Treatment of Men for “Low Testosterone”: A Systematic Review

Testosterone products are recommended by some prescribers in response to a diagnosis or presumption of “low testosterone” (low-T) for cardiovascular health, sexual function, muscle weakness or wasting, mood and behavior, and cognition. We performed a systematic review of 156 eligible randomized controlled trials in which testosterone was compared to placebo for one or more of these conditions. We included studies in bibliographic databases between January 1, 1950 and April 9, 2016, and excluded studies involving bodybuilding, contraceptive effectiveness, or treatment of any condition in women or children. Studies with multiple relevant endpoints were included in all relevant tables. Testosterone supplementation did not show consistent benefit for cardiovascular risk, sexual function, mood and behavior, or cognition. Studies that examined clinical cardiovascular endpoints have not favored testosterone therapy over placebo. Testosterone is ineffective in treating erectile dysfunction and controlled trials did not show a consistent effect on libido. Testosterone supplementation consistently increased muscle strength but did not have beneficial effects on physical function. Most studies on mood-related endpoints found no beneficial effect of testosterone treatment on personality, psychological well-being, or mood. The prescription of testosterone supplementation for low-T for cardiovascular health, sexual function, physical function, mood, or cognitive function is without support from randomized clinical trials.


Introduction
Testosterone and methyltestosterone are marketed in the United States for men with congenital or acquired hypogonadism. Some practitioners have used testosterone preparations to treat a variety of symptoms identified as those of "low testosterone" (low-T), a term that has not been uniformly defined. We present a systematic review of randomized controlled trials (RCTs) that evaluated the use of testosterone therapy against placebo or inactive comparator in adult men for cardiovascular health, sexual function, muscle weakness/wasting, mood and

Quality Assessment
We assessed quality of studies by a 5-point Jadad score. In order to be as inclusive as possible, we included all studies identified regardless of Jadad score. For clinical endpoints only (angina/ ischemia, congestive heart failure, and erectile dysfunction) we also included an analysis of studies restricted to Jadad scores of 4 or 5. We accepted whatever criteria were used by individual study authors to define low testosterone.

Results
Fig 1 lists the exclusion criteria used to select 226 qualifying papers from 11,417 reviewed abstracts. Although most studies were described by their authors as randomized, not all indicated the nature of the randomization procedures. Some studies included identical numbers of subjects in treatment and exposed conditions, suggesting that allocation was not random. After further examination, 70 papers did not meet our criteria, so the final data set included 156 papers. Table 1 summarizes extracted studies that focused on the effect of testosterone on cardiovascular endpoints, including 17 studies on ischemia/angina, 6 on congestive heart failure (CHF), 25 on lipids, and 11 on inflammatory and coagulation markers.

Cardiovascular Health
3.1.1 Coronary artery disease. In studies that investigated the effect of testosterone on patients with coronary artery disease (CAD), eligible men generally were identified based on stable angina, angiographic evidence of some degree of coronary artery occlusion, or a history of myocardial infarction (MI). Six studies involved men in whom the study authors reported evidence of hypogonadism either clinically [1] or based on plasma testosterone concentration [2][3][4][5][6]; the remainder included men without regard to plasma testosterone concentration. All but three of the studies evaluated ST-segment depression on an exercise stress test using a modification of the Bruce protocol. One of the studies not using the Bruce protocol evaluated findings on electrocardiography (ECG) and Holter monitoring, without specification of an exercise protocol. [7] Another study added single-photon emission computer tomography (SPECT) to evaluate for deficits in myocardial uptake of a labeled perfusion tracer. [8] One study used magnetic resonance imaging (MRI) estimates of myocardial perfusion. [3] One study evaluated change in coronary artery calcium score over time, showing no difference between testosterone and placebo. [6] Two studies evaluated brachial artery response to release of occlusion as an indicator of sensitivity to local vasodilators in men with CAD and did not directly address the coronary arteries; both reported results favorable to testosterone therapy. [9,10] Two studies in apparently healthy men with bioavailable testosterone <4.44 nM (128 ng/dL) or total testosterone 15 nM (432 ng/dL) found no change in brachial artery reactivity in response to transdermal testosterone or dihydrotestosterone therapy. [5,11] The study that used MRI showed no effect of 8 weeks of oral testosterone undecanoate therapy on myocardial perfusion, although there was increased perfusion of those segments supplied by an unobstructed coronary artery. [3] Three studies used acute treatments with intravenous (IV) testosterone just prior to exercise testing. Two of the studies showed favorable effects of treatment on time to ST-segment depression. [2,12] One study showed no effect on ECG or SPECT evidence of ischemia. [8] A year-long study showed benefits of testosterone treatment on ST-segment depression. [4] The remaining eight studies evaluated treatments of 2 to 24 weeks in duration. [1,3,10,11,[13][14][15][16][17] The 3 studies that looked at time to ST-segment depression found a benefit of testosterone supplementation. [1,14,16] Although 2 studies reported improvements in angina symptoms during or after testosterone treatment, [7,15] 4 studies showed no effect of treatment on angina. [1,4,8,16] Most studies did not report any measure of angina symptoms. A study on men with leg claudication or trophic ulcers attributed to arteriosclerosis did not show an improvement in subjective symptoms, walking, or plethysmographic estimation of blood flow endpoints after 3 months of testosterone therapy. [13] There was a decreased incidence of silent MI with testosterone treatment in 1 study. [15] Another study, designed to determine the effect of testosterone supplementation on lowerextremity strength and physical function in men 65 years of age and older, was stopped early by a Data and Safety Monitoring Board due to an excess of cardiovascular adverse events. [17] These adverse events included acute coronary syndrome (ACS), MI, ECG abnormalities, and arrhythmias, among others.
Eleven studies of coronary artery disease scored 4 or 5 on the Jadad scale. Of these, only one of five studies that included angina as an outcome found a benefit. Four of five studies that assessed ST segment depression found a benefit.
2.5 mg testosterone (T) or intravenous (IV) placebo over 5 min, 30 min prior to exercise test; tx switch at 2 days; randomization by computer; masking not described.
"Time to ST segment depression. #Maximum ST segment depression. #ST recovery. 3 Ong et al, 2000 [9] 22 men with CAD. 2.3 mg T or placebo IV over 10 min (n = 11) with switch after 1 week; 0.023-0.046 mg T or placebo IV over 10 min (n = 11) with tx switch after 1 week; randomization method not given.
"Percent change in brachial artery diameter after release of occlusion with high-dose T. No change in flow velocity in brachial artery after release of occlusion. Low-dose T had no effect. Interpreted as enhanced response to local effects of nitric oxide after T. 3 Thompson et al, 2002 [8] 34 men 69 ± 6 years old (mean ± SD) with CAD and exercise-or adenosineinducible ischemia.
T or placebo by bolus IV over 20 min with maintenance IV to increase basal serum T concentration 0, 2-, or 6-fold with each subject receiving all 3 conditions randomly 1 week apart; randomization method not given.

