Bhasin S, Storer TW, Javanbakht M, Berman N, Yarasheski KE, Phillips J, Dike M, Sinha-Hikim I, Shen R, Hays RD, Beall G. Testosterone Replacement and Resistance Exercise in HIV-Infected Men With Weight Loss and Low Testosterone Levels. JAMA. 2000;283(6):763-770. doi:10.1001/jama.283.6.763
Author Affiliations: Division of Endocrinology, Metabolism, and Molecular Medicine, Charles R. Drew University of Medicine and Science (Drs Bhasin, Dike, Sinha-Hikim, and Shen and Ms Javanbakht), and Department of General Internal Medicine and Health Service Research (Dr Hays), University of California, Los Angeles; Laboratory for Exercise Sciences, El Camino College (Dr Storer), and Departments of Pediatrics (Dr Berman), Radiology (Dr Phillips), and Medicine (Dr Beall), Harbor-University of California, Los Angeles Medical Center, Torrance; and Division of Metabolism, Endocrinology, and Diabetes, Washington University Medical School, St Louis, Mo (Dr Yarasheski).
Context Previous studies of testosterone supplementation in HIV-infected men
failed to demonstrate improvement in muscle strength. The effects of resistance
exercise combined with testosterone supplementation in HIV-infected men are
Objective To determine the effects of testosterone replacement with and without
resistance exercise on muscle strength and body composition in HIV-infected
men with low testosterone levels and weight loss.
Design and Setting Placebo-controlled, double-blind, randomized clinical trial conducted
from September 1995 to July 1998 at a general clinical research center.
Participants Sixty-one HIV-infected men aged 18 to 50 years with serum testosterone
levels of less than 12.1 nmol/L (349 ng/dL) and weight loss of 5% or more
in the previous 6 months, 49 of whom completed the study.
Interventions Participants were randomly assigned to 1 of 4 groups: placebo, no exercise
(n = 14); testosterone enanthate (100 mg/wk intramuscularly), no exercise
(n = 17); placebo and exercise (n = 15); or testosterone and exercise (n =
15). Treatment duration was 16 weeks.
Main Outcome Measures Changes in muscle strength, body weight, thigh muscle volume, and lean
body mass compared among the 4 treatment groups.
Results Body weight increased significantly by 2.6 kg (P<.001)
in men receiving testosterone alone and by 2.2 kg (P
= .02) in men who exercised alone but did not change in men receiving placebo
alone (−0.5 kg; P = .55) or testosterone and
exercise (0.7 kg; P = .08). Men treated with testosterone
alone, exercise alone, or both experienced significant increases in maximum
voluntary muscle strength in leg press (range, 22%-30%), leg curls (range,
18%-36%), bench press (range, 19%-33%), and latissimus pulls (range, 17%-33%).
Gains in strength in all exercise categories were greater in men assigned
to the testosterone-exercise group or to the exercise-alone group than in
those assigned to the placebo-alone group. There was a greater increase in
thigh muscle volume in men receiving testosterone alone (mean change, 40
cm3; P<.001 vs zero change)
or exercise alone (62 cm; P = .003) than in men receiving placebo alone (5
cm3; P = .70). Average lean body mass increased by 2.3 kg (P = .004) and 2.6 kg (P<.001),
respectively, in men who received testosterone alone or testosterone and exercise
but did not change in men receiving placebo alone (0.9 kg; P = .21).Hemoglobin levels increased in men receiving testosterone
but not in those receiving placebo.
Conclusion Our data suggest that testosterone and resistance exercise promote gains
in body weight, muscle mass, muscle strength, and lean body mass in HIV-infected
men with weight loss and low testosterone levels. Testosterone and exercise
together did not produce greater gains than either intervention alone.
