Effect of Testosterone Treatment on Glucose Metabolism in Men With Type 2 Diabetes: A Randomized Controlled Trial

Observational studies consistently show that 30–50% of men with type 2 diabetes (T2D) have lowered circulating testosterone levels, relative to references based on healthy young men (1–3). This association is independent of age and obesity (4,5), and low testosterone levels in men with T2D are independently associated with insulin resistance (3). However, it is not known whether low testosterone levels are a cause or a consequence of T2D or its associated clinical features.

Several lines of evidence lend support to the hypothesis that testosterone treatment decreases insulin resistance. Experimental evidence, reviewed in Grossmann et al. (6), suggests that testosterone not only promotes the commitment of pluripotent stem cells into the myogenic lineage and inhibits their differentiation into adipocytes (7), but also regulates the metabolic functions of mature adipocytes and myocytes in ways that reduce insulin resistance. Androgen deprivation therapy in men with prostate cancer leads to insulin resistance (8), and testosterone therapy reduces fat mass and increases lean body mass (9).

There is also evidence for reverse causality, demonstrating that low testosterone may be a consequence of dysglycemia and obesity. In prospective studies, the metabolic syndrome predicts low testosterone (10), and weight gain or development of T2D are major drivers of the age-related decline in testosterone levels (11,12). Moreover, weight loss increases testosterone levels in observational and in intervention studies (13–15).

Testosterone prescribing has increased markedly in the U.S. and globally in recent years (16). However, the effects of testosterone treatment on glucose metabolism and on other important outcomes, such as cardiovascular events, remain uncertain (16). Men with T2D who are frequently obese and commonly present with moderately low testosterone levels constitute a clinically important group. The question arises whether special considerations apply to treating such individuals with testosterone therapy and if testosterone offers specific benefits in improving glucose metabolism in such men.

We therefore conducted a randomized, placebo-controlled clinical trial to test the hypothesis that testosterone therapy decreases insulin resistance and improves glycemic control in men with T2D and lowered circulating testosterone levels. Specifically, we targeted the prevalent population of obese men with established, moderately well-controlled diabetes and moderately lowered testosterone levels. In addition to effects on glucose metabolism, we carefully monitored treatment-related changes in body composition and in individual abdominal fat compartments to dissect their influence on glycemic end points. Intensification of antiglycemic therapy was a strictly enforced protocol violation leading to exclusion from the study.

This 40-week, randomized, double-blind, placebo-controlled trial (ClinicalTrials.gov NCT00613782) was conducted at a tertiary referral center (Austin Health, Melbourne, Australia) between November 2009 and February 2013 and approved by the Human Research Ethics Committee, Austin Health. Each participant provided written informed consent prior to entering the study.

Study subjects were recruited from specialist diabetes clinics, primary care, and the general community. Men aged 35–70 years of age were eligible to participate in this trial if they had a history of T2D, and the total testosterone (TT) level (averaged from two fasting morning specimens) was ≤12.0 nmol/L (346 ng/dL), as measured by electrochemiluminescence immunoassay (ECLIA). Although TT was measured by both ECLIA and liquid chromatography-tandem mass spectroscopy (LCMS/MS), recruitment was based on ECLIA because the LCMS/MS assay (17) was not available for routine clinical use. Therefore, samples were batched and measured by LCMS/MS at study end.

Exclusion criteria included testosterone treatment within 5 years prior to randomization, established pituitary or testicular disorder, screening TT level of <5.0 nmol/L (144 ng/dL), luteinizing hormone (LH) level >1.5× upper limit of normal, or screening prostate-specific antigen (PSA) level > 4 μg/L, a history of urinary obstruction, prostate cancer, or breast cancer, hematocrit >0.50, uncontrolled hypertension (>160/90 mmHg despite treatment), untreated obstructive sleep apnea, estimated glomerular filtration rate < 30 mL/min, cardiac insufficiency (New York Heart Association score >2), active malignancy, unstable psychiatric disease, weight >135 kg (the weight limit for the dual-energy X-ray absorptiometry [DXA] scanner), current use of glucagon-like peptide-1 agonist therapy or very low-calorie diet, or an HbA_1c level >8.5% (69 mmol/mol).

Changes in oral hypoglycemic agents or initiation of insulin therapy constituted a violation of trial protocol, leading to withdrawal from the trial. Changes in insulin dose, antihypertensive, and lipid-lowering therapies were recorded at each study visit. All subjects received written recommendations regarding physical activity, food choices, and glycemic index (http://www.diabetesaustralia.com.au/en/NDSS-Content/Diabetes-Information-Sheets/).

