) | 0.94 | 0.24 | |
| Adjusted (n = 217) | 21.0 (19.0 - 23.2) | 26.3 (24.1 - 28.7) | 28.2 (25.6 - 31.1) |
| P value (vs. tertile 1) | 0.0023 | 0.0003 | |
| HOMA | |||
| Unadjusted (n = 319) | 22.6 (20.4 - 25.2) | 25.1 (22.6 - 28.0) | 34.3 (30.9 - 38.2) |
| P value (vs. tertile 1) | 0.29 | < 0.0001 | |
| Adjusted (n = 217) | 24.0 (21.9 - 26.2) | 22.1 (20.2 - 24.2) | 29.4 (26.9 - 32.2) |
| P value (vs. tertile 1) | 0.37 | 0.0036 | |
| Plasma adiponectin (μg/mL) | |||
| Unadjusted (n = 228) | 20.4 (18.1 - 23.0) | 25.9 (23.1 - 29.2) | 30.7 (27.3 - 34.6) |
| P value (vs. tertile 1) | 0.0088 | < 0.0001 | |
| Adjusted (n = 217) | 22.8 (20.8 - 25.0) | 26.5 (24.3 - 29.0) | 25.8 (23.5 - 28.2) |
| P value (vs. tertile 1) | 0.037 | 0.12 | |
| Urinary 2-OHE2/16α-OHE1 ratio | |||
| Unadjusted (n = 304) | 23.4 (20.9 - 26.2) | 27.8 (24.8 - 31.1) | 29.1 (26.0 - 32.6) |
| P value (vs. tertile 1) | 0.067 | 0.014 | |
| Adjusted (n = 217) | 24.2 (22.2 - 26.5) | 25.5 (23.3 - 27.9) | 25.2 (23.1 - 27.6) |
| P value (vs. tertile 1) | 0.60 | 0.73 | |
FIGURE 1.
Scatter plots for SHBG and 4 independent variables
DISCUSSION
The 18 independent variables included in the full multivariate linear regression model were chosen based on their biologically plausible relationships with conditions for which SHBG measurements may help assess risks. We included serum bioavailable T and E2 concentrations, as opposed to total or free concentrations, because the former provide better overall measures of androgen and estrogen action in most tissues (16). In any case, this discussion focuses on the independent variables included in the final model. Our study's cross-sectional design does not allow for direct delineation of causal relationships, but the large number of variables assessed does potentially allow for insight into which relationships are the most important, and which may be primarily on the basis of confounding.
Overall, the relationships we found were largely expected based on prior studies and current understanding of SHBG regulation, with some exceptions that may be explained by the complexity of the interactions involved. Circulating SHBG is exclusively produced by the liver, with its secretion influenced by a host of hormonal, metabolic, and nutritional factors. Increased SHBG concentrations are positively associated with increased hepatic SHBG mRNA levels (22), in large part on the basis of increased cellular activity of nuclear receptor hepatocyte nuclear factor 4-α (HNF4α), known to be affected by such factors (2, 9, 22). Many of the relationships between SHBG and these factors are bidirectional, and additionally there exist complex SHBG-independent interactions of the factors (23).
In our subjects, not surprisingly the most striking relationship between SHBG and all of the independent variables studied was the negative association with bioavailable T, as is evident in Tables 2 and 3, and in Figure 1. Although the precise underlying molecular mechanisms are not well understood, androgens clearly decrease hepatic SHBG secretion (9, 24). Furthermore, T has a very high affinity for SHBG (25), and so any effect of T to reduce SHBG secretion might be amplified by decreased SHBG increasing its bioavailability. Prior studies linking reduced SHBG to increased androgen action in premenopausal women have generally focused on women with polycystic ovarian syndrome and abnormal cycles (2). Our data show this relationship extends to normally cycling premenopausal women. Our findings are similar to those of a study of normal cycling women that found a negative correlation between luteal phase SHBG and total plasma T concentrations, although this correlation was not significant in a final regression model that included waist circumference (26).
Concentrations of SHBG were also negatively associated with bioavailable E2, although more weakly than with bioavailable T. This negative association might not be expected, because estrogens clearly increase hepatic secretion of SHBG (9). However, E2 also has a high affinity for SHBG (25), and thus any effect of E2 to increase SHBG might be at least partially offset by increased SHBG decreasing its bioavailability. Furthermore, other factors that tend to be negatively associated with decreased SHBG, particularly those related to body composition, may also be positively associated with increased bioavailable E2 and estrogen action in general, as seen, for example, in a study of cycling women showing a positive association between BMI and free E2 (27).
