It remains unclear whether endogenous sex hormones (ESH) are associated with risk of type 2 diabetes (T2D) in women. Data of 3,117 postmenopausal women participants of the Rotterdam Study were analyzed to examine whether ESH and sex hormone–binding globulin (SHBG) were associated with the risk of incident T2D. Additionally, we performed a systematic review and meta-analysis of studies assessing the prospective association of ESH and SHBG with T2D in women. During a median follow-up of 11.1 years, we identified 384 incident cases of T2D in the Rotterdam Study. No association was observed between total testosterone (TT) or bioavailable testosterone (BT) with T2D. SHBG was inversely associated with the risk of T2D, whereas total estradiol (TE) was associated with increased risk of T2D. Similarly, in the meta-analysis of 13 population-based prospective studies involving more than 1,912 incident T2D cases, low levels of SHBG and high levels of TE were associated with increased risk of T2D, whereas no associations were found for other hormones. The association of SHBG with T2D did not change by menopause status, whereas the associations of ESH and T2D were based only in postmenopausal women. SHBG and TE are independent risk factors for the development of T2D in women.

Menopause is an important transition in a woman’s life, not only for marking the end of reproductive life but also for being accompanied by an increased risk of cardiovascular disease and type 2 diabetes (T2D) (1,2). Changes in hormonal patterns in menopause, including the decline in endogenous estradiol (E) levels and the relative androgen excess, contribute to an increase in visceral adiposity that is associated with glycemic traits and therefore may influence the risk of T2D (3,4). Furthermore, polycystic ovary syndrome, a common disorder among women characterized by hyperandrogenism, has been identified as a significant nonmodifiable risk factor associated with T2D (5).

Although the relation between sex hormone–binding globulin (SHBG) and T2D has long been recognized (6,7), literature on the associations of steroid sex hormones, such as endogenous E and testosterone (T), with T2D is scarce. SHBG, T, and E have been associated with glucose metabolism and development of insulin resistance (69). Few epidemiological studies investigating the relation between sex hormones and T2D have yielded conflicting results (1012). These studies were limited by their cross-sectional design, selected samples, or insufficient adjustment for diabetes risk factors. To date, no large prospective cohort study has examined the association of T2D with SHBG, T, and E in healthy postmenopausal women. Thus, we aimed to investigate the association between SHBG, sex hormones, and T2D in postmenopausal women. Furthermore, to clarify the contradictory results, we systematically reviewed and meta-analyzed studies evaluating the association between SHBG, sex hormones, and T2D in women.

The Rotterdam Study

The Rotterdam Study is a prospective cohort study which started since 1990 in the Ommoord district, in the city of Rotterdam, the Netherlands. Details regarding the design, objectives, and methods of the Rotterdam Study have been described in detail elsewhere (13). In brief, in 1990, all inhabitants of a well-defined district of Rotterdam were invited, of whom 7,983 agreed (78.1%). In 2000, an additional 3,011 participants were enrolled (RS-II), consisting of all people living in the study district who had become 55 years of age. Follow-up examinations were performed periodically, approximately every 3–5 years (13). There were no eligibility criteria to enter the Rotterdam Study cohorts except the minimum age and residential area based on ZIP codes. The Rotterdam Study has been approved by the medical ethics committee according to the Population Screening Act: Rotterdam Study, executed by the Ministry of Health, Welfare and Sport of the Netherlands. All participants in the present analysis provided written informed consent to participate and to obtain information from their treating physicians.

Ascertainment of T2D

The participants were followed from the date of baseline center visit onwards. At baseline and during follow-up, cases of T2D were ascertained through active follow-up using general practitioners’ records, glucose hospital discharge letters, and glucose measurements from Rotterdam Study visits, which take place approximately every 4 years (14). T2D was defined according to recent World Health Organization guidelines, as a fasting blood glucose ≥7.0 mmol/L, a nonfasting blood glucose ≥11.1 mmol/L (when fasting samples were absent), or the use of blood glucose–lowering medication (15). Information regarding the use of blood glucose–lowering medication was derived from both structured home interviews and linkage to pharmacy records (14). At baseline, >95% of the Rotterdam Study population was covered by the pharmacies in the study area. All potential events of T2D were independently adjudicated by two study physicians. In case of disagreement, consensus was sought with an endocrinologist. Follow-up data were complete until 1 January 2012.

Sex Steroid Measurements

All blood samples were drawn in the morning (≤11:00 a.m.) and were fasting. Total estradiol (TE) levels were measured with a radioimmunoassay and SHBG with the Immulite platform (Diagnostic Products Corporation, Breda, the Netherlands). The minimum detection limit for E was 18.35 pmol/L. Undetectable E was scored as 18.35. Serum levels of total testosterone (TT) were measured with liquid chromatography–tandem mass spectrometry. The corresponding interassay coefficients of variations for TE, SHBG, and TT are <7%, <5%, and <5%. The free androgen index (FAI), calculated as (T/SHBG)*100, is used as a surrogate measure of bioavailable testosterone (BT) (16).

Population of Analysis

The current study used data from the third visit of the first cohort (RSI-3) and the baseline examinations of the second (RSII-1) cohort. Overall, there were 3,683 postmenopausal women eligible for blood measurements. Among them, 122 women did not come for a blood measurement at the research center and 32 did not have T2D follow-up data and were excluded from the analysis. Furthermore, 412 women with prevalent T2D were excluded, leaving 3,117 for our final analysis. Potential confounding variables are described in detail in Supplementary Appendix 1.

