The Effects of Pharmacologically Induced Hypogonadism on Mood in HealthyMen FREE

The age-related decline in circulating androgen levels progresses tohypogonadism in a substantial number of men^{1 - 2} andmay have a negative effect on bone metabolism, muscle mass, and, possibly,mood and behavior.^{3 - 5} Androgenshave been implicated in the regulation of mood by several types of evidence.First, in controlled clinical studies, supraphysiologic doses of androgens(eg, anabolic-androgenic steroids) induce prominent mood changes (hypomania,irritability) in approximately 5% of eugonadal men.^{6 - 8} Second,in placebo-controlled clinical trials, physiologic doses of testosterone produceantidepressant-like effects in hypogonadal men in some^{9 - 11} butnot all¹² studies. Clearly, the mood effectsof androgens are not uniformly observed in men, and the reported antidepressantresponse may reflect the reversal by testosterone of age-¹³ orillness-related sarcopenia (reductions in muscle mass) and fatigue¹⁰ instead of a direct action of androgen on mood. Moreover,several contextual factors could influence and, hence, complicate the interpretationof the inferences drawn about the relationship between androgens and mood,including impact of concomitant medical conditions, histories of psychiatricillness, substance abuse, and prior exposure to anabolic-androgenic steroids.^{6 ,14 - 16} Nostudy has examined the effects on mood of the acute withdrawal of testosteronein eugonadal healthy young men.

The purpose of this study was to address existing confounds by examiningthe effects on mood of the acute suppression of testosterone secretion withpharmacologically induced hypogonadism in young, healthy adult men with nohistory of psychiatric illness or substance or anabolic-androgenic steroidabuse.

Subjects were men aged 18 to 45 years recruited through advertisementsand referred from the National Institutes of Health (NIH) Normal VolunteerOffice. All were medication free, had no significant medical illness (currentlyor in the past 2 years), and had normal laboratory results. Specifically,complete blood cell counts, blood chemistry, thyroid function tests (thyroid-stimulatinghormone and free thyroxine), and prostate-specific antigen levels were withinnormal limits. Additionally, plasma total testosterone levels ranged from355 to 992 ng/dL (12.3-34.4 nmol/L) (normal range, 300-1200 ng/dL [10.4-41.6nmol/L]), and plasma prolactin levels were within normal limits (1-16 ng/mL)in all but 1 subject whose plasma prolactin was 29 ng/mL on repeat testing.The absence of current or past psychiatric illness was confirmed by a structuredpsychiatric diagnostic interview¹⁷ and dailysymptom self-ratings. Structured interviews were performed by 1 of 4 of us(P.J.S., M.A.D., N.A.H., or C.A.R.). Subjects were excluded from this studyif they had a past or present psychiatric illness or evidence of persistent(>3-5 days) clinically significant mood and behavioral symptoms of moderateseverity on the daily symptom rating form (see “Outcome Measures”)during their screening phase. The protocol was reviewed and approved by theNational Institute of Mental Health intramural research board, and oral andwritten informed consents were obtained from all subjects. All subjects werepaid for their participation in this protocol according to the guidelinesof the NIH Normal Volunteer Office.

This was a double-blind assessment of the effects of the acute inductionof hypogonadism and subsequent replacement in a crossover design with testosteroneand placebo. After a 2-month screening phase, men received 7.5 mg of intramuscular(IM) depot leuprolide acetate (Lupron; TAP Pharmaceuticals, Chicago, Ill)every 4 weeks for 3 months. Lupron alone was administered for the first 4weeks. Subjects then received, in addition to Lupron, 200 mg of IM testosteroneenanthate (Bristol-Myers Squibb, New York, NY) or 1.5 mL of IM placebo (sesameoil as color-matched vehicle) every 2 weeks for 1 month (ie, twice) and thencrossed-over to the other replacement. The order of replacement was randomlyassigned and counterbalanced. Men were seen at the National Institute of MentalHealth clinic every 2 weeks throughout the study. Blood samples were obtainedand symptom self-ratings were completed at each clinic visit on a biweeklybasis throughout the study. Both subjects and raters were blinded to the orderof replacement. All subjects were taking Lupron throughout the study, andtherefore their endogenous testosterone secretion was suppressed, obviatingthe need for collecting samples at a uniform time point. Blood samples weretaken solely to confirm hormone levels during each of the pharmacologicallyinduced hormone conditions. Each individual came into the clinic at approximatelythe same time during the study, but the time of day for visits varied acrossindividuals. Blood samples were centrifuged, aliquoted, and stored at −70°Cuntil time of assay.