The positive association between SHBG and estrone sulfate concentrations in the full and final multivariate linear regression analyses (Table 2) was not noted in the univariate analysis (Table 1). Similarly from Table 3 it is clear that only when other variables were taken into account is this association apparent. This speaks to the importance of confounding. Albumin is the only protein that binds estrone sulfate in plasma (28). Accordingly, for estrone sulfate, unlike for bioavailable E2, in the absence of confounding variables a strong positive association with SHBG would be expected, as increased hepatic SHBG production resulting from increased estrogen action would not be offset by decreased bioavailability. However, for estrone sulfate substantial confounding is expected. Unlike E2, estrone sulfate largely arises from peripheral conversion of androgens in fat, and not direct ovarian secretion. As a result, neglecting factors related to body composition, such as those in the full and final multivariate linear regression models, could obscure the described expected positive association between SHBG and estrone sulfate. Specifically, for example, obesity in and of itself would be expected to be associated with reduced hepatic SHBG secretion, but at the same time increased estrone sulfate production. Thus in the case of estrone sulfate, the disparity between the multivariate and univariate analysis results is likely a consequence of the multivariate models factoring in confounding variables that otherwise negate the expected positive association.
The positive association in our subjects between SHBG and HOMA was not expected. However, this association was not clear-cut, in that the multivariate-adjusted mean SHBG concentration in the first HOMA tertile was intermediate between that for the second and third HOMA tertiles. Studies of most populations have shown a significant negative relationship between SHBG and HOMA, or alternatively, a positive relationship between SHBG and measures of insulin sensitivity. However, a recent compilation of pertinent studies by Wallace et al. (23) included only a single study of premenopausal women (29, 30), in which only a nonsignificant negative association between SHBG and HOMA was found, similar to another study that found no significant association after controlling for BMI (31). In a similar vein, in a study of premenopausal women in which weight loss led to increased SHBG concentrations, it was concluded that this increase was mediated by decreased intra-abdominal fat and not changes in insulin (32). In any event, our results certainly imply that in normal cycling women there is no straightforward negative association between SHBG and HOMA.
As expected, we found a positive association between SHBG and adiponectin, a highly abundant plasma adipokine that upregulates intrahepatic SHBG production via increased HNF4α (33), and is tightly correlated with SHBG in both sexes (34, 35). In general low adiponectin concentrations are associated with markers of insulin resistance, with evidence for bidirectional cause and effect relationships between the two (35). Such effects and particularly that of T itself to at least indirectly decrease adiponectin concentrations (36, 37) are likely the main reasons for the weakening of the positive association of SHBG and adiponectin when other factors were taken into account, seen when comparing the pertinent univariate and multivariate regression results in Tables 1 and 2.
The urinary 2-OHE2/16α-OHE1 ratio was included in the full linear regression model, as higher ratio values may be inversely associated with breast cancer risk (38). The ratio was positively associated with SHBG in the univariate analysis, and included in the final model, but the association was lost in the multivariate analysis, suggesting that the univariate analysis result was on the basis of confounding.
The only body composition parameter included in the final model was the android/gynoid fat ratio. Android or abdominal fat as measured by DXA includes both visceral or intra-abdominal fat and subcutaneous abdominal fat, both of which, along with central obesity, have been shown to negatively associated with SHBG in premenopausal women (32, 39, 40). Thus the strong negative association between SHBG and the android/gynoid fat ratio in the univariate linear regression analysis (Table 1) was expected. Similarly both the unadjusted results for SHBG concentrations by android/gynoid ratio tertiles (Table 3) and Figure 1 indicate a clear-cut negative association between SHBG and the ratio when other factors are not taken into account. However, the association was not significant in the full or final multivariate linear regression model results (Table 2), similar to the adjusted results for SHBG concentrations by android/gynoid ratio tertiles. Similar results overall were found for BMI and body fat %, in that both were significantly correlated with SHBG in the univariate analyses, but not in the full model. This suggests that confounding factors, most obviously bioavailable T and adiponectin, largely mediate these associations. This speaks to the potential value of assessing many parameters when attempting to better understand the complex relationships involved, especially in a cross-sectional study such as ours.
