Androgenic sex steroids contribute to metabolic risk beyond intra-abdominal fat in overweight/obese black and white women

Authors


  • Disclosure: The authors declared no conflict of interest.

Correspondence: Arlette Perry (aperry@miami.edu)

Abstract

Objective

To determine the independent contribution of androgenic sex hormones beyond visceral adipose tissue (VAT) on metabolic risk.

Design and Methods

A cross-sectional evaluation of 66 (36 white and 30 black) premenopausal overweight/obese women using multiple regression analyses to determine the independent effects of sex hormone-binding globulin (SHBG), total testosterone (TT), and free testosterone using the free androgen index (FAI) on metabolic variables above VAT.

Results

SHBG contributed to the variance in insulin (P = 0.003), insulin resistance using HOMA-IR (P = 0.006), and high-density lipoprotein cholesterol2 (P = 0.029). TT contributed to the variance in systolic and diastolic blood pressure (P < 0.001), total cholesterol (P = 0.003), low-density lipoprotein cholesterol (P = 0.003), and apolipoprotein B (P = 0.004). FAI contributed to the variance in the greatest number of metabolic variables beyond VAT. There was also a significant race–FAI interaction for fasting glucose (P = 0.013). A Pearson's correlation coefficient showed a significant relationship between FAI and glucose in white women (r = 0.48, P = 0.003) while showing no relationship in black women (r = −0.01, P = 0.941).

Conclusions

Our study showed that androgenic sex steroids contributed significantly to the variance in metabolic variables associated with health risk. However, they do not provide sufficient information relevant to glucose status in black women.

Introduction

Past research has shown that the distribution of body fat is a better indicator of metabolic perturbations associated with diabetes and cardiovascular disease (CVD) than total obesity [1, 2]. Visceral adipose tissue (VAT) characterized by intra-abdominal adipose tissue is an independent predictor of the metabolic syndrome [3], diabetes [4, 5], and CVD [6]. The excessive accumulation of VAT observed in visceral obesity is reported to influence insulin resistance, compensatory hyperinsulinemia, and their association with a cluster of abnormalities that raise one's risk of coronary heart disease [7].

Several research studies have also shown that androgenic sex steroids are intricately associated with the metabolic perturbations observed in visceral obesity. In the SWAN study [10], it was shown that bioavailable testosterone was the strongest predictor of VAT, whereas sex hormone-binding globulin (SHBG) was just as strong as an inverse predictor of VAT, independent of race, total percent body fat, and other CVD risk factors. Given the strength of the relationship between VAT and bioavailable testosterone and its inverse relationship with SHBG, it would be important to determine if androgenic steroids have an independent effect on metabolic risk. Studies have already shown that elevated androgens and low SHBG are related to metabolic variables associated with CVD risk [11]. However, some studies have shown that upon controlling for VAT, the relationship between metabolic variables and SHBG and/or bioavailable testosterone is no longer found [14]. Furthermore, others have postulated that it is the bioavailable testosterone that regulates VAT accumulation associated with CVD risk, and may even directly contribute to CVD risk [10, 17].

Thus, it is unclear, as to whether VAT mediates the relationship between androgenic sex steroids and metabolic risk factors or whether androgenic sex steroids mediate the VAT–risk factor relationships. Overweight/obese premenopausal women have expanded fat depots and greater VAT accumulation and they represent an important group to examine early on. Furthermore, significant relationships between androgenic sex steroids and metabolic risk factors have been observed in younger women with no clinical evidence of disease [11, 13].

This study was done to examine whether SHBG, total testosterone (TT), and/or bioavailable testosterone using the free androgen index (FAI) can provide information relevant to metabolic risk beyond that of VAT in a biracial sample of overweight/obese premenopausal women.

