Several studies demonstrate a link between diabetes and sex steroid hormones, but the link with pre‐diabetes remains elusive. In this study, we hypothesize that pre‐diabetes, which is characterised by having impaired fasting glucose and/or impaired glucose tolerance and/or impaired HbA1C, may influence circulating sex steroid hormone concentrations in men. Thus, we investigated whether serum sex steroid hormone concentrations differ between men with and without pre‐diabetes. We analyzed data for 1139 men who were aged 20+ years when they participated in the Third National Health and Nutrition Examination Survey. We calculated adjusted geometric mean serum concentrations of total and estimated free testosterone, androstanediol glucuronide, total and estimated free estradiol, and sex hormone‐binding globulin (SHBG) in men with and without pre‐diabetes. Logistic regression was used to calculate adjusted odds ratios (OR) of lower concentrations of androgens and SHBG, and higher concentrations of estradiol by prediabetes status. Adjusting for age and race/ethnicity, total testosterone concentration was lower among men with (geometric mean: 4.68 ng/mL) than without (5.36 ng/mL, p = 0.01) pre‐diabetes. SHBG concentration was also lower in men with (31.67 nmol/L) than without (36.16 nmol/L; p = 0.01) pre‐diabetes. Concentrations of the other hormones did not differ between men with and without pre‐diabetes. After adjusting for demographic and lifestyle factors, pre‐diabetic men had a higher odds of lower testosterone (OR: 2.58; 95% CI: 1.54–4.29), higher free estradiol level (OR: 1.59; 95% CI: 1.14–2.22), and lower SHBG level (OR: 2.27; 95% CI: 1.32–3.92) compared to men without pre‐diabetes. These associations were attenuated after adjusting for adiposity (testosterone OR: 1.76; 95% CI 0.95–3.27, free estradiol OR: 1.29, 95% CI: 0.88–1.88, SHBG OR: 1.71; 95% CI 0.88–3.30). Our findings suggest that men with pre‐diabetes have lower circulating total testosterone and SHBG and higher free estradiol levels.
Pre‐diabetes is a relatively common metabolic disorder affecting approximately 37% adults in the United States (US) (Menke et al., 2015). Although mounting evidence from several epidemiological studies (Corona et al., 2011; Grossmann, 2011) suggest that diabetes affects sex steroid hormones levels in men, few studies have explored the link between pre‐diabetes, a precursor to diabetes characterized by impaired fasting glucose or tolerance, (American Diabetes Association 2015) and sex steroid hormones. Other metabolic disturbances such as obesity and insulin resistance, which are strongly associated with pre‐diabetes (Li et al., 2009; Neeland et al., 2012), are also known to influence circulating sex steroid hormones levels (Giovannucci et al., 2010). For example, several studies showed that obese men or those with insulin resistance tend to be androgen deficient (Pitteloud et al., 2005; Grossmann, 2011; Rohrmann et al., 2011a).
Biological evidence also corroborates a link between these metabolic disturbances and derangements in sex steroid hormone metabolism (Cohen, 2008; Allan & McLachlan, 2010; Ahn et al., 2013). Chronic perturbations in glucose metabolism and hyperinsulinemia, for instance, have been shown to impair the Leydig cells, thus causing testosterone deficiency in mice (Ahn et al., 2013). Both aberrations in glucose and sex steroid hormones metabolism have been implicated with the development of chronic diseases including cardiovascular diseases and certain types of cancer (Kasper & Giovannucci, 2006; Tsilidis et al., 2014; Arthur et al., 2016). It is postulated that the derangements in sex steroid hormones metabolism may play a role in modulating the association between pre‐diabetes and these chronic diseases but the link between pre‐diabetes and sex steroid hormones remains unclear.
Thus, to inform the unclear association between pre‐diabetes and circulating sex steroid hormone concentrations, we evaluated cross‐sectional differences in serum concentrations of sex steroid hormones between men with and without pre‐diabetes using data from the US nationally representative Third National Health and Nutrition Examination Survey (NHANES III).
Materials and Methods
A more detailed description of the Third National Health and Nutrition Examination Survey (NHANES III) can be found elsewhere (National Center for Health Statistics 1994). The survey was undertaken by National Center for Health Statistics between 1988 and 1994 and aimed to evaluate the health and nutritional status of adults and children in the US. NHANES III was designed as a cross‐sectional study using a multistage‐stratified, clustered probability sample of the US civilian non‐institutionalized population. Minority groups in the US population, including Mexican Americans and non‐Hispanic black men, were oversampled to ensure that there were adequate numbers in these subgroups. Participants were interviewed at home and underwent extensive physical examinations in mobile examination centers, where blood samples were collected. Blood was drawn after an overnight fast for participants in the morning sample during either an examination in the medical examination center or during an abbreviated examination at home. After centrifugation, the serum was aliquoted and stored at 70 °C.
