We conducted a cohort study in a community from June through November 2011 in Guangzhou, China. Subjects eligible for this study were taken from the Risk Evaluation of Cancers in Chinese Diabetic Individuals: A Longitudinal Study (the REACTION Study). More details of the REACTION Study have been published previously. A total of 10104 subjects aged ≥40 were recruited via home visits or examination notices and 9916 subjects provided written informed consent and enrolled in the project. The participation rate was 98.1%. Firstly, we excluded women from the study, leaving 2584 men in our analyses. Of these men, those who failed to provide information [questionnaire: n=1582; systolic blood pressure (SBP) at baseline: n=4; waist circumference (WC) at baseline: n=20; TG at baseline: n=6; HDL-c at baseline: n=1; fasting plasma glucose at baseline (FPG): n=3; SBP during follow up: n=2; WC during follow up: n=2] were excluded from the study. In addition, 344 subjects with MetS at baseline and 89 subjects who were lost to follow-up were also excluded. Eventually, our analyses included 531 eligible men. The flow chart of the selection of study subjects is shown in Figure 1 .

For each participant, personal information was collected via a well-established questionnaire that included demographic characteristics and lifestyle factors. There were three categories of smoking and drinking habits: ‘never’, ‘current’ (having smoked or drunk regularly during the recent half year), or ‘ever’ (having quit smoking or drinking for more than half a year). To estimate the frequency and duration of habitual physical activity, we used a short form of the International Physical Activity Questionnaire (IPAQ). Separate metabolic equivalent hours per week (MET-h/week) were computed to assess the level of physical activity. Standardized anthropometric measurements were completed by staff who had passed the training. Height (accurate to 0.1 cm) and weight (accurate to 0.1 kg) were measured, and body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m2). Each participant’s waist circumference (accurate to 0.1 cm) was measured at the umbilical level while standing. Blood pressure was determined as the average of 3 measurements by the same staff at a 5-min interval using a validated automated electronic device (OMRON, Omron Company, China). All blood samples were drawn after overnight fasting for at least 8 hours to measure sex hormone binding globulin (SHBG), total testosterone (TT), luteinizing hormone (LH), follicle-stimulating hormone (FSH), triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), fasting plasma glucose (FPG), glycosylated hemoglobin (HbA1c), fasting insulin, gamma-glutamyl transferase (γ-GGT), and serum creatinine. The collected venous blood samples were centrifuged at 25°C and stored at -80°C until assay. Considering the possible variation of testosterone and other hormones caused by circadian rhythms, individual blood samples were collected between 7:00 am and 9:00 am.

We used the chemiluminescence techniques to measure TT, LH, and FSH (IMMULITE 2000 Immunoassay System, Siemens Healthcare Diagnostics, USA), and the coefficients of variation were 4.7%–7.5% (intra-assay) and 2.5%–3.3% (inter-assay) for the evaluation of TT. SHBG was assessed by ELISA (DRG International, USA). The reference quartile cutoff points of TT, LH, FSH and SHBG are as follows: TT: [Q1:0–14.6; Q2:14.6–19.2; Q3:19.2–23.5; Q4:>23.5]; LH: [Q1:0–3.97; Q2:3.97–5.71; Q3:5.71–8.06; Q4:>8.06]; FSH: [Q1:0–7.32; Q2:7.32–10.54; Q3:10.54–15.58; Q4:>15.58]; SHBG: [Q1:0–33.1; Q2:33.1–52.2; Q3:52.2–101.1; Q4:>101.1].

An autoanalyzer (ARCHITECT ci16200 Integrated System, Abbott Laboratories, USA) was used for the measurement of TG, TC, HDL-C, LDL-C, FPG, and γ-GGT. HbA1c was determined via the HPLC method (VARIANT II TURBO, Bio-Rad, USA). Insulin was measured by chemiluminescence immunoassay (Centaur XP, SIEMENS, USA). The insulin resistance index (homeostasis model assessment of insulin resistance, HOMA-IR) was calculated using the following equation: HOMA-IR index= [fasting insulin(μIU/ml)] *[fasting glucose (mmol/L)]/22.5. Insulin resistance was estimated by a HOMA-IR index within the top quartile (greater than 2.18 in the present study) (23). Glomerular filtration rate (eGFR) was estimated using the abbreviated Modification of Diet in Renal Disease (MDRD; mL/min/1.73m2) formula: eGFR =186 × [serum creatinine × 0.011]-1.154 × [age]-0.203 × [0.742 if female] ×1.233, where serum creatinine was expressed as mmol/L and 1.233 was the calibration coefficient for Chinese population (24).