Weight loss during the course of human immunodeficiency virus (HIV)
infection is associated with increased mortality and adverse disease outcome.1- 4 Of the
various therapies being considered for the treatment of HIV-associated weight
loss, testosterone and exercise are attractive because they are relatively
inexpensive and safe. There is a high prevalence of low testosterone levels
in HIV-infected men.4- 11
Serum testosterone levels are lower in those with weight loss9
and correlate with deficits in muscle mass,11
low Karnofsky scores,12 and disease progression.10,13 Replacement doses of testosterone
augment lean body mass and muscle strength in healthy, hypogonadal men.14- 16 These data have led
to the hypothesis that testosterone replacement might also increase muscle
mass and strength in HIV-infected men with low testosterone levels. Some of
the studies17- 29
that have examined the effects of androgen administration on weight and body
composition in HIV-infected men were not placebo-controlled,23- 27
and most failed to control energy intake and exercise stimulus.23- 28
Two of the 4 placebo-controlled studies of testosterone supplementation of
HIV-infected men17,18 reported
gains in fat-free mass (FFM), while others19,20
found no change. None of the previous androgen studies in HIV-infected men
has demonstrated improvements in muscle strength.17- 20,23- 28
The effects of resistance exercise alone and in combination with testosterone
in HIV-infected men are unknown.
The objective of this study was to determine the effects of testosterone
replacement, with or without a program of resistance exercise, on muscle strength
and body composition in HIV-infected men with weight loss and low testosterone
levels. We hypothesized that testosterone and resistance exercise would each
increase muscle strength and FFM, and the 2 interventions, when administered
together, would produce greater gains in these measures than either intervention
This was a 16-week, double blind, placebo-controlled, randomized study,
conducted between September 1995 and July 1998. Written informed consent,
approved by the institutional review boards of Charles R. Drew University
of Medicine and Science, Los Angeles, and Research and Education Institute,
Torrance, Calif, was obtained from each volunteer.
The participants were HIV-infected men, between the ages of 18 and 50
years, with involuntary weight loss of at least 5% in the preceding 6 months
by self-report or documented from medical records and had serum total testosterone
levels less than 12.1 nmol/L (349 ng/dL). The subjects had been receiving
stable antiretroviral therapy for at least 12 weeks before enrollment. We
excluded patients with acute illness, prostate cancer, hypogonadism from a
known cause, or diarrhea. Patients with significant cardiovascular or respiratory
disease, diabetes, drug abuse, or heavy alcohol use in the last 6 months,
aspartate aminotransferase or alanine aminotransferase greater than 3 times
the upper limit of normal, bilirubin levels higher than 34.2 µmol/L
(2 mg/dL), and prostate-specific antigen (PSA) levels higher than 4 ng/mL
were excluded. The patients were excluded if they had received megestrol acetate,
androgen agonists or antagonists, growth hormone, or insulinlike growth factor
1 within 3 months of enrollment.
The sample size estimate was based on a previous study of testosterone
replacement in healthy, hypogonadal men14 in
which administration of 100 mg of testosterone enanthate weekly was associated
with mean (SE) FFM increase of 5.0 (0.7) kg and 22% (3%) increase in maximum
voluntary strength. Assuming that changes in HIV-infected men with low testosterone
levels would be similar to those in healthy hypogonadal men, we determined
that 12 subjects per group would give us 80% or more power to detect a similar
change in FFM and muscle strength. We assumed a 20% to 25% dropout rate and
therefore planned to enroll 60 men in the study.
Using randomization schedules, generated to create random numbers from
a uniform distribution on unit interval, 61 eligible men were assigned to
1 of 4 groups. Group 1 received placebo injections (sesame oil) but did not
exercise; group 2 received 100 mg/wk of testosterone enanthate but did not
exercise; group 3 received placebo and participated in a resistance exercise
program; and, group 4 received the testosterone treatment and participated
in an exercise program. Subjects were stratified by age (19-40 years and 41-60
years). A blocking size of 16 was used.
Testosterone replacement consisted of weekly intramuscular injections
of 100 mg of testosterone enanthate, administered in the General Clinical
Research Center, Torrance, Calif, to ensure compliance.
The exercise intensity was standardized based on initial 1-repetition
maximum (1-RM). In the first 4-weeks, the exercise regimen consisted of a
high volume (3 sets of 12-15 repetitions), low intensity (60% of initial 1-RM),
resistance–exercise 3 times weekly. This was followed by thrice weekly,
progressive, periodized, high-intensity (90% of 1-RM on heavy days, 80% on
medium days, and 70% on light days), low-volume (4 sets of 4-6 repetitions
each) resistance exercise during weeks 5 through 10. During weeks 11 through
16, the loads were increased by 7% for upper body and 12% for lower body exercises,
and the number of sets was increased to 5.
