Low total testosterone is associated with increased risk of incident type 2 diabetes mellitus in men: results from the Study of Health in Pomerania (SHIP)

Abstract

Objective. There is increasing evidence suggesting that low total testosterone concentration is associated with incident type 2 diabetes mellitus (T2DM) in men. The aim of this study was to evaluate the association between total testosterone and incident T2DM in a large population-based cohort.

Methods. Of 2117 men at baseline, 1589 were followed up 5 years later. Low total testosterone concentration at baseline determined by <10th percentile (10-year age-strata) were used as a risk factor for incident T2DM at follow-up. To evaluate for potential non-response bias, drop out weights were used in sensitivity analysis.

Results. From 1339 men eligible for analyses, 68 (5.1%) developed T2DM. Men with low total testosterone concentration had an increased risk of developing T2DM (odds ratio [OR] 3.4, 95% CI 1.9–6.1), even after adjustment for age, waist circumference and smoking, OR 3.0; (95% CI 1.6–5.7). Recalculated weighted models revealed almost identical estimates indicating no relevant non-response bias.

Discussion. Our prospective findings suggest that low total testosterone concentration is associated with incident T2DM in men and might represent a biomarker that might causally be involved in the risk of T2DM. This underlines the importance of measuring total testosterone in men as the predominant male sex hormone.

Keywords:

• Type 2 diabetes mellitus
• testosterone
• Study of Health in Pomerania
• population-based
• risk factor

Introduction

The public health burden of type 2 diabetes mellitus (T2DM) has dramatically increased [1]. By far, the most common cause of morbidity and mortality in T2DM is cardiovascular disease. Even pre-diabetes which carries a substantial risk of the disorder and consequently the cardiovascular risk starts many years before the development and diagnosis of T2DM [2]. Tremendous progress has been made in identifying risk factors to prevent or delay T2DM [3]. There is growing evidence suggesting that low total testosterone concentrations are associated with an increased risk of incident T2DM in men [4]. In a Southern California population-based cohort of 985 men aged 40–79 years in 1972–1974, 21% of patients with T2DM compared to 13% of men without T2DM had total testosterone concentration <12.1 nmol/l [5].

Testosterone may play a role in the pathogenesis of T2DM by increasing skeletal muscle tissue and decreasing visceral obesity and non-esterified fatty acids, which improve insulin sensitivity [6]. In addition, sex hormone-binding globuline may have impact on the development of T2DM at both the genomic and phenotypic levels [7]. Intervention studies showed reductions in visceral adiposity and insulin resistance during testosterone supplementation [8,9]. The Multiple Risk Factor Intervention Trial based on men at high risk for coronary heart disease suggested that low sex hormone-binding globuline and both free and total testosterone in addition with higher glucose and insulin levels and obesity are involved in the prediabetic phase after 5 years follow-up. The association was inconsistent over varying concentrations of total testosterone [10].

In population-based studies, total testosterone concentrations in men have consistently been associated with T2DM, but these studies are limited by their cross-sectional design [11–13]. Currently, data from few longitudinal observational studies with relatively small study populations are available. The Rancho Bernardo Study [14] investigating 294 surviving residents aged 55–89 years found low total testosterone concentrations associated with incident T2DM after 8 years follow-up. Moreover, the Massachusetts Male Aging Study [11] suggested that low total testosterone and sex hormone-binding globuline in 1156 men aged 40–70 years influence the development of insulin resistance and subsequent T2DM after 7–10 years follow-up [11]. However, previous findings of this study found an association between total testosterone as well as sex hormone-binding globuline as an independent predictor of incident T2DM [15]. In contrast, findings from the Tromsø Study showed that men with lower total testosterone and SHBG had an increased risk of T2DM which was dependent on waist circumference [16]. Furthermore, from a sample of 195 men from the Australian Longitudinal Study of Ageing, aged 70 years and older, 35 (5%) developed an incident T2DM after 8 years follow-up. Plasma total testosterone concentrations <8.0 nmol/l were inversely related to visceral obesity and components of the metabolic syndrome [17]. In a population-based study in Finland among 702 men aged 44–60 years with low total testosterone and SHBG, 147 developed the metabolic syndrome and 57 T2DM after 11 years [6]. Interestingly, this study suggested that the metabolic syndrome predicts development of hypogonadism defined by total testosterone concentration <11 nmol/l after 11 years of follow-up [18]. However, previous results from our study confirmed an increased risk of incident metabolic syndrome among men with low total testosterone concentrations, especially in younger men [19].