The diagnosis of MetS was based on the harmonized definition of a joint statement of organizations (25). We classified subjects with MetS if they met three or more of the following criteria: (1) raised triglycerides (≥1.7mmol/L) or previous hypertriglyceridemia treatment, (2) reduced HDL cholesterol (<1.0mmol/L in males and <1.3mmol/L in females), (3) high blood pressure (≥130 mmHg systolic and/or ≥85 mmHg diastolic) or under antihypertensive drug treatment, (4) high fasting plasma glucose (FPG ≥5.6mmol/L) or under treatment for glucose-lowering, and (5) central obesity(waist circumference ≥ 85 cm for men or ≥80 cm for women). The cutoff points for waist circumference were determined using Chinese population thresholds.

All the participants enrolled were free from MetS at baseline. We divided the study cohort into the non-MetS group and the newly-MetS group, depending on whether MetS was developing or not during the follow-up period. In the data analyses, the data were presented as the means ± SD, medians (interquartile ranges), or numbers (proportions) according to the different variable types. Gonadal hormones were expressed as quartiles, and linear regression models were run to test for trends across groups. Analysis of variance (ANOVA) was adopted for data comparisons among groups. Post hoc comparisons were performed using Bonferroni correction when continuous variables that conformed to the normal distribution were analyzed. The Chi-square test was used for the comparison of categorical variables between groups. Unadjusted and multivariate-adjusted logistic regression models were performed to assess the risk of prevalent MetS for each quartile of gonadal hormone levels. Odds ratios (OR) and their corresponding 95% confidence intervals (95% CI) were computed. Model 1 was unadjusted. Model 2 was adjusted for age, BMI, current smoking and drinking status, physical activity level, and SBP. Model 3 was adjusted for age, BMI, current smoking and drinking status, physical activity level, SBP, TG, eGFR, γ-GGT, HbA1c, LDL-C, and HOMA-IR. The relationship between the incidence of MetS and SHBG level was also examined in subgroup analysis and conducted within the age strata (≥ 60/<60years), BMI (normal/overweight and obesity), insulin resistance (yes/no), diabetes (yes/no) and eGFR (≥ 90/< 90ml/min per 1.73 m2). Tests for interaction were conducted by simultaneously including all strata factors, the quartiles of SHBG, and the respective interaction terms (strata factor multiplied by quartiles of SHBG) in the final model 3. All the statistical analyses were carried out with SAS V.9.3 (SAS Institute), and two-sided P-values <0.05 indicated statistical significance.

The baseline parameters of the cohort by the MetS status during follow-up were shown in Table 1 . A proportion of 20.7% of the participants developed MetS with an average follow-up time of 3.2 ± 0.4 years. The mean age of the 531 study subjects was 61.2 ± 7.2 years. Compared with the non-MetS group, subjects in the newly-MetS group had significantly higher BMI, WC, SBP, DBP, TG, TC, LDL-C, FPG, HbA1c, Fasting insulin, HOMA-IR, γ-GGT, and lower HDL-C, TT, LH and SHBG at baseline (all p<0.05).

The total population was assessed based on the different gonadal hormone levels. As shown in Figure 2 , the incidence of MetS according to elevated TT quartiles was 25.2%, 23.7%, 17.3% and 16.7% respectively (p for trend = 0.0425), the incidence of MetS was 25.0%, 21.8%, 21.8% and 14.3% respectively (p for trend =0.0411) according to elevated LH quartiles and the incidence of MetS according to elevated SHBG quartiles was 26.9%, 22.7%, 21.1% and 12.1% respectively (p for trend =0.0035). Nevertheless, no obvious trend differences were found according to the elevated FSH quartiles.

Univariate logistic regression analysis showed that subjects in quartile 1 of SHBG had significantly increased odds of incident MetS compared with those in quartile 4 ( Table 2 ). After multivariable adjustment for age, BMI, current smoking and drinking status, physical activity level, SBP, TG, eGFR, γ-GGT, HbA1c, LDL-C, and HOMA-IR (Model 3), the OR[CI] of incident MetS for Quartile 1 was 2.33[1.13–4.79] (p<0.05), which suggests that lower SHBG might be considered as a predictor of new-onset MetS and its progression. But there was no statistically significant difference among TT, LH, and FSH (p>0.05).

Figure 3 shows the incidence of each metabolic syndrome component based on the elevated SHBG quartiles. The incidence rates of hypertriglyceridemia, central obesity, and insulin resistance tended to decrease according to the elevated SHBG quartiles (all p for trend < 0.05). Nevertheless, no significant trend differences were found with low HDL-c, high glucose, and elevated blood pressure. To explore the internal conformance of the above findings, we further analyzed the direct relationship between gonadal hormone levels and the number of metabolic syndrome components. As shown in Figure 4 , TT, LH, and SHBG levels significantly decreased with the increasing number of metabolic syndrome components (all p for trend < 0.05).

Table 3 shows the multivariate-adjusted ORs of incident MetS according to decreased SHBG quartiles within different subgroups. The associations between SHBG level and incident MetS were inconsistent in subgroup analyses. Significant relation between SHBG level and Mets was detected in the subjects aged ≥ 60 years, without insulin resistance, with normal BMI, and with eGFR ≥90 mL/min per 1.73m2 (p < 0.05).