To assess the severity of mood symptoms, the following symptom ratingforms were completed at baseline and during each hormonal condition: (1) avisual analogue scale^{18 - 19} completednightly for all symptoms; scores range from 0 (symptoms present in the extreme)to 100 (symptoms not present) and reflect the subject’s symptoms atthe time the ratings were completed; and (2) the Daily Rating Form, a 6-pointLikert-type scale modified to include the symptoms measured in this study,²⁰ also completed nightly, to represent a compositerating for the previous 12 hours; scores range from 1 (symptoms not present)to 6 (symptoms present in the extreme). The Daily Rating Form symptoms consistof the following: avoidance of social activity; loss of enjoyment or interest;impaired function at work or home; irritability or anger; impaired concentrationor distractibility; mood swings; feeling depressed, sad, low, or blue; feelinganxious or nervous; decreased eating; increased eating; more sleep, naps,or lying in bed; low energy; loneliness or feeling rejected; feeling physicallyrestless or agitated; feeling powerful, emotionally charged, or pumped up;increased sexual interest; decreased sexual interest; disturbed sleep; drinkingalcohol or using nonprescribed drugs; impulse to hurt self; impulse to hurtsomeone else; acting on impulse to hurt someone; daytime hot flushes; andnighttime hot flushes. Thirteen symptoms recorded by the visual analogue scaleconsisted of the following: rapidly changing mood, increased appetite or cravings,a global feeling (best ever/worst ever), impulse to hurt others, low self-esteem,impulse to hurt self, sadness, irritability, low energy, functional impairment,anxiety, extreme physical discomfort, and isolation and social avoidance.Four men did not complete the Daily Rating Form ratings, and 2 men did notcomplete the visual analogue scale ratings. The following standardized ratingscales also were completed during each clinic visit: the Beck Depression Inventory(BDI), a measure of depression severity,²¹ andthe Spielberger State-Trait Anxiety Inventory, a measure of anxiety severity.²²

Because measures of hostility have been correlated with testosteronelevels in a variety of studies (albeit not uniformly),^{23 - 30} weattempted to determine whether induced hypogonadism would be associated witha reduction in self-ratings of aggression. A subsample of 20 men completedrating forms that assessed changes in the subjects’ experiences of aggression,anger, and impulsiveness as follows: (1) Buss-Durkee Hostility Inventory (a75-item scale measuring the subscales of assault, indirect hostility [subjectis not direct target of hostility], irritability, verbal hostility, guilt,suspicion, resentment, and negativity)^{31 - 32} ;(2) Anger, Irritability, and Assault Questionnaire³³ (a42-item scale assessing variables such as irritability, verbal assault, indirectassault, direct assault, and anger); and (3) Barratt Impulsiveness Scale version7B (a 48-item scale measuring risk taking, interpersonal behavior, motor behavior,self-assessment, and sensory stimulation).^{34 - 35} (All3 rating forms were modified to reflect a subject’s experience duringthe 2 weeks prior to completing the scales.)

Blood levels of testosterone, free testosterone, estradiol, and dihydrotestosteronewere measured by radioimmunoassay, as described previously^{36 - 41} (QuestDiagnostics, Baltimore, Md, and Covance Laboratories, Vienna, Va).

The 7-day averages of the daily symptom scores were calculated duringthe fourth week of the 3 experimental conditions: baseline, Lupron plus testosterone,and Lupron plus placebo. Cross-sectional rating scores (for the BDI, SpielbergerState-Trait Anxiety Inventory, Buss-Durkee Hostility Inventory, Anger, Irritability,and Assault Questionnaire, and Barratt Impulsiveness Scale version 7B) weretaken from the last week of each phase. The averaged daily and single cross-sectionalsymptom scores were compared by analysis of variance with repeated measures(ANOVA-R) (Systat; SPSS Inc, Chicago, Ill), with hormonal condition (baselinevs Lupron plus testosterone vs Lupron plus placebo) as the within-subjectsvariable. Analyses of variance were reperformed using subject age as a covariate.Symptom-rating data during the first month (Lupron alone) were not includedin the analysis because plasma testosterone levels increase (or flare) transientlyafter the first injection of Lupron. We examined differences in the numberof men meeting a severity criterion score for the BDI (BDI score ≥7 forclinically relevant depressive symptoms⁴² )across hormonal conditions with the Fisher exact test.