NHANES III was conducted in two phases (1988–1991 and 1991–1994). Within each phase, subjects were randomly assigned to participate in either the morning or afternoon/evening examination session. Of the 14,781 male participants, of whom 9282 were at least 12 years old, 2205 took part in the morning session of phase 1. Morning sample participants were chosen for the study of sex steroid hormones to reduce extraneous variation because of diurnal production of sex hormones. Among the 2205 men who participated in the morning session, 1637 had available remaining serum samples in the main NHANES III repository. After excluding men with doctor‐diagnosed diabetes and those with fasting glucose >125 mg/dL (6.9 mmol/L) and/or 2‐h plasma glucose concentration in the 75 g oral glucose tolerance test >199 mg/dL (11.0 mmol/L), and/or HbA1c >6.4%, our final study population comprised 1139 men aged 20 years or older who had serum measurements for total testosterone, total estradiol, sex hormone‐binding globulin (SHBG), androstanediol glucuronide (AAG), and had their percent body fat assessed (Fig. 1).
Schematic representation of analytic study population.
Institutional review board approvals
The protocols for the conduct of NHANES III were approved by the institutional review board of the National Center for Health Statistics, US Centers for Disease Control and Prevention. Informed consent was obtained from all participants. The assay of stored serum specimens for the Hormone Demonstration Program was approved by institutional review boards at the Johns Hopkins Bloomberg School of Public Health and the National Center for Health Statistics, US Centers for Disease Control and Prevention.
Exposure measurements and covariates
Information on age, race/ethnicity, cigarette smoking, alcohol consumption, and physical activity was collected during the interview. Race and ethnicity were combined into four race/ethnic groups: non‐Hispanic white, non‐Hispanic black, Mexican American, and other. Participants were classified as never, former, and current smokers (<20, 20–39, ≥40 cigarettes per day) based on the self‐reported smoking habits. Frequency of alcohol consumption was measured by a food frequency questionnaire and categorized by times per week. Moderate or vigorous physical activity was defined by the following activities: jogging or running; swimming or aerobics (for men 40 years or older); biking, dancing, gardening, and calisthenics (for men 65 years or older); and walking and lifting weights (for men 80 years and older), and was categorized by frequency. Height and weight were measured as part of the NHANES III examination. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. To determine the subject’s whole body electrical resistance, a Valhalla Scientific Body Composition Analyzer was used (model 1990B; Valhalla Scientific, Inc., San Diego, CA, USA) (Sun et al., 2003). Validated prediction equations using height and weight were then used to convert whole body electrical resistance at 50 kHz to percent body fat (CV: 8%) (Sun et al., 2003).
Men were defined as having pre‐diabetes if they met the American Diabetes Association criteria (American Diabetes Association 2015): (i) Impaired fasting glucose (IFG): fasting plasma glucose concentration from 100 mg/dL (5.6 mmol/L) to 125 mg/dL (6.9 mmol/L), (ii) Impaired glucose tolerance (IGT): 2‐h plasma glucose concentration in the 75 g oral glucose tolerance test from 140 mg/dL (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L), or (iii) Glycated hemoglobin (HbA1c) 5.7–6.4%. Men were defined as not having pre‐diabetes if (i) they had a fasting glucose concentration <5.6 mmol/L, a 2‐h plasma glucose concentration <7.8 mmol/L, and an HbA1c <5.7%, and (ii) they did not answer yes to the following question: ‘Have you ever been told by a doctor or other health professional you had diabetes or sugar diabetes?’.
Testosterone, estradiol, and SHBG concentrations were measured by competitive electrochemiluminescence immunoassays on the 2010 Elecsysautoanalyzer (Roche Diagnostics, Indianapolis, IN, USA), and androstanediol glucuronide concentration was measured by enzyme immunoassay (Diagnostic Systems Laboratories, Webster, TX, USA) previously using serum stored in the NHANES III repository (Rohrmann et al., 2011b). Concentrations of these analytes are stable after multiple freeze–thaw cycles (Wickings & Nieschlag, 1976; Comstock et al., 2001). Samples were randomly ordered for testing, and the laboratory technicians were blinded to participant characteristics. The lowest detection limits of the assays were 0.02 ng/mL testosterone, 5 pg/mL estradiol, 0.33 ng/mL androstanediol glucuronide, and 3 nmol/L SHBG. Coefficients of variation for blinded quality control specimens run interspersed with the analyses of NHANES III specimens, were as follows: testosterone 5.9% and 5.8% at 2.48 and 5.51 ng/mL, androstanediol glucuronide 9.5% and 5.0% at 2.9 and 10.1 ng/mL, estradiol 6.5% and 6.7% at 0.38 and 1.7 pg/mL, and SHBG 5.3% and 5.9% at 5.3 and 16.6 nmol/L (Rohrmann et al., 2011b). In a separate run, we tested quality control samples with a mean estradiol concentration of 144.6 pmol/L, which is in the range of typical adult male estradiol concentration; the CV% was 2.5%. Free testosterone concentration was estimated from serum total testosterone, SHBG, and albumin concentrations while free estradiol was estimated from total estradiol, SHBG, and albumin using mass action equations (Vermeulen et al., 1999; Rinaldi et al., 2002).
Phase I morning sampling weights for NHANES III were used to account for sampling variability and to adjust for differential probability of selection of persons (National Center for Health Statistics 1994). By pre‐diabetes status, we calculated mean age, age‐adjusted means for BMI and percent body fat and the age‐adjusted percentages for race/ethnicity, alcohol consumption, smoking, and physical activity; we used the age distribution of the US population according to the 2000 Census.
