| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Neuroendocrine Unit and Program in Nutritional Metabolism, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114
Address all correspondence and requests for reprints to: Dr. Steven Grinspoon, Program in Nutritional Metabolism, Longfellow 207, Massachusetts General Hospital, 55 Fruit Street, Boston, Massachusetts 02114. E-mail: sgrinspoon{at}partners.org.
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The subjects described represent a subgroup of patients recruited for a previously described protocol, in which we assessed testosterone effects over 6 months in a randomized, placebo-controlled study. Thirty-one HIV-positive female subjects [median (interquartile range) age, 41 (35, 43) yr; range, 18–45 yr] with wasting [weight of <90% ideal body weight (IBW) or >10% weight loss from pre-illness weight, or weight of <100% IBW with a weight loss of 
5% of the pre-illness weight] and decreased screening free testosterone levels [measured by equilibrium dialysis, <3.0 pg/ml (10.4 pmol/liter) (median of the Endocrine Science Reference Range)] were enrolled in a randomized, placebo-controlled study of testosterone administration. Patients were recruited between 1998 and 2001 from the Massachusetts General Hospital Multidisciplinary HIV Clinic and from community-based practices and through newspaper advertisements. Inclusion was based on documented HIV infection, by Western blot or ELISA, and was not limited based on CD4 count. Subjects were eligible only if an acceptable form of birth control was being used during the study, such as barrier contraception or intrauterine device, but excluding oral contraceptives or Depo-Provera. Subjects who were pregnant, actively seeking pregnancy, or breastfeeding were excluded from the study. Active substance abuse and alcoholism were also criteria for exclusion. Subjects requiring parenteral nutrition, pharmacologic glucocorticoid therapy, or any androgen, estrogen, progestational derivative, or megestrol acetate within 3 months of the study were excluded. Subjects with any new opportunistic infection diagnosed within 4 wk of the study or intractable diarrhea (defined as >6 stools/d) were excluded. In addition, patients receiving antiretroviral agents, including protease inhibitors, were required to be on a stable regimen for at least 6 wk before study entry. Other exclusion criteria included clinically significant liver disease (or serum glutamic oxaloacetic transaminase of >5x normal), renal disease [or creatinine of >2.0 mg/dl (176.8 µmol/liter)] or a hemoglobin of less than 8.0 g/dl (80 g/liter). All subjects gave written informed consent and the study was approved by the Human Research Committee of the Massachusetts General Hospital and the Committee on the Research of Human Subjects at the Massachusetts Institute of Technology. Eligibility was determined at the screening visit, at which free testosterone, serum glutamic oxaloacetic transaminase, urine human chorionic gonadotropin, weight, and weight history were determined.
Study design
Eligible patients were randomly assigned to receive either active testosterone transdermal delivery systems (Testosterone TTS; 4.1 mg/patch; estimated delivery rate, 150 µg/d; Watson Pharmaceuticals, Salt Lake City, UT) twice a week or identical placebo patches provided by the manufacturer. Randomization was stratified by weight of less than 90% or more than or equal to 90% IBW, in blocks of four, based on randomly generated numbers. The Massachusetts General Hospital Pharmacy performed randomization. All investigators and patients were blinded to drug assignment. Each patient underwent a directed history for testosterone-dependent effects, including symptoms of virilization and libido, and physical examination and evaluation for signs of virilization and hirsutism.
For women with regular menses, the baseline evaluation occurred within the early follicular phase of the menstrual cycle (d 1–7). Fasting blood was obtained at approximately 0830 h for testosterone, free testosterone, estradiol, and IGF-I. Weight was measured and body mass index (BMI) was determined. Caloric intake was determined by food record. Subjects underwent frequent sampling overnight to determine GH pulsatility. Subjects returned for monthly visits, which included physical examination, urine pregnancy test, fasting 0830-h serum total and free testosterone, and weight. Time of last testosterone application was recorded at each visit. Subjects recorded menses in a monthly menstrual diary. A new month’s supply of study drug was given out only after documentation of a negative pregnancy test and used patches were collected at each visit. Subjects returned at 6 months for assessment of weight, body composition, and hormone levels, including frequent sampling for GH overnight, in identical fashion to that obtained at baseline. Study drug compliance was confirmed by history and a used-patch count at each monthly visit.