Plasma hormone levels obtained during the last 2 weeks of each add-backcondition (testosterone or placebo) were averaged and compared by ANOVA-Rwith hormone condition as the within-subjects variable.

Seminal earlier studies^{43 - 44} demonstratedthat the behavioral response to hypogonadism and testosterone replacementin animals could be predicted by pretreatment levels of androgen-related behavior.To determine if these observations could be extended to humans, we assignedsubjects to 1 of 2 categories defined by their baseline symptom scores on2 behavioral symptoms, ie, sexual interest and feeling emotionally charged,identified by ANOVA as changing significantly across the 3 hormonal conditions.The 2 subject groups comprised the 10 subjects with the highest (high group)and lowest (low group) baseline scores on the 2 selected symptoms. Their scoresfor these 2 symptoms were reanalyzed by ANOVA-R with baseline symptom score(low vs high) as the between-subjects variable and hormonal condition as thewithin-subjects variable.

Post hoc Bonferroni t tests were performedwithin and between groups when indicated by significant ANOVAs. Two-tailed t tests were performed to compare baseline plasma levelsof testosterone and estradiol between high and low baseline symptom groups.

To examine potential order effects on significant symptom measures,we repeated the ANOVA-R with order of receiving testosterone as a between-groupfactor. Additionally, to avoid any potential confound of the crossover designon the observed effects of Lupron plus placebo, we reanalyzed the data byANOVA-R including baseline and only the first hormone treatment that eachsubject received (ie, pseudoparallel design).

Finally, Pearson correlation coefficients were performed to examinewhether changes in symptom scores between the testosterone-replaced and hypogonadalconditions were associated with changes in testosterone levels across theseconditions. Thus, for those symptoms that changed significantly across hormoneconditions (ie, BDI scores, hot flushes, sexual interest, and feeling emotionallycharged), correlations were performed with the following measures of testosterone:(1) the change in the average testosterone level recorded from the testosterone-replacedto the hypogonadal conditions and (2) the maximum change (calculated by identifyingthe highest level of testosterone obtained and subtracting it from the lowestplasma testosterone level during the hypogonadal state).

Thirty-one men ranged in age from 23 to 46 years (mean ± SD,30.8 ± 5.8 years). Fourteen men received Lupron plus testosteronefirst, and 17 received Lupron plus placebo first after 4 weeks of treatmentwith Lupron alone.

Lupron plus placebo was associated with significantly lower plasma levelsof testosterone, free testosterone, dihydrotestosterone, and estradiol thaneither the eugonadal (baseline) or Lupron plus testosterone conditions (Table 1). Plasma levels of both total testosteroneand free testosterone were significantly higher during testosterone replacementthan during baseline. No significant differences in plasma levels of dihydrotestosteroneor estradiol were observed between baseline and Lupron plus testosterone.

With the exception of hot flushes (both daytime and nighttime), libido,and feeling emotionally charged, there were no significant differences betweenthe Lupron plus placebo condition and either the baseline or the Lupron plustestosterone conditions in the symptoms measured, including sadness, anxiety,irritability, mood lability, anhedonia, and decreased energy (Table 2). Hot flush (daytime and nighttime) severity significantlyincreased and both sexual interest and feeling emotionally charged significantlydecreased during Lupron plus placebo compared with both baseline and Lupronplus testosterone. There were no symptom differences between baseline andLupron plus testosterone. Repeated ANOVA with age as a covariate resultedin an identical pattern of effects.