Hormone concentrations were not normally distributed, so we transformed them using the natural logarithm. We calculated age‐ (continuous) and race/ethnicity (non‐Hispanic black, non‐Hispanic white, Mexican American, other)‐adjusted geometric mean concentrations of the sex steroid hormones and their 95% confidence intervals (CI) by pre‐diabetes using linear regression (model 1). To evaluate the confounding effect of lifestyle factors on the difference in sex steroid hormone concentrations by prediabetes, we additionally adjusted for smoking (never, former, current), alcohol consumption (<2, 2–3, 4–6 times a week, or daily or more), physical activity (never, 1–2 times weekly, 3–6 times weekly, or daily or more) (model 2). As extent of body fat influences both glucose and hormonal homeostasis, and hormones compete for binding to SHBG, we decided to assess the effect of these major confounders on the association between pre‐diabetes status and the hormones in a separate model (model 2) plus percent body fat (continuous) and/or testosterone (continuous), and/or estradiol (continuous), and/or SHBG (continuous).
As the risk of certain chronic disorders associated with pre‐diabetes vary by ethnicity and extent of adiposity (Platz & Giovannucci, 2006; World Cancer Research Fund International/American Institute for Cancer Research, 2014), we conducted further analyses to assess whether hormone concentrations differ by pre‐diabetes stratified by race/ethnicity and by BMI, which we classified according to the World Health Organization Classification as normal: 18.5–24.9 kg/m2 (10 men with BMI <18.5 kg/m2 were excluded from this part of the analyses), overweight: 25.0–29.9 kg/m2, and obese: ≥30 kg/m2. Geometric mean hormone concentrations between groups were compared using anova. We performed a test for interaction for all the above stratified analyses by introducing an interaction term in the linear regression models and testing its coefficient by the Wald test.
We also evaluated the association between pre‐diabetes and the odds of low levels of total testosterone, free testosterone, and SHBG, and high levels of total estradiol and free estradiol using logistic regression to adjust for demographics, lifestyle factors, adiposity, and mutually for the hormones (models 1–3). To define low or high concentrations, we used clinically defined cutpoints where available, or extreme quantile cutpoints as follows: low total testosterone <3.25 ng/mL (Rohrmann et al., 2011b), low free testosterone <0.07 ng/mL (23), low SHBG <20 nmol/L (<10th percentile), high estradiol ≥50 pg/mL (Ramasamy et al., 2014), and high free estradiol ≥1.37 pg/mL (≥90th percentile). As the number of men with clinically abnormal hormone concentrations was small, we also evaluated the association between pre‐diabetes and lower or higher hormone concentrations defined as being in the bottom tertile [vs. higher (middle and highest tertiles combined)] or in the top tertile [vs. lower (middle and lowest tertiles combined)], respectively.
Finally, as the mechanism of action between the different glycemic categories which were used to define the main pre‐diabetes definition, and sex steroid hormones and SHBG may differ (Yudkin & Montori, 2014), in sensitivity analyses, we reran these models to assess any potential differences in the associations based on IFG only, IGT only, and elevated HbA1c only.
All analyses were conducted with sas release 9.2 (SAS Institute, Cary, NC, USA), and sudaan 9.0 software (Research Triangle Park, Durham, NC, USA) as implemented in SAS 9.2.
In this sample of men 20 + years old in NHANES III, pre‐diabetic men were slightly younger, had higher mean percent body fat and BMI were less likely to be current smokers;, drank alcohol less frequently, and were slightly more likely to engage in moderate or vigorous physical activity than men without pre‐diabetes (Table 1). The proportion of men with pre‐diabetes appeared to differ by race/ethnicity.
Table 1. Age‐adjusted (standardized to the 2000 US Census age distribution) weighted characteristics in men 20 + years old with and without prediabetes, NHANES III
Pre‐diabetesc Yes No
|Mean (SE)||38.3 (0.6)||47.5 (1.2)|
|Percent body fat|
|Mean (SE)||26.0 (0.6)||24.2 (0.3)|
|Mean (SE)||27.5 (0.6)||25.4 (0.2)|
|Cigarette smoking status (%)|
|Current (<20 cigarettes/day)||16.2||25.0|
|Current (20–40 cigarettes/day)||9.9||10 7|
|Current (≥ 40 cigarettes/day)||0.1||0.5|
|Alcohol intake frequency (%)|
|Up to once a week||19.4||16.6|
|2–3 times a week||19.9||16.3|
|4–6 times a week||11.7||20.8|
|Daily or more||17.5||18.0|
|Moderate or vigorous physical activity frequency, %|
|3 to 6 times/week||6.3||7.0|
- a Unweighted.
- b Not age‐adjusted.
- c Men with diabetes were excluded from the analysis. Pre‐diabetes is defined as having (1) Impaired fasting glucose: fasting plasma glucose concentration from 100 mg/dL (5.6 mmol/L) to 125 mg/dL (6.9 mmol/L), or (2) Impaired glucose tolerance: 2‐h plasma glucose concentration in the 75 g oral glucose tolerance test from 140 mg/dL (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L), or (3) HbA1c 5.7–6.4%.