Subjects and Methods

Study sample

All subjects consisted of women volunteers interested in a weight loss program. Participants were required to be premenopausal, weight stable for a minimum of 3 months, possessing a BMI above 25, and free of known metabolic disease such as diabetes, hyperlipidemia, hypertension, polycystic ovary syndrome, or cardiovascular disease. Subjects who were pregnant or lactating, amenorrheic, or taking medications that would affect BP, carbohydrate, or lipid metabolism were excluded from the study. All subjects taking oral contraceptives or medicines for thyroid disorders were excluded from the study. Subjects were required to be eumenorrheic having a normal monthly menstrual cycle for the proceeding 6 months. A total of 66 subjects (36 white and 30 black) met the criteria necessary for participation in the study. Subjects gave informed consent and completed all testing procedures in accordance with the Institutional Review Board guidelines for use of human subjects at the University of Miami.

Physical and anthropometric measurements

Body weight was measured to the nearest 0.1 kg. Height was measured to the nearest 0.5 cm, and BMI was calculated as weight (kg) divided by the square of height (m2). Waist circumference measurements were taken midway between the lower rib margin and iliac crest by the same investigator who recorded the mean of the two measurements to the nearest 1 mm.

Systolic and diastolic blood pressure (SBP and DBP, respectively) were measured using an aneroid sphygmomanometer validated against our automated system (General Electric Medical System Information Technologies, Milwaukee, WI) using a special, large adult latex-free pressure cuff with appropriate fit (Critikom, Tampa, FL). Measurements were taken after subjects were quiet and seated for a minimum of 5 min. Duplicate BP measurements were taken from the left upper arm, averaged, and recorded to the nearest 2 mm Hg with a 5-min interval separating measurements.

Magnetic resonance imaging

The abdominal region was examined using a magnetic resonance imaging with a 1.5 T instrument (Siemens Medical 50 Systems, Iselin, NJ). Spin-echo imaging was performed using a T1-weighted sequence with a 147-ms repetition time and 4.8-ms echo time. During a single breath hold, image thickness was 10 mm with a 2.5-mm gap between images. A total of seven images were obtained in each subject with the central slice of the acquisition center at the L4-L5 intervertebral disk space. The VAT was defined as adipose tissue contained within the boundaries of the rectus abdominus, internal obliques, quadratus lumborum, and long back muscles, whereas the subcutaneous adipose tissue (SAT) was the adipose tissue located between the skin and the same group of muscles. The VAT and SAT volume was computed by summing the VAT and SAT area, respectively, in each slice multiplied by the nominal slice thickness of 10 mm and converting to liters.

Serum measurements

Blood was withdrawn from the antecubital vein following a 12-h fast, immediately centrifuged to separate serum, and analyzed within 1 week of venipuncture. All serum measurements were taken while subjects were seated in a quiet position for at least 5 min. Total cholesterol [18], high-density lipoprotein (HDL) cholesterol [19], its subfractions [20], and triglycerides (TG) [21] were measured by the Diabetes Research Institute Lipid Laboratory, University of Miami. Serum standards used for calibration were developed by the Diabetes Research Institute and calibrated against serum samples from the Centers for Disease Control and Prevention Laboratory, Atlanta, Georgia. Low-density lipoprotein (LDL) cholesterol was calculated as total cholesterol minus HDL cholesterol minus (triglycerides ÷ 5) [22] as used by the National Cholesterol Program to establish cholesterol cut points [23].

Apolipoprotein B (Apo B) in serum was measured by turbidimetric immunoassay using a commercially available kit (Incstar, Stillwater, MN) according to procedures outlined by the manufacturers. All Apo B procedures have been developed in accordance with the guidelines set forth by the International Federation of Clinical Chemistry.

Fasting glucose levels were determined spectrophotometrically at a wavelength of 340 nm using a hexokinase reaction developed by Roche (Roche Diagnostic System, Nutley, NJ).

Serum insulin was measured by radioimmunoassay using a Coat-A-Count insulin procedure (Diagnostic Products, Los Angeles, CA). The insulin resistance was assessed by using the homeostasis model assessment (HOMA) index, which divides the product of fasting insulin and glucose by 22.5 [24].