Beck Depression Inventory but not Spielberger State-Trait Anxiety Inventoryscores significantly increased during Lupron plus placebo compared with baselineand Lupron plus testosterone. The increase in BDI score during Lupron plusplacebo (mean ± SD, 2.3 ± 3.2) was statisticallybut not clinically significant and reflected the effects of 3 men who scored7 or higher on the BDI during Lupron plus placebo compared with none duringLupron plus testosterone replacement (Fisher exact test, P not significant). One man met DSM-IV⁴⁵ criteriafor a major depressive episode during Lupron plus placebo, and his BDI scoreswere 0 at baseline, 14 during Lupron plus placebo, and 1 during Lupron plustestosterone. We observed significant effects of order of hormone administrationand a trend for a significant interaction between order of hormone administrationand hormone condition on the scores of the BDI (ANOVA-R: effects of order,F_2,58 = 7.0, P < .01;interaction, F_2,58 = 2.5, P = .09).A similar pattern was observed for the symptom of decreased sexual interest(ANOVA-R: effects of order, F_2,50 = 19.5, P < .001; interaction, F_2,50 = 4.1, P < .05). Men receiving Lupron plus placebofirst but not those receiving Lupron plus testosterone first experienced asignificant increase in BDI scores (and a decrease in sexual interest) duringLupron plus placebo compared with baseline conditions (t₅₈ = 3.4 and t₅₀ = 3.7, respectively; P < .01).

Total scores on the Buss-Durkee Hostility Inventory, Anger, Irritability,and Assault Questionnaire, and Barratt Impulsiveness Scale version 7B didnot change across hormonal conditions; however, several individual subscalescores changed significantly from baseline with study participation. Buss-DurkeeHostility Inventory assault scores decreased from baseline during both Lupronplus placebo and Lupron plus testosterone (with significant differences betweenbaseline and Lupron plus placebo). A similar pattern of change was observedfor the individual subscales of risk taking, motor behavior, and sensory stimulationon the Barratt Impulsiveness Scale version 7B (Table 3). There was no significant interaction between order ofhormone administration and hormone condition on the scores of either the Buss-DurkeeHostility Inventory assault or the Barratt Impulsiveness Scale version 7Bsubscales.

Men were assigned to high (n = 10) and low (n = 10)symptom groups for the baseline symptoms of sexual interest and feeling emotionallycharged. There were significant group × hormone condition interactions for bothsexual interest and feeling emotionally charged (ANOVA-R: F_2,36 = 5.9, P < .01, and F_2,36 = 11.6, P < .001, respectively). For both symptoms,scores for the high symptom group were significantly higher than those forthe low symptom group during baseline (by definition) and Lupron plus testosteronebut not during Lupron plus placebo (baseline: t₅₄ = 5.8, P < .01, and t₅₄ = 5.5, P < .01;Lupron plus testosterone: t₅₄ = 3.6, P < .01, and t₅₄ = 2.5, P = .05 [forincreased sexual interest and feeling emotionally charged, respectively]).Men with high baseline sexual interest had a significant decrease in sexualinterest during Lupron plus placebo (t₃₆ = 5.0, P < .01), and levels of sexual interest wererestored to baseline levels during Lupron plus testosterone (t₃₆ = 5.1, P < .01,compared with Lupron plus placebo) (mean ± SD, 2.3 ± 0.4[baseline], 1.2 ± 0.3 [Lupron plus placebo], and 2.3 ± 0.9[Lupron plus testosterone]). The low symptom group reported no significanteffects on sexual interest during Lupron plus placebo or during testosteronereplacement (mean ± SD, 1.0 ± 0.1 [baseline],1.0 ± 0.04 [Lupron plus placebo], and 1.5 ± 0.6[Lupron plus testosterone]). The average percent decrease in sexual interestfrom baseline to Lupron plus placebo was 46% in high and 2.3% in low baselinesymptom groups. A similar pattern was found for the symptom of feeling emotionallycharged; for the low baseline emotionally charged group, scores remained lowduring all 3 conditions, whereas symptoms for men with high baseline decreasedsignificantly from baseline to Lupron plus placebo (t₃₆ = 5.8, P < .01) andincreased nonsignificantly during Lupron plus testosterone but not to baselinelevels (mean ± SD, 2.3 ± 0.8 [baseline],1.5 ± 0.6 [Lupron plus placebo], and 1.8 ± 0.7[Lupron plus testosterone]).

