Kaplan-Meier survival analysisshowing time to diagnosed depression in hypogonadal vs eugonadal men. Thehypogonadal men had a significantly shorter time to diagnosed depressive illness(log-rank χ21 = 6.9, P = .008).
Two-year incidence of depressionin hypogonadal men, using different total testosterone threshold levels todefine hypogonadism. There is an increase in incident depression as the severityof hypogonadism increases. Asterisk signifies that at all total testosteronethresholds below 280 ng/dL, hypogonadal men showed a significant increasein incident depression compared with eugonadal men (P<.05).To convert testosterone to nanomoles per liter, multiply by 0.0347.
Shores MM, Sloan KL, Matsumoto AM, Moceri VM, Felker B, Kivlahan DR. Increased Incidence of Diagnosed Depressive Illness in HypogonadalOlder Men. Arch Gen Psychiatry. 2004;61(2):162-167. doi:10.1001/archpsyc.61.2.162
Age-associated hypogonadism (testosterone deficit) occurs in 30% of
men after the age of 55; it is associated with decreased muscle mass, bone
mineral density, and libido, and with anorexia, fatigue, and irritability.
Although some of these symptoms overlap with those of depression, the association
between the 2 disorders is unclear.
To determine if hypogonadal men have an increased incidence of depressive
illness compared with eugonadal men.
Historical cohort study using computerized medical records, followed
by a manual medical record review.
Veterans Affairs Puget Sound Health Care System.
Two hundred seventy-eight men 45 years and older, without prior diagnosed
depressive illness and with consistently normal or low testosterone levels
(total testosterone level ≤200 ng/dL [≤6.94 nmol/L]; or free testosterone
level ≤0.9 ng/dL [≤0.03 nmol/L]) at baseline and during a 2-year follow-up
Main Outcome Measures
Incidence of, and time to, a depression diagnosis.
The 2-year incidence of diagnosed depressive illness was 21.7% in hypogonadal
men vs 7.1% in others (χ21 = 6.0, P =
.01). A Kaplan-Meier survival analysis showed a significant difference between
hypogonadal and eugonadal men in time to diagnosed depression (log-rank test χ21 = 6.9, P = .008). We used Cox proportional
hazards regression models to examine the association of hypogonadism and time
to depression diagnosis, adjusting for age, race, number of clinic visits,
alcohol use disorders, prostate cancer, and overall medical comorbidity. The
unadjusted hazard ratio for depression with hypogonadism was 3.5 (95% confidence
interval, 1.3-9.4) (P = .01). Controlling for all covariates,
hypogonadism remained significantly associated with depression (adjusted hazard
ratio, 4.2; 95% confidence interval, 1.5-12.0) (P = .008).
Hypogonadal men showed an increased incidence of depressive illness
and a shorter time to diagnosis of depression. Further prospective studies
are needed to confirm these preliminary findings and to clarify the role of
testosterone in the treatment of depressive illness in older men.