2-24 Week Treatment Period
Dohn et al, 1968 [13] 44 men with leg claudication or ulcers attributed to arteriosclerosis (n = 43) or Buerger's diseas (n = 1). Two men did not complete study, but numbers in tables add to 86 subjects. Not possible to tell for sure how many men were analyzed. 300 mg aqueous T isobutyrate or placebo (meprobamate) every 14 days for 3 months; route of administration not given; double-blind; randomization method not given.
No effect on subject improvement, walking test, plethysmographic estimation of pulse volume, blood flow at rest or after exercise, or hyperemia after compression. "Skin temperature. 3 Ly et al, 2001 [11] 37 men mean age 68.2 years with plasma T concentration 15 nM (432 ng/dL); 4 dropouts were excluded from analysis. 70 mg dihydrotestosterone (DHT) gel (n = 18) or placebo (n = 19) applied daily for 3 months; randomization method not given.
No effect on flow-or nitroglycerinmediated dilatation of brachial artery.
"Flow-and nitroglycerin-mediated dilatation of brachial artery. 1 Malkin et al, 2004 [1] 12 men 60.8 ± 4.6 years old (mean ± SD) with CAD and "clinical need for T replacement." One man failed screening and another withdrew at unspecified point in the study.
100 mg T or placebo IM every 2 weeks for 4 weeks, 1 month washout, then tx switch; randomization by computer; single-blind.
"Time to ST segment depression. No significant change in Seattle Angina Score. #Beck Depression Inventory (BDI) score. #Incidence of silent myocardial infarction. #Total ischemic burden. #Number of anginal attacks/week. 5 English et al, 2000 [16] 50 men mean age 62 years with stable CAD.
5 mg daily T patch (n = 25) or placebo (n = 25) for 12 weeks; double-blind; randomization method not given. Three withdrawals from T arm and 1 withdrawal from placebo arm eliminated from analysis.  Webb et al, 2008 [3] 25 men 40-75 years old with angiographically proven CAD (!70% lesion in at least one major coronary artery or major branch), plasma T concentration 12 nM (346 ng/dL); 2 dropouts prior to medication tx were not analyzed. One subject had unanalyzable data.
160 mg/day T undecanoate or placebo PO for 8 weeks followed by tx switch. Method of randomization not given.
No change in global myocardial perfusion by magnetic resonance imaging (MRI). "Perfusion of segments supplied by coronary arteries without significant obstruction. "Left ventricular (LV) ejection fraction (EF) (by 2%). No change in stroke volume (SV), endsystolic volume (ESV), end-diastolic volume (EDV), or heart mass. 4 Basaria et al, 2010 [17] 209 men !65 years old with total serum T 100-350 ng/dL (3.5-12.1 nM) or free serum T <50 pg/mL (174 pM) and with mobility limitations. Analysis restricted to 176 men with a baseline assessment and at least one outcome assessment.
100 mg T (n = 106) or placebo (n = 103) gel daily for 6 months. After 2 weeks, dose level was increased or decreased by 50% based on serum T. Randomization was by age blocks but was not otherwise described.

Wu et al, 1993[7]
62 men 55-75 years old with angina. 120 mg T undecanoate PO QD × 2 weeks then 40 mg/day × 2 weeks or placebo for 2 weeks followed by 2-week washout, then tx switch; described as randomized, but there were 31 men in each group and randomization was not described.
#Ischemia on ECG and Holter recordings by subjective scoring system. #Angina by subjective scoring system.

!12-Month Treatment Period
Mathur et al, 2009 [4] 15 men 64.8 ± 7 years old (mean ± SD) with stable chronic angina, ST segment depression at baseline, and at least 2 early morning serum T concentrations < 12 nM (346 ng/dL); one man assigned to each arm withdrew at unspecified point in the study.
"Time to ST segment depression at 14, 28, and 52 weeks and increased level of exercise attained. No significant change in Seattle Angina Score (SAS). 4 Kenny et al, 2002 [5] 67 men 65-87 years old (mean 74) with bioavailable T <4.44 nM (128 ng/dL); 23 dropouts (10 T, 13 placebo) and 8 men with technical difficulties were not included in the analysis.
No change in vascular reactivity after occlusion.

3
Basaria et al, 2015 [6] Men aged 60 years or older, morning total T 100-400 ng/dL (3.5-14 nM) or free <50 testosterone pg/mL (1.7 pM). 1:1 concealed randomization with stratification by age dichotomized at 75 years and by site. Computer-generated randomization. All subjects receiving at least 1 medication dose were retained for analysis. 29 men 36-78 years old with a clinical indication for T replacement for hypogonadism; 2 subjects were withdrawn and 2 additional patients did not contribute analyzable sera.
100 mg T (n = 27) or placebo IM (n = 27) every 2 weeks; randomization using blocks of computer-generated numbers. A crossover design appears likely, although not explicit.
#Serum tumor necrosis factor-α (TNF-α). "Interleukin-10 (IL- 10). No change in IL-1β (identified as decreased by authors, but not statistically significant). 60 mg T or placebo given buccally followed the next day by tx switch. Described as randomized, but randomization method not given.
No effect on serum concentration of TNF-α with any of these T tx.
3 20 men with NY Heart Association class II or III CHF, mean age 63.9 years in the active group and 61.1 years in the placebo group.
100 mg T (n = 10) or placebo (n = 10) IM every 2 weeks for 12 weeks. Subjects said to be randomized, randomization method not given. No effect of tx on C-reactive protein, except " in men with baseline C-reactive protein concentration below the median.

Congestive heart failure.
Six studies evaluated effects of testosterone treatment on CHF. [18][19][20][21][22][23] In two papers from the same group, [19,20] it is not clear whether treatments were randomly assigned. Administration of testosterone by the buccal route was associated with beneficial effects on cardiac index and systemic vascular index in the acute catheterization setting, consistent with an acute vasodilatory effect.
[19] Intramuscular (IM) testosterone treatment for 12 weeks improved exercise capacity and reduced heart failure symptom scores without identifiable effects on left ventricular size or ejection fraction (EF).
[20] Another study of IM testosterone in men with CHF showed an improvement in oxygen consumption, respiratory efficiency (ventilation/carbon dioxide consumption), and distance walked in 10 minutes without changes in EF or left ventricular end-diastolic diameter. [18] The improvements in exercise function appeared attributable to the response of men with baseline plasma testosterone concentration <12 ng/mL (~4 nM). A study of a testosterone patch showed improvement in the shuttle walk test.
[23] Another study showed no effect of IM testosterone enanthate on ejection fraction, although there was an improvement in a Doppler-based myocardial performance index. [22] The only study that scored above a 3 on Jadad found a benefit on CHF measures. [18] 3.1.3 Lipids. Serum or plasma concentrations of cholesterol fractions, triglycerides, and lipoproteins have been used as surrogate endpoints for cardiovascular risk, although they should not be mistaken for markers of cardiovascular adverse events. In 25 studies, testosterone treatment was associated with favorable, unfavorable, or no effects on lipids as summarized in Table 1. Favorable effects in 11 studies included 5-11% decreases in total cholesterol concentration and variable and inconsistent decreases in triglycerides and low-density lipoprotein (LDL) cholesterol. One of the studies counted as showing a favorable effect did not demonstrate a change in total or high-density lipoprotein (HDL) cholesterol or triglycerides but reported a 13% reduction in LDL cholesterol.
[24] This finding was based on averages of several repeated measurements over the course of 1 year rather than a determination of improved lipid measurements at the end of the treatment period.
Unfavorable changes were reported in 2 studies [5,25] and included increases in total cholesterol, LDL cholesterol, and triglycerides and decreases in HDL cholesterol in men evaluated as hypogonadal prior to androgen therapy. Most of the studies that did not report favorable effects of testosterone on lipids reported no effects at all. One of these studies[26] reported a decrease in lipoprotein-a (LP-a), but this finding was transient and occurred in a study with multiple measurements at multiple time points in multiple patient subgroups without adjustment for multiple comparisons. Nine of the 11 studies that had favorable effects on lipids had Jadad scores of 4 or 5. Nine of the 14 studies that lacked favorable effects on lipids had Jadad scores of 4 or 5. The discordance between studies on the lipid effects of testosterone treatment did not appear to be route dependent. Seven of the 11 studies showing favorable effects used IM injection of testosterone enanthate, esters, or undecanoate. Five of the 14 studies not showing favorable effects on lipids used IM injection of testosterone esters, cypionate, or undecanoate.