Adjusting for age and race/ethnicity, mean total testosterone concentration was statistically significantly lower in men with pre‐diabetes compared to men without; free testosterone did not differ by pre‐diabetes status (Table 2). Mean difference in total testosterone level was attenuated but remained statistically significant after further adjustment for smoking, alcohol consumption, and physical activity. After additional adjustment for percent body fat, estradiol, and SHBG, total testosterone level did not differ significantly between men with and without pre‐diabetes. In age‐adjusted analyses stratified by race/ethnicity (Table 3), the difference in the total testosterone concentration between men with and without pre‐diabetes was apparent in all three race/ethnic groups, but was most pronounced in non‐Hispanic black men. However, in the multivariable‐adjusted model, total testosterone concentration did not differ between men with and without pre‐diabetes in any of the three race/ethnic groups. Total testosterone did not differ by pre‐diabetes status in any BMI category except possibly for lower total testosterone in men with pre‐diabetes than without among those with normal BMI. After multivariable adjustment, total testosterone concentration did not differ by pre‐diabetes status in any BMI category (Table 4). Patterns of free testosterone differences in men with and without pre‐diabetes were not apparent among the race/ethnic groups and BMI categories.
Table 2. Adjusteda geometric mean serum concentrations of sex steroid hormones and SHBG by pre‐diabetes status, men 20+ years old, NHANES III
Pre‐diabetes p‐value Yes No Geometric mean concentration (95% CI) Geometric mean concentration (95% CI)
|Total testosterone (ng/mL)|
|Model 1||4.68 (4.30–5.10)||5.36 (5.15–5.58)||0.01|
|Model 2||4.68 (4.28–5.11)||5.32 (5.13–5.53)||<0.01|
|Model 3||4.93 (4.59–5.30)||5.19 (5.00–5.38)||0.25|
|Free testosterone (ng/mL)|
|Model 1||0.098 (0.089–0.107)||0.104 (0.101–0.108)||0.21|
|Model 2||0.098 (0.090–0.108)||0.103 (0.100–0.107)||0.08|
|Model 3||0.099 (0.092–0.106)||0.103 (0.099–0.108)||0.37|
|Androstanediol glucuronide (ng/mL)b|
|Model 1||12.03 (10.94–13.23)||11.81 (11.14–12.51)||0.74|
|Model 2||12.02 (10.78–13.40)||11.89 (11.11–12.73)||0.79|
|Model 3||11.90 (10.67–13.28)||11.95 (11.18–12.77)||0.94|
|Total estradiol (pg/mL)|
|Model 1||35.18 (33.08–37.40)||36.13 (34.63–37.70)||0.26|
|Model 2||35.13 (33.22–37.15)||35.98 (34.69–37.31)||0.14|
|Model 3||35.45 (33.95–37.02)||35.81 (34.38–37.31)||0.56|
|Free estradiol (pg/mL)|
|Model 1||0.93 (0.87–0.99)||0.91 (0.86–0.95)||0.39|
|Model 2||0.93 (0.87–0.99)||0.90 (0.86–0.94)||0.20|
|Model 3||0.92 (0.88–0.96)||0.90 (0.86–0.95)||0.33|
|Model 1||31.67 (29.39–34.13)||36.16 (34.27–38.15)||0.01|
|Model 2||31.40 (29.00–33.99)||36.20 (34.37–38.13)||0.02|
|Model 3||32.78 (30.63–35.08)||35.42 (33.71–37.21)||0.08|
- a Model 1 – Adjusted for age and race/ethnicity, Model 2 – Also adjusted for smoking, physical activity, alcohol consumption, Model 3 – Also adjusted for percent body fat, and other hormones (testosterone – adjusted for estradiol and SHBG and vice versa, free testosterone – adjusted for free estradiol and vice versa).
- b Not adjusted for other hormones.