The SHBG was analyzed in serum by immunoradiometric assay (Siemens Health Care Diagnostics, Los Angeles, CA). To avoid interassay variation, all assays were analyzed in duplicate in a single assay for each hormone. The intra-assay and interassay coefficients were 2.7 and 4.5%, respectively. The TT was analyzed by COAT-A-COUNT Radioimmunoassay with a testosterone-specific antibody using an extraction step (Siemens Health Care Diagnostics). Quality control procedures for accuracy and reliability in the laboratory yielded intra-assay and interassay coefficients of variation of less than 4 and 7.5%, respectively, for TT. The FAI was used to estimate the amount of testosterone unbound from SHBG and therefore biologically active. FAI was derived from the ratio of TT/SHBG calculated as TT × 3.47 × 100/SHBG and quantified in nanomoles per liter [25]. As the active portion of testosterone includes the non-SHBG-bound fractions, FAI has been used as an accurate, reliable estimate of bioavailable testosterone [26].

Statistical analysis

All statistical analyses were completed using the Statistical Package for the Social Sciences, version 10.1. Means ± standard deviations of all variables were calculated. Natural log transformation was used for triglycerides, SHBG, and FAI, while the square root of TT was used to achieve normality of distribution. As one black woman had abnormally elevated blood glucose consistent with diabetes (3 standard deviations above the mean), her values were removed from the Student's t-test and subsequent regression analyses.

Separate multiple regression models were used to examine the relationship between the three androgenic sex steroids and each of the metabolic variables associated with health risk. In each model, VAT was first entered into the model to determine its relationship with each of the metabolic variables. This was followed by entering a second independent variable of interest (either SHBG, TT, or FAI), for each model. The change in F score was used to determine if any of the three independent variables (SHBG, TT, or FAI) added a statistically significant amount of explained variance to the model. Third, race was entered into the model to determine the independent effects of race on metabolic variables after controlling for VAT and either one of the three independent variables (SHBG, TT, or FAI). Finally, an interaction term of race and the independent variable of interest (SHBG, TT, or FAI) were entered into the model to determine if the influence of androgenic steroids on metabolic variables was significantly different in white and black women after controlling for VAT. If a significant race effect was found, the regression analysis was repeated after adding VAT and the independent variable of interest (SHBG, TT, or FAI) for white and black women separately. If a significant race interaction was found and VAT did not contribute significantly to the model, it was removed and Pearson's correlation coefficient was performed to determine the relationship between the variables in question by race. Significance was set at ≤0.05.

Results

Physical characteristics

A comparison of subject characteristics between white and black women is presented in Table 1. Means ± SD were reported for 66 subjects (36 white and 30 black women). Values for SHBG and TT were similar to other studies and within normal range [14, 15].

Table 1. Subject characteristics for the entire sample (N = 66)
CharacteristicsMean (X)Standard deviation
  1. Apo B, apolipoprotein B; BMI, body mass index; FAI, free androgen index, calculated as TT × 3.467 × 100/SHBG; HDL, high-density lipoprotein cholesterol; HDL2, high-density lipoprotein subfraction2; HOMA IR, insulin resistance calculated as (insulin, μU/ml × glucose, mmol/l)/22.5; LDL, low-density lipoprotein cholesterol; SAT, volume of subcutaneous adipose tissue; VAT, volume of visceral adipose tissue.
  2. aOne outlier was excluded from the analysis.
Age (years)40.138.67
BMI (kg/m2)35.014.91
VAT (l)1.000.46
SAT (l)4.591.32
Metabolic variables
Systolic blood pressure (mm Hg)125.0014.55
Diastolic blood pressure (mm Hg)82.159.43
Insulin (μU/ml)14.669.88
Glucose (mmol/l)a4.780.50
HOMA IR (ratio)3.161.62
Total cholesterol (mg/dl)202.8752.37
HDL (mg/dl)55.7613.92
HDL2 (mg/dl)14.669.88
LDL (mg/dl)120.4342.16
Apo B (mg/dl)104.7729.72
Triglycerides (mg/dl)127.8075.43
Total cholesterol/HDL (ratio)3.811.29
Androgenic steroids
Sex hormone-binding globulin (nmol/l)38.3927.02
Total testosterone (ng/ml)0.290.16
Free androgen index (ratio)4.184.15