Our study differs from most studies assessing relationships between SHBG and the other variables assessed in that our subjects were all well characterized women with ovulatory cycles. Its strengths include the large number of variables assessed, and that blood and urine samples were obtained at specific times of the menstrual cycle. Although it has been well established for decades that premenopausal women with conditions of androgen excess such as polycystic ovarian syndrome (PCOS) and linked ovulatory dysfunction have low SHBG concentrations (2), our results highlight the link between androgen action as reflected by bioavailable T and circulating SHBG concentrations in all premenopausal women. Our results also are consistent with the absence of an important negative relationship between SHBG and insulin resistance in the same population. Finally they underscore the importance of the relationship between SHBG and adiponeptin, and suggest that this relationship, which is likely bidirectional (34), is at least in part independent of androgen action.
Abstract capsule.
In normally cycling premenopausal women, sex hormone-binding globulin concentrations, independently of other parameters, are negatively associated with bioavailable T and positively associated with adiponectin.
Acknowledgments
The authors wish to acknowledge the General Clinical Research Center at the University of Minnesota, the study administrative staff, and study participants.
Grant Support: The WISER study was funded by the NIH/National Cancer Institute grant 1U54CA116849-010003, the Department of Defense/U.S. Army Medical Research and Materiel Command Congressionally Directed Medical Research Programs award #W81XWH-08-1-0301, and the NIH/National Center for Research Resources grant M01-RR00400.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
T.N.C. has nothing to disclose. A.Y.A. has nothing to disclose. M.S.K. has nothing to disclose. K.H.S. has nothing to disclose. W.R.P. has nothing to disclose.
Clinical Trial Registration: clinicaltrials.gov identifier: NCT00393172.
REFERENCES
- 1.Avvakumov GV, Cherkasov A, Muller YA, Hammond GL. Structural analyses of sex hormone-binding globulin reveal novel ligands and function. Mol Cell Endocrinol. 2010;316:13–23. doi: 10.1016/j.mce.2009.09.005. [DOI] [PubMed] [Google Scholar]
- 2.Hammond GL, Wu TS, Simard M. Evolving utility of sex hormone-binding globulin measurements in clinical medicine. Curr Opin Endocrinol Diabetes Obes. 2012;19:183–9. doi: 10.1097/MED.0b013e328353732f. [DOI] [PubMed] [Google Scholar]
- 3.Mendel CM. The free hormone hypothesis: a physiologically based mathematical model. Endocr Rev. 1989;10:232–74. doi: 10.1210/edrv-10-3-232. [DOI] [PubMed] [Google Scholar]
- 4.Rosner W, Hryb DJ, Kahn SM, Nakhla AM, Romas NA. Interactions of sex hormone-binding globulin with target cells. Mol Cell Endocrinol. 2010;316:79–85. doi: 10.1016/j.mce.2009.08.009. [DOI] [PubMed] [Google Scholar]
- 5.Willnow TE, Hammes A, Nykjaer A. Endocytosis of sex steroids: the hypothesis of free hormones revisited. Ann Endocrinol (Paris) 2008;69:101–2. doi: 10.1016/j.ando.2008.02.024. [DOI] [PubMed] [Google Scholar]
- 6.Anderson DC. Sex-hormone-binding globulin. Clin Endocrinol (Oxf) 1974;3:69–96. doi: 10.1111/j.1365-2265.1974.tb03298.x. [DOI] [PubMed] [Google Scholar]
- 7.Le TN, Nestler JE, Strauss JF, 3rd, Wickham EP., 3rd Sex hormone-binding globulin and type 2 diabetes mellitus. Trends Endocrinol Metab. 2012;23:32–40. doi: 10.1016/j.tem.2011.09.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Fortunati N, Catalano MG, Boccuzzi G, Frairia R. Sex Hormone-Binding Globulin (SHBG), estradiol and breast cancer. Mol Cell Endocrinol. 2010;316:86–92. doi: 10.1016/j.mce.2009.09.012. [DOI] [PubMed] [Google Scholar]
- 9.Pugeat M, Nader N, Hogeveen K, Raverot G, Dechaud H, Grenot C. Sex hormone-binding globulin gene expression in the liver: drugs and the metabolic syndrome. Mol Cell Endocrinol. 2010;316:53–9. doi: 10.1016/j.mce.2009.09.020. [DOI] [PubMed] [Google Scholar]