Eligible participants were randomly assigned in a concealed 1:1 allocation to either testosterone or placebo using permuted blocks with a block size of 4. The randomization sequence was generated by a statistician and implemented by the Austin Health clinical trials pharmacy. Pharmacists, trial investigators, and participants were blinded to intervention allocation. Intramuscular testosterone undecanoate 1,000 mg or a visually identical placebo injection (both in oily base) was injected into the upper outer quadrant of the buttock at 0, 6, 18, and 30 weeks.

The primary outcome measure was the change across groups and time from baseline in the homeostasis model assessment index of insulin resistance (HOMA-IR). The secondary outcome measure was the change across group and time in glycemic control as measured by HbA_1c. Other outcome measures were considered as explanatory variables.

At 0, 18, and 40 weeks, the following variables were assessed: drug treatment, adverse events, body weight, BMI, waist circumference, blood pressure (BP), HbA_1c, lipid profile, C-reactive protein (CRP), PSA, and a complete blood count. Baseline and, as per standard clinical practice, weeks 18 and 40 trough TT, sex hormone–binding globulin (SHBG), calculated free testosterone (cFT), and LH levels were measured. All blood samples were collected in the early morning in a fasted state prior to trial medication administration.

In men not on insulin treatment, fasting and post–75-g oral glucose tolerance test (OGTT) glucose and insulin levels (at 0, 90, and 120 min) were assessed at baseline, 18, and 40 weeks. HOMA-IR (3) and OGTT-based Insulin Sensitivity Index (18) were calculated. Oral hypoglycemic drugs were withheld for 48 h prior to measurement of glucose and insulin levels.

TT was measured by both ECLIA (3) and LCMS/MS (17), SHBG, and LH using ECLIA (3), and free testosterone was calculated according to Vermeulen’s formula as described (3). Glucose, insulin, HbA_1c (3), and adiponectin (19) were measured as described.

At 0 and 40 weeks, total body mass, fat mass, and lean mass were assessed using DXA (DXA Prodigy, Version 10.51; GE Lunar, Madison, WI) (8). In subjects who had no contraindications to magnetic resonance imaging (MRI) scanning, axial images of the abdominal region were obtained using a three-tesla MRI scanner (Siemens, Erlangen, Germany) and analyzed using the SliceOmatic program software (version 4.2; Tomovision, Montreal, Canada). For each subject and time point, five 10-mm slices were selected for analysis of subcutaneous and visceral adipose tissue volume, with one slice centered at the level of the L4 vertebral superior endplate, three slices distributed at 40-mm intervals superiorly, and one inferiorly. The measured areas (in mm²) were converted into volumes (in cm³). Analyses were performed by a single individual (E.J.G.) blinded to patient identity, date, and study visit. The intraobserver coefficients of variation for visceral and subcutaneous fat area were 3.7 and 2.3%, respectively.

The trial was designed to have a power of 90% to detect a mean difference in HOMA-IR change from baseline between testosterone treatment and placebo of 1.73 (SD 1) and of an HbA_1c of 0.37% (SD 0.4) at 40 weeks based on a crossover trial in men with T2D and lowered serum testosterone treated with intramuscular testosterone (20).

We used a generalized linear mixed model to compare the primary and secondary outcomes between the treatment and placebo group. The model included the fixed effects of baseline values of the variable assessed, treatment group (testosterone vs. placebo as a categorical variable), repeated measurements taken at the three visits (0, 18, and 40 weeks, used as categorical time points), and the interaction term of visit × treatment group. The random effect was repeated measure by participant. The primary comparison of interest was the change across groups over time represented by the interaction term in the model. The overall P value given refers to this comparison. As a quantitative measure, mean adjusted difference (MAD) plus 95% CI between the groups from baseline to week 40 are provided. Following an intention-to-treat protocol, the analysis included all randomized subjects who were enrolled in the trial. A sensitivity analysis was performed first, after imputing missing values by multivariate imputation with chained equations and secondly including only subjects who had completed the study and adhered to the protocol. We also assessed sensitivity of the main outcomes to relevant baseline variables. Separate models with similar characteristics were used to assess other outcome data and safety variables. Comparison of baseline characteristics was based on Wilcoxon rank-sum test or χ² test in case of categorical variables. In case of low numbers, the Fisher exact test was used. To compare repeated measurements of variables considered explanatory within groups between two time points, the Wilcoxon signed-rank test was used. Data shown are median plus interquartile range (IQR). All tests were two-tailed with P < 0.05 denoting statistical significance. No adjustments were made for multiple comparisons on explanatory variables. Analyses were conducted using R for Mac version 3.01 (21,22) and SPSS version 21 (SPSS Inc., Chicago, IL).