Various definitions of hypogonadism have been proposed and used in the literature making the comparison across studies complicated [20]. Thresholds for total testosterone indicating hypogonadism ranged from <6.9 nmol/l [21] to <10.4 nmol/l along with signs and symptoms of hypogonadism [22].

Taken together, the data addressing the association between total testosterone and incident T2DM so far are inconsistent and have been generated predominantly from cross-sectional, patient-based studies or relatively small longitudinal studies, often restricted to a specific age group or region. Because the components of the metabolic syndrome describe risk factors for cardiovascular disease and T2DM, the aim of our study was to investigate the longitudinal association between low total testosterone concentration and the risk of incident T2DM in men covering a wide age range with longitudinal data from the Study of Health in Pomerania (SHIP).

Methods

Study population

SHIP is a population-based cohort study in West Pomerania, a region in the northeast of Germany. The total population of West Pomerania selected for SHIP comprised 212,157 inhabitants. The sampling was performed from population registries, where all German citizens are registered. The net sample (without migrated or deceased people) consisted of 6267 eligible subjects; of which 4308 (2117 men) aged 20–79 (response 69%) participated in the baseline study (baseline). Data collection started in October 1997 and was finished in March of 2001. From March 2003 until July 2006, the first 5-year follow-up examination was performed (follow-up). The net sample (without migrated, deceased or non-responding people) then compromised 3300 subjects (response 84%). All participants gave written informed consent. The study conformed to the principles of the Declaration of Helsinki as reflected by an a priori approval of the Ethics Committee of the University of Greifswald.

From 2117 men at baseline, those without participation at follow-up (n = 528) were excluded. Furthermore, men with following characteristics were excluded from this study: type 1 diabetes mellitus (n = 4), known T2DM (n = 124), and unknown T2DM at baseline with concentrations within the diabetic range (n = 9); missing data for diabetes status at baseline (n = 3), and follow-up (n = 6), missing biomarker such as serum glucose (n = 7), serum total testosterone (n = 46); uninterpretable total testosterone concentrations <0.690 mg/dl (n = 2); prostate cancer (n = 6), receiving sexual hormones (anatomic-therapeutical-chemical code (ATC) G03, n = 1), testosterone 5α reductase inhibitors (ATC G04CB, n = 2), glucocorticoids (ATC R03BA; H02AB; n = 36) at baseline, none of the men reported to receive sexual hormone antagonists (ATC L02B) or anabolic steroids (ATC A14A), missing data in the covariates (n = 4). This resulted in a study population of 1339 men for the present analyses.

Measurement

A history of self-reported diabetes mellitus, socio-demographic information, data on health-related behaviour as well as concomitant medical history were assessed by standardised face-to-face computer-assisted personal interviews. The outcome variable for this analysis was incident T2DM between baseline and 5-year follow-up period. At baseline, we operationally defined known diabetes mellitus on the basis of (1) self-reported physician's diagnosis; (2) use of anti-diabetic medication (ATC code A10A, A10B); (3) age of disease onset. We definded type 1 diabetes mellitus if the onset of disease occurred before the age of 30 years and insulin was commenced less than 1 year after disease onset. All other diabetic subjects were defined as having T2DM. Unknown T2DM was defined based on a random plasma glucose of ≥11.1 mmol/l [23].

