Effect of 2 Years of Testosterone Replacement on Insulin Secretion, Insulin Action, Glucose Effectiveness, Hepatic Insulin Clearance, and Postprandial Glucose Turnover in Elderly Men

The study design and methods have been previously described in detail (20). In brief, men whose bioavailable testosterone (non–sex hormone–binding globulin [SHBG]-bound) concentration was <103 ng/dl and DHEA concentration <1.57 μg/ml (15th percentile for young men) were eligible for study and were considered as having relative testosterone deficiency. Volunteer characteristics are given in Table 1. The study was a randomized, placebo-controlled, double-blind, 2-year trial. Subjects were assigned to wear either a testosterone (5 mg/day) (D-TRANS; Alza, Mountain View, CA) or placebo patch each day. Baseline data and results of DHEA replacement have been published previously (3,4,11).

Subjects were admitted and given a standard 10 kcal/kg meal. No additional food was eaten until the next morning. On one occasion, a mixed meal (10 kcal/kg) consisting of scrambled eggs, Canadian bacon, and [1-¹³C]glucose Jell-O was consumed within 15 min (3,4). Infusion of [6-³H]glucose was started with the meal, and infusion of both [6-³H]glucose and [6,6-²H₂]glucose were altered, minimizing the change in plasma glucose enrichment (22). On another occasion, 0.33 g/kg glucose containing [6,6²H₂]glucose was injected at time 0 min, and 0.02 units/kg insulin was injected at time 20 min (4). Arterialized venous blood was collected at frequent intervals, as previously described (3,4).

Plasma glucose concentrations were measured using a glucose oxidase method (YSI, Yellow Springs, OH). Plasma insulin concentrations were measured by chemiluminescence (Access Assay; Beckman, Chaska, MN). Plasma C-peptide concentrations were measured by radioimmunoassay (Linco Research, St. Louis, MO). Levels of sulfated DHEA and total bioavailable testosterone and estrogen were measured by a competitive chemiluminescene immunoassay, and SHBG was measured by solid-phase chemiluminescene assay (Immulite; Diagnostic Products, Los Angeles, CA). In subjects with low testosterone levels, values were obtained with the use of a high-sensitivity chemiluminescence assay. Body composition was measured using dual-energy X-ray absorptiometry (DPX scanner; Lunar, Madison, WI).

The oral and intravenous glucose minimal models (23,24) were used to interpret the plasma glucose and insulin concentrations measured following meal ingestion or glucose injection (supplemental Fig. 3 found in an online appendix at http://dx.doi.org/10.2337/dc07-0359).

The systemic rates of meal appearance (R_aMEAL), endogenous glucose production, and glucose disappearance (R_d) were calculated using Radziuk's two-compartment model (25). Values from −30 to 0 min were averaged and considered as basal. Area above basal was calculated using the trapezoidal rule. Parameters of all models were estimated using SAAM II software (26). Measurement errors have been assumed to be independent and Gaussian, with 0 mean and variance for glucose and tracer glucose (as previously described [27]) and for C-peptide concentrations (as previously described [28]).

Data are presented as means ± SE. Area above basal was calculated using the trapezoidal rule. Changes from baseline (i.e., 24-month values minus baseline values) were compared in the testosterone and placebo groups using Student's t test. Demographic, hormonal, and biochemical data are presented as median and interquartile range. Multiple regression analysis was done in which the dependent variable was the change from baseline (with the use of a rank transformation), and the independent variables were study group, sex, age at the time of randomization, length of follow-up, and baseline values. A P value <0.05 was considered to be statistically significant.

The effects of testosterone replacement on plasma hormone concentrations and body composition have been described in detail elsewhere (20). In brief, plasma testosterone concentration was significantly higher during 2 years of treatment with testosterone versus placebo (P < 0.005) (Table 1). In contrast, plasma testosterone concentrations did not change following 2 years of treatment with placebo. Estrogen concentrations were not different following treatment with either testosterone or placebo. Two years of treatment with testosterone did not alter visceral or percentage body fat; however, it resulted in a slight but statistically significant increase (P < 0.01) in fat-free mass. Even so, treatment with testosterone did not alter peak Vo₂, leg isometric knee extension, double leg press, or chest press.