Statistical Analysis

Person-years of follow-up were calculated from study entrance (March 1997 to December 1999 for RSI-3 and February 2000 to December 2001 for RSII-1) to the date of diagnosis of T2D, death, or the censor date (date of last contact of the living), whichever occurred first. Follow-up was until 1 January 2012. Cox proportional hazards modeling was used to evaluate whether SHBG, TT, TE, and BT were associated with T2D. Relative risks (RRs) and 95% CIs were reported. All sex hormone variables were assessed in separate models, continuously and in tertiles. For E, first tertile included all women with levels of E lower than the detection limit (n = 992). To study the relations across increasing tertiles, trend tests were computed by entering the categorical variables as continuous variables in multivariable Cox proportional hazards models. To achieve approximately normal distribution, skewed variables (SHBG, TT, BT, plasma triglyceride, LDL cholesterol [LDL-C], C-reactive protein [CRP], thyroid-stimulating hormone [TSH], and insulin) were natural log transformed. In the base model (model 1), we adjusted for age, cohort (1,2), and fasting status (fasting sample vs. nonfasting sample). To examine whether the relations of sex hormones and SHBG with risk of T2D were independent of established risk factors for T2D, model 2 included the terms of model 1, BMI (continuous), glucose (continuous), and insulin (continuous). BMI and waist circumference were highly correlated (Pearson correlation coefficient = 0.81, P < 0.001), so only BMI was used as a measure of adiposity, consistent with previous studies (10,12). Model 3 included all covariates in model 2 and further potential intermediate factors, including metabolic risk factors (total cholesterol, systolic blood pressure [continuous], indication for hypertension [yes vs. no], and use of lipid-lowering medications [yes vs. no]), lifestyle factors (alcohol intake [continuous] and smoking status [current vs. former/never]), prevalent coronary heart disease (yes vs. no), age of menopause, hormone replacement therapy (yes vs. no), CRP (continuous), and sex hormones for each other. Effect modifications of sex hormones by BMI and years since menopause were tested in the final multivariable model in addition to performing stratified analysis. We also performed a series of sensitivity analyses. Since waist circumference is a better measure of visceral adiposity, an important risk factor for diabetes and of sex hormone levels after menopause, we performed the analysis substituting it with BMI. To account for the specific effects of lipid particles on diabetes, we substituted total cholesterol with HDL cholesterol, triglycerides, and LDL-C. TSH, physical activity, number of pregnancies, and type of menopause (nonnatural vs. natural) are associated with sex hormone levels and/or risk of T2D; therefore, the models were further adjusted for these factors. To explore potential reverse causation, we reran the analysis by excluding the first 3 years of follow-up. Multiple imputation procedure was used (n = 5 imputations) to adjust for potential bias associated with missing data. Rubin method was used for the pooled regression coefficients (β) and 95% CI (17). A P value of <0.05 was considered statistically significant. All analyses were done using SPSS statistical software (SPSS, version 21.0; SPSS, Inc., Chicago, IL).

Systematic Review and Meta-analysis

Data Sources and Search Strategy.

The review was conducted using a predefined protocol and in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) (18) and Meta-analysis Of Observational Studies in Epidemiology (MOOSE) (19) guidelines (Supplementary Appendices 2 and 3). Medline, Embase.com, Web of Science, the Cochrane Library, PubMed, and Google Scholar were searched from inception until 2 November 2015 (date last searched) with the assistance of an experienced biomedical information specialist. The computer-based searches combined terms related to the exposure (e.g., sex hormone binding globulin, T, and E) with outcomes (e.g., T2D), without any language restriction. Details on the search strategy are provided in Supplementary Appendix 4.

Study Selection and Eligibility Criteria.

Studies were included if they 1) were observational cohort, case-cohort, or prospective nested case-control studies; 2) had reported on at least one of the sex hormones as exposures (SHBG, TT, BT, TE, and bioavailable estradiol [BE]); and 3) had assessed associations with risk of T2D in women (pre- and postmenopausal). Two independent reviewers screened the titles and abstracts of all initially identified studies according to the selection criteria. Full texts were retrieved from studies that satisfied all selection criteria. Data extraction, quality assessment, and data synthesis and analysis are described in detail in Supplementary Appendix 5.

Table 1 summarizes the baseline characteristics of the participants included in the analysis. Of the 3,117 postmenopausal women without diabetes at baseline, 384 women developed diabetes over a median follow-up of 11.1 years.

Table 1

Selected characteristics of study participants, the Rotterdam Study

Women (n = 3,117)% missing values
Age (years) 69.7 ± 8.7 
Years since menopause (years) 20.9 ± 10.0 4.4 
Age of menopause (years) 48.9 ± 5.2 4.4 
Number of pregnancies of at least 6 months 2.3 ± 2 12.4 
Natural menopause, n (%) 2,433 (78.1) 
Current smokers, n (%) 218 (9.2) 1.8 
Alcohol intake (g/day) 1.3 (10)a 26.5 
BMI (kg/m227.0 ± 4.3 2.3 
Waist circumference (cm) 89.4 ± 11.6 5.8 
Prevalent coronary heart disease, n (%) 86 (2.8) 0.06 
E (pmol/L) 34.2 (41.62)a 
TT (nmol/L) 0.8 (0.56)a 
SHBG (nmol/L) 69.6 ± 33.0 
FAI 1.3 (1.1)a 
TSH (mU/L) 1.95 (1.7)a 0.03 
Hormone replacement therapy, n (%) 159 (5.3) 4.8 
Insulin (pmol/L) 67 (47)a 0.26 
Glucose (mmol/L) 5.5 ± 0.6 1.3 
CRP (mg/mL) 1.7 (2.93)a 3.7 
Total cholesterol (mmol/L) 6.0 ± 1.0 1.3 
LDL-C (mmol/L) 4.2 (1.22)a 2.5 
HDL cholesterol (mmol/L) 1.5 ± 0.4 2.3 
Statin use, n (%) 681 (14) 4.8 
Triglycerides (mmol/L) 1.27 (0.74)a 0.26 
Systolic blood pressure (mmHg) 142.0 ± 21.1 1.03 
Indication for hypertension, n (%) 794 (25.5) 1.03 
Incident T2D, n (%) 384 (12.3) 
Women (n = 3,117)% missing values
Age (years) 69.7 ± 8.7 
Years since menopause (years) 20.9 ± 10.0 4.4 
Age of menopause (years) 48.9 ± 5.2 4.4 
Number of pregnancies of at least 6 months 2.3 ± 2 12.4 
Natural menopause, n (%) 2,433 (78.1) 
Current smokers, n (%) 218 (9.2) 1.8 
Alcohol intake (g/day) 1.3 (10)a 26.5 
BMI (kg/m227.0 ± 4.3 2.3 
Waist circumference (cm) 89.4 ± 11.6 5.8 
Prevalent coronary heart disease, n (%) 86 (2.8) 0.06 
E (pmol/L) 34.2 (41.62)a 
TT (nmol/L) 0.8 (0.56)a 
SHBG (nmol/L) 69.6 ± 33.0 
FAI 1.3 (1.1)a 
TSH (mU/L) 1.95 (1.7)a 0.03 
Hormone replacement therapy, n (%) 159 (5.3) 4.8 
Insulin (pmol/L) 67 (47)a 0.26 
Glucose (mmol/L) 5.5 ± 0.6 1.3 
CRP (mg/mL) 1.7 (2.93)a 3.7 
Total cholesterol (mmol/L) 6.0 ± 1.0 1.3 
LDL-C (mmol/L) 4.2 (1.22)a 2.5 
HDL cholesterol (mmol/L) 1.5 ± 0.4 2.3 
Statin use, n (%) 681 (14) 4.8 
Triglycerides (mmol/L) 1.27 (0.74)a 0.26 
Systolic blood pressure (mmHg) 142.0 ± 21.1 1.03 
Indication for hypertension, n (%) 794 (25.5) 1.03 
Incident T2D, n (%) 384 (12.3) 