Figure 1 presents the relationship between two different measures of central obesity, SAT and VAT, with metabolic variables associated with CVD risk. SAT contributed significantly to the variance in one metabolic variable, SBP (R2 = 10.06%; P = 0.001). In contrast, VAT contributed to the variance in five metabolic variables: SBP (R2 = 16.6%; P = 0.003), DBP (R2 = 8.5; P = 0.018), insulin (R2 = 14.4; P < 0.0020), HOMA IR (R2 =17.9; P = 0.001), and TGs (R2 = 13.2%; P = 0.003).

Figure 1.

The contribution of visceral (VAT) and subcutaneous (SAT) adipose tissue to the variance in metabolic variables for the entire sample (n = 66). SBP, systolic blood pressure; DBP, diastolic blood pressure; HOMA IR, insulin resistance calculated as (insulin × glucose) ÷ 22.5; T-Chol, total cholesterol; HDL, high-density lipoprotein cholesterol; HDL2, high-density lipoprotein subfraction2; LDL, low-density lipoprotein cholesterol; Apo B, apolipoprotein B; TG, triglycerides. *P ≤ 0.05; **P ≤ 0.01; †P ≤ 0.001.

Shown in Figure 2 are the contributions of SHBG to the metabolic variables associated with health risk beyond that of VAT. SHBP was significantly associated with three metabolic variables: insulin (R2 = 10.8; P = 0.001), HOMA IR (R2 = 11.9%; P = 0.001), and HDL2 (R2 = 7.3%; P = 0.029).

Figure 2.

The contribution of sex hormone-binding globulin (SHBG) to the variance in metabolic variables after controlling for visceral adipose tissue (VAT) for the entire sample (n = 66). SBP, systolic blood pressure; DBP, diastolic blood pressure; HOMA IR, insulin resistance calculated as (insulin × glucose) ÷ 22.5; T-Chol, total cholesterol; HDL, high-density lipoprotein cholesterol; HDL2, high-density lipoprotein subfraction2; LDL, low-density lipoprotein cholesterol; Apo B, apolipoprotein B; TG, triglycerides. *P ≤ 0.05; **P ≤ 0.01; †P ≤ 0.001.

Although there were no significant SHBG–race interactions, there was a significant race effect for insulin (R2 = 5.0%; P = 0.038) and HOMA IR (R2 = 7.6%; P = 0.011). After controlling for VAT, SHBG contributed 14.6% (P = 0.004) and 9.9% (P = 0.013) to the variance in insulin and HOMA IR, respectively, for white women. In black women, SHBG contributed 11.5% (P = 0.07) and 11.2% (P = 0.094) to the variance in insulin and HOMA IR, respectively.

Shown in Figure 3 is the contribution of TT to metabolic variable associated with health risk after controlling for VAT. TT was significantly associated with five metabolic variables: SBP (R2 = 14.8%; P = 0.00), DBP (R2 = 20.2%; P = 0.001), total cholesterol (R2 = 13.3%; P = 0.003), LDL cholesterol (R2 = 12.0%; P = 0.003), and Apo B (R2 = 12.1%; P = 0.004). There was also a race effect for HOMA IR (R2 = 5.9%; P = 0.035). Upon further analysis, TT showed a small nonsignificant relationship with HOMA IR in white women (R2 = 3.9%; P = 0.127) and no relationship at all in black women (R2 = 0; P = 0.979).

Figure 3.