The change (i.e., 24 months minus baseline) of fasting glucose, insulin, and C-peptide concentrations present on the day of the mixed-meal study did not differ following 2 years of treatment with testosterone or placebo (Fig. 1). The change in the postprandial increments (i.e., area above basal) of glucose, insulin, and C-peptide concentrations also did not differ following treatment with testosterone or placebo.

The change in fasting rates of endogenous glucose production and glucose disappearance did not differ following 2 years of treatment with testosterone from that observed following 2 years of placebo (Fig. 2). The change in the postprandial increment in meal appearance and glucose disappearance or postprandial decrement in endogenous glucose production also did not differ following 2 years of treatment with testosterone or placebo.

The change in fasting glucose, insulin, and C-peptide concentrations present on the day of the intravenous glucose tolerance test did not differ in the testosterone and placebo groups (supplemental Fig. 3). Similarly, the change in glucose, insulin, and C-peptide concentrations observed following intravenous injection of glucose did not differ following 2 years of treatment with testosterone or placebo.

The change from baseline in net insulin action (S_i) measured with either the unlabeled oral or unlabeled “oral” or “intravenous” glucose minimal models did not differ following 2 years of treatment with either testosterone or placebo. The change in the ability of insulin to stimulate glucose uptake (S_i*) measured with either the labeled oral or labeled intravenous minimal models also did not differ in the testosterone and placebo groups (Table 2).

The change from baseline in meal indexes (Phi_dynamic, Phi_static, and Phi_total) intravenous glucose indexes (Phi₁, Phi₂, and Phi_total) of insulin secretion did not differ following 2 years of treatment with testosterone or placebo (Table 2). This resulted in no difference in the change from baseline in either the meal (DI_dynamic, DI_static, and DI_total) or intravenous glucose (DI₁, DI₂, and DI_total) disposition indexes. The change in net glucose effectiveness (GE), the ability of glucose to stimulate its own uptake (GE*), and hepatic insulin clearance measured following intravenous glucose injection also did not differ in the testosterone and placebo groups.

The present studies indicate that 2 years of testosterone replacement has no detectible effect on insulin secretion, insulin action, glucose effectiveness, hepatic insulin clearance, or postprandial glucose turnover in elderly men. These data strongly imply that relative or absolute testosterone deficiency does not contribute to insulin resistance or the impairment in insulin secretion that is commonly observed in elderly men. It also exemplifies that testosterone replacement does not delay or reverse age-associated deterioration in glucose tolerance.

We have previously reported that postprandial glucose disposal is lower in elderly men than in young men (3). The present data indicate that this defect is not reversed by testosterone replacement, since rates of postprandial glucose disposal were virtually identical before and after 2 years of treatment with testosterone and did not differ from those following placebo. The increment in glucose concentrations following a mixed meal is also influenced by the rate at which the ingested glucose enters the systemic circulation and by the rate at which endogenous glucose production is suppressed. Neither differed from those observed in young men or were influenced by testosterone replacement. Thus, testosterone replacement did not improve glucose tolerance in elderly men and did not restore postprandial rates of glucose disappearance to those observed in healthy younger subjects.

Elderly men commonly have low testosterone concentrations and are insulin resistant (5,8–11). Testosterone receptors are present on muscle (29), and some (6,13,14), but not all (15,16,19), studies have suggested that treatment with androgens can improve insulin action in rats (6) or middle-aged humans (13,14). We have previously reported that testosterone replacement in elderly men does not improve net insulin action measured with the unlabeled oral minimal model (20). The present data extend these observations by showing that testosterone replacement also had no effect on net insulin action (S_i) measured with the unlabeled intravenous glucose tolerance test or any effect on insulin measured with either the labeled oral or intravenous-labeled minimal models. The latter observation is particularly noteworthy since the labeled minimal model specifically measures the ability of insulin to stimulate glucose disposal (S_i*). Furthermore, the oral and intravenous glucose tolerance tests were performed on two separate occasions, and insulin action measured with the unlabeled and labeled minimal models provided independent assessments of insulin action. Therefore, whereas it is possible that an effect of testosterone replacement on insulin action was missed, it must have been small and likely of limited biologic significance. Glucose uptake is regulated by glucose and insulin concentrations (30,31). The former is commonly referred to as glucose effectiveness. Net glucose effectiveness and the ability of glucose to stimulate its own uptake can be measured with the unlabeled and labeled minimal models, respectively.