Table 3. Adjusteda geometric mean (95% CI) serum sex steroid hormone concentrations in men 20+ years old with and without pre‐diabetes stratified by race/ethnicity, NHANES III
Race/ethnicity Pre‐diabetes p‐value p for interaction Pre‐diabetes p‐value p for interaction Yes No Yes No Age‐adjusted geometric mean concentration (95% CI) Multivariable‐adjusteda geometric mean concentration (95% CI)a
|Total testosterone (ng/mL)||Non‐Hispanic white||4.95 (4.60–5.32)||5.31 (5.04–5.59)||0.10||0.04||5.16 (4.77–5.57)||5.17 (4.98–5.37)||0.97||0.07|
|Non‐Hispanic black||4.70 (4.16–5.31)||5.67 (5.17–6.23)||0.02||5.23 (4.75–5.76)||5.53 (5.17–5.92)||0.26|
|Mexican American||5.05 (4.60–5.54)||5.56 (5.30–5.83)||0.16||5.31 (5.03–5.61)||5.40 (5.22–5.60)||0.66|
|Free testosterone (ng/mL)||Non‐Hispanic white||0.102 (0.093–0.111)||0.102 (0.098–0.107)||0.94||0.047||0.103 (0.096–0.111)||0.101 (0.096–0.106)||0.70||0.07|
|Non‐Hispanic black||0.102 (0.091–0.115)||0.112 (0.102–0.122)||0.16||0.100 (0.089–0.113)||0.113 (0.104–0.122)||0.06|
|Mexican American||0.114 (0.108–0.122)||0.114 (0.111–0.117)||0.93||0.111 (0.105–0.117)||0.114 (0.111–0.118)||0.38|
|Androstanediol glucuronide (ng/mL)||Non‐Hispanic white||12.47 (11.13–13.98)||12.52 (11.69–13.40)||0.95||0.29||12.50 (11.11–14.07)||12.65 (11.82–13.53)||0.86||0.19|
|Non‐Hispanic black||10.60 (8.60–13.07)||10.62 (9.77–11.55)||0.99||10.53 (8.57–12.95)||10.71 (9.97–11.50)||0.89|
|Mexican American||12.14 (10.72–13.76)||11.34 (10.06–12.79)||0.52||12.28 (10.56–14.29)||11.14 (10.10–12.28)||0.38|
|Total estradiol (pg/mL)||Non‐Hispanic white||35.24 (33.06–37.57)||35.48 (33.69–37.36)||0.78||0.03||35.10 (33.33–36.95)||35.40 (33.81–37.07)||0.66||0.29|
|Non‐Hispanic black||40.89 (37.97–44.04)||41.21 (39.47–43.04)||0.85||41.65 (39.09–44.38)||40.87 (39.33–42.47)||0.57|
|Mexican American||34.49 (31.78–37.44)||34.00 (31.77–36.38)||0.77||34.73 (32.34–37.30)||33.82 (31.98–35.77)||0.46|
|Free estradiol (pg/mL)||Non‐Hispanic white||0.92 (0.85–0.99)||0.88 (0.84–0.94)||0.15||0.04||0.90 (0.86–0.96)||0.89 (0.85–0.93)||0.46||0.41|
|Non‐Hispanic black||1.11 (1.03–1.20)||1.05 (0.99–1.11)||0.25||1.12 (1.04–1.20)||1.04 (0.99–1.09)||0.10|
|Mexican American||0.94 (0.84–1.05)||0.87 (0.82–0.94)||0.24||0.94 (0.85–1.03)||0.87 (0.83–0.91)||0.11|
|SHBG (nmol/L)||Non‐Hispanic white||32.92 (30.36–35.70)||36.41 (34.28–38.68)||0.06||0.20||33.67 (31.12–36.43)||35.72 (33.94–37.59)||0.23||0.71|
|Non‐Hispanic black||29.29 (26.27–32.66)||37.38 (35.68–39.17)||<0.01||32.01 (29.72–34.48)||36.61 (35.39–37.86)||<0.01|
|Mexican American||26.84 (23.45–30.71)||33.32 (30.29–36.65)||0.06||28.98 (27.26–30.81)||32.58 (30.08–35.28)||0.10|
- a Adjusted for age, smoking, physical activity, alcohol consumption, percent body fat, and as follows: testosterone was also adjusted for estradiol and SHGB and vice versa, and free testosterone was adjusted for free estradiol and vice versa.
Table 4. Adjusteda geometric mean (95% CI) of serum sex steroid hormone concentrations in men 20 + years old with and without pre‐diabetes stratified by BMI, NHANES III
BMI (kg/m2) Pre‐diabetes p‐value p for interaction Pre‐diabetes p‐value p for interaction Yes No Yes No Age‐ and race/ethnicity‐adjusted geometric mean concentration (95% CI) Multivariable‐adjusted geometric mean concentration (95% CI)a
|Total testosterone (ng/mL)||18.5–24.99||4.83 (3.90–5.99)||6.00 (5.68–6.33)||0.08||0.14||5.21 (4.30–6.32)||5.77 (5.46–6.11)||0.36||0.32|
|25–29.99||4.86 (4.46–5.30)||4.93 (4.65–5.23)||0.78||4.94 (4.64–5.26)||4.90 (4.68–5.14)||0.84|
|30+||4.19 (3.80–4.63)||4.17 (3.94–4.41)||0.92||4.15 (3.95–4.37)||4.23 (4.02–4.45)||0.69|
|Free testosterone (ng/mL)||18.5–24.99||0.093 (0.074–0.118)||0.111 (0.106–0.116)||0.17||0.22||0.098 (0.081–0.118)||0.108 (0.101–0.115)||0.38||0.83|
|25–29.99||0.104 (0.095–0.114)||0.100 (0.094–0.107)||0.49||0.103 (0.096–0.110)||0.101 (0.095–0.107)||0.70|
|30+||0.094 (0.084–0.106)||0.092 (0.086–0.098)||0.78||0.092 (0.087–0.098)||0.094 (0.089–0.100)||0.70|