GH pulse and deconvolution analysis: PULSE and CLUSTER programs
Subjects underwent frequent GH sampling performed every 20 min from 2200–0800 h at the baseline visit and at 6 months. To assess GH pulsatility, we used CLUSTER, a largely model-free computerized pulse analysis algorithm to identify statistically significant pulses in relation to dose-dependent measurement error in each hormone time series (4). In performing the analysis, we specified individual test cluster sizes for the nadir and peak width of 2 (2 x 2), a minimum and maximum intraseries coefficient of variation (CV), a t statistic to identify significant increase, and a t statistic to define a significant decrease (5). A CV of 4.4%, the intraassay CV for our GH assay, was used in the settings of the program. Information about the secretion of the hormone into the serum and the elimination of the hormone from the serum was obtained from PULSE 2 and PULSE 4 deconvolution and pulse detection algorithms.
Biochemical and immunological assays
All samples from the same patient were run in duplicate in the same assay. The free testosterone concentration was determined as the product of the percent free testosterone, measured by equilibrium dialysis, and the total testosterone concentration (Endocrine Sciences, Calabasas Hills, CA). The intraassay CV of free testosterone is 6.9%, and the intraassay CV for total testosterone is less than 8.1%. The intraassay CV were developed using pooled sera covering the range of the assay. The normal range for total testosterone is 10–55 ng/dl (0.35–1.91 nmol/liter) and for free testosterone is 1.1–6.3 pg/ml (3.8–21.8 pmol/liter) in adult females. The interassay CV for testosterone is 8–15% and for free testosterone, 8.9–11.9%. The sensitivity of the total testosterone assay is 3 ng/dl (0.10 nmol/liter). The sensitivity of the determination of percent free testosterone by this method is 0.1%. RIAs were performed for LH (intraassay CV, 2.6%; interassay CV, 4.5–5.4%) (Nichols Institute, San Juan Capistrano, CA) and FSH (intraassay CV, 1.6–2.3%; interassay CV, 3.2–3.8%) (Nichols Institute). SHBG was performed by immunoradiometric assay with an intraassay CV of less than 4% and an interassay CV of 7.8–10.6% (Endocrine Sciences). Serum estradiol was measured by RIA kit (Diagnostic Systems Laboratories, Inc., Webster, TX) with an intraassay CV of 6.5–8.9%. Serum IGF-I was measured by RIA kit (Diagnostics Systems Laboratories, Inc.) with an intraassay CV of 3.9–7%. GH levels were assessed by a single measurement at each time point during the overnight frequent sampling in each subject. Serum GH levels were run in duplicate using a two-site radioimmunometric assay with an intraassay CV average of 1.6–8.1% (Nichols Institute Diagnostics). The sensitivity of the assay was determined to be 0.01, based on serial dilutions.
The CD4 count was determined by flow cytometry (BD Immunocytochemistry Systems, San Jose, CA), and the HIV viral load was determined by ultrasensitive assay (Amplicor HIV-1 Monitor Assay; Roche Molecular Systems, Indianapolis, IN), with limits of detection of 50–75,000 copies/ml.
Caloric intake
Subjects were instructed to complete a 4-d food record that was analyzed for total calorie content (Minnesota Nutrition Data Systems, version 8A/2.6; Minneapolis, MN).
Statistical analysis
Baseline comparisons were made by the Wilcoxon test between the groups. Categorical variables were compared by 
2 analysis. Treatment effect was determined by comparison of change between baseline and 6 months using analysis of covariance (ANCOVA). The measurement obtained at 6 months was the outcome variable, treatment assignment was the main effect, and baseline measurement of the variable was used as a covariate. The factors associated with baseline IGF-I, 6-month IGF-I, and the change in IGF-I in response to testosterone were determined in separate multivariate regression analyses. Outlier analysis was performed using the criteria of Dixon and Massey (6). Statistical analyses were performed using SAS JMP software (SAS Institute, Inc., Cary, NC). Statistical significance was defined as a two-tailed 
value of P 
0.05. Results are median (interquartile range) unless otherwise indicated. The Data Safety Monitoring Board met every 3 months to review any adverse events occurring in the study.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The baseline clinical characteristics of the study subjects are shown in Tables 1
and 2. Baseline clinical characteristics including age, BMI, CD4 count, and viral load were not significantly different between the patients randomized to testosterone or placebo.