Plus/minus values are mean ± SD.

aMedian (interquartile range).

Sex Hormones and the Risk of Developing T2D

In models adjusted for age, cohort effect, and fasting status, lower SHBG levels (third vs. first tertile: RR 0.33 [95% CI 0.25–0.43], P trend <0.001) and higher levels of BT (third vs. first tertile: RR 2.01 [95% CI 1.55–2.60], P trend <0.001) and TE (third vs. first tertile: RR 2.02 [95% CI 1.50–2.70], P trend <0.001) were associated with an increased risk of T2D (Table 2). Further adjustments for BMI, insulin, and glucose attenuated but did not abolish the association between SHBG (third vs. first tertile: RR 0.56 [95% CI 0.41–0.77], P trend <0.001) or TE and incident T2D (third vs. first tertile: RR 1.39 [95% CI 1.004–1.93], P trend = 0.07). On the other hand, adjustment for obesity and glycemic traits weakened the associations of BT with T2D such that they were no longer statistically significant (Table 2). Controlling for metabolic risk factors, lifestyle factors, inflammatory markers, and prevalent coronary heart disease did not materially affect these associations (Table 2). No association was found between TT and incident T2D in any of the models (Table 2).

Table 2

Associations of SHBG, TT, FAI, and TE with the risk of T2D in postmenopausal women, the Rotterdam Study (n = 3,117)

SHBG
 Tertile 1 Tertile 2 Tertile 3 Continuous P trend 
Case subjects 191 119 74   
Model 1, HR (95% CI) 1.00 0.56 (0.45–0.71) 0.33 (0.25–0.43) 0.37 (0.30–0.46) <0.001 
Model 2, HR (95% CI) 1.00 0.82 (0.64–1.04) 0.56 (0.41–0.77) 0.63 (0.49–0.81) <0.001 
Model 3, HR (95% CI)
 
1.00
 
0.82 (0.64–1.05)
 
0.56 (0.40–0.79)
 
0.66 (0.51–0.86)
 
0.001
 
 TT
 
  

 
Tertile 1
 
Tertile 2
 
Tertile 3
 
Continuous
 
P trend
 
Case subjects 126 139 119   
Model 1, HR (95% CI) 1.00 1.04 (0.82–1.32) 0.90 (0.69–1.16) 0.91 (0.75–1.10) 0.40 
Model 2, HR (95% CI) 1.00 0.94 (0.74–1.20) 0.82 (0.63–1.07) 0.87 (0.71–1.07) 0.15 
Model 3, HR (95% CI)
 
1.00
 
0.96 (0.75–1.24)
 
0.88 (0.67–1.16)
 
0.93 (0.76–1.14)
 
0.36
 
 FAI
 
  

 
Tertile 1
 
Tertile 2
 
Tertile 3
 
Continuous
 
P trend
 
Case subjects 87 124 173   
Model 1, HR (95% CI) 1.00 1.39 (1.05–1.82) 2.01 (1.55–2.60) 1.54 (1.32–1.79) <0.001 
Model 2, HR (95% CI) 1.00 1.06 (0.79–1.42) 1.17 (0.87–1.57) 1.13 (0.94–1.36) 0.28 
Model 3, HR (95% CI)
 
1.00
 
1.05 (0.78–1.42)
 
1.15 (0.85–1.54)
 
1.10 (0.92–1.32)
 
0.34
 
 TE
 
  

 
Tertile 1
 
Tertile 2
 
Tertile 3
 
Continuous
 
P trend
 
Case subjects 109 132 143   
Model 1, HR (95% CI) 1.00 1.28 (0.99–1.65) 2.02 (1.50–2.70) 1.003 (1.001–1.004) <0.001 
Model 2, HR (95% CI) 1.00 1.00 (0.74–1.34) 1.39 (1.004–1.93) 1.003 (1.001–1.004) 0.07 
Model 3, HR (95% CI) 1.00 1.05 (0.78–1.41) 1.42 (1.01–2.00) 1.002 (1.001–1.004) 0.055 
SHBG
 Tertile 1 Tertile 2 Tertile 3 Continuous P trend 
Case subjects 191 119 74   
Model 1, HR (95% CI) 1.00 0.56 (0.45–0.71) 0.33 (0.25–0.43) 0.37 (0.30–0.46) <0.001 
Model 2, HR (95% CI) 1.00 0.82 (0.64–1.04) 0.56 (0.41–0.77) 0.63 (0.49–0.81) <0.001 
Model 3, HR (95% CI)
 
1.00
 
0.82 (0.64–1.05)
 
0.56 (0.40–0.79)
 
0.66 (0.51–0.86)
 
0.001
 
 TT
 
  