Inflammatory or coagulation markers.
Eleven studies were identified in which markers that have been associated with atherosclerotic cardiovascular disease risk were measured in men using testosterone or dihydrotestosterone therapy. Three studies reported favorable effects of testosterone on tumor necrosis factor-α (TNF-α), a marker of inflammation. [1,3,27] One study in men with CHF showed no effect of testosterone treatment by buccal, IM, or transdermal routes on serum concentration of TNF-α.
[21] One of the studies asserted that there was a decrease in the inflammatory marker interleukin-1β (IL-1β), but a statistically significant effect was not shown. [1] Another study showed a decrease in interleukin-6 (IL-6) and C-reactive protein, additional inflammatory markers.
[28] Two studies performed in elderly men who were largely without a diagnosis of CAD showed no beneficial effect of testosterone therapy on C-reactive protein [29,30] as did two studies of men with type 2 diabetes mellitus.
[31, 32] Transdermal dihydrotestosterone did not affect inflammatory markers in men with low total pretreatment testosterone concentrations. [33] No change in fibrinogen, plasminogen activator inhibitor-1, or tissue plasminogen activator was shown in men with CAD who used testosterone patches or oral doses. [3,34]
Improved sexual function in 9 Ttreated (n = 19) and 2 placebo-treated (n = 17) men (Fisher P = 0.03 calculated by us, not by authors). No difference from placebo in ability to have an erection or sexual satisfaction (both groups improved).
Crossover, within-subject, 3 phase design. T enanthate 0, 100, or 400 mg IM each administered once, a month apart. Tx were administered in "arbitrarily chosen order and "varied at random within and among subjects." 400 mg dose improved total erections, nocturnal erections, and coital frequency. No effect on incidence of orgasm or masturbation. 3

Skakkebaek et al, 1981[45]
12 men 22-48 years old diagnosed as androgen deficient; one man subsequently found to be normal and excluded.
160 mg T undecanoate or placebo daily by mouth for 2 months followed by tx switch; randomization of order of tx not described but said to be balanced. 200 mg T cypionate or placebo IM every 2 weeks for 12 weeks followed by the opposite tx; no washout period mentioned, randomization of tx order not described.
"Libido and potency based in part on scores assigned to questionnaire responses analyzed using t-test; insufficient detail provided to permit nonparametric analysis of ranked data. 200 mg T enanthate or placebo IM every 2 weeks for 6 weeks followed by 4-week washout followed by the opposite tx; order was randomized by unstated method; 7 subjects received placebo first, 5 subjects received T first.
No effect on sexual satisfaction, frequency of sexual desire, masturbation, sex with partner, morning erections, or degree of erections. "Ejaculations.
10 mg T or taste-matched placebo (pseudoephedrine 3 mg) buccal tablet daily for 4 weeks; doubling of dose permitted after end of first 4 weeks; randomization method not given.
No effect on total scores of Watts Sexual Function Questionnaire (WSFQ) after 8 weeks; "Frequency of sexual desire and morning erections, but not erections in general; "Maximum rigidity and duration of full nocturnal penile tumescence assessed in a sleep lab. 3

Rabkin et al, 2000[35]
74 HIV-positive men, mean age 39, with serum T <17.4 nM (501 ng/dL) with sexual dysfunction and at least one "hypogonadal" mood symptom; 3 subjects in the placebo group dropped out, 1 in T group was excluded from analysis because of medication error. Among the 70 subjects analyzed, 26 had major depressive disorder (MDD), dysthymia, "minor" depression, or MDD in remission.
Randomization in blocks of 4 by computer-generated numbers to T cypionate (n = 38 overall, 26 with depression diagnosis) or placebo (n = 32 overall, 7 with a depression diagnosis) injected biweekly (presumably IM) for 6 weeks; the first T dose was 200 mg, subsequent doses were 400 mg. Openlabel phase with T followed the 6-week double-blind study and is not summarized here.
"Libido and morning erections. Testosterone improved ED by expanded CGI (Clinical Global Index) among completers with initial ED. 4

Howell et al, 2001[24]
35 men mean age 40.9 years with some degree of testicular dysfunction after cytotoxic cancer therapy; blood LH concentration !8 mIU/L and T concentration <20 nM (576 ng/dL); 2 dropouts, 1 in each group, were excluded from analysis.
2.5 or 5.0 mg T patch (n = 16) or placebo patch (n = 19) daily for 12 months; randomization method not described, study described as single-blind.
No effect on interest in sex, sexual activity, or frequency of erections. 3

Park et al, 2003[66]
39 men with sexual dysfunction, infertility, symptoms of hypogonadism and T <400 ng/dL (13.9 nM); 4 subjects in the T group dropped out.
160 mg T undecanoate (n = 33) or placebo (n = 6) daily by mouth for 3 months; described as single-blind, randomization method not given.
Between-group analyses not reported.
Fischer's exact test as performed by us showed no significant difference between groups in showing sexual function improvement. Androgen Deficiency in the Aging Male (ADAM) questionnaire scores reported improvement only in the treated group, but total scores not provided. 5 mg transdermal T (n = 10) or placebo (n = 10) patch daily for 1 month; all subjects received sildenafil 100 mg to use on demand; randomization method not given.
Improved overall erectile function domain score, intercourse satisfaction, overall satisfaction, number of acts of sexual intercourse, and percentage of successful intercourse attempts. Improved erections in 8 men on T (n = 10) and 1 on placebo (n = 8). No change in sexual desire and orgasmic function. 150 men >50 years old with symptoms consistent with hypogonadism and free T <6 pg/mL (21 pM); 20 dropouts analyzed by ITT.
"Erectile function at 3 and 6 months, sexual desire at 3 but not 6 months, and sexual intercourse satisfaction at 6 but not 3 months. No change in general sexual well-being or orgasm. Carnitine superior to T for erectile function at 3 and 6 months and for orgasm and general sexual well-being at 6 but not 3 months.

4
O'Connor et al, 2004 [67] 28 healthy eugonadal men 22-44 years old; 4 subjects treated with T withdrew and were excluded. 1000 mg T undecanoate or placebo IM at the beginning of an 8-week phase, followed by 8-week washout, followed by the opposite tx, randomization of tx order not described.
No effect on frequency of sexual intercourse, masturbation, sexual desire, enjoyment of intercourse, or overall satisfaction with sexual experience on weekly logs.  No difference between groups in sexual function assessed by a 5-question scale: 1) "I found it easy to achieve an erection when I wanted to"; 2) "I have lost interest in sex"; 3) "I found it difficult to sustain an erection when I wanted to"; 4) "I had problems achieving an orgasm"; and 5) "I am generally satisfied with the sex that I have." 160 mg T undecanoate (n = 120) or placebo (n = 117) in divided daily oral doses for 26 weeks; randomization computer-generated using blocks of 6 (Emmelot-Vonk et al [91]).
No effect on sexual functioning including frequency of sexual activity, quality of sexual functioning, and ability to achieve or maintain erection.  At 30 weeks, "erectile function, intercourse satisfaction, sexual desire, orgasm; no effect on overall satisfaction; no changes in AMS.