Bayer Pharma AG (Berlin, Germany) provided testosterone, placebo, and financial support to conduct investigations, but had no role in trial design, data analysis, or writing the manuscript.

Of 965 men approached, 263 men proceeded to screening investigations, 88 of whom were randomized (testosterone, 45; placebo, 43) and included in the primary analysis. Seventy-five men completed the trial. The most common reason for noncompletion, occurring in 8 out of the 13 men who did not complete the study, was intensification of oral hypoglycemic agents or commencement of insulin therapy, which constituted predefined protocol violations (Supplementary Fig. 1). While changes in insulin dose in the 20% of men receiving insulin therapy (Table 1) was not an exclusion criterion, none of the testosterone-treated men had a reduction in the total daily dose (TDD) of insulin during the study. In addition, patients receiving placebo were not more likely to have an increase in TDD in insulin: three men receiving placebo and three men receiving testosterone treatment had such an increase.

Baseline characteristics were comparable between the groups except that subjects in the testosterone-treated group had lower body weight, lower fasting glucose, and slightly greater insulin sensitivity as measured by the OGTT-based Insulin Sensitivity Index (Table 1).

TT measured by ECLIA averaged 8.7 nmol/L (251 ng/dL) and 8.5 nmol/L (245 ng/dL) at baseline in the testosterone and placebo groups, respectively (P = 0.6) (Table 1). Baseline LCMS/MS-derived TT levels were, as expected (16,23), higher than ECLIA values at 10.6 nmol/L (306 ng/dL) in the testosterone group and 11.0 nmol/L (317 ng/dL) in the placebo group (P = 0.76). At 40 weeks, both ECLIA TT and LCMS/MS TT, as well as ECLIA cFT and LCMS/MS cFT, increased significantly in the testosterone group, while there was no significant change in the placebo group (Supplementary Fig. 2). The MADs in change over 40 weeks across the two groups were: LCMS/MS TT, 5.9 nmol/L (170 ng/dL [95% CI 3.9–7.9 (112–228)]; P < 0.001) and LCMS/MS cFT, 183 pmol/L (53 pg/mL [138–228 (40–66)]; P < 0.001). LH decreased (MAD −5.1 IU/L [95% CI −13 to −2.9]; P < 0.001), as did SHBG (MAD −3.7 nmol/L [−5.9 to −1.5]; P < 0.001) (Supplementary Fig. 2).

Testosterone treatment did not improve the primary outcome, HOMA-IR over 40 weeks, compared with placebo (MAD −0.08 [95% CI −0.31 to 0.47]; P = 0.23) or the secondary outcome, HbA_1c (MAD 0.36% [0.0–0.7]; P = 0.05) (Table 2).

There was no significant change in OGTT-based Insulin Sensitivity Index, fasting glucose, C-peptide, or insulin levels between the testosterone or placebo group (Table 3). In the men treated with insulin therapy, there was no significant difference in TDD of insulin between the two groups from baseline to 40 weeks (P = 0.32).

While there was no change in body weight, BMI, or waist circumference between groups, testosterone treatment significantly decreased total fat mass (MAD −2,377 g [95% CI −3,096 to −1,658]; P < 0.001) and increased lean mass (MAD 2,078 g [1,519–2,637]; P < 0.001) (Table 3). There was a significant decrease in abdominal subcutaneous adipose tissue (MAD −320 cm³ [−477 to −163]; P < 0.001) but interestingly not in visceral adipose tissue (MAD 140 cm³ [−89 to 369]; P = 0.90) or adiponectin levels (MAD −3.3 μg/mL [−8.2 to 1.6]; P = 0.42) (Table 3).

In subjects with no change in their lipid-lowering agents, testosterone treatment decreased total cholesterol (MAD −0.45 mmol/L [−17.4 mg/dL] [95% CI −0.7 to −0.2 (−27 to 7.7)]; P < 0.001), LDL cholesterol (MAD −0.26 mmol/L [−10.0 mg/dL] [−0.46 to −0.06 (−17.8 to −2.3)]; P = 0.01), and HDL cholesterol (MAD −0.11 mmol/L [−4.3 mg/dL] [−0.19 to −0.03 (−7.3 to −1.2)]; P = 0.002). There was no significant between-group change in prevalence of the metabolic syndrome, CRP levels, or, in participants without changes in antihypertensive agents, in BP (Table 3).

