HIV Articles  
Back 
 
 
Low Testosterone Levels are Common and Associated with Insulin Resistance in Men with Diabetes
 
 
  March 4, 2008
Journal of Clinical Endocrinology & Metabolism, doi:10.1210/jc.2007-2177
 
Mathis Grossmann MD, PhD*, Merlin C Thomas MBChB, PhD, Sianna Panagiotopoulos PhD, Ken Sharpe PhD, Richard J MacIsaac MBBS, PhD, Sophie Clarke MBBS, Jeffrey D Zajac MBBS, PhD, and George Jerums MBBS, MD
 
Department of Endocrinology and Medicine, University of Melbourne, Austin Health, Australia, Baker Heart Research Institute, Melbourne, Australia; Department of Mathematics and Statistics, University of Melbourne, Australia
 
Context: Low testosterone levels are common in men with type 2 diabetes and may be associated with insulin resistance.
 
Objective: We investigated prevalence of testosterone deficiency and the relationship between testosterone and insulin resistance in a large cohort of men with type 2 and type 1 diabetes.
 
Design: Cross-sectional survey of 580 men with type 2 diabetes and 69 men with type 1 diabetes. A subgroup of 262 men with type 2 diabetes was then reassessed after a median of six months.
 
Results: Forty-three percent of men with type 2 diabetes had a reduced total testosterone (TT), and 57% had a reduced calculated free testosterone (cFT). Only 7% of men with type 1 diabetes had low TT. By contrast, 20.3% of men with type 1 diabetes had low cFT, similar to that observed in type 2 diabetes (age-BMI adjusted odds ratio 1.4, 95% CI 0.7-2.9). Low testosterone levels were independently associated with insulin resistance in men with type 1 diabetes, as well as type 2 diabetes. Serial measurements also revealed an inverse relationship between changes in testosterone levels and insulin resistance.
 
Conclusions: Testosterone deficiency is common in men with diabetes, regardless of the type. Testosterone levels are partly influenced by insulin resistance, which may represent an important avenue for intervention, while the utility of testosterone replacement remains to be established in prospective trials.
 
"While there is a strong rationale for testosterone replacement, the balance of benefits and risks is currently unknown and still to be defined by large and longterm clinical trials. Certainly, testosterone replacement can improve performance, mood, and libido in men with hypogonadism (22), and augments insulin sensitivity (11, 15). However, testosterone may have deleterious actions on prostate disease, sleep apnoea and possibly cardiovascular risk (35). While insulin sensitivity is associated with testosterone deficiency, there is no evidence that insulin sensitisers, including metformin and thiazolidinediones are able to elevate testosterone levels in men with diabetes. Exercise and weight loss appears to be effective, but such lifestyle modifications should already be employed for a range of other reasons. Consequently the appropriate clinical response to this emerging problem remains to be determined."
 
Introduction

Testosterone deficiency is common in men with type 2 diabetes (1) in whom it may contribute to impaired performance, mood and libido (2). While a direct relationship between testosterone deficiency and cardiovascular risk remains controversial (3, 4), there is evidence that testosterone levels are inversely associated with insulin resistance (5), a potent risk factor for both micro- and macrovascular complications of diabetes (6). In particular, reduced total testosterone (TT) levels have been associated with insulin resistance and subsequent risk for developing type 2 diabetes (2, 7-10). Moreover, short-term studies in men have shown that testosterone supplementation may improve insulin sensitivity (11-15).
 
In contrast to studies in men with type 2 diabetes, relatively little is known about testosterone status in type 1 diabetes. A recent small study suggested that testosterone levels were lower in men with type 2 diabetes than in type 1 diabetes, concluding that low testosterone levels may be specific to type 2 diabetes (16). However, low levels of sex hormone binding globulin (SHBG), the main carrier protein of TT in the circulation, may be independently associated with the risk of type 2 diabetes (17, 18), potentially confounding this relationship. Moreover, insulin resistance is common in treated individuals with type 1 diabetes, and strongly associated with adverse outcomes (19).
 
In the present study, we examine the prevalence and predictors of testosterone deficiency, as estimated by both TT and SHBG-adjusted calculated free testosterone (cFT) levels (20), in a large, unselected cohort of men with type 1 or type 2 diabetes presenting at a single centre. In addition, we explore the potential determinants of testosterone levels in both populations, including insulin resistance and systemic inflammation (5, 8).
 