A trend for an interaction between group (high vs low baseline sexualinterest scores) and hormone condition (F_2,36 = 3.1, P = .06) was observed for BDI scores. Despitecomparable baseline BDI scores, men with high baseline sexual interest, butnot those with low baseline sexual interest, had a significant increase inBDI scores during Lupron plus placebo compared with their baseline scores(t₃₆ = 3.1, P < .05).During Lupron plus placebo, BDI scores in the high group also were significantlyhigher than the low group scores (mean ± SD, 3.9 ± 4.3vs 0.8 ± 1.0, t₅₄ = 3.1, P < .01). A comparable number of men who received placebofirst (ie, more likely to develop mood symptoms) were in the high baselinegroup (n = 7) as were in the low baseline group (n = 6).

Baseline plasma testosterone and estradiol levels did not differ betweenmen with high baseline symptoms and those with low baseline symptoms.

Finally, we observed a significant correlation between BDI scores andseverity of nighttime hot flushes (r = 0.7, P < .001). There were no significant correlationsbetween symptom scores during hypogonadism and changes in testosterone levels(r = 0.1-0.3, P notsignificant).

Few subjects in this study developed negative mood symptoms during anotherwise dramatic albeit brief (4-week) withdrawal and replacement of testosteroneunder double-blind conditions. Some measures of mood did worsen during Lupronplus placebo, but the response within individuals varied considerably. Forexample, while BDI scores significantly increased during Lupron plus placebo,this increase was almost entirely due to the emergence in 3 men of symptomsof depression (BDI scores, 7-14) lasting 7 to 14 days during Lupron plus placebo.The relevance of changes in testosterone levels to mood for only a subgroupof men was similarly demonstrated in 4 studies of anabolic-androgenic steroids,reporting that approximately 5% of healthy male volunteers (none of whom werehypogonadal or drug abusers) experienced clinically significant mood symptoms,such as hypomania, when administered supraphysiologic doses of testosterone.^{6 - 8 ,15} Indeed,in our group of carefully screened healthy volunteers, we observed the onsetof clinically significant mood symptoms during Lupron plus placebo (hypogonadism),in contrast to during supraphysiologic levels of androgens, in a small percentage(approximately 10%) of the sample. Similarly, substantial effects in a smallsubset of patients appear responsible for observed mood-elevating effectsof testosterone in samples of depressed men.¹¹ Thus,a clinically significant mood response to induced hypogonadism or testosteronereplacement in men appears to reflect a differential behavioral response toalterations in reproductive hormones, as previously reported in women,⁴⁶ which may be mediated by genetic factors.⁴⁷ In samples in which rates of physical and psychiatricillness are higher than those in our sample and similar to those found inthe general population, it is conceivable that negative mood response to inducedhypogonadism could be more prevalent.

It is possible that, as suggested in perimenopausal women, depressionmay occur secondary to severe hot flushes. Indeed, in this study, BDI scoreswere significantly correlated with ratings of hot flush severity but not sleepdisturbance. However, the high prevalence rates of hot flushes (67% in ouracutely hypogonadal men and approximately 70% in perimenopausal women)⁴⁸ stand in contrast to the relatively low rate of depressionin these samples. Thus, although the severity of both hot flushes and depressionmay be correlated, the presence of severe hot flushes is not sufficient toproduce depression. Men and women who develop depression in the context ofhot flushes, therefore, display a differential sensitivity to the negativeeffects of either declining testosterone (or estradiol) levels or hot flusheson mood.

Baseline level of symptomatology appeared to differentiate the men’sresponses to hormone withdrawal and replacement. Men with higher baselinelevels of sexual interest (and feeling emotionally charged) experienced significantdeclines in these measures during Lupron plus placebo. Neither differencesin baseline symptom scores nor responses to Lupron plus placebo were predictedby baseline or treatment-related plasma testosterone or estradiol levels.High baseline levels of sexual interest also identified a group of men whoexperienced a significant increase in BDI mood scores during Lupron plus placebo.At baseline, BDI scores in the high sexual interest group did not differ fromthose in the low sexual interest group; however, the high but not the lowsexual interest group experienced a significant increase in BDI scores duringLupron plus placebo. Our findings are similar to those of Grunt and Young⁴³ and Moore,⁴⁴ in whichanimals with the highest levels of sexual activity at baseline experiencedthe greatest declines and increases in sexual activity during hypogonadismand testosterone replacement, respectively. A floor effect may have preventeddetection of a further decline in symptoms during Lupron plus placebo in thelow symptoms group. Nonetheless, the restoration of significantly elevatedsexual interest in the high group but not the low group, despite comparableelevations of testosterone, suggests that testosterone is critical for sexualinterest but only in some men; that is, men differ in the sensitivity to thiseffect of testosterone. Thus, our data suggest that a behavioral phenotypecan predict a disparate response to declining testosterone secretion, andthis phenotype may contribute to the substantial variation in the observedsymptomatic response to aging and reproductive senescence.