Age-associated testosterone deficiency is a common condition in oldermen, occurring in 30% after the age of 55.1 Totaltestosterone levels peak in early adulthood, and then decrease by approximately1% per year after the age of 40.2 Age-associatedhypogonadism reflects a decline in hypothalamic and testicular function. Inaddition, severe illness, malnutrition, and drugs (such as corticosteroidsand alcohol) may also decrease testosterone levels.3- 5 Althoughthere is no uniformly accepted threshold level for testosterone in older men,experts in geriatric andrology suggest that most men with age-associated hypogonadismhave total testosterone levels between 150 and 350 ng/dL (5.20-12.14 nmol/L).4,6
Symptoms of hypogonadism include diminished muscle mass and strength,decreased bone mineral density, anorexia, decreased libido, fatigue, dysphoria,and irritability.3- 6 Someof these symptoms overlap with those of depressive illness. However, the associationbetween hypogonadism and depression is unclear. Prior studies have revealedmixed findings on the relation of testosterone and mood in older men. Endocrinologicstudies7- 10 oftestosterone replacement have shown an improvement in general well-being inolder hypogonadal men. However, these studies focused primarily on the effectson muscle, bone, and sexual function. Moreover, they used samples from specialtyclinics and used nonstandardized mood measures.11,12 Psychiatricstudies that have specifically examined the association of testosterone andmood have yielded conflicting results. One large cross-sectional study13 of older men found that testosterone levels wereinversely associated with scores on the Beck Depression Inventory. Anotherstudy14 found that low testosterone levelswere associated with dysthymic disorder, but other reports15- 17 drewno correlation between testosterone level and depression in young or middle-agedmen. Testosterone treatment trials in depressed subjects have shown similarlyconflicting results. Several studies18,19 foundthat depressed hypogonadal human immunodeficiency virus–positive menwere effectively treated with testosterone. Two small studies20,21 foundthat testosterone effectively augmented antidepressant treatment in hypogonadalmen with refractory depression. In contrast, a randomized, double-blind, placebo-controlledtrial22 found no difference in treatment responsebetween testosterone and placebo in older, depressed, hypogonadal men.
We, therefore, sought to examine the longitudinal relation of hypogonadismand incident depression in older men. We hypothesized that, compared withmen with normal testosterone levels, those with low testosterone levels wouldhave an increased incidence of depressive illness and a correspondingly shortenedtime to the development of depression.
We examined computerized clinical records of older male patients fromthe Veterans Affairs Puget Sound Health Care System to assess the relationof testosterone level and 2-year incidence of diagnosed depression. The recordscontained demographic information, laboratory and pharmacy data, and inpatientand outpatient International Classification of Diseases,9th Revision, Clinical Modification (ICD-9-CM),diagnostic codes.23 We first identified allmale patients meeting the following inclusion criteria: (a) 45 years or older as of January 1, 1998, (b)seen at the medical center at least twice a year, (c)available baseline (January 1, 1995–December 31, 1997) and follow-up(January 1, 1998–December 31, 1999) testosterone levels, and (d) testosterone levels stable at values either above orbelow specified threshold levels of total or free testosterone. We excludedmen treated with antiandrogens and those with diagnoses of depressive illnessbefore 1998 (ICD-9-CM codes 296.2-296.9 [major depressivedisorder], 300.4 [dysthymic disorder], and 311.0 [depressive disorder nototherwise specified]). These ICD-9-CM codes wererecorded by clinicians following routine clinical care in outpatient and inpatientsettings.
To ensure that men with low testosterone levels would have meaningfulhypogonadism, likely to benefit from testosterone replacement,24 weused a stringent cutoff value for repeated testosterone levels (≤200 ng/dL[≤6.94 nmol/L] for total testosterone or ≤0.9 ng/dL [≤0.03 nmol/L]for free testosterone) to define hypogonadism. In instances of disagreementin classification by total vs free testosterone, we accepted the more reliabletotal testosterone categorization as definitive. Because testosterone levelshave a circadian fluctuation, we compared the time of day for plasma acquisitionfrom hypogonadal and eugonadal men.
We then ascertained the occurrence of clinically diagnosed depressiveillness during an analytic period from January 1, 1998, through December 31,1999. Using the same ICD-9-CM diagnostic codes previouslydescribed, we noted the dates when clinical diagnoses of depressive illnesswere first entered. From the same medical records, we abstracted potentiallyrelevant covariates, including age, race, total number of clinic visits, alcoholuse disorders, prostate cancer, and the Chronic Disease Score (CDS).25 The CDS, which is obtained algorithmically from theVeterans Affairs Computerized Patient Record System, represents a simple countof 29 different chronic medical conditions, yielding a score of 0 to 29. Ithas been validated as an index of general medical comorbidity.25,26
A research assistant who was blinded to testosterone level categorynext performed a manual medical record review, confirming diagnoses of depressiveillness and noting the indication for testosterone assays and the specialtyof the practitioners who ordered these assays. The research assistant hadbeen instructed to consult a research psychiatrist (M.M.S.) (also blindedto testosterone level category) regarding any ambiguities in depression diagnosesand indications for a testosterone assay. Then, the psychiatrist performeda more comprehensive medical record review of men with depression diagnoses,further evaluating and corroborating these diagnoses and their first recordeddates. In 17 (6.1%) of the instances of disagreement between the psychiatrist'sreview and original computerized data acquisition, we used the psychiatrist'sreview for our analysis.