Results
 
Cohort characteristics
 
The initial cross-sectional survey included 580 men with type 2 diabetes and 69 men with type 1 diabetes. Six men with type 2 diabetes who produced a standardised residual greater than four were excluded from the statistical analysis. Clinical characteristics of these individuals are described in Table 1. Notably, most participants had longstanding diabetes and the prevalence of diabetic complications was high. One third of men with type 2 diabetes had documented macrovascular disease, and two thirds (66%) had microvascular complications. Of the patients with type 2 diabetes, 24% received metformin, 12% received a sulfonylurea, 31% were on both metformin and a sulfonylurea, and 10% received a thiazolidinedione. 40% of men also received insulin in combination with oral hypoglycaemic therapy.
 
The prevalence of testosterone deficiency in men with type 2 diabetes
 
Forty-three percent of all men (n= 249) with type 2 diabetes in our clinic had low TT levels (<10 nmol/L) (figure 1). In this cohort, TT levels were inversely related to age (figure 2a). In men with type 2 diabetes younger than 40 years, the prevalence of low TT levels was 20%, 29% in men aged 40-49 years, 37% in 50-59 year olds, 43% in 60-69 year-olds, 46% in 70-79 year-olds, and 61% in men aged 80 years or older.
 
Fifty-seven percent (n= 326) of all men with type 2 diabetes in our clinic had low cFT levels (<0.23 nmol/L), most of who also had low TT (63%), (figure 1). More than 85% of men with low TT levels also had a low cFT. The inverse association between cFT levels and age (figure 2b) was significantly stronger that that observed for TT (p<0.01), possibly because of the age-associated rise in SHBG (figure 2c). The prevalence of low cFT levels was 13% in men with type 2 diabetes younger than 40 years, 19% in men aged 40-49 years, 45% in 50-59 year-olds, 60% in 60-69 year-olds, 67% in 70-79 year olds, and 76% in men with type 2 diabetes aged 80 years or older. For every decade increase in age, the prevalence of low cFT levels effectively doubled (adjusted odds ratio 2.0, 95% CI 1.4 to 2.4).
 
The prevalence of testosterone deficiency in men with type 1 diabetes
 
Few men with type 1 diabetes had low TT (7.2%, p<0.001, vs type 2 diabetes) (figure 1). This frequency of low TT levels in men with type 1 diabetes was not significantly greater than observed in reproductively normal young men (21). After adjusting for age, BMI and other confounding factors, the frequency of low TT levels remained significantly higher in men with type 2 diabetes, when compared to those with type 1 diabetes (adjusted odds ratio 4.0, 95% CI 1.5 to 10.7, p=0.005). However, when SHBG was included into the model, this difference between type 1 and type 2 diabetes was eliminated (p=0.16).
 
By contrast, one in five men with type 1 diabetes had low cFT levels (20.3%) (figure 1), a prevalence significantly higher than normally observed in healthy men (21). This frequency was statistically similar to that observed in age- and BMI-matched men with type 2 diabetes (adjusted odds ratio 1.4, 95% CI 0.7 -2.9), reflecting the difference in significant SHBG levels in the two groups (figure 2c, p<0.001). As in men with type 2 diabetes, the major predictor of cFT levels in individuals with type 1 diabetes was age (figure 2b). For each decade of life, the prevalence of low cFT levels effectively doubled (adjusted odds ratio for low cFT levels 2.4, 95% CI 1.4 to 3.9).
 
Testosterone indices and insulin Resistance
 
In men with type 2 diabetes in our clinic, insulin resistance (as estimated by the HOMA-IR equation (23)) was independently associated with low TT levels (Odds ratio 1.2, 95% CI 1.0 to 1.4), after adjusting for age, BMI, treatment regimens and other potentially confounding variables. Individuals with low TT levels were also more likely to have a BMI >30kg/m2 (55% vs 35%, p<0.001), elevated triglycerides >1.7 mmol/L (45% vs 28%, p<0.001), reduced HDL cholesterol levels (28% vs 17%, p<0.001), and higher hs-CRP levels (median 7.7 vs 4.5, p<0.01). However, there was no difference in glycaemic control, blood pressure levels or the frequency or intensity of antihypertensive treatments between those with and without low testosterone levels. TT levels were also correlated with the HOMA-estimated insulin resistance within the normal range of both parameters after adjusting for age and obesity (partial correlation coefficient -0.13, p= 0.002).
 