Outcomes were essentially unchanged after imputation of missing values or when analyzing participants who adhered to protocol only. MADs were as follows: HOMA-IR, −0.17 (95% CI −0.52 to 0.22), P = 0.25, n = 88; and HbA_1c, 0.22% (−0.05 to 0.49), P = 0.01; n = 88 in the imputed sample; and in the per-protocol data: HOMA-IR, −0.15 (−0.52 to 0.22), P = 0.40, n = 56; and HbA_1c, 0.20% (−0.09 to 0.49), P = 0.38, n = 75. Outcomes also remained unchanged when the three men in each group with changes in their TDD of insulin were excluded: HOMA-IR, −0.08 (−0.31 to 0.47), P = 0.23; and HbA_1c, 0.22% (−0.15 to 0.59), P = 0.32. Further, the change in HOMA-IR and HbA_1c over the treatment period was not related to relevant baseline characteristics, such as age, BMI, duration of diabetes, insulin use, or testosterone levels. Moreover, in a post hoc analysis, a higher baseline HbA_1c did not predict a better glycemic response (P = 0.42). Similarly, a lower baseline total (P = 0.42) or cFT (P = 0.30) did not predict a better glycemic response. There was no correlation between the change of either TT or free testosterone levels during treatment and measures of glucose metabolism. In contrast, the fat change in testosterone-treated men was related both to the delta (difference between week 40 and week 0) LCMS/MS TT (−0.26; P < 0.02) and LCMS/MS TT achieved at week 40 (−0.27; P < 0.02), but not to the baseline LCMS/MS TT (−0.01; P = 0.91). Moreover, differences in MAD for HOMA-IR and HbA_1c remained nonsignificant if men with a baseline LCMS/MS TT of <12 and ≥12 nmol/L were analyzed separately. By contrast, the MAD for fat mass remained significant in both these two groups (data not shown).

The frequency of adverse events is shown in Table 4. Serious adverse events were few, and incidence was not significantly different between groups.

There was a significant increase in hematocrit (0.04 [IQR 0.02–0.07]; P < 0.001) and hemoglobin (14 g/L [IQR 4–19]; P < 0.001) in the testosterone-treated patients at 40 weeks, but no change in the placebo group hematocrit (0 [−0.02 to 0.01]; P = 0.98) or hemoglobin (−1 g/L [−6 to 3]; P = 0.23) (Supplementary Fig. 3). The MADs across groups over 40 weeks were: hematocrit, 0.04 (95% CI 0.03–0.05), P < 0.001, and hemoglobin, 14.7 g/L (11.4–18.4), P < 0.001 (Table 3). In the testosterone group, three participants (6.6%) developed transient increases in the hematocrit to >0.54. One participant in the testosterone group was withdrawn, as his hematocrit was >0.54 prior to his 30-week injection. There was a significant increase in PSA (0.33 μg/L [IQR 0.11–0.67]; P < 0.001) in the testosterone-treated patients at 40 weeks, but no change in the placebo group (0.01 μg/L [IQR −0.04 to 0.13]; P = 0.39) (Supplementary Fig. 3). The MAD in PSA levels across groups over 40 weeks was 0.55 μg/L (95% CI 0.25–0.84); P < 0.001 (Table 3). There were no significant differences between groups with regard to new diagnosis of ischemic heart disease or significant congestive cardiac failure (Table 4).

The hypothesis of the trial that testosterone treatment may confer a specific glycemic benefit to obese men with reasonably well-controlled T2D and moderately lowered testosterone levels was not confirmed in this randomized controlled trial (RCT). Testosterone therapy did not improve measures of basal or dynamic insulin resistance and did not improve glycemic control in such men. These data therefore do not support the addition of testosterone treatment to antidiabetic medications to improve glucose metabolism in such men.