The primary efficacy variable was change during treatment in 1-RM strength.
We also measured changes in thigh muscle volume by magnetic resonance imaging,
FFM by dual-energy x-ray absorptiometry (DEXA) and deuterium oxide dilution,
body weight, and health-related quality of life (HRQOL). Serum total and free
testosterone, dihydrotestosterone, luteinizing hormone (LH), follicle-stimulating
hormone (FSH), and sex hormone–binding globulin (SHBG) levels were measured
on several occasions during the control and treatment periods. Absolute and
percentage CD4+ and CD8+ cell counts and plasma HIV–copy
number were measured at baseline and during week 16. Adverse experiences were
recorded every 2 weeks.
We measured effort-dependent strength in the leg press, bench press,
leg curls, latissimus pulls, and overhead press exercises using the 1-RM method.
All subjects underwent an instructional period, in which lifting mechanics
were demonstrated. The resistance was progressively increased until the men
could not complete the lift; the maximum amount of weight that they were able
to lift was recorded as the 1-RM strength. To minimize the confounding influence
of the learning effect, the participants returned for reassessment of strength
within 7 days of the initial assessment. The higher of the 2 1-RM values was
recorded when the difference between the 2 measurements was less than 5%.
Body weight was measured every 2 weeks. Body composition was assessed
by DEXA scan (Hologic 4500; Waltham, Mass), deuterium oxide dilution, and
magnetic resonance imaging. For estimation of total body water, the men ingested
20 g of deuterium oxide14 and plasma samples
were drawn at −15, 0, 120, 180, and 240 minutes. We measured deuterium
abundance in plasma by nuclear magnetic resonance spectroscopy,30
using a correction factor of 0.985 for exchangeable hydrogen.
Energy and protein intake were standardized at 168 J/kg (40 kcal) per
day and 1.5 g/kg per day. The participants were given dietary instructions
2 weeks before treatment initiation; these instructions were reinforced every
2 weeks. The nutritional intake was verified by analysis of 3-day food records
during weeks 1, 8, and 16, and 24-hour food recalls every 4 weeks, by using
the Minnesota Nutritional Software.
Serum total testosterone levels were measured by an immunoassay31,32 and free testosterone levels by equilibrium
dialysis.32 The sensitivities of the total
and free testosterone assays were 0.02 nmol/L (0.58 ng/dL) and 0.2 pmol/L
(0.058 pg/mL), respectively. Intra-assay and interassay coefficients of variation
for the total and free testosterone assays were 8.2% and 13.2%, and 4.2%,
and 12.3%, respectively. Serum LH, FSH, and SHBG levels were measured by immunofluorometric
assays,31 with sensitivities of 0.05 IU/L,
0.15 IU/L, and 6.25 nmol/L, respectively. Plasma HIV RNA copy number was measured
by reverse transcriptase polymerase chain reaction (Amplicor; Roche Diagnostic
Systems Inc, Sommerville, NJ).
The HRQOL survey17- 33
included multi-item measures of physical functioning, role limitations due
to physical problems, general health perceptions, emotional well-being, role
limitations due to emotional problems and social, cognitive and sexual functioning.
The survey also included a 22-item symptom checklist, a disability-days item,
an overall health rating item, and a self-reported time trade-off preference
assessment. Data were analyzed using linear regression. Dummy variables were
used for group assignment, and baseline status and case mix variables were
included in the model.