|Androstanediol glucuronide (ng/mL)||18.5–24.99||10.97 (8.84–13.62)||11.84 (11.00–12.74)||0.51||0.75||11.04 (8.79–13.86)||11.93 (11.10–12.83)||0.53||0.74|
|25–29.99||13.32 (11.86–14.96)||11.92 (10.88–13.05)||0.22||13.17 (11.76–14.74)||12.00 (10.93–13.16)||0.30|
|30+||11.44 (9.71–13.48)||11.89 (10.22–13.83)||0.76||11.70 (10.10–13.56)||11.66 (9.96–13.65)||0.98|
|Total estradiol (pg/mL)||18.5–24.99||33.85 (31.40–36.48)||36.88 (35.13–38.73)||0.02||0.09||34.47 (32.84–36.18)||36.29 (34.36–38.33)||0.13||0.58|
|25–29.99||36.15 (33.19–39.38)||35.04 (33.08–37.12)||0.46||36.08 (33.66–38.67)||35.03 (33.51–36.62)||0.25|
|30+||26.66 (24.74–28.73)||27.62 (25.31–30.14)||0.40||35.18 (33.15–37.34)||36.60 (34.80–38.49)||0.26|
|Free estradiol (pg/mL)||18.5–24.99||0.85 (0.78–0.92)||0.89 (0.84–0.95)||0.23||0.31||0.87 (0.82–0.93)||0.88 (0.82–0.94)||0.87||0.72|
|25–29.99||0.96 (0.88–1.06)||0.90 (0.86–0.95)||0.11||0.95 (0.89–1.02)||0.91 (0.87–0.95)||0.08|
|30+||0.97 (0.91–1.04)||1.00 (0.94–1.07)||0.48||0.97 (0.91–1.03)||1.01 (0.95–1.07)||0.32|
|SHBG (nmol/mL)||18.5–24.99||36.75 (32.51–41.55)||40.68 (37.83–43.75)||0.22||0.19||36.97 (33.19–41.18)||40.59 (38.33–42.99)||0.18||0.67|
|25–29.99||30.97 (28.74–33.38)||33.02 (30.62–35.62)||0.25||30.72 (28.53–33.08)||33.16 (31.05–35.42)||0.12|
|30+||26.66 (24.74–28.73)||27.62 (25.31–30.14)||0.54||27.24 (25.34–29.28)||27.00 (25.28–28.83)||0.88|
- a Adjusted for age, race/ethnicity, smoking, physical activity, alcohol consumption, percent body fat, and as follows: testosterone was adjusted for estradiol and SHBG and vice versa, and free testosterone was adjusted for free estradiol and vice versa.
When considering a clinical cutpoint, pre‐diabetes was not associated with low (<3.25 ng/mL) total testosterone irrespective of adjustment. However, when considering lower total testosterone defined as the bottom tertile, pre‐diabetes was positively associated with lower total testosterone after adjustment for age and race/ethnicity (model 1) and additionally for lifestyle factors (model 2) (Table 5). After additional adjustment for percent body fat, estradiol, and SHBG (model 3), the association was attenuated and was no longer statistically significant. Pre‐diabetes was not statistically significantly associated with low free testosterone level based on either the clinical cutpoint or the bottom tertile (Table 5).
Table 5. Adjusteda odds ratios (OR) and 95% confidence intervals (CI) of low levels of testosterone, free testosterone, and SHBG, and high levels of estradiol and high free estradiol by pre‐diabetes status in men 20+ years old, NHANES III
OR (95% CI) of low serum concentration associated with pre‐diabetes Model 1 Model 2 Model 3
|Clinically low <3.25 vs. ≥3.25||1.30 (0.74–2.28)||1.07 (0.68–1.77)||0.90 (0.51–1.59)|
|Lowest tertile vs. higherb||2.69 (1.65–4.39)||2.58 (1.54–4.29)||1.39 (0.70–2.76)|
|Free testosterone (ng/mL)|
|Clinically low <0.07 vs. ≥0.07||0.90 (0.48–1.66)||0.79 (0.45–1.40)||0.81 (0.45–1.47)|
|Lowest tertile vs. higherb||1.15 (0.60–2.20)||1.05 (0.57–1.93)||0.89 (0.45–1.74)|
|Total estradiol (pg/mL)|
|Clinically high ≥50.0 vs. < 50.0||0.75 (0.51–1.11)||0.88 (0.59–1.32)||1.04 (0.71–1.53)|
|Highest tertile vs. lowerb||0.79 (0.58–1.08)||0.93 (0.68–1.26)||0.95 (0.62–1.45)|
|Free estradiol (pg/mL)|
|Above 90th percentile ≥1.37 vs. <1.37||0.96 (0.59–1.56)||1.18 (0.68–2.07)||1.14 (0.64–2.04)|
|Highest tertile vs. lowerb||1.30 (0.96–1.74)||1.59 (1.14–2.22)||1.38 (0.83–2.29)|
|Below 90th percentile <20 vs. ≥20||1.47 (0.79–2.76)||1.19 (0.60–2.40)||0.91 (0.49–1.73)|
|Lowest tertile vs. higherb||2.30 (1.33–3.96)||2.27 (1.32–3.92)||1.49 (0.70–3.16)|
- a Model 1 – Adjusted for age and race/ethnicity, Model 2 – Also adjusted for smoking, physical activity, alcohol consumption, Model 3 – Also adjusted for percent body fat, and other hormones (testosterone – adjusted for estradiol and SHBG and vice versa, free testosterone – adjusted for free estradiol and vice versa.