We examined differences between groups with independent sample t tests for continuous measures and χ2 testsfor categorical observations. We used a Kaplan-Meier survival analysis tocompare time to diagnosed depressive illness.27 Survivalcurves and median time to diagnosed depression were compared using the log-rank χ2 test.28 Cox proportional hazards regressionmodels were used to examine the association of testosterone category and timeto diagnosed depression while adjusting for the influence of covariates.29 These analyses included all named covariates (age,race, number of clinic visits, alcohol use disorders, prostate cancer, andCDS), although CDS was the only significant factor in these models.
From the computerized clinical data, we identified 398 men 45 yearsor older who had testosterone level results before and after January 1, 1998,but who showed no diagnosis of a depressive disorder before this date. Ofthese men, 294 (73.9%) had repeated low or repeated normal testosterone levels,as previously described. Because acute illness, medications, and intermittenttestosterone treatment could cause testosterone fluctuations that would bedifficult to identify using historical methods, our primary analyses excludedthe 104 men (26.1%) with inconsistent testosterone level categories. We alsoexcluded 16 subjects following the manual medical record review, 15 becauseof a preexisting diagnosis of depression and 1 because of antiandrogen treatment.
Of the remaining 278 men, 23 had repeated low testosterone levels while255 had repeated normal levels. Table 1 demonstrates a lack of significant difference between the hypogonadaland eugonadal men for age, ethnicity, alcohol use disorders, prostate cancer,CDS, or clinic visits. Surprisingly, the eugonadal men had more clinic visits,although this was not statistically significant.
Testosterone assays were obtained by clinicians for the following indications:evaluation of sexual dysfunction (31.6%), osteoporosis (21.6%), current testosteronetreatment or follow-up of a prior low testosterone level (15.4%), geriatricrehabilitation (10.4%), other genitourinary conditions (9.0%), cancer (3.2%),endocrine conditions (3.0%), and other or unknown reasons (5.8%). Most cliniciansordering testosterone levels were in primary care (60.2%), followed by urology(10.3%), endocrinology (8.6%), geriatrics (7.3%), oncology, rheumatology,or neurology (6.5%), and psychiatry (0.3%). (Data were missing in 6.8% ofthe cases.) There were no significant (P = .98) differencesin the phlebotomy times for the men with low vs normal testosterone levels(data not shown).
During the 2-year follow-up period, the hypogonadal men had a significantlyincreased occurrence of diagnosed depressive illness (crude rates, 21.7% vs7.1%; χ21 = 6.0, P = .01).The 104 men with inconsistent testosterone level results showed an intermediatecrude occurrence of depression (15.1%). Small differences were observed indepressed vs nondepressed men for total number of clinic visits (P = .98) and total number of mental health visits (P = .69), but these differences did not reach statistical significance.
A Kaplan-Meier survival analysis (Figure1) showed that hypogonadal men had a significantly shorter timeto diagnosed depression (log-rank χ21 = 6.9, P = .008). In Cox proportional hazards regression models,the unadjusted hazard ratio for diagnosed depression with a low testosteronelevel was 3.5 (95% confidence interval, 1.3-9.4) (P =.01). Overall medical comorbidity proved to be the only covariate significantlyassociated with time to diagnosed depression. After adjustment for all covariates,the adjusted hazard ratio was 4.2 (95% confidence interval, 1.5-12.0) (P = .008).