At follow-up, diagnosed diabetes was operationally defined on the basis of (1) self-reported physician's diagnosis and (2) use of anti-diabetic medication. We defined type 1 diabetes mellitus if the onset of disease occurred before the age of 35 years and insulin was commenced. All other diabetic subjects were defined as having incident T2DM. Variables that might influence the pathway between total testosterone and incident T2DM were assessed at baseline. Age was considered as a continuous variable. Waist circumference was measured to the nearest 0.1 cm using an inelastic tape midway between the lower rib margin and the iliac crest in the horizontal plane, with the subject standing comfortably with weight distributed evenly on both feet. Height was measured to the nearest 1 cm using a digital ultrasound instrument and weight was measured to the nearest 0.1 kg in light clothing and without shoes using standard digital scales (Soehnle-Waagen GmbH, Nassau, Germany). Body mass index was calculated as body weight divided by body height squared (kg/m²). Smoking status was assessed by dividing subjects into categories of non-smokers [never or occasionally (<1 cigarette/day)], former smokers and current smokers (1–14 cigarettes/day, ≥15 cigarettes/day). No consideration was given to subjects who smoked cigars or pipes because of low proportions (<1.5%). Physical activity was estimated by recall of the past week's activity, including duration and intensity. Men who participated in physical activities during summer or winter for at least 1 h a week were classified as being physically active. A history of prostate cancer was based on subject's self-reports.

Assays

The exposure variable of primary interest was baseline serum total testosterone concentration. Non-fasting blood samples were drawn from the cubital vein in the supine position. Samples were taken between 07:00 a.m. and 04:00 p.m. Serum aliquots were prepared for immediate analysis and for storage at −80°C for further analysis. Total testosterone concentrations were measured from frozen serum aliquots using competitive chemiluminescent enzyme immunoassays on an Immulite 2500 analyzer (Siemens Immulite 2500 Total Testosterone, ref. L5KTW, lot 110; Siemens Healthcare Diagnostics GmbH, Eschborn, Germany). An aliquot of two alternating levels of a third party commercial control material (Bio-Rad Lyphochek Immunoassay Plus Control, lot 40151 and lot 40152; Bio-Rad, Munich, Germany) was included in each series in single determination. For total testosterone, the inter-assay coefficient of variation was 13.2% with a systematic deviation of +2.3% at the 3.2 nmol/l concentration, and 8.9% with a systematic deviation of +0.24% at the 22.5 nmol/l concentration.

Low concentrations of total testosterone were defined <10th percentile in every 10-year age-group [6,24,25]. Because there is no widely accepted cut-off for the definition of low total testosterone concentration due to the diurnal and inter-individual variability in total testosterone concentration in the healthy male during the day, as well as among assays [26], additional analyses were done with clinically used thresholds of total testosterone concentration of <8 nmol, <10.4 and <12 nmol/l. Glucose was determined enzymatically on a Hitachi 704 analyzer (Roche Diagnostics, Mannheim, Germany). Glycated haemoglobin A1c (HbA1c) at baseline was determined as per high-performance liquid chromatography (Bio-Rad Diamat, Munich,Germany). All assays were performed according to the manufacturers' recommendations by skilled technical personnal. The laboratory takes part in official quarterly German external proficiency testing programs. In addition, internal quality controls were analysed daily.