Men enrolled in this RCT had only modestly lowered baseline testosterone levels. As expected (16,23), values measured by LCMS/MS were higher than those obtained by immunoassay. Previous RCTs in men with diabetes (Supplementary Table 1) have measured testosterone only by immunoassay, and baseline levels were similar or higher compared with the immunoassay testosterone levels in our cohort. Profound testosterone deficiency (<1 nmol/L) induced by androgen-deprivation therapy has been shown to induce insulin resistance in observational studies (8). However, current evidence does not support a consistent threshold concentration of circulating testosterone required to maintain insulin sensitivity: in the general population, insulin resistance (24) or prevalence of diabetes (25) was increased in men with circulating testosterone levels <8 nmol/L (24) and <10 nmol/L (25). By contrast, in men with diabetes, the inverse relation between testosterone levels and insulin resistance was present even in men with testosterone levels extending well into the normal range (3). A study of experimentally induced hypogonadism in healthy men showed that reducing testosterone levels to 6.1 nmol/L had no effect on insulin sensitivity (26). Collectively, these data do not exclude the possibility that men with T2D and substantially lowered testosterone levels may derive a glycemic benefit with testosterone therapy. However, marked reductions of testosterone are uncommon in men with diabetes: in meta-analyses, mean pooled differences in TT relative to men without diabetes ranged from −1.61 nmol/L (4) to −2.99 nmol/L (27). Moreover, in our post hoc analyses, lower baseline TT and free testosterone levels did not predict a better glycemic response. Men with unequivocal reductions in their testosterone levels, whether they have T2D or not, should be considered for testosterone treatment irrespective of potential effects on glucose metabolism, given its general benefits in such men (28).

We cannot directly answer the question whether testosterone therapy could be more effective in men with less well-controlled diabetes. We considered this clinically less important given the availability of effective antiglycemic therapies. It can be noted that none of our study subjects required a reduction in their oral antidiabetic medications or insulin doses. A higher baseline HbA_1c did not predict a better glycemic response.

Interestingly and somewhat paradoxically, while we did document an expected reduction in fat mass and an increase in lean mass with testosterone therapy, the changes in body composition failed to translate into an improvement in glycemic outcomes. Importantly, the changes in fat mass in the testosterone-treated men were related to the testosterone level achieved during treatment, but not to the baseline testosterone level, arguing, consistent with our findings for glucose metabolism, against a testosterone threshold for this biological response. Our finding is consistent with studies in men with experimentally induced hypogonadism (26) and in mice lacking the androgen receptor (29), which also showed that the effects of testosterone on body composition and glucose metabolism can be dissociated. In our study, this dissociation may have occurred because testosterone treatment did not decrease visceral adipose tissue (Table 3), the fat compartment most closely associated with insulin resistance. Consistent with this, we did not observe an increase in adiponectin, an insulin-sensitizing adipokine inversely associated with visceral adipose tissue mass in T2D (30). Indeed, our findings are consistent with recent experimental castration studies in men: Finkelstein et al. (31) showed that testosterone add-back prevents gain of subcutaneous but not of visceral fat, although effects on glucose metabolism were not reported. Juang et al. (32) provided evidence, using insulin clamps, that testosterone add-back prevents gain of total fat mass, but does not improve insulin sensitivity, although effects on regional fat mass were not reported. Effects on adiponectin were not measured in either study.

Lack of compliance can be excluded for the negative outcome of the trial because the study drug was administered by intramuscular injection. Failure to achieve adequate testosterone levels with treatment is also unlikely, given an average increase of 5.9 nmol/L (170 ng/L) in circulating trough TT levels and the suppression of trough LH levels in testosterone-treated men (Supplementary Fig. 2). In addition, a robust increase in HCT and PSA levels was observed (Supplementary Fig. 3).

Statistically, it appears also unlikely that the trial may have missed a meaningful testosterone-related improvement in glucose metabolism. The CIs of our analyses make it unlikely (<2.5%) that there was an effect of a >0.31 decrease in HOMA-IR or an effect of a >0.0 decrease in HbA_1c. Indeed, the CI for HBA_1c was favoring placebo, suggesting that the HbA_1c may actually have deteriorated slightly in the testosterone-treated group (P = 0.05).

It should be noted that men in the placebo group had a higher body weight compared with men in the testosterone group (P = 0.04), while measures of total, subcutaneous, and visceral abdominal fat mass as well as percent fat and lean mass were not different between groups. In addition, statistical analyses were adjusted for all baseline differences by putting the respective baseline values as a covariate in the mixed model.

Another important variable, change of antidiabetic medication, was closely monitored and resulted in exclusion. Indeed, of the 13 participants who did not complete the trial, 8 were discontinued because intensification of oral hypoglycemic agents or commencement of insulin therapy, which constituted predefined protocol violations.

Consistent with a previous meta-analysis (33), changes in lipid levels or BP, arguably more important cardiovascular risk factors than glycemia, were minor or nonsignificant in this RCT (Table 3).

