Journal of Clinical Endocrinology and Metabolism
Volume 84 • Number 2 • February 1, 1999
Copyright © 1999 The Endocrine Society


573

Bioavailable Testosterone and Depressed Mood in Older Men: The Rancho Bernardo Study



ELIZABETH BARRETT-CONNOR 1
DENISE G. VON MUHLEN 1
DONNA KRITZ-SILVERSTEIN 1


1 Department of Family and Preventive Medicine, School of Medicine, University of California, San Diego, La Jolla, California 92093-0607


Received August 19, 1998.
Revision received November 9, 1998.
Accepted November 12, 1998.

Address all correspondence and requests for reprints to: Dr. Elizabeth Barrett-Connor, Department of Family and Preventive Medicine, University of California, San Diego, La Jolla, California 92093-0607.

* This research was supported by the National Institute of Diabetes and Digestive and Kidney Disease Grant DK31801 and the National Institute of Aging Grant AG07181.

ABSTRACT

A cross-sectional population-based study examined the association between endogenous sex hormones and depressed mood in community-dwelling older men. Participants included 856 men, ages 50-89 yr, who attended a clinic visit between 1984-87. Total and bioavailable testosterone, total and bioavailable estradiol, and dihydrotestosterone levels were measured by radioimmunoassay in an endocrinology research laboratory. Depressed mood was assessed with the Beck Depression Inventory (BDI). Levels of bioavailable testosterone and bioavailable estradiol decreased with age, but total testosterone, dihydrotestosterone, and total estradiol did not. BDI scores increased with age. Low bioavailable testosterone levels and high BDI scores were associated with weight loss and lack of physical activity, but not with cigarette smoking or alcohol intake. By linear regression or quartile analysis the BDI score was significantly and inversely associated with bioavailable testosterone (both P's = 0.007), independent of age, weight change, and physical activity; similar associations were seen for dihydrotestosterone ( P = 0.048 and P = 0.09, respectively). Bioavailable testosterone levels were 17% lower for the 25 men with categorically defined depression than levels observed in all other men ( P = 0.01). Neither total nor bioavailable estradiol was associated with depressed mood. These results suggest that testosterone treatment might improve depressed mood in older men who have low levels of bioavailable testosterone. A clinical trial is necessary to test this hypothesis. ( J Clin Endocrinol Metab 84: 573-577, 1999)


BIOAVAILABLE testosterone levels in men decrease by as much as 40% between the ages of 40 and 70 [1] [2] . The prevalence of depressed mood also increases in old age [3] [4] . The possibility that the testosterone-mood association is causal has been examined in studies of hypogonadal men. Burris and colleagues [5] reported that six profoundly hypogonadal men scored significantly higher on self-reported depression scores than eugonadal men; with androgen treatment the mood scores decreased, although they did not return to normal. More recently, Wang et al. [6] reported that testosterone therapy significantly improved mood scores in 51 hypogonadal men. Testosterone levels were associated with mood scores only in the baseline period when testosterone levels were below normal.

This paper reports our investigations on whether plasma testosterone levels were related to depressed mood or categorical depression in 856 community-dwelling older men unselected for hypogonadism.



Subjects and Methods

Between 1972 and 1974, 82% of adult residents in Rancho Bernardo, a southern California community, were enrolled in a study of heart disease risk factors. In 1984-1987, 82% of the surviving community-dwelling residents attended a clinic visit primarily designed to study diabetes. The present report includes the 856 men aged 50-89 yr at the time of the 1984-1987 visit who had fasted the requested 12 h, who were not using sex hormone therapy, whose endogenous sex hormones were measured, and who completed the Beck Depression Inventory (BDI) [7] . This includes 81% of all age-eligible participants. All men were ambulatory, and all gave written, informed consent.

A standardized interview was used to obtain information on current lifestyle, including physical activity (exercise three or more times per week), cigarette smoking (yes/no), and alcohol consumption (number of drinks of beer, wine, or liquor converted to grams of alcohol per week). Previous studies have indirectly validated these reported behaviors by showing a negative correlation between cigarette smoking and pulmonary function [8] , positive correlations between alcohol use and high density lipoprotein (HDL)-cholesterol and liver function tests [9] , and positive correlations between reported physical activity and high density lipoprotein (HDL)-cholesterol [10] .

In the clinic, height, weight, and waist and hip circumference were measured, with the participants wearing light clothing and no shoes. Body mass index [weight (Kg)/height (m2 )] was used as an estimate of obesity. The waist-to-hip (cm/cm) ratio was used as an estimate of central adiposity. Weight change was calculated as weight measured in 1972-1974 visit subtracted from weight measured in 1984-1987.

Information on depressed mood was obtained using 18 of the 21 items of the BDI [7] . Three items (guilt, expectation of punishment, and selfhate) were excluded from the questionnaire because studies have suggested that as many as three fourths of the items from highly reliable measures can be dropped without much loss in sensitivity or specificity [11] , and other studies have reported on the validity of a 13-item shortform BDI [12] . Therefore, the use of an 18-item scale should not compromise its validity. Total scores on the BDI were computed by summing the responses to each question. Higher scores are indicative of depressed mood. These scores were then proportionally adjusted to correspond to scores and cutpoints previously established for the full 21-item scale. In this cohort, reliability as assessed by Cronbach's alpha was 0.75, which is comparable to the reliability obtained using samples of elderly community volunteers (alpha = 0.76) and depressed outpatients (alpha = 0.73) [13] . The BDI was completed before the venipuncture.