|
|
Effects of testosterone administration
Hormonal indices, immune function, IGF-I, and GH pulse parameters were not significantly different between the groups at baseline (Tables 2
and 3). Total and free testosterone increased significantly in the testosterone-treated compared with control subjects (Table 2). Estradiol levels tended to increase more in the testosterone- compared with the placebo-treated patient (Table 2). No evidence of virilization or androgen-related side effects were observed. Viral load, CD4, and weight did not change significantly between the groups (Table 2). Antiretroviral treatment remained stable in both groups.
|
; Fig. 1] [4.8 (–0.5, 9.6) vs. –3.9 (–12.8, 5.1) nmol/liter], and this result remained significant, controlling for changes in body composition including BMI, VAT, and lean mass (data not shown). Similar effects on IGF-I were observed in an analysis of the larger cohort, for whom GH pulsatility was not measured (3). GH pulse frequency increased significantly in the testosterone-treated compared with the placebo-treated subjects [1.0 (1.0, 2.0) vs. 0.0 (–0.5, 1.5) peaks per 12 h, testosterone vs. placebo; P = 0.02] (Table 3). In contrast, the average valley mean [–0.1 (–0.5, 0.3) vs. –0.0 (–0.2, 0.8) ng/ml, P = 0.07] [–0.1 (–0.5, 0.3) vs. –0.0 (–0.2, 0.8) µg/liter] tended to decrease in the testosterone-treated patients compared with the placebo-treated patients (Table 3). The changes in pulse amplitude, interpulse interval, and mean GH were not significant between the groups. IGF binding protein (IGFBP)-1 tended to decrease more in the testosterone-treated compared with the placebo-treated group, whereas no effect of testosterone on IGFBP-3 was seen (Table 3).
|
At baseline, before testosterone administration, GH pulse amplitude was significantly related to IGF-I in univariate and multivariate regression analysis. Neither free testosterone, estradiol, nor BMI were significant predictors of baseline IGF-I (Table 4). In a second analysis, substituting mean GH as a surrogate for overall GH secretion, neither free testosterone, mean GH, nor estradiol were significant in predicting IGF-I (Table 5). Controlling for VAT and lean body mass did not affect the models (data not shown).
|
|
The change in IGF-I was significantly related to the change in free testosterone in univariate regression analysis (Fig. 2) and multivariate regression analysis. In contrast, change in estradiol was not related to change in IGF-I in either univariate or multivariate regression analysis (Fig. 2). There was no relationship between absolute levels of free testosterone, estradiol, and IGF-I at 6 months, in contrast to the data relating changes in response to testosterone (data not shown). In multivariate regression analysis, there was a 19 ng/ml (2.5 nmol/liter) increase in IGF-I for each 1.0 pg/ml (3.5 pmol/liter) unit increase in free testosterone, controlling for estradiol, BMI, and GH pulse parameters. The model explained 64% of the variability in IGF-I (Table 6). In a second analysis, substituting in mean GH as a surrogate for overall GH secretion for individual pulse parameters, the free testosterone remained a significant predictor of IGF-I (Table 7). Controlling for VAT and lean body mass did not affect the models (data not shown).
|
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
At baseline, before testosterone administration, mean GH from frequent sampling was inversely correlated with BMI and visceral fat. Prior studies in normal weight and viscerally obese non-HIV-infected women also show significant relationships between GH and body composition parameters, including weight and visceral fat (8, 9). In contrast, IGF-I was not related to body composition, but was related to GH pulse amplitude in multivariate regression analysis before testosterone administration in our study. Although body composition affected overall GH secretion, basal IGF-I was largely determined by GH pulse amplitude, controlling for sex steroid levels, weight, and other GH pulse parameters.