Insulin sensitivity was also independently associated with cFT levels in individuals with type 2 diabetes, such that that men with low cFT levels also tended to be more insulin resistant after adjusting for age, obesity, treatment regimens and other potentially confounding variables (partial correlation coefficient -0.10, p=0.02). Individuals with low cFT levels were also more likely to have a BMI >30kg/m2 (49% vs 39%, p=0.03) and reduced HDL cholesterol levels (25% vs 17%, p=0.03).
 
Levels of SHBG were not associated insulin resistance in men with type 2 diabetes after adjusting for age and BMI (p=0.281). However, low SHBG levels were independently associated with poorer glycemic control (HbA1c) after adjustment for age and BMI (p=0.04). SHBG levels were not associated with type of oral hypoglycemic therapies or the use of statins.
 
In men with type 1 diabetes, cFT levels were also independently associated with effective glucose disposal date (eGDR), a marker of insulin sensitivity in individuals with type 1 diabetes (24), after adjusting for age (p=0.04). There was no statistically significant association between eGDR and TT or SHBG levels.
 
Changes in testosterone levels over time.
 
In a randomly selected subgroup of 262 men with type 2 diabetes, TT and cFT determinations were repeated at their next routine clinic appointment (median of 6 months; range 1-15). The clinical characteristics of this subgroup in which testosterone levels were re-tested were not significantly different to those of the total cohort of men with type 2 diabetes (table 1). None of these men received testosterone therapy.
 
At second analysis, the prevalence of testosterone deficiency defined by low TT levels (<10 nmol/L) was not significantly different than obtained at the first estimation (39 vs 42%, p=0.2), and there was a strong correlation between estimations (R2=0.73, figure 3a). Seventy-three percent of individuals with low TT level on the first estimation had low TT levels on repeat testing. Most of those rising above 10 had borderline levels (8-10 nmol/L) on initial testing. Although the variability between samples was small, there was some evidence of eregression to the mean', with the lowest samples showing a mean increase in levels (figure 3b). However, after adjusting for each subject's their baseline parameters, age and the time between clinic visits, we were able to demonstrate a significant inverse relationship between the change in TT level and the change in HOMA-IR during the same follow-up period (p=0.01). In addition, the change in TT level was also independently correlated with the change in HbA1c during the same follow-up period (p=0.02). This was partly explained by the association between SHBG and HbA1c (p<0.01).
 
At second analysis, there was also a good correlation between cFT levels obtained at the two time points (R2=0.57, figure 3c). However, the frequency of testosterone deficiency defined by low cFT levels (<0.230 nmol/L, 48%) at the second time-point was significantly lower than obtained at the first estimation (60%, difference p<0.001). Two thirds (66%) of patients with low testosterone levels at the first estimation, continued to have low levels at the second reading. There was some evidence of eregression to the mean' (figure 3d). Nonetheless, after adjusting for each subject's baseline measurements, the change in cFT was inversely correlated with the change in HOMA-IR during the same follow-up period (p=0.03). There was no significant relationship between the change in cFT level during follow-up and the change in HbA1c.
 
Discussion
 
In this cross-sectional analysis of a large unselected cohort of men presenting to a single tertiary referral centre, we found that 43% of men with type 2 diabetes had low TT levels and 57% had low cFT levels. In addition, 20% of men with type 1 diabetes also had low cFT levels, an age-adjusted rate not significantly different from that observed in type 2 diabetes. In our cohort, age and BMI were major factors influencing both TT and cFT levels, consistent with reports from previous studies (2, 7, 25). As testosterone deficiency may contribute to impaired performance, mood and libido (2), as well as have adverse impact on cardiovascular risk (3-5), these findings demonstrate the presence of a significant and unrecognised problem.
 
These findings are consistent with a smaller clinic-based study showing a 33% prevalence of low testosterone in men with type 2 diabetes (1) and population-studies in which reduced testosterone levels are more common in men with type 2 diabetes than in the age-matched general population (8). However, this study in the first to demonstrate an similar prevalence of low cFT levels in individuals with type 1 diabetes, in contrast to previous findings (16), despite otherwise similar demographic and biochemical patient characteristics.
 