 
Tertile 1
 
Tertile 2
 
Tertile 3
 
Continuous
 
P trend
 
Case subjects 126 139 119   
Model 1, HR (95% CI) 1.00 1.04 (0.82–1.32) 0.90 (0.69–1.16) 0.91 (0.75–1.10) 0.40 
Model 2, HR (95% CI) 1.00 0.94 (0.74–1.20) 0.82 (0.63–1.07) 0.87 (0.71–1.07) 0.15 
Model 3, HR (95% CI)
 
1.00
 
0.96 (0.75–1.24)
 
0.88 (0.67–1.16)
 
0.93 (0.76–1.14)
 
0.36
 
 FAI
 
  

 
Tertile 1
 
Tertile 2
 
Tertile 3
 
Continuous
 
P trend
 
Case subjects 87 124 173   
Model 1, HR (95% CI) 1.00 1.39 (1.05–1.82) 2.01 (1.55–2.60) 1.54 (1.32–1.79) <0.001 
Model 2, HR (95% CI) 1.00 1.06 (0.79–1.42) 1.17 (0.87–1.57) 1.13 (0.94–1.36) 0.28 
Model 3, HR (95% CI)
 
1.00
 
1.05 (0.78–1.42)
 
1.15 (0.85–1.54)
 
1.10 (0.92–1.32)
 
0.34
 
 TE
 
  

 
Tertile 1
 
Tertile 2
 
Tertile 3
 
Continuous
 
P trend
 
Case subjects 109 132 143   
Model 1, HR (95% CI) 1.00 1.28 (0.99–1.65) 2.02 (1.50–2.70) 1.003 (1.001–1.004) <0.001 
Model 2, HR (95% CI) 1.00 1.00 (0.74–1.34) 1.39 (1.004–1.93) 1.003 (1.001–1.004) 0.07 
Model 3, HR (95% CI) 1.00 1.05 (0.78–1.41) 1.42 (1.01–2.00) 1.002 (1.001–1.004) 0.055 

Significant association (P < 0.05) indicated by boldface type. Model 1: adjusted for age, cohort, fasting status; model 2: model 1 + insulin, glucose, and BMI; model 3: model 2 + alcohol intake, smoking status, coronary heart disease, serum total cholesterol, statin use, systolic blood pressure, treatment for hypertension, hormone replacement therapy, age of menopause, CRP, and sex hormones for each other.

Because associations of continuous hormone variables with T2D in model 1 appeared linear, RRs stratified and sensitivity analyses were expressed per unit log or unit increase in hormone biomarkers. In the sensitivity analyses, substituting BMI with waist circumference as a measure of adiposity, substituting total cholesterol for other blood lipids, adjusting further for serum TSH, physical activity, number of pregnancies of at least 6 months, or menopause type, and excluding the first 3 years of follow-up did not affect any of the associations (Supplementary Table 1). Also, in the stratification analysis, no significant interactions were found for SHBG and TE with BMI or years since menopause (Supplementary Table 1). Significant interaction terms were found for TT (P interaction = 0.019) and FAI (P interaction = 0.03) with years since menopause. However, no association was found between these hormones and T2D after stratification for time since menopause (Supplementary Table 1). Also, no effect modification by BMI was found for TT and BT (Supplementary Table 1).

Systematic Review and Meta-analysis

Literature Search, Characteristics, and Quality of Eligible Studies

The initial search identified 3,209 potentially relevant citations. After screening and detailed assessment, 15 articles based on 12 unique studies were included (Supplementary Fig. 1 and Supplementary Appendix 5). Therefore, we meta-analyzed estimates from 13 studies (including the current study) involving a total of 14,902 pre- and postmenopausal women with 1,912 incident T2D cases, reporting on the association between sex hormones and T2D risk. Detailed characteristics of these studies and quality assessment have been summarized in Supplementary Table 2. All studies were medium to high quality except one.

Sex Hormones and T2D in Pooled Analysis

The meta-analyses for BT, TE, and BE are based only on studies examining postmenopausal women; the meta-analysis for TT is based on four studies including postmenopausal women and one study including pre- and postmenopausal women, whereas the findings for SHBG derive from studies including premenopausal women (n = 2), postmenopausal women (n = 4), and combined (n = 3). The pooled RRs for T2D adjusted for several metabolic risk factors comparing third versus first tertile of SHBG, TT, BT, TE, and BE were 0.44 (95% CI 0.30–0.66, I2 = 77.9%, P < 0.001), 1.32 (95% CI 0.79–2.21, I2 = 53.8%, P = 0.07), 1.75 (95% CI 0.92–3.33, I2 = 80.7%, P = 0.001), 1.99 (95% CI 1.21–3.27, I2 = 55.1%, P = 0.06), and 3.58 (95% CI 0.86–14.84, I2 = 81.0%, P = 0.02) (Figs. 13). There was evidence of between-study heterogeneity for all these analyses, with the possible exception of the meta-analysis on the association between TE and the risk of T2D (Figs. 13). For SHBG, heterogeneity was not explained by any of the study-level characteristics assessed, such as menopause status, location, and number of participants (Supplementary Table 4). For TT, the level of heterogeneity was largely explained by location (Supplementary Table 4). Five studies could not be included in the meta-analyses. Soriguer found that in pre- and postmenopausal women, per one unit log increase in SHBG, TT, and BT, the corresponding RRs were 0.23 (95% CI 0.1–0.53), 1.04 (0.59–1.83), and 1.12 (0.59–2.13), respectively (20). Boyd-Woschinko et al. (21) reported a fivefold increase in T2D incidence in the lowest quintile of SHBG. Similarly, Lindstedt et al. (22) found that among patients in the low SHBG tertile, 18% converted to T2D as compared with 5% in the mid SHBG tertile and 2.5% in the high SHBG tertile. Okubo et al. (23) reported lower levels of SHBG in T2D converters (59.7 ± 8.4 nmol/L) than nonconverters (69.5 ± 2.5 nmol/L) during 3 years of follow-up, but that was not significantly different after adjusting for age, BMI, and waist-to-hip ratio. Sex steroids and SHBG were not associated with diabetes outcomes in pre- and postmenopausal women in the study of Mather et al. (24).