- 10.Ding EL, Song Y, Malik VS, Liu S. Sex differences of endogenous sex hormones and risk of type 2 diabetes: a systematic review and meta-analysis. JAMA. 2006;295:1288–99. doi: 10.1001/jama.295.11.1288. [DOI] [PubMed] [Google Scholar]
- 11.Arikawa AY, O'Dougherty M, Kaufman BC, Smith AJ, Thomas W, Warren M, et al. Women in Steady Exercise Research (WISER): study design and methods. Contemp Clin Trials. 2010;31:457–65. doi: 10.1016/j.cct.2010.05.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Arikawa AY, Thomas W, Gross M, Smith A, Phipps WR, Kurzer MS, et al. Aerobic training reduces systemic oxidative stress in young women with elevated levels of F2-isoprostanes. Contemp Clin Trials. 2013;34:212–7. doi: 10.1016/j.cct.2012.11.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Smith AJ, Phipps WR, Arikawa AY, O'Dougherty M, Kaufman B, Thomas W, et al. Effects of aerobic exercise on premenopausal sex hormone levels: results of the WISER study, a randomized clinical trial in healthy, sedentary, eumenorrheic women. Cancer Epidemiol Biomarkers Prev. 2011;20:1098–106. doi: 10.1158/1055-9965.EPI-10-1219. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Smith AJ, Phipps WR, Thomas W, Schmitz KH, Kurzer MS. The effects of aerobic exercise on estrogen metabolism in healthy premenopausal women. Cancer Epidemiol Biomarkers Prev. 2013;22:756–64. doi: 10.1158/1055-9965.EPI-12-1325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.American College of Sports Medicine . Guidelines for Exercise Testing and Prescription. 6th ed. Lippincott Williams & Wilkins; Philadelphia, PA: 2000. [Google Scholar]
- 16.Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of simple methods for the estimation of free testosterone in serum. J Clin Endocrinol Metab. 1999;84:3666–72. doi: 10.1210/jcem.84.10.6079. [DOI] [PubMed] [Google Scholar]
- 17.Mazer NA. A novel spreadsheet method for calculating the free serum concentrations of testosterone, dihydrotestosterone, estradiol, estrone and cortisol: with illustrative examples from male and female populations. Steroids. 2009;74:512–9. doi: 10.1016/j.steroids.2009.01.008. [DOI] [PubMed] [Google Scholar]
- 18.Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia. 1985;28:412–9. doi: 10.1007/BF00280883. [DOI] [PubMed] [Google Scholar]
- 19.Arikawa AY, Kurzer MS, Thomas W, Schmitz KH. No effect of exercise on insulin-like growth factor-I, insulin, and glucose in young women participating in a 16-week randomized controlled trial. Cancer Epidemiol Biomarkers Prev. 2010;19:2987–90. doi: 10.1158/1055-9965.EPI-10-0828. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Arikawa AY, Thomas W, Schmitz KH, Kurzer MS. Sixteen weeks of exercise reduces C-reactive protein levels in young women. Med Sci Sports Exerc. 2011;43:1002–9. doi: 10.1249/MSS.0b013e3182059eda. [DOI] [PubMed] [Google Scholar]
- 21.Mallows CL. Some comments on Cp. Technometrics. 1973;15:661–75.22. [Google Scholar]
- 22.Winters SJ, Gogineni J, Karegar M, Scoggins C, Wunderlich CA, Baumgartner R, et al. Sex hormone-binding globulin gene expression and insulin resistance. J Clin Endocrinol Metab. 2014;99:E2780–8. doi: 10.1210/jc.2014-2640. [DOI] [PubMed] [Google Scholar]
- 23.Wallace IR, McKinley MC, Bell PM, Hunter SJ. Sex hormone binding globulin and insulin resistance. Clin Endocrinol (Oxf) 2013;78:321–9. doi: 10.1111/cen.12086. [DOI] [PubMed] [Google Scholar]
- 24.Polderman KH, Gooren LJ, Asscheman H, Bakker A, Heine RJ. Induction of insulin resistance by androgens and estrogens. J Clin Endocrinol Metab. 1994;79:265–71. doi: 10.1210/jcem.79.1.8027240. [DOI] [PubMed] [Google Scholar]
- 25.Dunn JF, Nisula BC, Rodbard D. Transport of steroid hormones: binding of 21 endogenous steroids to both testosterone-binding globulin and corticosteroid-binding globulin in human plasma. J Clin Endocrinol Metab. 1981;53:58–68. doi: 10.1210/jcem-53-1-58. [DOI] [PubMed] [Google Scholar]