Statistical analyses

Continuous variables were given as mean ± standard deviation (SD), except for variables with skewed distribution that were given as median (25th, 75th). Categorical variables were given as percentages. For continuous data, comparisons between groups were performed using the Mann–Whitney U-test. For nominal data, the χ² test was applied. The study population was compared with respect to total testosterone concentrations <10th percentile and ≥10th percentile in every 10-year age-group. Respondents and non-respondents at follow-up were compared regarding basic baseline characteristics. The theory of directed acyclic graphs [27] was applied to select confounders for adjustment. According to our graph, the variables age (continous, years), waist circumference (continuous, cm) and smoking (non-smokers, ex-smokers, current smokers with 1–14 cigarettes/day, or 15 or more cigarettes per day) belonged to the minimally sufficient adjustment set. We used these variables to adjust our effect estimation. To assess the association between total testosterone and these covariates at baseline with incident T2DM at follow-up, we used logistic regression models with incident T2DM as dependent variable. The models included dichotomized total testosterone concentrations and all variables that belonged to the minimally sufficient adjustment set. For additional analyses, established clinically thresholds for low concentrations of total testosterone (<8 nmol/l, <10.4 nmol/l, <12 nmol/l) were used as well as a threshold (<16 nmol/l) beyond which the risk of T2DM did not increase. Furthermore, non-response analysis was conducted to evaluate possible bias due to missing data. We applied weights because individuals commonly differ in their propensity to drop-out of surveys. This propensity depends on the individuals' characteristics and can be expressed as a probability. By taking the inverse of this probability, we can assume how many persons of the baseline sample are represented by each participating individual at follow-up. For this purpose, parameter estimates from a logistic regression model were repeated using statistical weights that accounted for drop out from baseline to follow-up using sex, age, marital status, education, smoking, alcohol intake, and subjective health as predictors, to derive inverse probability weights that account for selective non-response. The incidence odds ratio (OR) was presented together with 95% confidence interval (CI). All statistical analyses were performed with SAS release 9.1 (SAS Institute, Cary, NC).

Results

Among 1339 men, 130 (13.1%) had low total testosterone concentration at baseline (Table I). Men with low total testosterone concentration had a higher waist circumference and body mass index. Men with low total testosterone concentration were less frequently smokers and physically active. During follow-up, 68 men (5.1%) developed incident T2DM.

Table I. Baseline characteristics of men with total testosterone concentrations below and above the 10th percentile.

	Testosterone <10th percentile (n = 130)	Testosterone ≥10th percentile (n = 1209)
Age (years)	50 (15)	49 (15)
Waist circumference (cm)	99.7 (12.0)	94.5 (11.0)*
Body mass index (kg/m^2)	28.9 (4.2)	27.3 (3.8)*
Smoking^† (%)
Never smoker	30.0	26.6
Former smoker	46.9	42.9
1–14	7.7	9.6
≥15	15.4	21.0
Physical activity (%)	38.5	45.0
Serum total testosterone (nmol/l)	9.1 (8.1; 9.6)	16.6 (13.8; 20.4)^‡
Random plasma Glucose (mmol/l)	5.4 (5.0; 6.0)	5.3 (4.9; 5.8)
Haemoglobin A1c (%)	5.4 (4.8; 5.8)	5.3 (5.0; 5.7)
Cholesterol (mmol/l)	5.7 (5.0; 6.5)	5.7 (5.0; 6.5)
Triglycerides (mmol/l)	2.2 (1.4; 3.1)	1.7 (1.1; 2.5)*
HDL cholesterol (mmol/l)	1.2 (1.0; 1.5)	1.3 (1.1; 1.5)
LDL cholesterol (mmol/l)	3.5 (2.7; 4.3)	3.6 (2.9; 4.3)
Incident type 2 diabetes mellitus (%)	13.1	4.2*

Data are percentages, mean (SD) or median (25th; 75th).

*p < 0.05 for tested differences between men with total testosterone concentrations below and above the 10th percentile based on the χ²-test (nominal data) or the Mann–Whitney U-test (continuous data).

^†Never smoking include occasional smoking (<1 cigarette per day).To convert testosterone from SI units (nmol/l) into conventional units (ng/dl), multiply by 28.82; Glucose from mmol/l into mg/dl by 0.0555.

^‡No test for comparison.

The baseline comparison between respondents and non-respondents revealed that respondents were younger, had less often diagnosed T2DM, were more frequent never- and former smoker, and more physical active in comparison to non-respondents (Table II). Regarding overall total testosterone concentration, no differences were observed between respondents and non-respondents.