Figure 1

RRs of T2D comparing top vs. bottom thirds of baseline plasma SHBG. The summary estimates presented were calculated using random-effects models (D+L) and fixed effects (I-V). The sizes of the data markers are proportional to the inverse of the variance of the odds ratio; the CIs are represented by the bars. X2 = 36.2. I2 = 77.9%. P < 0.001.

Figure 1

RRs of T2D comparing top vs. bottom thirds of baseline plasma SHBG. The summary estimates presented were calculated using random-effects models (D+L) and fixed effects (I-V). The sizes of the data markers are proportional to the inverse of the variance of the odds ratio; the CIs are represented by the bars. X2 = 36.2. I2 = 77.9%. P < 0.001.

Figure 2

RRs of T2D comparing top vs. bottom thirds of baseline plasma TT and FT levels. The summary estimates presented were calculated using random-effects models (D+L) and fixed effects (I-V). The sizes of the data markers are proportional to the inverse of the variance of the odds ratio; the CIs are represented by the bars. A: X2 = 8.6. I2 = 53.8%. P = 0.07. B: X2 = 15.5. I2 = 80.7%. P = 0.001.

Figure 2

RRs of T2D comparing top vs. bottom thirds of baseline plasma TT and FT levels. The summary estimates presented were calculated using random-effects models (D+L) and fixed effects (I-V). The sizes of the data markers are proportional to the inverse of the variance of the odds ratio; the CIs are represented by the bars. A: X2 = 8.6. I2 = 53.8%. P = 0.07. B: X2 = 15.5. I2 = 80.7%. P = 0.001.

Figure 3

RRs of T2D comparing top vs. bottom thirds of baseline plasma TE and free E levels. The summary estimates presented were calculated using random-effects models (D+L) and fixed effects (I-V). The sizes of the data markers are proportional to the inverse of the variance of the odds ratio; the CIs are represented by the bars. A: X2 = 8.91. I2 = 55.1%. P = 0.06. B: X2 = 5.26. I2 = 81.0%. P = 0.02.

Figure 3

RRs of T2D comparing top vs. bottom thirds of baseline plasma TE and free E levels. The summary estimates presented were calculated using random-effects models (D+L) and fixed effects (I-V). The sizes of the data markers are proportional to the inverse of the variance of the odds ratio; the CIs are represented by the bars. A: X2 = 8.91. I2 = 55.1%. P = 0.06. B: X2 = 5.26. I2 = 81.0%. P = 0.02.

Publication Bias

The appearance of funnel plots was asymmetrical for the analysis on SHBG and T2D, and Egger test results were significant (P = 0.014) (Supplementary Fig. 2). This suggested that publication bias may be present. After exclusion of the four studies that included 50 or fewer case subjects with T2D, findings were not statistically significant (Egger test, P = 0.93, data not shown). No evidence of publication bias was observed for the analysis of TT or TE and T2D (Supplementary Fig. 2).

In this large population–based study of postmenopausal women free of T2D at baseline, we showed that that lower levels of SHBG and higher levels of TE were associated with the risk of T2D, independent of established risk factors for T2D, including BMI, glucose, and insulin. In contrast, the association between T and the risk of T2D was explained by BMI, glucose, and insulin. Pooled results from the systematic meta-analysis of 13 studies reinforce the validity and generalizability of our findings, suggesting that SHBG and TE are robust risk markers of T2D in women.

Unlike the previous meta-analysis by Ding et al. (25), which was based mainly on studies with cross-sectional design and examined only mean differences between case subjects with T2D and control subjects without T2D, our current pooled analysis is based on the findings from 13 prospective studies (only 2 studies included in the previous review were eligible), including 14,902 participants with 1,912 case subjects with T2D. Therefore, our meta-analysis provides a more detailed assessment of the nature and magnitude of the association between sex hormones and T2D in women.

SHBG levels have been associated with metabolic syndrome, glucose, and insulin levels, established risk factors for T2D (7,8,26). Also, women with polycystic ovary syndrome, a condition of anovulation and hyperandrogenism, are at increased risk of T2D, and levels of SHBG are decreased in these women (27). The complex biological mechanisms that explain the association between circulating SHBG levels and the risk for T2D are not fully understood. Classically, the primary function of SHBG was thought to be the binding of circulating hormones in order to regulate free sex hormone bioavailability to target tissues. Therefore, it has been hypothesized that the relation between SHBG and T2D may result from the indirect influence of alterations in SHBG on sex hormone bioavailability. However, in our study, the association between SHBG and T2D risk remains significant after adjustment for TT, BE, and TE, implicating SHBG levels as a risk factor for T2D independent of serum androgen levels. Additional evidence in support of an independent effect of SHBG on T2D comes from recent studies that have found several polymorphisms in the SHBG to associate with insulin resistance and T2D, suggesting that altered SHBG physiology may be a primary defect in the pathogenesis of disease (2831). Furthermore, a growing body of evidence shows that SHBG may directly mediate cell-surface signaling, cellular delivery, and biological action of sex hormones via activation of a specific plasma receptor (3234). At the target tissue level, the fraction of SHBG that is not bound to sex steroid has the ability to bind plasma membrane high-affinity receptors (RSHBG) (32). Sex steroids of variable biological potency can activate the anchored SHBG-RSHBG complex, and the activated complex can have either an agonist or antagonist effect. For example, SHBG-RSHBG complex can have direct cellular antagonistic properties against estrogen; SHBG may interact with cellular estrogen receptors, which can trigger a biological antiestrogenic response (32). Specific downstream effects of the SHBG-receptor complex merit further investigation since they may help to clarify the underlying mechanisms linking SHBG to T2D.