5
At 18 weeks, "intercourse satisfaction, sexual desire, and orgasm; no effect on ED, overall satisfaction; or AMS.

Basaria et al, 2015[6]
Men aged 60 years or older, morning total T 100-400 ng/dL (3.5-14 nM) or free T <50 pg/mL (1.7 pM). 1:1 concealed randomization with stratification by age dichotomized at 75 years and by site. Computer-generated randomization. All subjects receiving at least 1 medication dose were retained for analysis.  [50] reported a benefit of both testosterone and placebo compared to baseline; however, our analysis did not show a difference between treatment groups. Limiting analysis to the 17 studies with Jadad scores of 4 or 5 yielded similar results; 9 were positive and 8 were negative. Twelve studies included men with ED; 8 found no benefit of testosterone over placebo, [55,59,62,64,68,72,74,76] and 4 found a benefit. [35,46,47,49] One negative study found that testosterone reduced erectile function when compared to placebo; however, there was no change when each group was compared to its baseline. [76] Of 23 studies that specifically reported changes in libido, 13 found that testosterone treatment increased libido, [26,35,37,45,46,54,56,60,63,65,70,77,79] eight found no effect, [24,38,39,47,59,67,75,76] and 1 found an effect after 3 but not 6 months of treatment. [48] Hackett et al [58] found that testosterone improved sexual desire in a group with initial testosterone 8.0 nM but not in a group with initial testosterone 8.1-12 nM. Eleven studies used the Aging Males' Symptoms scale, which includes 3 questions on libido and sexual function. Five studies found no difference between testosterone and placebo on total scores,[26, 54,57,72,76] and 4 studies found a benefit of testosterone. [52,57,58,66] One paper [53] reported only sexual subscales but not total AMS scores. On the sexual subscale of the AMS scale, this study reported a benefit, Ho et al [73] found no benefit, and Legros et al [57] found a benefit of testosterone on the AMS sexual subscale only in the middle (160 mg) of 3 dose levels at 3 of 4 time points. Hackett et al [58] found that testosterone improved AMS scores in a group with initial testosterone 8.0 nM but not in a group with initial testosterone 8.1-12 nM. Subjects were block randomized to receive testosterone undecanoate 1000 (n = 97) or 0 (n = 102) mg IM at week 0, 6, and 18. Improvement in erectile function, intercourse satisfaction, sexual desire, and AMS in group starting with total T 8.0 nM; Improved AMS score but no improvement in sexual function in group starting with total T between 8.1 and 12 nM. Ten of 13 of the studies on libido or desire with a Jadad score of 4 or 5 found a benefit. Seven of 12 studies on erectile dysfunction with a Jadad score of 4 or 5 found a benefit. Table 3 summarizes 39 studies that evaluated the effect of testosterone on physical function, muscle strength, or HIV-associated muscle wasting, including 19 in men assessed as having low serum testosterone, 9 on HIV-negative men with normal serum testosterone, 1 on healthy men with normal serum testosterone, and 10 on HIV-positive men. Studies that measured testosterone effects only on body composition (other than in HIV-associated wasting) without measures of physical function or muscle strength were excluded. Subjects included both those defined by the authors as hypogonadal and those considered to have normal testosterone concentrations. Common measures of muscle strength included grip strength dynamometry and the 1-repetition maximum for exercises including the bench press and leg press. Physical function was often measured by the 6-minute walk test, the time and number of steps required to walk 25 feet, and the get-up-and-go test, which evaluates the ability to rise from a chair, walk a short distance, and return to sitting.

Muscle Weakness/Wasting
Twenty studies evaluated subjects described as hypogonadal, with 11 of those evaluating healthy subjects. Five studies examined the effects of testosterone supplementation on physical frailty, functional limitations, or a categorization as "sedentary, " [53,[80][81][82][83] and single studies evaluated subjects with COPD, [84] advanced cancer, [75] and Parkinson's disease. [85] Ten studies evaluated subjects considered to have normal testosterone concentrations; 1 study included healthy, elderly men, and the remainder included subjects with planned knee replacement surgery, [86] stable CHF, [18,22,87] leg claudication or ulcers, [13] long-term glucocorticoid therapy, [88] myotonic dystrophy, [89] arterial insufficiency, [40] COPD, [43] or who were planning or undergoing physical rehabilitation. [90] Ten studies evaluated subjects with HIV; 8 of those studies included subjects with HIVwasting, 1 included subjects with abdominal obesity, and 1 did not use weight criteria. Most of these papers studied older men. Few studies investigated the use of testosterone supplementation in men younger than 60 years.
Some studies did not measure muscle and fat mass specifically but used other body composition endpoints. Two studies showed no changes in body weight or BMI, [22,104] but another showed an increase in body weight and BMI. [18] One study, with a Jadad score of 3, showed no change in weight or estimates of body fat (triceps and scapula skinfold thickness). [105] In studies of HIV-positive men with weight loss, 3 of 6 studies (all of which had Jadad scores of 4 or 5) showed an increase in weight with testosterone treatment, [99,101,106] and all 4 studies that measured muscle mass showed an increase. [41,[99][100][101] Of the 30 studies that assessed muscle strength as a primary or secondary endpoint, 13 studies (43%) reported an improvement in at least 1 measure of muscle strength. [11,18,53,81,83,84,88,93,94,101,102,104,107] Eleven of 13 of these studies had a Jadad score of 4 or 5. Three of these 12 studies (all with Jadad scores of 4 or 5) reported improvements in fewer than 25% of the measurements. [11,53,81] In studies of men without HIV, 11 of 24 studies (45.8%)  [92] 10 men 60-78 years old with morning serum T 200-450 ng/dL (6.9-15.6 nM), serum prolactin below 25 μg/L, LH and FSH below 20 IU/L, IGF-I below 200 μg/ L. T patch 5 mg/day or placebo x 1 month with each subject serving as his own control; subjects also received growth hormone with or without T (not discussed here); randomization method not given.
T enanthate 100 mg (n = 23) or placebo (n = 24) IM weekly with or without resistance training x10 weeks; randomization method not given.
T + no training group: Compared to placebo + training: "trunk lean mass, #% fat. Compared to placebo groups: "arm and total lean mass; #leg and total fat. Compared to placebo + no training: "leg lean mass and leg press strength; #leg press fatigue and peak work rate. No change: arm and trunk fat, maximal inspiratory pressure, peak O 2 uptake, lactic acidosis threshold, constant work rate duration.