All men who dropped out did so before week 6, and did not undergo posttreatment
DEXA scan or strength measurements. The analyses of muscle strength and body
composition, therefore, were performed on all randomized patients for whom
these efficacy data were available. For secondary variables, the last values
carried forward for withdrawn patients were analyzed. Continuous data are
reported as mean (SEM) and categorical data as frequency tabulations. All
variables were examined for their distribution characteristics. Variables
that did not meet the assumption of a normal distribution were log-transformed
and retested. If the assumption of normality could not be met by transformation,
then nonparametric methods of statistical comparison were used. An analysis
of variance (ANOVA) model was used to compare change from baseline between
the 4 groups for muscle strength, muscle volume, body composition measures,
and hormone levels. Student-Newman-Keuls method was used for pairwise comparisons
following a significant ANOVA. All outcome measures were also analyzed using
paired t test to detect a nonzero change from baseline
at week 16 within each treatment group.
Of the 61 patients who were enrolled, 49 completed the study, 12 in
group 1, 15 in group 2, 11 in group 3, and 11 in group 4. No discontinuations
were attributed to adverse experiences (Figure
1). The 4 groups did not significantly differ in age, height, weight,
CD4+ cell counts, prior weight loss, and baseline testosterone
levels (Table 1). The percentage
of men receiving antiretroviral therapy and protease inhibitors was not significantly
different between groups. The men who completed the study did not differ from
those who dropped out in body weight, ethnic composition, testosterone levels,
and CD4+ cell counts but were older than the latter group (35.5
vs 41.4 years, P = .01).
All evaluable subjects received more than 90% of their injections. Of
the 11 men in the combined testosterone and exercise group, 9 attended more
than 90% and 2 attended 75% to 89% of their scheduled exercise sessions. Of
the 11 men in the exercise alone group, 7 attended 90% to 100%, 3 attended
75% to 89%, and 1 attended 70% of the scheduled sessions.
Daily energy intake and percentage of energy derived from protein, carbohydrate,
and fat were not significantly different between the 4 groups at baseline
and did not significantly change during treatment. Specific daily energy and
macronutrient intake are provided in Appendix
Serum total and free testosterone levels, measured 1 week after injection,
increased significantly from baseline in the testosterone groups and did not
change in those receiving placebo. Serum LH, FSH, and SHBG levels decreased
significantly in the testosterone-treated men but not in men treated with
placebo (Table 2).
Among those who were in the placebo alone group, muscle strength did
not change in any of the 5 exercises (−0.3%-−4.0%) (Table 3, Figure 2 and Appendix 2.
This indicates that this strategy was effective in minimizing the influence
of the learning effect. Persons in the exercise-alone group increased muscle
strength in the 5 exercises by 29% to 36%, and those in the testosterone-alone
group increased it by 17% to 28%. Although those in the testosterone-exercise
group increased their muscle strength by 10% to 32%, it was not significantly
greater than either intervention alone. The change in leg press strength correlated
with change in muscle volume (R = 0.44, P = .003) and change in FFM (R = 0.55, P<.001).
The thigh muscle volume did not significantly change in those taking
placebo alone (change, 5 cm3; P = .696) but increased among those taking testosterone(change, 40
cm3; P = <.001),
exercising alone (change, 62 cm3; P = .003), or taking testosterone and
exercising (change 44
cm3; P = .001) (Figure 3). The exercise-alone, and testosterone-alone groups experienced
greater increases in muscle volume than the placebo-alone group (P = .05). Specific tension, defined as muscle strength in the leg press
exercise per unit quadriceps volume, did not change significantly in men who
did not exercise regardless of whether they received placebo or testosterone
but increased in both groups of men who exercised.
Body weight was stable in those taking placebo alone during treatment
(mean change, –0.5 kg; P = .546). Those taking
testosterone alone (change, 2.6 kg; P<.001) and
those who exercised alone (change, 2.2 kg, P = .023)
experienced significant increases in body weight. The weight gain in these
2 groups was greater than that in the placebo-alone group (P<.05). The combination of testosterone and exercise training did
not increase body weight more than testosterone alone or exercise alone. Weight
gain correlated with change in FFM measured by deuterium oxide dilution (R = 0.75, P<.001) (Figure 3).