- 26.Nayeem F, Nagamani M, Anderson KE, Huang Y, Grady JJ, Lu LJ. Dietary beta-tocopherol and linoleic acid, serum insulin, and waist circumference predict circulating sex hormone-binding globulin in premenopausal women. J Nutr. 2009;139:1135–42. doi: 10.3945/jn.108.103291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Yeung EH, Zhang C, Albert PS, Mumford SL, Ye A, Perkins NJ, et al. Adiposity and sex hormones across the menstrual cycle: the BioCycle Study. Int J Obes (Lond) 2013;37:237–43. doi: 10.1038/ijo.2012.9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Rosenthal HE, Pietrzak E, Slaunwhite WR, Jr., Sandberg AA. Binding of estrone sulfate in human plasma. J Clin Endocrinol Metab. 1972;34:805–13. doi: 10.1210/jcem-34-5-805. [DOI] [PubMed] [Google Scholar]
- 29.Akin F, Bastemir M, Alkis E. Effect of insulin sensitivity on SHBG levels in premenopausal versus postmenopausal obese women. Adv Ther. 2007;24:1210–20. doi: 10.1007/BF02877767. [DOI] [PubMed] [Google Scholar]
- 30.Akin F, Bastemir M, Alkis E, Kaptanoglu B. SHBG levels correlate with insulin resistance in postmenopausal women. Eur J Intern Med. 2009;20:162–7. doi: 10.1016/j.ejim.2007.09.023. [DOI] [PubMed] [Google Scholar]
- 31.Yeung EH, Zhang C, Mumford SL, Ye A, Trevisan M, Chen L, et al. Longitudinal study of insulin resistance and sex hormones over the menstrual cycle: the BioCycle Study. J Clin Endocrinol Metab. 2010;95:5435–42. doi: 10.1210/jc.2010-0702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Azrad M, Gower BA, Hunter GR, Nagy TR. Intra-abdominal adipose tissue is independently associated with sex-hormone binding globulin in premenopausal women. Obesity (Silver Spring) 2012;20:1012–5. doi: 10.1038/oby.2011.375. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Simo R, Saez-Lopez C, Lecube A, Hernandez C, Fort JM, Selva DM. Adiponectin Upregulates SHBG Production: Molecular Mechanisms and Potential Implications. Endocrinology. 2014;155:2820–30. doi: 10.1210/en.2014-1072. [DOI] [PubMed] [Google Scholar]
- 34.Bonnet F, Balkau B, Malecot JM, Picard P, Lange C, Fumeron F, et al. Sex hormone-binding globulin predicts the incidence of hyperglycemia in women: interactions with adiponectin levels. Eur J Endocrinol. 2009;161:81–5. doi: 10.1530/EJE-09-0202. [DOI] [PubMed] [Google Scholar]
- 35.Cook JR, Semple RK. Hypoadiponectinemia--cause or consequence of human “insulin resistance”? J Clin Endocrinol Metab. 2010;95:1544–54. doi: 10.1210/jc.2009-2286. [DOI] [PubMed] [Google Scholar]
- 36.Nishizawa H, Shimomura I, Kishida K, Maeda N, Kuriyama H, Nagaretani H, et al. Androgens decrease plasma adiponectin, an insulin-sensitizing adipocyte-derived protein. Diabetes. 2002;51:2734–41. doi: 10.2337/diabetes.51.9.2734. [DOI] [PubMed] [Google Scholar]
- 37.Horenburg S, Fischer-Posovszky P, Debatin KM, Wabitsch M. Influence of sex hormones on adiponectin expression in human adipocytes. Horm Metab Res. 2008;40:779–86. doi: 10.1055/s-0028-1083780. [DOI] [PubMed] [Google Scholar]
- 38.Bradlow HL, Hershcopf RJ, Martucci CP, Fishman J. Estradiol 16 alpha-hydroxylation in the mouse correlates with mammary tumor incidence and presence of murine mammary tumor virus: a possible model for the hormonal etiology of breast cancer in humans. Proc Natl Acad Sci U S A. 1985;82:6295–9. doi: 10.1073/pnas.82.18.6295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Tchernof A, Toth MJ, Poehlman ET. Sex hormone-binding globulin levels in middle-aged premenopausal women. Associations with visceral obesity and metabolic profile. Diabetes Care. 1999;22:1875–81. doi: 10.2337/diacare.22.11.1875. [DOI] [PubMed] [Google Scholar]
- 40.Evans DJ, Hoffmann RG, Kalkhoff RK, Kissebah AH. Relationship of androgenic activity to body fat topography, fat cell morphology, and metabolic aberrations in premenopausal women. J Clin Endocrinol Metab. 1983;57:304–10. doi: 10.1210/jcem-57-2-304. [DOI] [PubMed] [Google Scholar]