- b Cutoff points for tertile 1 (T1) – Testosterone (ng/mL): <4.43, free testosterone (ng/mL): <0.09, and SHBG (nmol/L): <29.46. Cutoff points for tertile 3 (T3) – Estradiol (pg/mL): >40.45, and free estradiol (pg/mL): >1.03.
Androstanediol glucuronide concentration did not differ between men with and without pre‐diabetes in any of the analyses performed (Tables 2-4).
Mean concentration of total or free estradiol did not differ between men with and without pre‐diabetes overall (Table 2) or when stratified by race/ethnicity (Table 3). In normal weight men (BMI 18.5 to <24.9 kg/m2), total estradiol concentration was statistically significantly lower in men with pre‐diabetes than men without after age and race/ethnicity adjustment (Table 4); this difference was attenuated in the multivariable‐adjusted model (model 4).
When considering the clinical cutpoint for high total estradiol (>50 pg/mL) and top 10th percentile for free estradiol, pre‐diabetes was not associated with high levels (Table 5). When defining higher levels based on the top tertile (vs. lower), pre‐diabetes was not associated with higher total estradiol irrespective of adjustments. However, after adjustment for age, race/ethnicity, and lifestyle factors, pre‐diabetes (model 2) was statistically significantly associated with higher odds of free estradiol in the top tertile (vs. lower). The associations were attenuated after further adjustment for adiposity (model 3), and additionally for free testosterone (model 4) but ORs remained in the positive direction.
Sex hormone‐binding globulin
SHBG concentration was statistically significantly lower in men with than without pre‐diabetes, a difference that was only slightly attenuated after adjustment for percent body fat and after further adjustment for total testosterone and total estradiol (Table 2). After age adjustment, SHBG concentration was lower among men with than without pre‐diabetes in all three race/ethnic groups, but especially among non‐Hispanic black men. After multivariable adjustment, the difference remained statistically significant only in non‐Hispanic black men. The difference in SHBG concentration by pre‐diabetes did not differ by BMI category after age and race/ethnicity or after multivariable adjustment (Table 4).
Pre‐diabetes was not associated with low SHBG, defined as below the 10th percentile (<20 nmol/L), irrespective of adjustment. However, pre‐diabetic men had higher odds of lower SHBG concentration defined as the bottom tertile vs. higher after adjusting for age and race/ethnicity (model 1) (Table 5). This association remained after further adjustment for the lifestyle factors (model 2), but was attenuated after additionally adjusting for percent body fat, total testosterone, and total estradiol (models 3 and 4), although ORs remained in the positive direction.
In sensitivity analyses (Table S1), the association of pre‐diabetes as defined by IFG with lower (bottom tertile vs. higher) or higher (highest tertile vs. lower) levels of sex steroid hormones were similar to those for the main pre‐diabetes definition. Using IGT to define pre‐diabetes, the associations for lower total testosterone and lower SHBG were generally similar to those for the main pre‐diabetes definition. Using HbA1c in the ‘at risk for diabetes’ range to define pre‐diabetes, some associations were similar, but others differed somewhat from those for the main definition. Unlike the main definition, the association with lower free testosterone was in the positive direction in each model. In line with the main definition, the associations were generally inverse for higher total estradiol. Unlike for the main definition, associations with higher free estradiol were inverse in each model.
In this US nationally representative group of men 20+ years old, we observed that men with pre‐diabetes had lower mean serum concentrations of total testosterone and SHBG than men without pre‐diabetes. Pre‐diabetic men also had higher odds of having lower serum concentrations of total testosterone and SHBG, and higher serum concentrations of free estradiol. Free testosterone, AAG, and total estradiol levels did not vary between men with and without pre‐diabetes.
Consistent with studies comparing persons with and without glucose perturbations (diabetes, IFG, IGT) (Selvin et al., 2007; Colangelo et al., 2009; Grossmann, 2011; Shin et al., 2012; Kim & Halter, 2014), we found weak evidence suggesting that pre‐diabetic men had lower total testosterone levels than men without pre‐diabetes. This observed inverse, but statistically non‐significant, association was also seen when defining pre‐diabetes based on IFG, IGT, and HbA1c. Similar to another study, we did not find any evidence to suggest that race/ethnicity modifies the association between pre‐diabetes and total testosterone (Colangelo et al., 2009). There was also a lack of association between pre‐diabetes and testosterone by strata of BMI. Nevertheless, the attenuating effect that percent body fat, in addition to competing hormones, had on the association between pre‐diabetes and the hormones implies that the associations are mediated by these variables, and hence, the link between pre‐diabetes and testosterone may be indirect (Vermeulen et al., 1993; Pitteloud et al., 2005; Yeap et al., 2009; Grossmann, 2011). While we found some evidence indicating a link between pre‐diabetes and testosterone, there was no association between pre‐diabetes and free testosterone.