The contribution of total testosterone (TT) to the variance in metabolic variables after controlling for visceral adipose tissue (VAT) for the entire sample (n = 66). SBP, systolic blood pressure; DBP, diastolic blood pressure; HOMA IR, insulin resistance calculated as (insulin × glucose) ÷ 22.5; T-Chol, total cholesterol; HDL, high-density lipoprotein cholesterol; HDL2, high-density lipoprotein subfraction2; LDL, low-density lipoprotein cholesterol; Apo B, apolipoprotein B; TG, triglycerides. *P ≤ 0.05; **P ≤ 0.01; †P ≤ 0.001.

Shown in Figure 4 is the contribution of FAI to metabolic variables beyond VAT. After controlling for VAT, FAI contributed significantly to 10 metabolic variables including: SBP (R2 = 10.1%; P = 0.006), DBP (R2 = 12.0%; P = 0.003), insulin (R2 = 10.7%; P = 0.004), glucose (R2 = 7.4%; P = 0.03), HOMA IR (R2 = 9.5%; P = 0.007), total cholesterol (R2 = 6.9%; P = 0.034), HDL cholesterol (R2 = 7.5%; P = 0.025), HDL2 (R2 = 8.5%; P = 0.018), LDL cholesterol (R2 = 7.1%; P = 0.032), and Apo B (R2 = 5.7%; P = 0.051).

Figure 4.

The contribution of free androgen index (FAI) to the variance in metabolic variables after controlling for visceral adipose tissue (VAT) for the entire sample (n = 66). SBP, systolic blood pressure; DBP, diastolic blood pressure; HOMA IR, insulin resistance calculated as (insulin × glucose) ÷ 22.5; T-Chol, total cholesterol; HDL, high-density lipoprotein cholesterol; HDL2, high-density lipoprotein subfraction2; LDL, low-density lipoprotein cholesterol; Apo B, apolipoprotein B; TG, triglycerides. *P ≤ 0.05; **P ≤ 0.01; †P ≤ 0.001.

There was also a race effect for HOMA IR (F Change = 5.120; R2 = 5.9%; P = 0.027) and a trend toward a race effect for insulin (F Change = 3.003; R2 = 3.5%; P = 0.088). Upon further analysis, FAI contributed positively 11.3% to the variance in HOMA IR (P = 0.007) and 13.4% to the variance in insulin (P = 0.006) in white women. In black women, FAI contributed positively 6.7% to the variance in HOMA IR (P = 0.176) and 5.1% to the variance in insulin (P = 0.265).

A significant FAI–race interaction was also demonstrated for glucose (F Change = 4.930; R2 change = 0.070%; P = 0.030). Upon controlling for VAT, there was a significant positive relationship between FAI and glucose in white women (R2 = 8.48; P = 0.006) and a small negative relationship in black women (R2 = 6.7%; P = 0.176). As VAT did not significantly contribute to the variance in glucose (4.1%), the relationship between FAI and glucose was analyzed again using a simple Pearson's correlation coefficient. Shown in Figure 5 is the significant and positive correlation between FAI and glucose in white women (r = 0.484; P = 0.003) and the absence of any correlation in black women (r = −0.015; P = 0.941).

Figure 5.

The relationship between the free androgen index (FAI) and blood glucose levels in white and black women.

Discussion

Several relevant findings were noted with respect to central obesity measures, androgenic sex steroids, and its association with metabolic variables in overweight/obese women. As expected, VAT contributed more so than SAT, to variables associated with health risk. In an elegant review, Desprès et al. [27] noted the increased lipolytic activity of VAT exposing the liver to high concentrations of free fatty acids thereby resulting in elevated endogenous synthesis of TGs. Increased flux of free fatty acids may inhibit the action of insulin setting the stage for insulin resistance. VAT is also responsible for the secretion of proinflammatory adipokines that influence lipid metabolism, insulin resistance, and blood pressure [28]. Thus, it is not surprising that VAT more so than SAT was positively associated with a constellation of metabolic variables characteristic of the metabolic syndrome and CVD risk. Although SAT does have some of the same metabolic effects as VAT, it is not as strongly correlated with proinflammatory, hyperinsulinemic markers [29, 30]. Of the metabolic variables examined, SAT had a significant relationship with SBP only. This finding may also be the result of larger quantity of fat contained in SAT scores.