5
T + strength training group: Compared to placebo groups: "arm, trunk, leg, and total lean mass; #leg, total, and % fat.
Compared to non-training groups: "leg press strength and peak oxygen uptake; #leg press fatigue. Compared to T + no training: "peak work rate. Compared to placebo + no training: "lactic acidosis threshold. No change: arm and trunk fat, maximum inspiratory pressure, constant work rate duration.
T enanthate 200 mg (n = 7) or placebo (n = 7) IM every 2 weeks x 12 weeks, with muscle testing every 4 weeks; randomization method not given.
No change in handgrip, knee extensor, and knee flexor strength, leg extensor power, step height. 237 men 60-80 years old with total T <395 ng/dL (13.7 nM); 30 dropouts were excluded from analysis.
T undecanoate 80 mg (n = 120) or placebo (n = 117) capsules twice daily x 6 months; a randomization list without stratification using blocks of 6 was computer-generated using the ADLS.  167 men over 60 years old (mean age 66) with total T 200-350 ng/dL (6.9-12.1 nM); 24 dropouts were excluded from analysis.
T gel 1% 2.5, 5.0 g (n = 111), or placebo (n = 56) daily x 12 months, with or without progressive resistance exercise training; randomization was performed using permuted block randomization with random block sizes.
T treated groups analyzed in aggregate. In subjects with resistance training, no change in physical functional performance tests: upper body strength and flexibility, lower body strength, balance, and endurance, stair climb speed and 6-min walk, strength (bench press, incline press, overhead pull-down, seated row, average upper body, grip strength, knee extension, knee flexion, seated leg press, average lower body), power (leg extensor). #Fat mass; "overall and arm fat free mass. No change in weight, BMI, trunk fat mass, appendicular and leg fat-free mass, waist and hip circumferenceIn subjects without PRT: "Lower body strength, strength for bench and incline press, average upper body and grip strength. No change in physical functional performance test total score, upper body strength and flexibility, lower body balance and endurance, stair climb speed and 6-min walk, strength for overhead pull-down, seated row, knee extension/flexion, seated leg press, and average lower body, power (leg extensor). "Appendicular and arm fat free mass. #Overall and trunk fat mass, waist circumference. No change in weight, BMI, overall and leg fat-free mass, hip circumference 5 Kenny et al, 2001 [166] 67 men >65 years old (mean age 76) with bioavailable T <128 ng/dL (4.44 nM); 23 dropouts were excluded from analysis.
T patch 5 mg (n = 34) or placebo (n = 33) daily x 1 year; all received 500 mg calcium and 400 IU vitamin D supplements; randomization method not given.
No change in BMI, % body fat or lean mass. No change in PASE (Physical Activity Scale for the Elderly) score. PASE includes 8-ft walk speed, chair rise, single-leg stance, supine-to-stand, and get up and go time. Kenny et al, 2010 [80] 131 men !60 years old (mean age 77.1) with T <350 ng/dL (12.1 nM), physical frailty, and BMD T-score at the hip -2.0 or a nontraumatic fracture within last 5 years; delays in recruitment and lack of funds resulted in !16 months of follow-up with analysis performed at 12 months; 69 dropouts excluded from analysis.
T gel 5 mg (n = 69) or placebo (n = 62) daily x 16 months; all subjects maintained calcium intake of 1500 mg/day and were given 1000 IU cholecalciferol/day; randomization with block sizes (2 or 4) stratified by frailty status "Total body lean mass, appendicular skeletal muscle mass; No change in total body fat %. No change in hand grip/ leg press strength or PASE  T enanthate 100 mg (n = 37) or placebo (n = 34) IM weekly with low-or highintensity resistance strength training x 12 weeks; randomization stratified by arbitrary score of physical ability; within each stratum, subjects randomized with blocks sizes in multiples of 4 and randomly varied No change in strength for chest and leg press, performance testing score (sit-tostand maneuver, habitual and maximal safe gait speed tests, stair climb). "Midthigh cross-sectional muscle area 5

Travison et al, 2011[83]
209 men !65 years old with total serum T 100-350 ng/dL (3.5-12.1 nM) or free serum T <50 pg/mL (174 pM) with mobility limitations; analysis restricted to 165 men with a baseline assessment and at least one outcome assessment.
T 100 mg (n = 106) or placebo (n = 103) gel daily x 6 months; after 2 weeks, dose level was increased or decreased by 50% based on serum T; randomization by computer-generated table in blocks of 6 stratified by age "Strength for leg and chest press, loaded stair-climbing power, loaded walk speed. No change in grip strength, unloaded walking speed, unloaded stair-climbing power. "Total lean mass, appendicular skeletal muscle mass. #Total fat mass, appendicular fat mass. "Cardiovascularrelated AEs with T (23 vs 5) Testosterone, compared to finasteride or placebo, increased strength (leg press, knee flexion and extension, chest press, triceps extension, and grip strength), body fat-free mass. Total fat mass reduced 3.87 kg. Lumbar spine and hip BMD increased 5 [86] 25 men 58-86 years old undergoing knee replacement surgery; 3 dropouts were excluded from analysis.

Amory et al, 2002
T enanthate 600 mg (n = 10) or placebo (n = 12) IM 21, 14, 7, and 1 day before surgery; randomized using a random number sequence. 15 men !65 years old who had been admitted to the Geriatric Evaluation and Management unit for rehabilitation; 1 subject in the T group died unexpectedly 2 weeks after admission on the day of his planned discharge probably as the result of a cardiovascular event.
T enanthate 100 mg (n = 9) or placebo (n = 6) IM weekly until discharge or for a maximum of 8 weeks; subjects received rehabilitation therapy as appropriate to their needs; randomization method not given.
"Functional Independence Measure and grip strength compared to baseline; difference compared to placebo not reported, but no difference in either endpoint compared to placebo by t-test performed by us.

4, 5
Dohn et al, 1968 [13] 44 men with leg claudication or ulcers attributed to arteriosclerosis in 43 men and to Buerger's disease in 1 man; age not given. Two men did not complete study. Numbers in tables add to 86 subjects. Not possible to tell for sure how many men were analyzed.
Aqueous T isobutyrate 300 mg or meprobamate as placebo every 14 days for 3 months; route of administration not given; double-blinded, randomization not discussed No effect on subject improvement, walking test 3
T mixed esters 200 mg (n = 18) or placebo (n = 16) IM every 2 weeks x 12 months; all received calcium carbonate 600 mg daily; randomization method not given. Nandrolone not discussed.

3
Griggs et al, 1989 [89] 40 men 18-65 years old (mean age 33.3 years for T and 41.5 years for placebo) with myotonic dystrophy and a typical distribution of weakness (but ambulatory); 3 dropouts were excluded from analysis. T enanthate or placebo 3 mg/kg IM weekly x 12 months; number of subjects/group not given; evaluations every 3 months; randomization performed using computergenerated random numbers, stratified by grip strength and previous T tx "Time to climb stairs. No change in arm volume, weight, manual muscle testing, myometry, grip dynamometry, forced vital capacity (VC), maximum voluntary ventilation, maximum expiratory pressure, time to walk 30 feet, time to cut standard square. "Creatinine excretion ("muscle mass) 4
200 mg T (n = 19) or placebo (n = 17) IM weekly for 3 weeks, then once every second week for 6 months; consecutive patients assigned to tx or placebo by record numbers.
No change in grip strength, walking distance.

Svartberg et al, 2004[43]
29 men 54-75 years old (mean age 66 years) with moderate to severe COPD; 2 dropouts were excluded from analysis.   No difference between groups in muscle strength or body weight. Both groups improved in 6-minute walking distance; no post-intervention effect was seen. 3

Giannoulis et al, 2006[168]
Healthy men 65-80 years old with circulating IGF-I <50 th percentile for the local age-specific range; of those meeting criteria, the 80 subjects with the lowest T concentration but without T deficiency, mean 397.6 ng/dL (13.8 nM); 5 subjects were excluded because PSA concentrations met revised exclusion criteria and 6 dropouts were excluded from analysis.
T 5 mg or placebo transdermal patch daily x 6 months; randomization was performed using computer-generated pre-allocated study numbers; growth hormone arm not discussed here No change in knee flexion/extension peak torque/force, hand-grip peak force, VO 2 max. No change in BMI, hip/waist ratio, lean body mass, subcutaneous/ cross-sectional abdominal visceral/total body fat, mid-thigh muscle cross sectional area. reported an improvement in at least 1 measure of muscle strength. In studies of men with HIV, 2 [101,107] of 5 studies reported an improvement in at least 1 measure of muscle strength; 3 showed no effect. [41,99,103] Twenty-four studies evaluated the effects of testosterone treatment on physical function endpoints and, of these, 5 found an improvement in at least 1 measure of function. [18,22,81,83,86] Neither of the 2 studies of HIV patients measuring physical function showed an improvement in function. [41,100] Six of these studies had a Jadad score of 4 or 5. [18,41,81,83,86,100] In summary, the majority of studies show increased muscle mass but no effect of testosterone on muscle strength or function.

Mood and Behavior
Forty-five studies evaluating the effect of testosterone on mood and behavior are summarized in Table 4. Twenty-nine of these studies focused on men without psychiatric disorders, and 16 on men with psychiatric disorders.