Finally, we performed a sensitivity analysis to assess whether our resultswould vary using different testosterone threshold levels to define hypogonadaland eugonadal status. Figure 2 showsthat lower testosterone thresholds were associated with an increase in incidentdepression. The 2-year incidence of depression in men with a total testosteronelevel below 150 ng/dL (5.20 nmol/L) was 29%. By contrast, use of the leaststringent threshold of 350 ng/dL (12.14 nmol/L) reduced the incidence of depressionto 13%. At all total testosterone thresholds below 280 ng/dL (9.72 nmol/L),hypogonadal men showed a significant increase in incident depression comparedwith eugonadal men.
We observed an increased incidence of depressive illness in hypogonadalolder men. Others have examined correlations between testosterone levels andmood13,16,17 or theutility of testosterone treatment or testosterone augmentation for depression.19- 22 Toour knowledge, this is the first study to examine the longitudinal relationof hypogonadism and incident depression in older men. Compared with eugonadalpatients, hypogonadal men with total testosterone levels of 200 ng/dL or less(≤6.94 nmol/L) showed an approximate 4-fold increase in the risk of incidentdepression. Post hoc dose-response analyses showed that depression risk wasinversely related to testosterone level, with statistically significant findingsobserved at testosterone levels lower than 280 ng/dL (9.72 nmol/L). Negativeor mixed results from prior testosterone treatment trials22,30 ofdepression in elderly men may have been due to the inclusion of eugonadalmen or men with less marked hypogonadism with testosterone levels above 280ng/dL.
Several plausible biological mechanisms may explain the associationbetween hypogonadism and depressive illness. Hypogonadism may directly causesymptoms such as muscle wasting, anorexia, fatigue, and decreased libido,which may have an effect on mood. Hypogonadism might also cause depressiveillness directly through alterations in central neurotransmitter function,because testosterone is known to have multiple central effects.31 Forexample, in animal models, testosterone increases cortical serotonin2A receptor binding densities32,33;and in humans, cortical serotonin2A receptors decrease with depression34 and aging.35 Thus,a low testosterone level may cause depression via decreased serotonin2A receptor density. If so, older men would be particularly vulnerableto these effects, because serotonin2A receptors are already decreasedfrom normal aging.35 Another potential biologicalmechanism may be that testosterone influences affective illness similar tothe manner in which thyroid hormone modulates affective illness (ie, markedthyroid deficiency can precipitate or exacerbate a depressive illness, andthyroid augmentation may enhance antidepressant response in patients withtreatment-refractory depression).36
Given our reliance on medical records, these results should be viewedas preliminary. Our methods may be subject to several forms of bias. Theseinclude selection bias, detection bias, exposure (testosterone level) misclassification,incomplete control on phlebotomy times, and susceptibility to other unsuspectedor unmeasured confounding factors. We discuss these briefly, in turn notingsome arguments on why they do not explain our results in full.
Selection bias might have occurred if testosterone levels had been orderedpreferentially in subjects who seemed depressed. Testosterone levels wererarely ordered for the evaluation of a mood disorder, however. Most testosteronelevels were ordered because of sexual dysfunction, osteoporosis, rehabilitationassessment, and endocrine illness. Detection bias might have occurred becausedepression is frequently unrecognized in primary care settings. Thus, it islikely that depression was not detected in many cases. During the period coveredby the study, depression screening (2 questions about dysphoria and anhedonia)was mandated in Veterans Affairs primary care settings. Such depression screeningwould likely increase the detection of depressive illness and minimize theunderdiagnosis of depression. In contrast, another type of detection biascould occur if hypogonadal men had more frequent clinic visits than eugonadalmen. In this case, hypogonadal men could have spurious increases in depressionrates that might simply reflect increased surveillance. In fact, althoughthe number of clinic visits was comparable between the 2 groups, the eugonadalmen had more clinic visits than the hypogonadal men. Thus, oversurveillanceof the hypogonadal group does not seem to be a factor in the higher rate ofdepression diagnoses in that group. Even so, the similar symptoms of hypogonadismand depression may have led to greater recognition of depressive illness inhypogonadal subjects. In future studies, this concern might be addressed byexamining all participants for depression using standardized techniques.