Table II. Baseline characteristics of respondents and non-respondents.

	Respondents (n = 1339)	Non-respondents (n = 528)
Age (years)	50 (37; 61)	58 (35; 70)*
Diagnosed diabetes mellitus (%)
Type 1	0.2	0.2
Type 2	7.8	11.0
Undiagnosed^†	0.2	0.6*
Serum total testosterone (nmol/l)	16.0 (12.9; 19.9)	16.4 (12.7; 21.1)
Serum total testosterone (nmol/l)
<10th percentile	13.0	11.8
20–29	12.5	10.5
30–39	10.7	8.4
40–49	9.6	8.6
50–59	10.3	9.4
60–69	10.2	8.0
>70
Random plasma glucose (mmol/l)	5.3 (5.0; 5.8)	5.5 (5.0: 6.3)*
Hemoglobin A1c (%)	5.3 (4.9; 5.7)	5.4 (5.0; 5.9)*
Waist circumference (cm)	94.5 (87.0; 102.2)	95.5 (87.1; 103.8)
Body mass index (kg/m^2)	27.2 (24.8; 29.7)	27.6 (24.7; 30.0)
Smoking (cigarettes per day) (%)
Never smoker^‡	26.9	18.4
Former smoker	43.2	42.2
1–14	9.4	13.6
≥15	20.5	24.4*
Physical activity (%)	44.4	37.7*

Data are percentages, mean (SD) or median (25th; 75th).

*p < 0.05 for tested differences between respondents and non-respondents based on the χ²-test (nominal data) or the Mann–Whitney U-test (continuous data).

^†Undiagnosed T2DM is defined by random plasma glucose ≥11.1 mmol/l (WHO) [23].

^‡Never smoker include occasional (<1 cigarette per day). To convert testosterone from SI units (nmol/l) into conventional units (ng/dl), multiply by 28.82; Glucose from mmol/l into mg/dl by 0.0555.

Table III presents the estimated crude and adjusted effect estimates for the association between total testosterone concentration and T2DM after 5-years of follow-up. Low baseline total testosterone concentrations apart from <16 nmol/l, were associated with a considerably increased odds of incident T2DM. The comparison of crude and adjusted incidence odds ratios revealed that the confounders produced some overestimation of the strength of association.

Table III. Estimated crude and adjusted odds ratios for the association of total testosterone associated with incident type 2 diabetes mellitus.

	Testosterone <10th percentile*	Testosterone <8 nmol/l^†	Testosterone <10.4 nmol/l^†	Testosterone <12 nmol/l^†	Testosterone <16 nmol/l^†
Crude Model	3.4 (1.9; 6.1)	5.7 (2.4; 13.7)	3.5 (2.0–6.2)	2.7 (1.6; 4.6)	2.4 (1.4–4.0)
Adjusted Model^‡	3.0 (1.6; 5.7)	4.5 (1.6; 12.4)	2.5 (1.3–4.7)	1.9 (1.1; 3.4)	1.6 (0.9–3.0)

*Total testosterone concentrations <10th percentile vs. ≥10th percentile by 10 year age strata.

^†Total testosterone concentration vs. >8 nmol/l; vs. ≥8 nmol/l; >14.4 nmol/l; vs. ≥14.4 nmol/l; >12 nmol/l; vs. ≥12 nmol/l; >16 nmol/l; vs. ≥16 nmol/l.

^‡Model: adjusted for age (continuous), WC (continuous), smoking (non smoking as reference; exsmoker, 1–14 cigarettes/die, ≥15 cigarettes/die). To convert testosterone from SI units (nmol/l) into conventional units (ng/dl), multiply by 28.82.

To estimate possible bias due to missing data, we repeated our analyses by using statistical weights that accounted for drop out from baseline to follow-up (Figure 1). The associations between low total testosterone concentration and incident T2DM remained almost identical indicating no relevant bias due to non-response.