Our result for a positive relation between E and T2D is in contrast with the results from previous large randomized control trials of oral estrogen therapy, which showed a lower risk of T2D among postmenopausal women assigned to estrogen treatment (3537). However, due to the observational design, our study does not provide causality. Mendelian randomization experiments are warranted to investigate the potential causal implications of E on T2D. Exogenous estrogen may have different physiological effects depending on type, route, duration, and dose of estrogen therapy (3841). For example, opposing effects of oral estrogen on fasting glucose versus glucose tolerance have been reported (38,39). Also, in a randomized trial of postmenopausal women, oral estrogen elevated CRP levels up to 12 months of treatment but not transdermal E (40). Moreover, a bimodal relationship of estrogen dose may exist. In a clinical trial of postmenopausal women, a lower dose of estrogen therapy increased insulin sensitivity whereas a higher dose had the opposite effect (42).

In postmenopausal women, endogenous E may be associated with diabetes risk through its relation to glucose, insulin, obesity, and inflammation. Indeed, previous cross-sectional studies have linked both BE and TE with higher glucose and insulin resistance levels in postmenopausal women, independent of obesity (6,9,43). Also, whereas animal studies suggest that E regulates body composition, many studies in postmenopausal women have failed to show a consistent beneficial role of E in weight loss and in the distribution of body fat (44). However, in our study, the association between TE and T2D, although attenuated, remained significant after adjustment for plasma levels of glucose and insulin, BMI, and CRP, suggesting that E may play a direct role in the pathophysiology of T2D in postmenopausal women. Furthermore, additional adjustment for TT did not affect this association, suggesting that E may be more than just a marker of increased aromatase conversion. Explicit mechanisms of estrogen in relation to T2D require further study.

Our study showed no association between TT and the risk of T2D, whereas a suggestive positive association was observed between BT and T2D. The lack of association between free testosterone (FT) and the risk of T2D in our study might be due to the lack of a direct measure of BT in the blood, which could have biased our results toward the null. These findings are in line with previous studies reporting higher levels of insulin resistance with increasing levels of BT in postmenopausal women, whereas no association has been observed between TT and insulin resistance (6,14). Similarly, BT has been related to increased odds of having impaired fasting glucose (14).

The strengths of our study include its prospective design, the long follow-up, and adequate adjustments for a broad range of possible confounders. We also performed several sensitivity analyses, such as excluding the first 3 years of follow-up to avoid potential bias of undiagnosed disease at baseline. Furthermore, our study included, in addition to an analysis of primary data, a systematic review of all available published prospective cohorts, which is the first-ever quantitative synthesis of these associations thus far in women. Also, most of the studies included in our meta-analysis adjusted for potential confounding. However, there are several limitations that need to be taken into account. First, we did not have measures of BE in the Rotterdam Study, which could have strengthened our results. Also, TE was measured using an immunoassay with a detection limit of 18.35 pmol/L, which is considered suboptimal, particularly in postmenopausal women. However, the observed effect remained the same while analyzing TE continuously and categorically. Second, FT levels were not measured directly in the blood and therefore have to be interpreted with caution. Nevertheless, FT levels in this study were derived from the ratio of T to SHBG, which is considered a precise proxy for BT (45). Third, we observed a moderate to high level of heterogeneity across the included studies. Different assays (Supplementary Table 3) used to assess the levels of sex hormones and SHBG contributed to the observed heterogeneity. However, since the number of available studies included in each meta-analysis was generally small, it precluded our ability to conduct subgroup analyses involving various study-level characteristics (such as age). Fourth, there was evidence of publication bias for the association between SHBG and the risk of T2D, so it is possible that our results constitute an overestimation of the performance of the test. However, when we excluded small studies, differences were not statistically significant, and therefore, the effect of publication bias may be only minor. Fifth, except for SHBG, the other findings come from studies conducted mainly in postmenopausal women, and thus, these results cannot be extended to pre- or perimenopausal women. Finally, contrary to the results of random-effects models, the fixed-effects models showed a significant association of both BE and BT with the risk of T2D. The differences in random- versus fixed-effects models might be explained by the substantial heterogeneity observed between studies (for example, in the association of BT), which could be better captured under the random-effects model (46). For BE, the small size of the studies might undermine the precision of the estimate under a fixed-effects model. However, in light of these observations, the overall results of this study should be interpreted with caution.

In conclusion, lower levels of SHBG and higher levels of TE are independently associated with risk of T2D in postmenopausal women. Further studies are needed to establish hormone thresholds at which diabetes risk is increased, because this may aid in identifying high-risk postmenopausal women in the clinical setting.

See accompanying article, p. 568.

Acknowledgments. The authors thank Dr. Wanes Kazanjian (Erasmus MC) for reviewing the abstract.

Funding. This study was sponsored and funded by Metagenics Inc.

Metagenics Inc. had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. The funder/sponsor did not have the ability to veto the publication of study results.

Duality of Interest. T.M., L.J., and O.H.F. work at ErasmusAGE, a center for aging research across the life course funded by Nestlé Nutrition (Nestec Ltd.), Metagenics Inc., and AXA. T.M. and L.J. reported receiving research support from Metagenics Inc. J.N. has been financially supported by Erasmus Mundus Western Balkans (ERAWEB), a project funded by the European Commission. M.K. is supported by the AXA Research Fund. O.H.F. reported receiving grants or research support from Metagenics Inc. No other potential conflicts of interest relevant to this article were reported.