With testosterone administration, IGF-I increased significantly in this randomized, placebo-controlled study. In addition, we show a significant effect of testosterone on lean body mass, but not on visceral fat or gross functional measures such as the 6-min walk in this substudy. In contrast to IGF-I, significant changes in IGFBP-1 and IGFBP-3 were not seen, although there was a trend for a greater decrease in IGFBP-1 in response to testosterone compared with placebo.
Prior cross-sectional studies suggest gender and significant sex steroid influences on the GH-IGF-I axis. Van Den Berg et al. (10) used cluster analysis to reveal significantly higher mean serum GH concentrations over 24 h, increased GH peak amplitude, longer peak duration, and increased mean area under the GH concentration peak in premenopausal women compared with men. In contrast, GH peak frequency was similar in women and men. Similar findings were demonstrated by Ho et al. (11) in which 24-h integrated GH concentration was significantly greater in healthy young women compared with men. However, mean pulse frequency was not affected by gender. Serum concentration of free estradiol, not free testosterone, correlated with 24-h integrated GH concentration, pulse amplitude, and fraction of GH secreted in pulses during the 24-h period.
A limited number of studies have also investigated the effects of androgen administration on GH secretion. These prior studies have shown different results depending on gender, age, and pubertal status of the subjects; the type, dose, and route of androgen administration; and degree of aromatization. Azziz et al. (12) investigated seven oophorectomized women who each received 3 wk of testosterone propionate im (20 mg, three times a week) and conjugated oral estrogen (0.625 or 1.25 mg). Azziz et al. (12) demonstrated a significant increase in IGF-I, from 55 ± 23 to 124 ± 37 ng/ml (7.2 ± 3.0 to 16.2 ± 4.8 nmol/liter) and a significant decrease in fasting GH over 4 wk, suggesting potential direct effects of testosterone on IGF-I synthesis. In our study, we show similar effects of transdermal testosterone without simultaneous estrogen or progesterone administration on IGF-I.
Studies in prepubertal boys have shown that integrated GH as well as IGF-I increase significantly with testosterone treatment (13, 14, 15, 16, 17, 18, 19, 20). A few of these studies show that nonaromatizable androgens do not enhance GH secretion in the same way that testosterone does (16, 19, 20), suggesting that the aromatization of testosterone to estrogen is the proximal stimulus that amplifies GH secretion in puberty. In girls with Turner’s syndrome, treatment with low-dose estrogen therapy has been shown to result in significant increases in IGF-I levels (21) as well as GH concentration. In contrast, an earlier study in short, normal prepubertal children receiving low-dose estrogen, demonstrated increased GH secretion without a change in IGF-I levels (22).
In men with hypogonadism, treatment with long-term testosterone results in increased mean 24-h GH levels, GH pulse amplitude, area under the curve, and IGF-I level compared with pretreatment values, without a change in GH pulse frequency (23). In contrast, estrogen blockade prevents many of the effects of testosterone on the GH-IGF-I axis, indicating that the testosterone effect is at least partly due to aromatization to estradiol (24). In our study, there was a nonsignificant trend toward increased estradiol in response to testosterone administration, presumably from aromatization of the transdermal testosterone, and we controlled for estradiol in all analyses. In women with hypothalamic amenorrhea, the role of estrogens on GH secretion was demonstrated by Genazzani et al. (25) who showed that administration of estradiol through transdermal patch decreased GH pulse frequency, but increased integrated GH concentration and pulse amplitude significantly. In contrast, studies of oral estrogen administration demonstrate decreased IGF-I and increased GH in postmenopausal women (26, 27, 28). We controlled for changes in estradiol and GH pulse parameters in regression modeling, and only the change in free testosterone, not estradiol, contributed to the change in IGF-I.
GH and IGF-I levels may also change throughout the menstrual cycle relating to changes in endogenous steroid levels. In normally menstruating women, serum GH concentration in late follicular phase is higher than in early follicular phase. Progesterone is thought to blunt this estrogen-associated effect, resulting in an intermediate GH concentration during the midluteal phase (29). Ovesen et al. (30) demonstrated increased GH secretory bursts, mean integrated 24-h GH concentration and IGF-I, but normal GH secretion rate, GH pulse mass, and amplitude in the periovulatory phase compared with the early follicular phase. In our study, subjects were studied in the early follicular phase to minimize changes associated with the menstrual cycle on testosterone levels.