Fat-free mass, measured by deuterium oxide dilution, increased significantly
from baseline in those taking testosterone alone and those exercising alone
, but it did not change significantly in the placebo group (Table 4). However, the change in FFM was not significantly greater
in those taking testosterone alone and exercising alone than in those taking
placebo alone. The effects of 2 interventions together were not significantly
greater than those produced by either intervention alone. Total body water
increased significantly in those taking testosterone alone and those exercising
alone, but it did not increase in those taking placebo alone. The ratio of
FFM by DEXA and total body water did not significantly change in any treatment
group, indicating that the apparent increases in FFM were not due to fluid
retention in excess of protein accretion. Specific body composition assessed
by DEXA is provided in Appendix 3.
Fat-free mass measured by DEXA correlated highly with total body water
(R = 0.746, P<.001).
Fat-free mass, measured by DEXA, did not change in the placebo-treated groups,
but increased significantly in the 2 testosterone-treated groups without (change,
2.3 kg; P = .004) and with exercise (change, 2.6
kg; P<.001). The lean mass in the arms, legs,
and trunk increased significantly in those who were treated with testosterone
but not in those taking placebo. Truncal and whole body fat mass, and bone
mineral content did not change in any group.
There was no association between the change in HRQOL measures and testosterone
administration or exercise in any group.
We analyzed the body weight and hemoglobin data carrying the last available
value forward. The mean (SE) change in body weight was significantly greater
(overall ANOVA, P = .004) in men treated with testosterone
alone (change, 2.9 [0.8] kg; P = .002) or exercise
alone (change, 1.7 [0.6] kg; P = .02) than in those
receiving placebo alone (0.7 [0.7] kg; P = .39).
The mean (SE) change in hemoglobin was significantly greater in those taking
testosterone alone (13.5 [5.6] g/L; P = .03) or in
those taking testosterone and exercising (7.2 [2.5] g/L; P = .01) than in those in taking placebo alone ( −5.6 [2.7] g/L)
or those exercising alone (1.8 [5.2] g/L) (overall ANOVA, P = .032).
Hemoglobin levels increased in testosterone-treated men regardless of
exercise (change, P<.05). For those who received
testosterone and exercised, it increased by 6%. For those who did not, it
increased by 14% (P = .04). It remained unchanged
in those taking placebo. The changes in CD4+ and CD8+cell
counts, and HIV RNA, bilirubin, alanine aminotransferase, aspartate aminotransferase,
plasma triglycerides, low-density lipoprotein and high-density lipoprotein
cholesterol, and PSA levels in testosterone-treated men were not significantly
different from those in the placebo groups. One person receiving testosterone
and 1 receiving placebo developed acne; 1 testosterone-treated man experienced
In HIV-infected men with moderate weight loss and low testosterone levels,
testosterone replacement and resistance exercise each was associated with
significant gains in muscle strength, muscle size, and body weight. Fat-free
mass did not significantly change in any group. Most of the increase in body
weight in testosterone and exercise groups was due to lean mass accretion.
The increase in muscle strength and FFM were correlated with increase in muscle
volume. The effects of combining testosterone and exercise were not additive.
We were unable to detect improvements in HRQOL; a larger sample size might
be needed to detect changes in HRQOL.
There is uncertainty about what magnitude of increase in FFM is clinically
significant. Testosterone and resistance exercise were each associated with
gains in FFM in excess of 2 kg and significant increases in muscle strength.
Therefore, we posit that these changes are clinically significant.
The testosterone regimen increased serum testosterone levels by approximately
4.3 nmol/L (124 ng/dL) and produced significant suppression of LH and FSH
levels, providing evidence of androgenic effects. Because serum testosterone
levels were measured 7 days after the previous injection, this reflects the
smallest increment in testosterone levels that occurred following an injection.
There were no significant differences in the change in FFM or muscle
strength during testosterone treatment between men who had baseline testosterone
less than 9.5 nmol/L (<275 ng/dL) and those with baseline testosterone
between 9.5 and 12.1 nmol/L (275-350 ng/dL). Baseline serum testosterone levels
did not correlate with change in FFM. Further studies are needed to determine
whether HIV-infected men with low normal testosterone levels respond differently
to testosterone supplementation than those who are truly androgen deficient.
The treatment regimen was safe and well tolerated. Testosterone treatment
was associated with significant increases in hemoglobin. Changes in serum
bilirubin, alanine aminotransferase and aspartate aminotransferase, plasma
lipids, and PSA levels were not significantly different from those observed
among those taking placebo.