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

Author Contributions. T.M. conceived and designed the study, ran the analysis, screened the title and abstract, obtained the full text, determined the eligibility of articles, participated in data extraction, participated in data synthesis, analysis, and interpretation, drafted the final manuscript, and contributed to the critical revision of the manuscript and approved the final version. J.N. screened the title and abstract, obtained the full text, determined the eligibility of articles, participated in data extraction, and contributed to the critical revision of the manuscript and approved the final version. L.J., C.M., A.H., A.D., M.K., and J.S.E.L. contributed to the critical revision of the manuscript and approved the final version. W.M.B. designed and executed the search strategies and contributed to the critical revision of the manuscript and approved the final version. O.H.F. conceived and designed the study, drafted the final manuscript, and contributed to the critical revision of the manuscript and approved the final version. T.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Atsma
F
,
Bartelink
ML
,
Grobbee
DE
,
van der Schouw
YT
.
Postmenopausal status and early menopause as independent risk factors for cardiovascular disease: a meta-analysis
.
Menopause
2006
;
13
:
265
279
[PubMed]
2.
Gambacciani
M
,
Ciaponi
M
,
Cappagli
B
, et al
.
Body weight, body fat distribution, and hormonal replacement therapy in early postmenopausal women
.
J Clin Endocrinol Metab
1997
;
82
:
414
417
3.
Liu
Y
,
Ding
J
,
Bush
TL
, et al
.
Relative androgen excess and increased cardiovascular risk after menopause: a hypothesized relation
.
Am J Epidemiol
2001
;
154
:
489
494
[PubMed]
4.
Brand
JS
,
van der Schouw
YT
,
Onland-Moret
NC
, et al
.
Age at menopause, reproductive life span, and type 2 diabetes risk: results from the EPIC-InterAct study
.
Diabetes Care
2013
;
36
:
1012
1019
[PubMed]
5.
Gambineri
A
,
Patton
L
,
Altieri
P
, et al
.
Polycystic ovary syndrome is a risk factor for type 2 diabetes: results from a long-term prospective study
.
Diabetes
2012
;
61
:
2369
2374
[PubMed]
6.
Kalish
GM
,
Barrett-Connor
E
,
Laughlin
GA
,
Gulanski
BI
;
Postmenopausal Estrogen/Progestin Intervention Trial
.
Association of endogenous sex hormones and insulin resistance among postmenopausal women: results from the Postmenopausal Estrogen/Progestin Intervention Trial
.
J Clin Endocrinol Metab
2003
;
88
:
1646
1652
[PubMed]
7.
Brand
JS
,
van der Schouw
YT
.
Testosterone, SHBG and cardiovascular health in postmenopausal women
.
Int J Impot Res
2010
;
22
:
91
104
[PubMed]
8.
Brand
JS
,
van der Tweel
I
,
Grobbee
DE
,
Emmelot-Vonk
MH
,
van der Schouw
YT
.
Testosterone, sex hormone-binding globulin and the metabolic syndrome: a systematic review and meta-analysis of observational studies
.
Int J Epidemiol
2011
;
40
:
189
207
[PubMed]
9.
Golden
SH
,
Dobs
AS
,
Vaidya
D
, et al
.
Endogenous sex hormones and glucose tolerance status in postmenopausal women
.
J Clin Endocrinol Metab
2007
;
92
:
1289
1295
[PubMed]
10.
Ding
EL
,
Song
Y
,
Manson
JE
,
Rifai
N
,
Buring
JE
,
Liu
S
.
Plasma sex steroid hormones and risk of developing type 2 diabetes in women: a prospective study
.
Diabetologia
2007
;
50
:
2076
2084
[PubMed]
11.
Oh
JY
,
Barrett-Connor
E
,
Wedick
NM
,
Wingard
DL
;
Rancho Bernardo Study
.
Endogenous sex hormones and the development of type 2 diabetes in older men and women: the Rancho Bernardo study
.
Diabetes Care
2002
;
25
:
55
60
[PubMed]
12.
Kalyani
RR
,
Franco
M
,
Dobs
AS
, et al
.
The association of endogenous sex hormones, adiposity, and insulin resistance with incident diabetes in postmenopausal women
.
J Clin Endocrinol Metab
2009
;
94
:
4127
4135
[PubMed]
13.
Hofman
A
,
Darwish Murad
S
,
van Duijn
CM
, et al
.
The Rotterdam Study: 2014 objectives and design update
.
Eur J Epidemiol
2013
;
28
:
889
926
[PubMed]
14.
Leening
MJG
,
Kavousi
M
,
Heeringa
J
, et al
.
Methods of data collection and definitions of cardiac outcomes in the Rotterdam Study
.
Eur J Epidemiol
2012
;
27
:
173
185
[PubMed]
15.
World Health Organization
.
Definition and Diagnosis of Diabetes Mellitus and Intermediate Hyperglycemia: Report of a WHO/IDF Consultation
.
Geneva
,
World Health Org.
,
2006
, p.
1
50
16.
Rosner
W
,
Auchus
RJ
,
Azziz
R
,
Sluss
PM
,
Raff
H
.
Position statement: utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement
.
J Clin Endocrinol Metab
2007
;
92
:
405
413
[PubMed]
17.
Rubin
BD
.
Multiple imputation for nonresponse in surveys
.
Invest Radiol
1987
18.
Rubin
RB
.
Multiple Imputation for Nonresponse in Surveys
.
New York
,
John Wiley & Sons
, 1987
19.
Stroup
DF
,
Berlin
JA
,
Morton
SC
, et al
.
Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group
.
JAMA
2000
;
283
:
2008
2012
[PubMed]
20.
Soriguer
F
,
Rubio-Martin
E
,
Fernandez
D
,
Valdes
S
, et al
.
Testosterone, SHBG and risk of type 2 diabetes in the second evaluation of the Pizarra cohort study
.
Eur J Invest
2012
;
42
:
79
85
[PubMed]
21.
Boyd-Woschinko
G
,
Kushner
H
,
Falkner
B
.
Androgen excess is associated with insulin resistance and the development of diabetes in African American women
.
J Cardiometab Syndr
2007
;
2
:
254
259
[PubMed]
22.
Lindstedt
G
,
Lundberg
PA
,
Lapidus
L
,
Lundgren
H
,
Bengtsson
C
,
Bjorntorp
P
.