An increase in hematocrit was the most frequent adverse effect of testosterone therapy, and a significant proportion (6.6%) of men experienced transient increases above 54%, the recommended threshold for testosterone therapy cessation (28). This marked stimulatory effect of testosterone on erythropoiesis is likely because of the older age of our participants, as testosterone has a more pronounced effect on erythropoiesis in older compared with young men (34). Testosterone therapy increased PSA (Table 3) but was not associated with prostate-associated serious adverse events.

Previous RCTs of testosterone therapy in men with diabetes or the metabolic syndrome, summarized in Supplementary Table 1, yielded inconsistent outcomes (20,35–37). The largest RCT, although unpublished, in 180 men with T2D did not report any significant changes in HOMA-IR or HbA_1c (accessed at http://www.solvaypharmaceuticals.com/static/wma/pdf/1/3/4/4/2/S176.2.101.pdf). However, limited information regarding this unpublished trial is available, and baseline and on-treatment testosterone levels were not reported. A 12-month RCT of 220 men, 62% of whom had T2D, showed a 15% reduction of HOMA-IR at 6 months with no change in total body fat (37). However, this was measured by bioelectrical impedance, which is less accurate than DXA and abdominal MRI scanning used in our study. After 6 months, adjustments to antiglycemic mediations were allowed, and the improvement in HOMA-IR was maintained until study end (37). Testosterone treatment improved HbA_1c at 9 months (P = 0.035), but not at 6 or 12 months. Dropout rates were 29% at 6 months and 46% at 12 months (37). A 30-week RCT of testosterone therapy conducted in 184 men with the metabolic syndrome, 56 of whom had T2D, showed a 31% improvement in HOMA-IR with testosterone treatment, in association with significant weight loss (−4.3 kg), although changes in body composition were not assessed (36). The findings in these RCTs are consistent with the possibility that testosterone treatment may have more favorable effects on glucose metabolism in men without established diabetes.

Strengths of our study, compared with the previous RCTs in this population (Supplementary Table 1), include exclusive focus on men with established T2D, the long duration of testosterone treatment, a low dropout rate, and, importantly, the rigorous enforcement of changes in antidiabetic medications as a criterion for withdrawal from the trial. While, similar to previous RCTs, we did not use clamp studies, we assessed glucose metabolism by two different methods that have reasonably good agreement with clamp studies (17,38,39). Importantly, in contrast to previous trials, we rigorously evaluated changes in body composition and assessed the effects of testosterone treatment on abdominal adipose tissue compartments. Indeed, our observation that testosterone treatment did not decrease visceral adipose tissue or increase adiponectin levels provides novel insights into the paradox that testosterone treatment did not improve glucose metabolism, despite metabolically favorable effects on total fat mass (decrease) and lean mass (increase). Given that visceral adipose tissue and adiponectin contribute to cardiovascular risk, the cardiovascular implications of our findings require further study, especially in the light of recent evidence, although controversial, that testosterone treatment may increase cardiovascular events (40).

In conclusion, in this RCT of men with moderately well-controlled diabetes and modest reductions in testosterone levels typical of the majority of men with T2D, testosterone therapy did not improve glucose metabolism. Indications for testosterone therapy in such men should be no different than men without T2D and be reserved for patients with clinically significant symptoms with persistently low testosterone (28,29). Testosterone therapy should not be routinely given to such men until a favorable risk-to-benefit ratio is confirmed by well-conducted clinical trials. Instead, the first response should focus on the optimization of lifestyle measures and use of established therapies with high-level evidence of outcome benefit.

Funding. Bayer Pharma AG (Berlin, Germany) provided testosterone and placebo. E.J.G. was supported by a National Health and Medical Research Council Postgraduate Fellowship (1018196), a Cardiovascular-Lipid Research Grant from Pfizer Australia (CR05.09), a Royal Australasian College of Physicians/Osteoporosis Australia Research Entry Grant, and a University of Melbourne Postgraduate Scholarship. M.G. was supported by a National Health and Medical Research Council of Australia Career Development Fellowship (1024139).

Duality of Interest. Bayer Pharma AG provided financial support to conduct investigations. No other potential conflicts of interest relevant to this article were reported.

The funder played no role in trial design, data analysis, or in writing the manuscript.

Author Contributions. E.J.G. and M.G. researched data and wrote the manuscript. P.D. and J.M.W. researched data and reviewed and edited the manuscript. R.H. researched data, contributed to the discussion, and reviewed and edited the manuscript. B.J.S. and J.D.Z. contributed to the discussion and reviewed and edited the manuscript. E.J.G. and M.G. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.