These findings may be partly elucidated by the inhibition of SHBG production, resulting from hyperinsulinemia in men with pre‐diabetes, and subsequent reduction in testosterone concentration (Pasquali, 2006). Moreover, other studies reported that hyperinsulinemia may cause androgen deficiency by impairing testosterone‐producing Leydig cells or through the induction of genes which regulate testicular steroidogenesis (Pitteloud et al., 2005; Giovannucci et al., 2010).
Regarding AAG (Rohrmann et al., 2011a), we did not observe any differences in their concentration by pre‐diabetes status. To the best of our knowledge, no previous studies have reported differences in the levels of AAG among persons with and without pre‐diabetes.
As reported in two other studies, we did not see any variations in estradiol levels between men with and without pre‐diabetes (Haffner et al., 1994; Shin et al., 2012). Conversely, in the Multi‐Ethnic Study of Atherosclerosis, a positive association between pre‐diabetes and estradiol was reported (Colangelo et al., 2009) while two other studies reported an inverse relation between estradiol and fasting glucose (Shono et al., 1996; Dhindsa et al., 2011). When free estradiol was taken into account, we did observe a positive but non‐significant association after additional adjustments for adiposity and other hormones. Although we did not find any differences in the levels of total or free estradiol concentration between men with and without pre‐diabetes by BMI category, the attenuating effect of adiposity on the associations suggest that any existing associations between pre‐diabetes and estradiol is modulated by adiposity.
Similar to other studies, we observed that pre‐diabetic men had lower SHBG levels than men without pre‐diabetes (Haffner et al., 1994; Colangelo et al., 2009; Hong et al., 2013). Our stratified analyses suggested that race/ethnicity, is not an effect modifier for the association between pre‐diabetes and SHBG. However, our analyses provide evidence corroborating the view that body fat levels may mediate the association between pre‐diabetes and SHBG. In our study, while SHBG levels appeared to be lower in pre‐diabetic men who were normal or overweight than their counterparts, among those who are obese, serum SHBG levels did not differ. This observed lower SHBG concentration may be partly explained by the inhibitory effect of hyperinsulinemia on SHBG production (Pasquali, 2006). Moreover, another study showed that glucose and fructose inhibited human SHBG production in the liver cells by inhibiting production of hepatic HNF‐4 levels, while triggering an increase in cellular palmitate levels (Selva et al., 2007).
Strengths and limitations
This is one of the largest studies assessing the concentrations of pre‐diabetes in relation to sex steroid hormones. The sampling design of NHANES makes our results generalizable to the US population of civilian non‐institutionalized men aged 20 or older in the US in 1990. We were also able to adjust for a wide range of potential confounding factors including lifestyle factors and adiposity and to assess effect modification by race/ethnicity and BMI.
We conducted a cross‐sectional evaluation of the link between pre‐diabetes and testosterone, and it is therefore, not possible to conclude about temporality of the association. We did not have information on the duration of pre‐diabetes in this cross‐sectional study, and thus could not address the influence of a prolonged duration of pre‐diabetes on hormone concentrations. Studies on sex steroid hormones are limited by high inter‐ and intra‐subject variability in the production, circulating levels, and metabolic clearance rates of steroid hormones resulting from changes in the diurnal rhythm (Araujo & Wittert, 2011). Here, we aimed to minimize the influence of variation resulting from diurnal production of the hormones by selecting participants with morning samples. Furthermore, our study is based on a single measurement of hormones and SHBG and lacked information on symptoms of hypogonadism, and/or gonadotropin levels. Hormonal and SHBG measurements were also conducted using immunoassay, which is less reliable and less specific than more recent methods such as the liquid chromatography tandem mass spectrometry that is used to determine estradiol levels (Dhindsa et al., 2011). It is therefore possible that the outcome in our study was not accurately classified. Information on some covariates such as physical activity and alcohol consumption were self‐reported and this may have introduced reporting bias and measurement error.
Although our findings may be partly explained by the effect of body fat mass and competing hormones, the results suggest a weak inverse association between pre‐diabetes and total testosterone, SHBG concentrations, and a positive association with free estradiol in males. As both pre‐diabetes and these hormonal factors may influence risk of chronic diseases such as cardiovascular diseases, such knowledge can help us further understand how these variables interplay to influence risk of these diseases. However, further studies are needed to substantiate the link between pre‐diabetes and hormonal/ SHBG levels.
Funding: This is the 32nd paper from the Hormone Demonstration Program funded by the Maryland Cigarette Restitution Fund at Johns Hopkins (Nelson). This work was also supported by NCI P30 CA006973 (Nelson). The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the Maryland Department of Health and Mental Hygiene or the National Institutes of Health.
Rhonda Arthur was responsible for analyses and interpretation of the data, drafting the manuscript, and final approval of the version to be published. Mieke Van Hemelrijck, Sabine Rohrmann, and Henrik Moller contributed to the analyses and interpretation of the data, revising the draft, and giving final approval. Elizabeth Selvin, Adrian S. Dobs, Norma F. Kanarek, William G. Nelson, and Elizabeth A. Platz contributed to the conception and design, acquisition of data and/or analysis and interpretation of data, revising the manuscript, and giving final approval.
Conflict of Interest