Figure 1.  Odds ratio of the association between low total testosterone concentrations and incident type 2 diabetes mellitus for the responding full cohort (N = 1339 men/n = 68 cases) using different cut-offs [<10th percentile (N = 130/n = 17); <8 nmol/l (N = 32/n = 7); <10.4 nmol/l (N = 136/n = 18)]; <12nmol/l (N = 247/n = 25)] in comparison to weighted analyses accounted for drop out from baseline to follow-up.

Discussion

In the present study, we found that low concentrations of total testosterone were associated with incident T2DM in men in a 5-year follow-up after adjustment for baseline confounders. This longitudinal result, to our knowledge one of the first in Europe, is in line with previous findings from other longitudinal population based studies [6,11,14,15]. In addition to these results, our findings suggest that low total testosterone plays a role in the development of T2DM in men and therefore, represent a biomarker that might also causally be involved in the risk of incident T2DM. This is supported by previous results from SHIP suggesting that low concentrations of total testosterone predict development of the metabolic syndrome consisting of risk factors for cardiovascular disease and T2DM [19]. In contrast to this direction of association, longitudinal results showed that the metabolic syndrome predisposes to development of hypogonadism defined by total testosterone concentrations <11 nmol/l [6]. Results from the Australian Longitudinal Study of Ageing do not support a predictive or causative role for decreasing total testosterone in the development of incident T2DM. Although total testosterone concentrations were inversely related to visceral obesity and components of the metabolic syndrome, low concentrations of total testosterone may be a consequence rather than a cause of poor metabolic status [17].

However, there exists no widely accepted consensus as to what constitutes the concentrations of total testosterone to be low. In the present study, total testosterone concentrations <10th percentile across age in decades were associated with a similar risk of incident T2DM in comparison to the recommended threshold of total testosterone concentrations <10.4 nmol/l as low by the Endocrine Society [22]. Thus, it can be assumed that the present results reflect a reliable estimate of the risk of developing T2DM during a 5-year follow up. The risk of developing T2DM did not increase at a threshold of total testosterone concentrations <16 nmol/l in our study.

Moreover, the application of specific cut-offs identified in one study population implies that different studies will provide different cut-offs which makes comparison difficult [28]. There are some specificities regarding testosterone which need to be taken into account when interpreting results. As age is a main determinant of testosterone levels in men, low levels of total testosterone should not be based on threshold values that are constant across the whole age spectrum [24]. Instead, age-specific threshold values should be used as applied in our study [25].

Strengths of the present study include the random, population-based sample of men in a wide age range from a defined geographic area. A further strength of our study is its longitudinal design. Thus, it was possible to present incidence proportions, which, unlike prevalence estimates, reflect the risk of T2DM. Further strengths are the population representativeness, the high level of quality assurance, particularly in standardisation of non-invasive examination methods and data management [29].

Although the present study enabled the investigation between total testosterone concentration and the development of T2DM, some weaknesses should be noted. The operational definition of diabetes mellitus was based on self-reported data of known diabetes, and hence not restricted to diagnosed diabetes. The reliance on self-report of diabetes and medication use raises the possibility that some undetected T2DM or type 1 diabetes mellitus cases were included in the sample. Both are unlikely because of using a random plasma glucose level for undetected T2DM (≥11.1 mmol/l) [23] and the latter because of the age of the group examined. Because such misclassification is a bias toward the null, the results of the present study may not compromise. However, using self-reported data to diagnose diabetes is a common practice in many epidemiologic studies. Reliable evidence indicates that there is substantial agreement between self-reported diabetes and medical record data and in interview data [30].