Low sex-hormone-binding globulin concentration as independent risk factor for development of NIDDM. 12-yr follow-up of population study of women in Gothenburg, Sweden
.
Diabetes
1991
;
40
:
123
128
[PubMed]
23.
Okubo
M
,
Tokui
M
,
Egusa
G
,
Yamakido
M
.
Association of sex hormone-binding globulin and insulin resistance among Japanese-American subjects
.
Diabetes Res Clin Pract
2000
;
47
:
71
75
[PubMed]
24.
Mather
KJKC
,
Kim
C
,
Christophi
CA
, et al.;
Diabetes Prevention Program
.
Steroid sex hormones, sex hormone-binding globulin, and diabetes incidence in the Diabetes Prevention Program
.
J Clin Endocrinol Metab
2015
;
100
:
3778
3786
[PubMed]
25.
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
1299
[PubMed]
26.
Weinberg
ME
,
Manson
JE
,
Buring
JE
, et al
.
Low sex hormone-binding globulin is associated with the metabolic syndrome in postmenopausal women
.
Metabolism
2006
;
55
:
1473
1480
[PubMed]
27.
Nestler
JE
,
Powers
LP
,
Matt
DW
, et al
.
A direct effect of hyperinsulinemia on serum sex hormone-binding globulin levels in obese women with the polycystic ovary syndrome
.
J Clin Endocrinol Metab
1991
;
72
:
83
89
[PubMed]
28.
Ding
EL
,
Song
Y
,
Manson
JE
, et al
.
Sex hormone-binding globulin and risk of type 2 diabetes in women and men
.
N Engl J Med
2009
;
361
:
1152
1163
[PubMed]
29.
Wang
Q
,
Kangas
AJ
,
Soininen
P
, et al
.
Sex hormone-binding globulin associations with circulating lipids and metabolites and the risk for type 2 diabetes: observational and causal effect estimates
.
Int J Epidemiol
2015
;
44
:
623
637
[PubMed]
30.
Zhao
JL
,
Chen
ZJ
,
Zhao
YR
, et al
.
Study on the (TAAAA)n repeat polymorphism in sex hormone-binding globulin gene and the SHBG serum levels in putative association with the glucose metabolic status of Chinese patients suffering from polycystic ovarian syndrome in Shandong province
.
Zhonghua Yi Xue Yi Chuan Xue Za Zhi
2005
;
22
:
644
647
[in Chinese]
[PubMed]
31.
Perry
JR
,
Weedon
MN
,
Langenberg
C
, et al.;
MAGIC
.
Genetic evidence that raised sex hormone binding globulin (SHBG) levels reduce the risk of type 2 diabetes
.
Hum Mol Genet
2010
;
19
:
535
544
[PubMed]
32.
Rosner
W
,
Hryb
DJ
,
Khan
MS
,
Nakhla
AM
,
Romas
NA
.
Sex hormone-binding globulin mediates steroid hormone signal transduction at the plasma membrane
.
J Steroid Biochem Mol Biol
1999
;
69
:
481
485
[PubMed]
33.
Catalano
MG
,
Frairia
R
,
Boccuzzi
G
,
Fortunati
N
.
Sex hormone-binding globulin antagonizes the anti-apoptotic effect of estradiol in breast cancer cells
.
Mol Cell Endocrinol
2005
;
230
:
31
37
[PubMed]
34.
Fortunati
N
,
Catalano
MG
,
Boccuzzi
G
,
Frairia
R
.
Sex hormone-binding globulin (SHBG), estradiol and breast cancer
.
Mol Cell Endocrinol
2010
;
316
:
86
92
[PubMed]
35.
Kanaya
AM
,
Herrington
D
,
Vittinghoff
E
, et al.;
Heart and Estrogen/progestin Replacement Study
.
Glycemic effects of postmenopausal hormone therapy: the Heart and Estrogen/Progestin Replacement Study. A randomized, double-blind, placebo-controlled trial
.
Ann Intern Med
2003
;
138
:
1
9
[PubMed]
36.
Margolis
KL
,
Bonds
DE
,
Rodabough
RJ
, et al.;
Women’s Health Initiative Investigators
.
Effect of oestrogen plus progestin on the incidence of diabetes in postmenopausal women: results from the Women’s Health Initiative Hormone Trial
.
Diabetologia
2004
;
47
:
1175
1187
[PubMed]
37.
Bonds
DE
,
Lasser
N
,
Qi
L
, et al
.
The effect of conjugated equine oestrogen on diabetes incidence: the Women’s Health Initiative randomised trial
.
Diabetologia
2006
;
49
:
459
468
[PubMed]
38.
Espeland
MA
,
Hogan
PE
,
Fineberg
SE
, et al
.
Effect of postmenopausal hormone therapy on glucose and insulin concentrations. PEPI Investigators. Postmenopausal Estrogen/Progestin Interventions
.
Diabetes Care
1998
;
21
:
1589
1595
[PubMed]
39.
Zhang
Y
,
Howard
BV
,
Cowan
LD
, et al
.
The effect of estrogen use on levels of glucose and insulin and the risk of type 2 diabetes in American Indian postmenopausal women: the Strong Heart Study
.
Diabetes Care
2002
;
25
:
500
504
[PubMed]
40.
Strandberg
TE
,
Ylikorkala
O
,
Tikkanen
MJ
.
Differing effects of oral and transdermal hormone replacement therapy on cardiovascular risk factors in healthy postmenopausal women
.
Am J Cardiol
2003
;
92
:
212
214
[PubMed]
41.
Decensi
A
,
Omodei
U
,
Robertson
C
, et al
.
Effect of transdermal estradiol and oral conjugated estrogen on C-reactive protein in retinoid-placebo trial in healthy women
.
Circulation
2002
;
106
:
1224
1228
[PubMed]
42.
Lindheim
SR
,
Presser
SC
,
Ditkoff
EC
,
Vijod
MA
,
Stanczyk
FZ
,
Lobo
RA
.
A possible bimodal effect of estrogen on insulin sensitivity in postmenopausal women and the attenuating effect of added progestin
.
Fertil Steril
1993
;
60
:
664
667
[PubMed]
43.
Goodman-Gruen
D
,
Barrett-Connor
E
.
Sex differences in the association of endogenous sex hormone levels and glucose tolerance status in older men and women
.
Diabetes Care
2000
;
23
:
912
918
[PubMed]
44.
Davis
SR
,
Castelo-Branco
C
,
Chedraui
P
, et al.;
Writing Group of the International Menopause Society for World Menopause Day 2012
.
Understanding weight gain at menopause
.
Climacteric
2012
;
15
:
419
429
[PubMed]
45.
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
3672
[PubMed]
46.
Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to Meta-Analysis. West Sussex, U.K., Wiley, 2009
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.

Supplementary data