Our findings showed that SHBG contributed 7.5-10% of the variance in insulin, HOMA-IR, and HDL2, respectively, beyond that of VAT. Although this contribution may appear low, we performed independent linear regressions, which showed that SHBG contributed 20.7% to the variance in insulin, almost 20% to the variance in HOMA IR, and over 8.2% to the variance in HDL2. This is in contrast to VAT, which contributed 14.4, 17.9, and less than 1% to the variance in insulin, HOMA IR, and HDL2, respectively. The shared variances were much smaller, highlighting the greater independent contribution of VAT and SHBG to metabolic variables.

SHBG is a pivotal determinant of the bioavailability of testosterone and may mediate the effects of androgenic sex steroids on the metabolic profile [31]. The independent association between SHBG and HDL2 is related to the biosynthesis of hepatic lipase, which increases the conversion rate of HDL2 and HDL3. Hepatic lipase is stimulated by androgens; thus, by regulating the androgenic milieu, SHBG acts as a modulator of hepatic lipase synthesis, which influences HDL subfractions and reduces HDL2 [32]. Furthermore, insulin is reported to directly reduce SHBG secretion from the liver [33]; thus, it is not surprising that SHBG showed significant independent associations with insulin, HOMA IR, and HDL2 beyond VAT.

Interestingly, upon controlling for VAT, the relationships between SHBG and insulin as well as HOMA IR were blunted in black women. This could be related to the smaller number of black women (30 vs. 36) in our study. In addition, the relationships between SHBG and insulin as well as HOMA IR in white women were stronger and significant to begin with (r = −0.52, −0.54; P = 0.001), whereas in black women they were lower and not significant (r = −0.33; P = 0.07-0.08 for both). Thus, after controlling for VAT, the relationships are expected to decline. Our findings of a race effect for the relationship between insulin and HOMA IR with SHBG were similar to the results found in a small group of postmenopausal white and black women by Berman et al. [34] but in contrast to the SWAN Study [10, 12].

TT contributed to the variance in more metabolic variables than SHBG, accounting for an additional 15-20% of the variance in SBP and DBP, respectively. Using independent regressions, TT contributed 21.3 and 16% to the variance in SBP and DBP, respectively, while VAT explained a lesser amount of variance (12.6 and 8.5% for SBP and DBP, respectively). This may be mediated by the effect of testosterone on vasoconstriction, which upregulates the renin, angiotensin, and aldosterone pathways thereby serving to increase blood pressure [35]. TT also contributed 12-13% of the variance in atherogenic lipids including total cholesterol, LDL cholesterol, and Apo B. This occurred in the absence of any significant contribution to these metabolic variables by VAT alone.

As TT represents a combination of free and bound androgens, there is controversy regarding its utility in relation to CVD risk. In contrast to others [13, 36], our results demonstrated the added value of examining TT in relation to metabolic variables associated with CVD risk and one that is independent of VAT. This was particularly evident for SBP, DBP, and serum lipoproteins but not for insulin/glucose variables. Our results may be due, in part, to the strong correlation between TT and FAI (r = 0.7; P < 0.001).

FAI contributed information relevant to the greatest number of metabolic variables associated with health risk including SBP, DBP, insulin/glucose status, and serum markers of Apo B and lipoproteins. Even after controlling for VAT, FAI contributed a greater explained variance for DBP, serum glucose, and serum lipoproteins, which was confirmed by our independent linear regressions. In contrast, our linear regressions showed that VAT contributed more information relevant to SBP, TG, insulin, and HOMA IR with the smallest shared variance for TG. Thus, it appears that androgenic sex steroids may act directly on metabolic variables, particularly serum lipoproteins and glucose, whereas VAT appears to act directly on TGs contributing strongly to the variance in that lipid.