In this historical study, depression diagnoses do not have the rigorousvalidity of a prospective study. We attempted to address this problem in partthrough a manual medical record review that yielded greater clinical detailthan the computerized database. Nevertheless, a systematic prospective studywould yield more robust diagnoses.
Testosterone levels are known to show circadian variation, particularlyin younger men. We, therefore, examined the phlebotomy times for systematicdifferences between the hypogonadal and the eugonadal men, but found none.Furthermore, in older men, the circadian variation of testosterone secretionis markedly diminished or absent.37 Thus, webelieve that the effect of variable phlebotomy times is minimal, because therewere no significant differences in phlebotomy times between the 2 groups andthe circadian secretion of testosterone is markedly attenuated in older men.
It is possible that low testosterone and depression are associated onlybecause both are related to some other common factor. For example, prostatecancer, sexual dysfunction, and overall medical comorbidity are associatedwith hypogonadism and depression.38- 40 We,therefore, repeated our analyses, excluding men with prostate cancer, butfound no changes in our results. Furthermore, there were no appreciable differencesin prostate cancer, overall medical morbidity, ethnicity, or sexual dysfunction;and the prevalence of sexual dysfunction was actually lower in the hypogonadalvs the eugonadal group (30.4% vs 35.7%; P = .61).Thus, it does not seem that our results can be explained by an associationwith these covariates, because there were no significant differences in thembetween groups. Our results could also be confounded by ethnicity, becauseAfrican American men have a lower incidence of depression41 and,at least at younger ages, a higher testosterone level.42 However,there were no differences in ethnicity between groups, and we repeated ouranalyses excluding African American men, but found no changes in our results(data not shown).
Despite its limitations, our study suggests a relationship between testosteronelevel and depression that could have significant public health implications.Even without concomitant depression, hypogonadism has several probable deleteriouseffects and is a common disorder in older men.4 Ifhypogonadism also provokes an increased risk of depression, it would thenhave significant resulting implications for morbidity, mortality, and qualityof life.43 Because depression is a major riskfactor for suicide, and older men have the highest suicide rate of any agegroup in the United States,44 identifying conditions(such as low testosterone) that increase the risk of depressive illness couldgenerate important opportunities for early intervention and treatment. Ifthese findings are substantiated, the detection of low testosterone couldaid in the early identification and treatment of depressive illness in oldermen—a group noted to experience underdetection and undertreatment ofdepressive illness.45
Our results do not speak to the possible effectiveness of testosteronetreatment or augmentation in patients with depressive disorders, but it wouldbe of interest to know whether such treatment might be beneficial in particulardiagnostic subgroups (eg, those with dysthymic disorder) or other categories(eg, the oldest-old). In addition, although testosterone replacement in hypogonadalmen has not been shown to increase the risk of prostate cancer,4- 6 thisremains a concern. Thus, future large prospective studies of testosteronereplacement are needed to clarify if potential benefits outweigh the risks,and to systematically study the effects of testosterone replacement on thehealth and welfare of older men.
Corresponding author and reprints: Molly M. Shores, MD, VeteransAffairs Puget Sound Health Care System, 1660 S Columbian Way, S-182 GRECC,Seattle, WA 98108 (e-mail: firstname.lastname@example.org).
Submitted for publication September 26, 2002; final revision receivedJune 4, 2003; accepted June 12, 2003.
This study was supported in part by the Geriatric Research Educationand Clinical Center (GRECC) and the Epidemiology Research and InformationCenter (ERIC) at the Veterans Affairs Puget Sound Health Care System; by theRoyalty Research Fund from the University of Washington, Seattle; and by aVeterans Affairs Merit Review Grant (Dr Matsumoto).
We thank John Breitner, MD, MPH, for his extensive review, insightfulcomments, and generous contribution of time; Edward Gottheil, MD, PhD, forhis thoughtful and timely commentary; and Tatiana Sadak for her excellenttechnical support.