Importantly, we show that free testosterone levels were independently correlated with indices of insulin resistance in men with type 2 diabetes, as well as those with type 1 diabetes. Moreover, the change in testosterone levels over time was also independently correlated with changes in insulin sensitivity in the subgroup of men with type 2 diabetes followed longitudinally. Longitudinal changes in testosterone levels in patients with type 1 diabetes remain to be explored in a larger cohort of patients. This data is consistent with findings in non-diabetic men, where cFT levels have been shown to be inversely associated with insulin levels (8), HOMA (25), and visceral adiposity (8, 25). In addition, these findings support the hypothesis that circulating testosterone levels in men with diabetes may be influenced by insulin sensitivity, and vice versa. While none of the men in our study received testosterone therapy, short-term studies in men have shown that testosterone supplementation may improve insulin sensitivity (11-13, 15). Male mice with a targeted deletion of the androgen receptor have increased blood glucose levels due to insulin resistance (26). Men with Klinefelter syndrome have increased insulin resistance (27), and androgen deprivation therapy in men with prostate cancer increases the risk of developing the metabolic syndrome as well as that of incident diabetes (28, 29).
 
Conversely, interventions to improve insulin sensitivity may also significantly impact on testosterone levels. In particular, visceral adiposity is an important cause of insulin resistance, and also decreases testosterone concentrations through conversion to estradiol by aromatase (5). In our study, as well as others (9, 10), testosterone levels in men with type 2 diabetes were correlated with BMI. Although BMI and weight are suboptimal markers of visceral adiposity, previous studies have reported an association of loss of weight in obese insulin resistant men with increased testosterone levels (30).
 
Although not specifically employed in our study, improved lifestyle factors or altered pharmacological management that contributed to improved insulin sensitivity may also have contributed to an increase testosterone levels observed in our patients. Similarly, in patients where insulin sensitivity declined, testosterone levels fell on average.
The four-fold higher prevalence of reduced TT levels observed in men with type 2 diabetes when compared to those with type 1 diabetes was largely driven by reduced levels of SHBG (figure 2c), which was present at all ages and across all levels of BMI. When SHBG was included into the model, this difference between type 1 and type 2 diabetes was eliminated (p=0.16). Insulin is known to inhibit hepatic production of SHBG (31), and SHBG levels fall acutely during hyperglycemic-euglycemic clamp studies (32). Indeed, reduced SHBG has been suggested as a surrogate marker for insulin resistance (33). While insulin resistance in individuals with type 2 diabetes may explain both the high frequency of both low SHBG levels and TT levels in our study, there was no independent association of SHBG levels with the HOMA index of insulin resistance in men with type 2 diabetes. Moreover, cFT levels, a SHBG-independent testosterone parameter, were still reduced in individuals with both type 2 and type 1 diabetes, and independently associated with insulin resistance.
 
The strengths of this study include the large number subjects, inclusion of a large cohort of patients with type 1 diabetes, measurement of serum testosterone levels at the appropriate time of day (early morning), accurate assays for total and free testosterone levels, and longitudinal follow-up in a substantial number of men with type 2 diabetes. However, it remains to be established whether the biochemical testosterone deficiency observed in this study represents a true hypogonadal state. Guidelines recommend that a diagnosis of hypogonadism be made "only in men with consistent signs and symptoms and unequivocally low serum testosterone levels" (22). Interpretation of our study is therefore limited because we did not obtain a detailed record of symptomatology history. This said, generalised symptomatology in individuals with longstanding diabetes is almost impossible to distinguish from those of hypogonadism.
 
Findings from the cross-sectional component of our study are also limited because a single low testosterone level is inadequate for making the diagnosis of hypogonadism, given the variability in serum testosterone levels that can result from circadian rhythms, the pulsatile nature of its secretion, use of concomitant medications and measurement variations (22). To assess this variability, we repeated testosterone determinations in a representative subset of individuals with type 2 diabetes, finding that 27% and 33% of individuals with low TT and cFT levels respectively had levels in the normal range when retested. This is consistent with reports in non-diabetic men, where as many as 30% of men will have normal levels when repeated (34).
 
While there is a strong rationale for testosterone replacement, the balance of benefits and risks is currently unknown and still to be defined by large and longterm clinical trials. Certainly, testosterone replacement can improve performance, mood, and libido in men with hypogonadism (22), and augments insulin sensitivity (11, 15). However, testosterone may have deleterious actions on prostate disease, sleep apnoea and possibly cardiovascular risk (35). While insulin sensitivity is associated with testosterone deficiency, there is no evidence that insulin sensitisers, including metformin and thiazolidinediones are able to elevate testosterone levels in men with diabetes. Exercise and weight loss appears to be effective, but such lifestyle modifications should already be employed for a range of other reasons. Consequently the appropriate clinical response to this emerging problem remains to be determined.
 
 
 
 
  icon paper stack View older Articles   Back to top   www.natap.org