Experiments in animals suggest that testosterone can regulate islet cell function (7). We have previously reported that insulin secretion is impaired in elderly men (3). The present data indicate that treatment with testosterone does not ameliorate this impairment. Insulin secretion whether evaluated qualitatively by comparing the area above basal of insulin and C-peptide concentrations after meal ingestion or glucose injection or quantitatively by using C-peptide models did not differ following treatment with testosterone or placebo. Since there were virtually no changes over time in insulin or C-peptide concentrations, these data also indicate that hepatic insulin clearance did not change. The oral and intravenous glucose tolerance tests evaluate different aspects of insulin secretion (32). The former assesses glucose- and nutrient-induced stimulation of insulin secretion, whereas the latter evaluates the response to glucose alone. Similarly, the intravenous Phi₁ is believed to reflect the acute release of insulin from previously docked insulin granules that occurs during the few minutes following glucose injection. In contrast, Phi_dynamic measures the increment in insulin secretion that occurs in response to the progressive increase in glucose concentrations observed during 30–60 min following meal ingestion. Therefore, it presumably is modulated by additional intra-islet events, which include translocation, priming, and docking of insulin granules. In contrast, the intravenous Phi₂ and the meal Phi_static are believed to be influenced by the still earlier events in the insulin secretory cascade including insulin synthesis and processing. Taken together, the present data indicate that testosterone replacement had no detectible effect on any aspect of β-cell function, whether considered as the absolute values of the various insulin secretion indexes or in light of the prevailing level of insulin action by calculation of disposition indexes.

The present study has certain limitations. We recruited men whose bioavailable testosterone concentration (not bound to SHBG) was less than the 15th percentile of healthy young men (33). However, none of the men had undetectable levels. Therefore, it is possible and even likely that testosterone treatment would improve carbohydrate tolerance in men who are overtly hypogonadal. The BMI of the subjects averaged ∼28 kg/m². Therefore, few of the subjects were overtly obese. Of interest, the effects of testosterone on insulin action (S_i and S_i*) and insulin secretion (DIs) did not differ when the subjects whose BMI or bioavailable testosterone in the upper tertile was compared with those in the lower tertile. Due to concern regarding the potential adverse effects of long-term treatment of elderly men with androgens, we sought to raise bioavailable testosterone concentrations to the lower range of normal for healthy young men. While we accomplished this goal, it is possible that larger amounts of testosterone would have altered glucose tolerance. However, use of high-dose androgens has multiple safety concerns and may decrease rather than increase insulin action (18,19). It is possible that 2 years of treatment with androgens was insufficient time to alter carbohydrate metabolism. We doubt this is the case since when such effects have been observed in animals or younger humans, they were detectible within days to weeks (7,13,14).

In summary, the present studies indicate that treatment of elderly men with relative testosterone deficiency for 2 years with doses sufficient to raise their bioavailable testosterone concentrations to the low normal range for healthy young men had no detectible effect on fasting or postprandial glucose concentrations. Testosterone replacement also did not alter meal appearance, postprandial suppression of endogenous glucose production, or stimulation of glucose uptake. Furthermore, testosterone replacement did not alter insulin secretion, insulin action, glucose effectiveness, or hepatic insulin clearance. Thus, testosterone deficiency is unlikely to be the cause of the age-associated deterioration in glucose tolerance commonly observed in elderly men. These data also argue against the premise that testosterone replacement will delay or prevent the progression of the age-associated deterioration in glucose tolerance that is commonly observed in elderly men.

This study was supported by grants from the National Institutes of Health (P01 AG-14283, RR00585, and DK29953) and the Mayo Clinic. R.A.R. is supported by the Earl and Annette McDonough Professorship.

We thank R. Rood, B. Dicke, L. Heins, J. Feehan, B. Norby, P. Hellwig, T. Hammer, and L. Wahlstrom for technical assistance and assistance in recruiting the subjects; R. Rood for assistance with graphics; M. Davis for assistance in the preparation of the manuscript; and the staff of the Mayo General Clinical Research Center for assistance in performing the studies. We also thank our coinvestigators on the program project, including Drs. Peter O'Brien and Donald Tindall, for thoughtful comments and suggestions.