Time for blood drawing is associated with total testosterone expecting higher concentrations in the morning [13,31]. Assessments ought to be made between 08:00 h and 10:00 h [22]. In SHIP, the blood samples were taken between 07:00 h and 16:00 h. But, as the diurnal variation is prominent in younger men and diminishes in older men, and T2DM did not occur in younger men in this study, this would only have weakened the described associations [13,31]. In the present study, only a single blood sample was drawn, which might have introduced random measurement errors in determining total testosterone concentration. The lack of repeated measurements may have led to an underestimation of the observed associations. However, total testosterone showed a high intraclass correlation, indicating that a single measure reliably characterises an individual [14].

The total testosterone concentration is largely determined by circulating sex hormone-binding globuline concentrations [32]. As a result of possible residual confounding, it was not feasible to control simultaneously for sex hormone-binding globuline because of the unmeasured condition in our study. Thus, it was not possible to clarify the extent to which the risk of low concentrations of total testosterone for developing T2DM is explainable by sex hormone-binding globuline. However, previous results demonstrated that total testosterone and sexual hormone-binding globuline are independently associated with incident T2DM, and the latter may contribute to the risk of T2DM through non-androgenic mechanisms providing novel targets for preventing T2DM [15]. Further potential limitations may also arise from the lack of free androgen index, and estradiol; especially because additional roles of these hormones in the development of T2DM have also been suggested [7,16]. Moreover, total testosterone was measured in our study using a chemiluminescent enzyme immunoassay which has some weaknesses in measuring low total testosterone concentration with a small variance as normally found in children and women. Whereas total testosterone in men as in our study has a wide range making the use of chemiluminescent enzyme immunoassays justifiable [33].

Non-response is a common bias in prospective cohort studies. Because the analyses are based on a follow-up examination, it can be argued whether or not the respondents are representative of the initial sample. In general, non-respondents in risk factor surveys are less healthy. Non-participants were more likely to have known and unknown T2DM. Therefore, the prevalence of known T2DM could have been underestimated. In the present study, respondents were younger than non-respondents; their total testosterone concentrations across age in decades differed from those seen at the follow-up visit, probably attributed to age. However, the bias was partly controlled, because non-response analysis was conducted to evaluate possible bias due to missing data. The logistic regression analysis was repeated by using statistical weights that accounted for drop out from baseline to follow-up. The associations remained almost identical indicating no bias of relevance due to missing data. Thus, it could be assumed that the present results give a reliable estimate of the association between total testosterone concentration and developing T2DM in the north-east German population. However, a more complete understanding of the pathogenesis leading to the development of T2DM might enable the identification of men at higher risk followed by more targeted preventive measures.

Taken together, this longitudinal population-based data demonstrated that total testosterone (<10th percentile and <10.4 nmol/l) seems to be an appropriate measurement for estimating the risk of developing T2DM during a 5-year follow-up, even though it was one single measurement randomly during the day. However, most of these men having low testosterone concentrations will not come to clinical attention as testosterone concentrations are not routinely measured for clinical practice [22,34,35]. Notably, asymptomatic men with low testosterone concentrations are at increased risk for several comorbid conditions including metabolic syndrome, dyslipidemia, erectile dysfunction or even T2DM [21]. Our findings underline the importance of measuring total testosterone concentrations in men as the predominant male sex hormone [36].

Acknowledgements

SHIP is part of the Community Medicine Research net (http://www.community-medicine.de) at the University of Greifswald, Germany. Funding was provided by grants from the German Federal Ministry of Education and Research; the Ministry for Education, Research, and Cultural Affairs; and the Ministry for Social Affairs of the Federal State of Mecklenburg–West Pomerania. Statistical analyses were further supported by the Competence Network Diabetes mellitus of Germany Federal Ministry of Education and Research (BMBF, grant 01GI0805-07). The testosterone reagents used were sponsored by Siemens Healthcare Diagnostics GmbH, Eschborn, formerly DPC Biermann GmbH, Bad Nauheim, Germany. Novo Nordisc provided partial grant support for the determination of plasma samples and data analysis. The contributions to data collection made by field workers, study physicians, ultrasound technicians, interviewers, and computer assistants are gratefully acknowledged.