Subject compliance with the treatment regimen was high. One hundred
percent of the evaluable patients received greater than 90% of their prescribed
injections. The compliance with the exercise regimen was less complete; this
may have affected the magnitude of strength gains achieved during exercise
A number of studies have examined the effects of androgen supplementation
on body composition in HIV-infected men,17- 29
although only some were placebo-controlled, blinded, and randomized trials.
The dietary intake was not controlled in some of the studies, while others
did not standardize the exercise stimulus. The methods used to assess body
composition also differed among studies. Of the 4 placebo-controlled studies
of testosterone,17- 20
only 217,18 demonstrated significant
increases in FFM. The 2 studies that did demonstrate significant increments
in FFM included patients with low testosterone levels. Studies using oxandrolone
and nandrolone decanoate have reported gains in FFM but not muscle strength.21,22,27,28
Our study is the first to demonstrate significant improvements in muscle
strength during a pharmacological intervention in HIV-infected men. The failure
to control the exercise stimulus and learning effect might have confounded
the results of other studies. We minimized the confounding influence of the
learning effect by using multiple training sessions during the control period.
This appears to have been effective because muscle strength was unchanged
among those who received placebo and did not exercise. Our findings indicate
that when confounding factors such as the learning effect are minimized and
the exercise stimulus and nutritional intake are standardized, testosterone
replacement and resistance exercise significantly increase muscle size and
strength in HIV-infected men with low testosterone levels.
Some of the initial strength gains from resistance training result from
neuromuscular learning. Because we used the same equipment and weight-lifting
exercises for strength testing as well as strength training, this strategy
may have favored the exercise intervention over androgen treatment in terms
of apparent strength gains.
Aerobic exercise improves cardiovascular fitness in HIV-infected patients34,35 but does not produce substantial
skeletal muscle hypertrophy. The resistance exercise regimen used in this
study resulted in significant muscle hypertrophy, lean body mass accretion,
and increments in muscle strength. Similar gains in FFM and strength following
resistance training were reported by Roubenoff et al,36
although that study did not have a nonexercising control group. While testosterone
and exercise both induced increase in muscle volume, only men who underwent
exercise experienced an increase in specific tension. These effects of strength
training were achieved in the setting of a motivated patient population, through
a supervised program administered by trainers in an exercise laboratory. Even
in this setting, the compliance with the exercise regimen was not perfect.
Similar degrees of compliance and strength gains may not be achievable with
exercise in a clinical setting.
The effects of testosterone and exercise training combined were not
additive in this study. The strength training has been shown to augment the
anabolic effects of supraphysiologic doses of androgen in healthy, eugonadal
men31 and HIV-infected men with weight loss.22,29 We do not know whether the failure
to demonstrate additive effects in this study was related to testosterone
dose, less than perfect compliance with the exercise regimen, or lack of prior
In comparison with other anabolic agents, such as recombinant human
growth hormone37,38 and megesterol
acetate,39 testosterone and exercise are relatively
inexpensive. Megesterol induces androgen deficiency and may induce a decrease
in lean mass. Recombinant human growth hormone administration does not improve
muscle strength in HIV-infected individuals.37,38
We conclude that testosterone induces FFM accretion and skeletal muscle
hypertrophy and increases effort-dependent muscle strength in HIV-infected
men with low testosterone levels and moderate weight loss. We can not extrapolate
these data to HIV-infected men with normal testosterone levels or severe wasting.
It would be premature to claim that testosterone administration improves muscle
function because maximum voluntary strength is only one aspect of muscle function;
power, endurance, and task-specific performance are other attributes that
were not examined in this study. Muscle strength per unit of muscle mass (specific
tension) increased in men undergoing resistance exercise, but it did not increase
in those receiving testosterone alone. This implies that resistance training
improves the contractile quality of skeletal muscle in HIV-infected men but
that testosterone administration without exercise training does not. Further
studies are needed to evaluate whether testosterone and exercise can induce
clinically useful changes in muscle function and HIV-related disease outcomes.