Healthy men.
Some studies of mood and behavior were designed to evaluate the potential adverse effects of anabolic steroid abuse. For example, men abusing anabolic steroids have been described as having "Roid Rage. " We did not evaluate steroid abuse studies, but we reviewed studies on testosterone preparations and their association with anger, aggression, and other mood alterations. There was little consistency among the studies we reviewed.
Five studies reported treatment-associated increases in anger, aggression, or hostility. [67,[108][109][110][111] Only two of these studies had a Jadad score of 4 or 5. [108,111] One study, [110] with a Jadad score of 3, determined that testosterone gel applied to the skin increased hostility based on evaluations by 2 undergraduate judges of a free-text paragraph written by each subject to describe his mood at the end of treatment. We do not know the reliability of this assessment. Two studies (Jadad score 3 and 5) reported a decrease in anxiety after testosterone treatment. [45,112] Seventeen of 29 studies reported no effect of testosterone treatment on personality, psychological well-being, or mood. [22,24,38,59,63,85,104,[113][114][115][116][117][118][119][120][121] Seven of 17 studies had a Jadad score of 4 or 5. One of these studies could not be evaluated because only a composite score for mood and sexual function was reported. [38] The study that used hostility assessments by undergraduate judges found no change in personality as assessed by the Gough and Heilbrun Adjective Check List. [110] Another study in this group reported that elevation of testosterone serum concentrations above normal using testosterone gel was associated with an increase in selfishness on a computer game that evaluated the willingness to give away small  [41] 61 HIV-positive men 18-60 years old (mean age 43) with unintentional weight loss or BMI <20 and serum T <400 ng/ dL (13.9 nM); 4 placebo-assigned and 9 T-assigned subjects dropped out and were analyzed using the last observation carried forward.
T enanthate 300 mg (n = 30) or placebo (n = 31) IM weekly x 16 weeks; computergenerated randomization list with a block size of 6 No change in physical function measures (stair-climbing power, walking speed, load-carrying ability), muscle performance (leg press strength, leg press power, leg press fatigability). "Fatfree mass. No change in fat mass, total weight, % body fat.
T undecanoate 160 mg or placebo daily for 2 months followed by opposite tx; randomization of tx order not described but said to be balanced. Improved self-rated anxiety and tension; no statistically significant effect on depression, anger, vigor, or fatigue (authors concluded otherwise based on P = .10).
T enanthate 200 mg IM weekly for 8 (n = 16) or 4 weeks with placebo IM weekly for the remaining 4 weeks (n = 15); randomization not discussed. Described as single-blinded.
No effect on self-assessment of mood (cheerful, lethargic, relaxed, tense, energetic, unhappy, irritable, ready to fight, easily angered) using an unvalidated ranking scale.

Pope et al, 2000[169]
53 presumed healthy men aged 20-50 years; an additional 13 men were recruited but not randomized or were randomized and not evaluable.
T or placebo IM every 2 weeks. T dose was 150 mg/week for 2 weeks, 300 mg/ week for 2 weeks, and 600 mg/week for 2 weeks. After a 6-week washout period, opposite tx given. Method of randomization not given

Daly et al, 2001[109]
20 healthy men aged 18-42 years. PO placebo daily for 3 days, then PO methyltestosterone 40 mg/day for 3 days, then PO methyltestosterone 240 mg/day for 3 days, then PO placebo for 3 days inpatient. Schedule fixed but unknown to subjects and raters (not clear who the raters were inasmuch as subjects self-rated).
High-dose methyltestosterone associated with visual analogue scale (VAS) self-rating of "Cognitive ability, "Distractibility, "Energy, "Sexual arousal, "Aggression, "Irritability.  [117] 30 healthy eugonadal men aged 23-40 years; 8 hypogonadal men included but given only active tx and so not summarized here. One subject dropped out; it is not clear if he was retained in the analysis.

(Continued)
Testosterone enanthate 200 mg IM or placebo weekly for 8 weeks (n = 15/ group). Method of randomization not given The authors reported a lack of effect of testosterone on mood and aggression in eugonadal men, but the analyses shown in the paper were chiefly between hypogonadal men and eugonadal men. Kunelius et al, 2002 [118] 120 men, age 50-70 years, with "andropause" symptoms, serum T <15 nM (432 ng/dL). Six subjects dropped out; it is not clear whether they were included in the analysis.
"Randomized" by sealed envelope to 2.5% DHT gel (125 mg/day DHT) or placebo (60 subjects per group) for 30 days after which dose was adjusted on serum DHT concentration to 187.5 or 250 mg/day DHT. Placebo dose was randomly adjusted. Tx duration was 6 months.
No effect on well-being or mood as assessed by questionnaire.
T undecanoate 1000 mg or placebo IM at the beginning of an 8-week tx phase followed by 8-week washout followed by the opposite tx. Self-assessment each week and psychometric assessments during week 4 of each tx phase "Anger-hostility (during first 2 weeks after injection) and #fatigue-inertia scores on POMS.  T enanthate 125 (n = 14) or 0 (n = 16) mg/week IM for 23 months. There also were finasteride + T arms that we ignore here.
T-associated reduction of a mean of 0.74 items on the 15-item Geriatric Depression Scale short form Subjects were block randomized to receive testosterone undecanoate 1000 (n = 97) or 0 (n = 102) mg IM at week 0, 6, and 18.
No improvement in Hospital Anxiety and Depression score (HADS) in men with baseline T 8.0 nM; "HADS-depression score but not HADS-anxiety in men with baseline T 8.1-12 nM 5

Malkin et al, 2004[1]
12 men 60.8 ± 4.6 years old (mean ± SD) with CAD and "clinical need for T replacement." One man failed screening and another withdrew at unspecified point in the study.
100 mg T or placebo IM every 2 weeks for 4 weeks, 1 month washout, then tx switch; randomization by computer; single-blinded #BDI score 5 (Continued)
Improved BDI by a mean of 5.8 points; improvement was associated with increased weight. BDI did not change with placebo. Men with BDI >18 at baseline were not analyzed separately. 3
Randomization in blocks of 4 by computer-generated numbers to T cypionate (n = 38 overall, 26 with depression diagnosis) or placebo (n = 32 overall, 7 with a depression diagnosis) injected biweekly (presumably IM) for 6 weeks; the first T dose was 200 mg, subsequent doses were 400 mg. Openlabel phase with T followed the 6-week double-blind study, not summarized here Clinical Global Impression (CGI) scale response (not defined) in 74% T and 19% placebo (P < 0.001). No effect of tx on CGI in subjects with a depression diagnosis (P = .08). Improvement in total and vegetative scores on HAM-D, but not the affective scale. No change in the BDI (P = .052) 4 [130] 123 HIV-positive men, mean age 41, with MDD or dysthymia; 33 men dropped out and were retained for ITT analysis.

Rabkin et al, 2004
Randomization in blocks of 6 by computer-generated numbers to T cypionate IM biweekly + daily PO placebo (n = 38), placebo IM biweekly + fluoxetine PO (20-40 mg) daily (n = 46), or placebo IM biweekly + placebo PO daily (n = 39). Initial T dose was 200 mg, subsequent doses were 400 mg. Within tx groups, subjects were also randomized to fluoxetine or placebo daily by mouth. Tx were given for 8 weeks. Eight of the subjects were randomized only to fluoxetine or placebo due to transient unavailability of T.
No difference in response on HAM-D or BDI by ITT and by considering only completers. Improvement in Chalder Fatigue Scale compared to placebo and to fluoxetine 4  T cypionate 200 mg (n = 13) or placebo (n = 10) IM every 10 days for 6 weeks; randomization by computer Greater improvement in HAM-D and BDI than with placebo. "Remission (CGI of 1 or 2 or HAM-D < 8) for 7 men on T (n = 13) compared to 1 man on placebo (n = 10).