In contrast to the baseline data demonstrating that IGF-I was related to GH pulse amplitude, the change in IGF-I in response to testosterone administration was strongly related to the change in free testosterone concentration in our study subjects. Free testosterone levels achieved at the end of the study were within the physiologic range, in contrast to other studies using higher dose, oral, or injectable testosterone. A number of potential mechanisms might contribute to changes in IGF-I in our specific study population, which was low weight and relatively androgen deficient. First, testosterone may directly increase IGF-I, as suggested in prior studies of men and women without HIV disease (10, 11, 12, 13, 14, 17, 18, 19, 23, 24, 31). In support of this potential mechanism is the observation of a significant relationship between the changes in testosterone and IGF-I in univariate and multivariate regression modeling, controlling for changes in GH pulse parameters.
A second possibility is that testosterone may increase IGF-I through effects on the GH axis. In this regard, we show that testosterone increases GH pulsatility, but, at the same time, tends to decrease other pulse parameters, including average valley mean. In this regard, our data are similar to those of Azziz et al. (12), showing stimulatory effects of testosterone on IGF-I and inhibitory effects on specific GH pulse parameters. It is possible that a complex interaction occurs on the GH-IGF-I axis with stimulatory effects on pulsatility and negative effects on other parameters due to feedback inhibition of increased IGF-I. Indeed, the relationship between change in IGF-I and mean GH was negative in multivariate regression analysis, suggesting that the increase in IGF-I in response to testosterone exhibited feedback inhibition on overall mean GH, despite a stimulatory effect of testosterone on GH peaks. Further studies will be necessary to determine the complex effects of testosterone on the GH-IGF-I axis in HIV-infected women.
A third possibility is that IGF-I increased due to nutritional effects of testosterone. In AIDS wasting and other conditions of chronic undernutrition, IGF-I is decreased and GH is increased due to resistance to the action of GH on IGF-I transcription in the liver (32). Reversal of undernutrition could therefore result in increased IGF-I and decreased GH. Over the 6 months of this study, testosterone did not increase weight or nutritional intake. Therefore, it is not likely that major nutritional changes were occurring in association with testosterone. Furthermore, we controlled for BMI in all models, and free testosterone remained significantly related to IGF-I. Our data argue against a nutritional effect of testosterone on the GH-IGF-I axis. However, our subjects were low weight at baseline and changes in GH in response to testosterone seen in this study may not be directly applicable to normal-weight women.
A fourth possibility is that changes in estrogen due to aromatization from testosterone administration might affect the GH-IGF-I axis. In this study, we saw a nonsignificant trend toward increased estradiol resulting from testosterone administration. However, estradiol was not related to the changes in IGF-I in multivariate modeling, suggesting that other factors related to testosterone administration affected IGF-I. Subjects in this study were low weight and may not experience as much aromatization as in more obese subjects or in response to other forms of testosterone administration. Aromatization to estradiol might play a greater role in changes in the GH-IGF-I axis resulting from testosterone administration in other populations.
In conclusion, physiologic testosterone administration in relatively androgen-deficient HIV-infected women resulted in a significant increase in IGF-I and mixed effects on GH pulse parameters (i.e. stimulatory effects on GH pulsatility and inhibitory effects on GH interpulse mean). Our data suggest direct effects of testosterone on IGF-I, independent of the effects on the GH axis. Increased IGF-I may be an important benefit and may contribute to increased muscle function in response to testosterone in this population (3). Further studies are needed to investigate the effects of physiologic, natural testosterone on the GH-IGF-I axis in HIV-infected women and other populations of androgen-deficient women.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Abbreviations: ANCOVA, Analysis of covariance; BMI, body mass index; CV, coefficient of variation; IBW, ideal body weight; IGFBP, IGF binding protein; VAT, visceral adipose tissue.
Received December 10, 2003.
Accepted March 18, 2004.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