Lupron plus placebo or testosterone (sufficient to achieve physiologiclevels) did not influence aggression and impulsivity in these healthy men,consistent with the findings of Tricker et al.¹⁴ However,had we studied a group of men with high baseline levels of aggression or impulsivity,or had supraphysiologic levels of testosterone been achieved more uniformly,it is possible that a larger change in these symptoms would have been observed.^{23 - 25} Nonetheless, evenwith doses substantially higher than those we used, Pope et al⁷ observedonly a low frequency (5%-10%) of idiosyncratic hypomanic reactions ratherthan frequent induction of aggression or impulsivity. Several subscale itemsdid decline during both Lupron plus placebo and Lupron plus testosterone;consequently, we cannot rule out the possibility that a factor present atbaseline, suppressed with Lupron and not replaced with testosterone, couldbe mediating changes in those subscale items. Conversely, decreased symptomscould have occurred consequent to study participation, although clearly othermeasures moved in the opposite direction (eg, BDI).

It is possible that a longer duration of hypogonadism than we inducedmay have a greater negative impact on mood than we observed. In support ofthis possibility, we did observe that the men who received Lupron plus placebofirst had greater changes in BDI and sexual interest scores during hypogonadismcompared with those receiving Lupron plus testosterone first. Those men randomizedto first receive placebo were hypogonadal for 5 to 6 consecutive weeks (the1-2 weeks following the flare during month 1 and 4 weeks in month 2), whereasthose men first receiving Lupron plus testosterone were hypogonadal duringonly 4 consecutive weeks in month 3. However, even in those men receivingplacebo first, mood changes during Lupron plus placebo were nonuniform andwere mild in the majority of subjects. Further, 4 weeks has been reportedby prior studies to be sufficient to result in mood changes (improvement)in hypogonadal men after testosterone replacement,^{13 ,49} andit was obviously sufficient to induce hot flushes and lower libido in themen participating in this study. Finally, our findings are comparable to thosein the existing literature on the relationship between depression and hypogonadismin men; that is, despite the prolonged hypogonadism associated with aging,the majority of men do not develop a depression. Between the ages of 60 and70 years, 20% to 30% of men may meet criteria for hypogonadism,¹ yetno comparable increase in the onset of depressive illness has been reported.Nevertheless, we should be cautious about generalizing our findings in youngmen, whose testosterone levels were pharmacologically manipulated, to changesin mood associated with reproductive aging in older men. Overall, our datasuggest that acute androgen withdrawal, while associated with decreased libidoand hot flushes, is not sufficient to uniformly alter mood in healthy, youngadult men.

Our data fail to support a uniform adverse effect on mood of inducedhypogonadism in healthy young men. Nonetheless, our findings are consistentwith a literature suggesting that some men are differentially sensitive toalterations in androgenic steroids, such that they experience disturbed moodin association with marked increases or decreases in these steroids. Evenfor a behavior (sexual function) that is more clearly linked to changes intestosterone, our demonstration that a behavioral phenotype (level of libido)can predict the degree of improvement in libido following testosterone replacementsuggests that testosterone will not serve as a panacea for age- or hypogonadism-relatedsexual disturbance. An examination of healthy, psychologically normal youngmen served our intent to isolate the behavioral effects of testosterone, sothat they would not be confounded by other factors (past psychiatric history,medical illness, or behavioral maladaptation) that may impact the relationshipbetween mood and hypogonadism. Consequently, the generalizability of our observationsto other (eg, aging) populations awaits further study.

Correspondence: Peter J. Schmidt, MD, NationalInstitute of Mental Health, Bldg 10, Room 3N238, 10 Center Dr MSC 1276, Bethesda,MD 20892-1276 (schmidtp@intra.nimh.nih.gov).

Accepted for Publication: May 2, 2004.