5
Shores et al, 2009 [126] 33 men !50 years old (mean age 59 years) with dysthymia or "minor depression" and T concentration 280 ng/dL (9.7 nM); 6 subjects discontinued and were included in ITT analysis. T gel 7.5 g (n = 17) or placebo (n = 16) for 12 weeks; "randomization" by computer using 1:1 ratio  amounts of money. [122] Two additional studies from the same group in non-depressed men with CHF did not show an effect of testosterone on the Beck Depression Inventory (BDI), [20,27] although the earlier of these studies concluded otherwise based on a finding that was not statistically significant. A study in non-depressed men with metabolic syndrome reported an improvement in the BDI in testosterone-treated compared to placebo-treated subjects. [52] A study found non-depressed men older than 60 years to have a mean 5% decrease in a geriatric depression scale when administered testosterone. [102] This study had a Jadad score of 1.
Another study [58] found that testosterone treatment had no effect on the Hospital Anxiety Depression score (HADS) in men with testosterone 8.0 nM but improved the depression subset of the HADS in men with testosterone of 8.1-12 nM. Malkin et al. [1] found that 100 mg testosterone every 2 weeks improved the BDI score. This study had a Jadad score of 5.

Men with psychiatric diagnoses.
Twelve studies (3 in HIV-positive men) evaluated testosterone supplementation in men with a diagnosis of depression or dysthymia (sometimes also called, "minor depression"), 1 study evaluated the use of testosterone in men with schizophrenia, and 2 studies were conducted in men with Alzheimer disease or cognitive impairment. The study in schizophrenic men used testosterone or placebo gel in addition to whatever treatment the subject was already using. [123] There were improvements in the negative symptom scores on a standardized scale but no change in the Calgary Depression Scale for Schizophrenia. The authors used an intention-to-treat (ITT) analysis and implied that better results were seen among subjects who completed the study; however, there were no significant differences in the depression scores between testosterone and placebo among completers. Two studies in men with cognitive impairment or Alzheimer disease (Jadad score 3) found no effect of treatment on neuropsychiatric symptoms, depression, behavior, or quality of life (QoL). [42,124] Caregiver-assessed QoL was improved in 1 of these studies. [42] The response of depression and dysthymia to testosterone was mixed and inconsistent. Among HIV-negative men, four studies (all with a Jadad score of 4 or 5) showed testosteroneassociated improvements in standard scoring systems for depression and/or in the proportion of subjects who achieved remission of their psychiatric disorder. [37,51,125,126] Four other studies (2 with a Jadad score of 4 or 5) showed no improvement in depression or dysthymia with testosterone compared to placebo. [36,111,127,128] One study (Jadad score 4) showed a transient improvement in depression and melancholia after 3 months of treatment that was no longer apparent after 6 months of treatment. [48] Because it has been noted that HIV-positive men can be depressed and "hypogonadal, " 3 studies administered testosterone to HIV-positive men with depression or dysthymia. [35,129,130] Two of the studies had a Jadad score of 4[35, 130] and one study had a Jadad score of 3.
[129] Testosterone treatment had inconsistent effects on measures of depression; one study showed a 5.8-point improvement in the Beck Depression Inventory (BDI) in men with HIVassociated wasting, although the improvement may have been explained by an increase in weight. [129] Another study showed a testosterone-associated improvement in HIV-positive men overall in the Clinical Global Impression (CGI) scale but not among subjects with a depression diagnosis. [35] This study also showed improvement in the total and vegetative symptom scores of the Hamilton Rating Scale for Depression (HAM-D) but not in the affective scale, and there was no significant change in BDI scores. A subsequent, larger study by the same group showed no difference in response of depression measured by HAM-D or BDI in men given testosterone compared to placebo. [130] Authors attributed the mixed responses in the literature to the considerable placebo response in most studies and to the possibility of an idiosyncratic response to testosterone, with putative subgroups of responders who were difficult to identify a priori. [111,127] The studies, however, did not show consistent responses in subgroups of men who had low serum testosterone concentrations, depression resistant to standard therapy, or men characterized as middle-aged or elderly. In studies in which serum testosterone concentrations were measured on therapy (both with a Jadad score of 5), response of depression or dysthymia was not consistently associated with serum hormone concentration.

Cognition
Twenty-two studies evaluating the effects of testosterone on cognition are summarized in Table 5. Seventeen focused on men without cognitive impairment and 4 focused on men with cognitive impairment.
3.5.1 Men described as normal. Ten studies evaluated the effects of testosterone treatment on cognitive endpoints in healthy men. Spatial cognition/memory was reported to be improved with testosterone supplementation in 3 studies, [114,131,132] unchanged in 2 studies, [42,133] and poorer with supplementation in 1 study. [134] Although 1 study reported improved working memory [135] and 1 study found improved verbal fluency, [134] most other studies found no improvement in verbal fluency, memory, or other cognitive endpoints in healthy men given testosterone. [42,91,120,131,133,134,136] Two of 5 studies that showed improvement and 4 of 7 of the studies that showed no improvement had a Jadad score of 4 or 5.
3.5.2 men described as hypogonadal. Hypogonadal men, variously defined, were found in 1 study to have better verbal learning and reversal of digits on number sequencing with testosterone supplementation, [112] but no effect on the same domain was found in another study. [137] The study showing an advantage used injected testosterone enanthate 200 mg while the negative study used a daily 5 mg patch. Another injection study found no effect of supplementation on memory in hypogonadal men. [104] One study reported a possible disadvantage of treatment with dihydrotestosterone compared to placebo in performance on the Modified Mini-Mental State Examination (MMSE), [11] but data were not shown and the putative difference could not be evaluated. Another study showed no improvement in visuospatial cognition or MMSE with testosterone treatment for 12 months. [119] All studies had Jadad scores of 4 or 5.
3.5.3 Men with cognitive impairment. Treatment of men with suspected or diagnosed Alzheimer disease or cognitive impairment was reported in five studies, two of which had a Jadad score above 3. Although 1 injection study found an improvement on the Alzheimer Disease Assessment Scale-Cognitive Subscale (ADAS-COG), [138] another study using testosterone gel found no effect on the same instrument or on other cognitive function tests. [42] Spatial and verbal memory were improved after 6 weekly injections of testosterone enanthate in 1 study, but the effect did not persist during a 6-week washout period without treatment. [132] A fourth study found no effect of testosterone injections on behavior, activities of daily living (ADLs), or cognition. [124] The fifth study found that transdermal testosterone gel was not associated with statistically significant changes in measures of cognition, mood, or quality of life. [139] 3.5.4 Proposed explanations for inconsistent results. Because study results have been varied and inconsistent, some authors have proposed that testosterone is not the only factor or even the most important factor in cognitive function. Janowsky et al [114] found improved spatial cognition in men treated with scrotal testosterone patches, but there was an imbalance between placebo and testosterone groups in baseline blood concentrations of 17β-estradiol, which these authors attributed to chance. The effect of testosterone and 17β-estradiol on spatial cognition testing was explored using post-hoc testing, and the putative testosterone effect on spatial cognition appeared to be associated with suppression of 17β-estradiol by testosterone supplementation rather than a direct effect of testosterone. This study had a Jadad score of 3.  Most authors with an interest in 17β-estradiol have suggested that the effectiveness of testosterone, when it has shown effectiveness, is due to aromatization to 17β-estradiol. Cherrier et al [131] measured testosterone and 17β-estradiol concentrations after injection of testosterone supplements in healthy men and reported that both testosterone and 17β-estradiol concentrations were associated with recall of a test story, but only 17β-estradiol concentrations were associated with performance on the Stroop test. In another study, Cherrier et al reported that only men with an increase in 17β-estradiol concentration after testosterone supplementation showed improvements in verbal memory testing. [136] This study used administration of the aromatase inhibitor anastrozole to differentiate between effects attributable to testosterone and those that might be due to 17β-estradiol. Another study without anastrozole found 17β-estradiol serum concentrations after testosterone therapy to be a significant predictor of performance on verbal memory testing. [140] All three studies had a Jadad score of 5.
It has also been suggested that testosterone supplementation has produced inconsistent results in cognitive function studies because the blood concentrations achieved by supplementation need to be in an optimum range for effectiveness. Under this hypothesis, over-supplementation is as ineffective as under-supplementation. Cherrier et al [140] administered testosterone enanthate IM at 0, 50, 100, or 300 mg weekly and administered tests of verbal and spatial memory. Results were not reported according to treatment group; rather, subjects were divided into those with no, moderate, or large increases in serum testosterone concentrations over baseline. These response categories were defined based on 1 standard deviation above the control response and 1 standard deviation above the response to 100 mg. Subjects with a moderate increase in serum testosterone concentration over baseline (defined as 11-50 nM) performed better on cognitive testing than those with "no increase" (0-10 nM) or a large increase (>51 nM). Seventeen of the 22 men in this moderate-increase group had received testosterone 100 mg/week with the balance evenly divided between the 50 mg and 300 mg doses. The authors explained that they did not use tertiles or quartiles, because using quartiles or tertiles resulted in some placebo patients with significant changes from baseline, raising the question of whether men on placebo with an increase in their serum testosterone concentration were distinguishable on cognitive testing from men who received testosterone supplementation. This study had a Jadad score of 5.
In summary, there is no support for the use of testosterone to enhance cognition in normal or cognitively impaired men. 22 men age 60-90 with mild cognitive impairment and total T concentration <300 ng/dL (10.4 nM) and AUA symptom score <19 Transdermal testosterone gel 50 to 100 mg/day, with a target total T level of 500 to 900 ng/dL (n = 10) or placebo gel (n = 12)