The relationship of androgenic sex steroids with insulin and HOMA IR is more complex. SHBG is an indirect marker of the androgen/estrogen balance and given its strong inverse relationship with FAI (r = −0.841; P < 0.001), it may play an important role in moderating the relationship between FAI and insulin as well as HOMA IR. Insulin too may moderate SHBG [33] and VAT accumulation [28], thereby making it difficult to separate the relationship between androgenic sex steroids and VAT apart from aforementioned metabolic variables.

Although FAI showed a significant relationship with the greatest number of metabolic variables, the relationship between FAI and insulin/glucose status was significantly stronger in white women. After controlling for VAT, FAI contributed over 13% to the variance in insulin in white women while contributing exactly one-half that amount in black women. A similar finding was noted for HOMA IR.

The significant FAI by race interaction for glucose highlights racial differences with regard to glucose levels. FAI contributed to over 23% (P = 0.003) of the variance in fasting glucose in white women. For each 1 SD unit increase in FAI in white women, there was a 0.48 SD unit increase in glucose, whereas in black women, almost no relationship at all was found between FAI and glucose levels (Figure 5). These findings coincide with previous research showing significant differences between white and black women in insulin sensitivity, the way glucose is utilized, and diabetes risk [37]. Our results show that FAI is not a useful marker of the evaluation of insulin/glucose status in black women. Significant associations observed between FAI and glucose as well as insulin and HOMA IR were driven by the strength of these relationships in white participants.

It can be argued that statistically significant predictors of metabolic variables in young healthy women are not as relevant as examining clinical outcomes. Ding et al. [38] have shown that in those already possessing type 2 diabetes, SHBG is lower and TT, as well as bioactive testosterone, is higher. We maintain that the identification of statistically significant contributors to metabolic variables is a forerunner of relevant clinical outcomes. This is particularly true for overweight/obese women who may be more vulnerable to metabolic perturbations associated with the metabolic syndrome and CVD risk later on.

It should be noted that there were several limitations with respect to this study. First, all relationships were based upon a single blood drawn at one time point. As this was a cross-sectional study, it was difficult to determine any temporal or causal relationships between sex steroids, SHBG, and metabolic variables. A few women in both groups were in their mid- to late 40s and may have been perimenopausal. This transition period prior to menopause may be linked to a progressive reduction in ovarian function, reduced estrogen levels, and stable androgen levels. Furthermore, samples were taken randomly across the menstrual cycle. Total and LDL cholesterol may show variability across follicular and luteal phases of the cycle [39]; thus, it would have been helpful to sample all women at the same time in their menstrual cycle. As TT and bioactive testosterone are very low in women, there is controversy regarding the best methods to accurately measure androgens. This study used high-affinity antibodies and radioactive labels along with an extraction step, which is reported to provide a more accurate detection of low concentrations of testosterone [40]. The lower limit of quantification for the testosterone assay was 0.15 ng/ml and in this study all women exceeded this value. Finally, although many of the correlations between androgenic sex steroids and metabolic variables were significant, they were small. This is expected because of the small sample size of only 66 subjects and the fact that there were only four independent predictors used in the multiple regression analysis (VAT, androgenic sex steroids, race, and an interaction term). Adding more variables to the regression, such as physical activity levels, total calorie and nutrient intake, inflammatory markers, and stress markers, would have certainly increased the explained variance in each of our models.

In conclusion, this study confirmed the fact that VAT more so than SAT contributed significantly to metabolic variables associated with health risk in overweight/obese black and white women. Furthermore, our results make a strong case for the independent contribution of androgenic sex steroids to the variance in several metabolic variables associated with CVD risk and metabolic syndrome beyond that of VAT accumulation. It should be noted, however, that androgenic sex steroids do not provide sufficient information relevant to insulin/glucose status in black women. Thus, further research is necessary to examine the underlying mechanisms responsible for race differences with regards to these variables.

Ancillary