Discussion
This systematic review examined published RCTs of testosterone supplementation for cardiovascular disease or surrogates of cardiovascular disease, sexual function, muscle strength, mood, and cognition. The review was limited to published studies in English and to trials indexed before April 9, 2016. The evidence supporting the use of testosterone for preventing or treating cardiovascular disease is inconsistent and, on balance, unconvincing. Some evidence supported an acute and chronic effect of testosterone therapy on increasing time to ST-segment depression, and there is evidence of improvement in some measures of congestive heart failure. Most studies showed no effect of testosterone therapy on inflammatory markers, and the effects on lipids were inconsistent.
Studies that examined clinical effects have not favored testosterone therapy over placebo. Two of 3 studies that assessed angina showed no effect. Three studies from the same group found a benefit for symptoms associated with CHF. One study was stopped early for cardiovascular adverse effects.
Testosterone supplementation did not demonstrate consistent effectiveness for improving sexual function or satisfaction. Testosterone is ineffective in treating ED. Controlled trials were mixed on libido, with more positive than negative studies. Substantial evidence supports a favorable effect of testosterone treatment on muscle mass in both healthy men and men with HIV, and a majority of studies showed a decrease in fat mass. Testosterone did not affect most measures of muscle strength. While decreasing frailty and increasing strength in older men might be beneficial, testosterone supplementation does not improve physical function in older men.
Most studies on mood-related endpoints found no beneficial effect of testosterone treatment on personality, psychological well-being, or mood. Although 2 studies showed decreased anxiety, treatment of depression showed mixed and inconsistent results. Even if testosterone did benefit mood, social adverse events might ensue; 5 studies noted treatmentrelated increases in anger, aggression, or hostility. Testosterone did not benefit cognitive impairment or Alzheimer disease; neither did it benefit verbal fluency, memory, or other cognitive endpoints in normal men.
In summary, evidence from RCTs does not support treatment of so-called low-T for improving physical function, sexual function, mood, or cognition. Testosterone increases muscle mass, but not strength, and while some improvement is seen in some surrogate markers of cardiovascular risk, there is little evidence of clinical benefit.
There is conflicting evidence on the association between testosterone supplementation and cardiovascular events. RCTs have reported increased cardiovascular risk with testosterone therapy. One such trial that specifically examined cardiovascular disease and mortality endpoints was stopped early because of an increased risk of cardiovascular events. [17] A metaanalysis of 2994 men in 27 randomized controlled trials through 2012 found that testosterone therapy increased the risk of cardiovascular events (OR, 1.54; 95% CI, 1.09-2.18). [141] Observational studies examining the effect of testosterone treatment have shown conflicting results on risk. A Veterans Administration study evaluated men who had undergone coronary angiography and had a total testosterone concentration (presumably plasma) less than 300 ng/ dL (10.4 nM). [142] Men who were treated with testosterone had an increased risk of all-cause mortality, MI, and stroke compared to men who did not use testosterone (HR, 1.29; 95% CI, 1.05-1.58), based on a mean of 27.5 months of follow-up. Another retrospective cohort study using Veterans Administration data showed a lower rate of all-cause mortality, myocardial infarction, and stroke among testosterone-treated men whose testosterone concentrations "normalized" after treatment. [143] Another observational study of men in a large, integrated health care organization found that death rates were reduced over 3 years, but there was no effect on myocardial infarction or stroke. [144] A Medicare-based study identified testosterone exposures and MI outcomes using claims data and matched testosterone-treated with untreated subjects using an empirically derived propensity score and found no increased risk. [145] The adjusted HR for testosterone therapy and MI was 0.84 (95% CI 0.69-1.02). Analysis of subjects in the highest quartile propensity score range suggested a protective effect of testosterone treatment, with a HR of 0.69 (95% CI 0.53-0.92). An observational study in men with low testosterone found that treatment was associated with reduced mortality; [146] another in diabetics [147] reported benefit on all-cause mortality but excluded men who had received testosterone for less than one year and excluded deaths occurring before six months. A large cohort study found that myocardial infarction rates were significantly increased within three months of testosterone treatment initiation; testosterone-treated men over 65 experienced double the rate of myocardial infarctions compared to men who did not received testosterone. [148] Testosterone treatment has been considered for disease prevention because men who are obese, diabetic, hypertensive, or chronically ill have lower plasma concentrations of testosterone. [149] However, the direction of causality is unclear; it is possible that obesity or lack of exercise and chronic disease lower testosterone rather than low testosterone concentrations causing disease. It is also possible that another mechanism both lowers testosterone concentrations and increases the risk of some diseases. Observational studies attributing positive health effects to testosterone may be affected by an increased likelihood of healthier men being prescribed testosterone rather than testosterone improving health.
There are parallels between the recommendation of testosterone and of menopausal hormone therapy in women. Physicians prescribed estrogen and estrogen-progestin preparations to menopausal women to prevent cardiovascular disease because observational studies showed that women who took menopausal hormones had less heart disease than women who did not. RCTs, however, showed that menopausal hormone therapy increased the risk of heart attacks and stroke. [150][151][152][153] It is likely that healthier women chose to take menopausal hormone therapy, but menopausal hormone administration did not improve health.
In 2012, sales for testosterone therapies exceeded $2 billion, and sales continue to grow in dozens of countries. [154] To the extent that this increase in use of testosterone supplementation is based on anticipated improvements in cardiovascular health, sexual function, physical functioning, mood, or cognition, we suggest that it might represent therapy without adequate clinical trial support. We identified no population of normal men for whom the benefits of testosterone use outweigh its risk. Given the known risks of testosterone therapy and the lack of evidence for clinical benefits in normal men, we do not think further trials of testosterone are necessary.