Blood was obtained in the morning, between 0730 and 1100 h, after a requested 12-h fast. Samples of plasma were frozen in polypropylene tubes at 70 C, specifically for later assay of sex hormones. The season of plasma collection was noted. Between 1992-1993, sex hormones were


574
measured by radioimmunoassay in the endocrinology research laboratory of S. S. C. Yen (University of California, San Diego, La Jolla, CA) using first-thawed specimens. Duration of storage was calculated as year when samples were assayed minus year of clinical visit (when samples were collected and frozen). Bioavailable testosterone and bioavailable estradiol were determined using a method modified from Tremblay and Dube [14] . The sensitivity and the inter- and intraassay coefficients of variation in this laboratory were 0.13 nmol/L, 4% and 6.8% for testosterone; 22.03 pmol/L, 5.9% and 7.4% for estradiol; 0.12 nmol/L, 7.5% and 7.5% for dihydrotestosterone (DHT); 0.13 nmol/L, 5.8% and 7.6% for bioavailable testosterone; and 22.03 pmol/L, 3.7% and 5.2% for bioavailable estradiol. Nine men had estradiol and bioavailable estradiol values below the sensitivity of the assay; five men had values below the sensitivity of the assay for testosterone and its bioavailable fraction. Men with undetectable levels were assigned levels just below the sensitivity of the assay for this analysis.



Results

Data were analyzed with SPSS and SAS [15] [16] . The BDI score was used as a continuous measure of depressed mood and as a categorical variable measure of clinical depression, defined as a score = 13. All sex hormones were normally distributed and did not require log transformation. Analyses were performed with and without 22 subjects whose sex hormone levels were more than 3 SD from the mean, with similar results; these subjects were included in the analyses shown. Body mass index and weight change were highly positively correlated (r = 0.84, P < 0.001); weight change was used in these analyses because it was more highly correlated with BDI. Age-specific mean sex hormone levels were calculated, and linear trends were tested across age categories using analyses of variance and linear contrasts. Age-adjusted and multiply-adjusted (for age, weight change or current BMI, exercise, smoking, and alcohol intake) sex hormone concentrations and covariates were compared by quartile of BDI using analyses of covariance. Age-adjusted and multiply-adjusted (for age, weight change, exercise, smoking, and alcohol intake) mean BDI scores were compared by quartile of sex hormone levels using analyses of covariance. Linear trends across quartiles were tested using linear contrasts. Pearson's age-adjusted partial correlation coefficients were calculated to assess the association between sex hormones and age, BMI, weight change, cigarettes per day, and BDI. Multiple linear regression was used to examine the association between sex hormones and scores on the Beck Depression Inventory after adjusting for potentially confounding covariates. All statistical tests were two-tailed.

The mean age of these 856 men was 70.2 yr ( SD = 9.2). As shown in Table 1 , bioavailable testosterone and bioavailable estradiol decreased significantly with increasing age, while total testosterone, total estradiol, and DHT did not. Mean BDI scores increased with age.

Table 2 shows the distributions of potentially confounding covariates by BDI quartile. In addition to age, loss of body weight and lack of regular exercise were each associated with higher BDI scores ( P < 0.01). Smoking, alcohol intake, BMI, and waist-to-hip ratio did not vary by BDI quartile. Among these same variables, only weight loss and regular exercise varied significantly with sex hormones ( P's < 0.05, data not shown).

Table 3 shows the age-adjusted and multiply-adjusted mean hormone levels by quartile of BDI. After adjustment for age, a significant negative trend in BDI score was seen only for bioavailable testosterone; the association persisted after additional adjustment for change of body weight and regular exercise. In similarly adjusted models, decreasing total testosterone and dihydrotestosterone levels were weakly associated with increasing BDI scores. Total estradiol and its bioavailable fraction did not vary significantly with BDI in any analysis.

Figure 1 shows mean BDI score by quartile of each sex hormone adjusted for age, weight change, and lack of physical activity. A significant negative trend in BDI score was observed for testosterone ( P =.04), bioavailable testosterone ( P =.005), and DHT ( P =.02). No significant trends were observed for mean BDI score and estradiol or bioavailable estradiol.

Multiple linear regression analysis adjusting for age, change in body weight, and regular exercise also showed a significant inverse association of BDI with bioavailable testosterone ( P < 0.007) and DHT ( P < 0.048) ( Table 4 ). BDI was not significantly associated with total testosterone, total estradiol, or bioavailable estradiol.

Only 25 men in this cohort had clinical depression as defined by BDI, with a score of = 13 and/or current use of antidepressants; these men had lower levels of bioavailable testosterone than all other men (2.83 nmol/L ( SD = 1.2) vs. 3.41 nmol/L ( SD = 1.1), respectively, P = 0.01).

Season was not a confounder in these analyses. Only estradiol and bioavailable estradiol levels varied by season when plasma was collected, with higher levels in fall and winter; depressed mood did not vary by season of testing in this cohort (data not shown). Sex hormone levels in this study did not vary by duration of plasma storage.



Discussion

As reported previously from this cohort of community-dwelling older men [17] and elsewhere [18] [19] , bioavailable testosterone levels decrease much more with age than does total testosterone or estradiol levels. The decrease in bioavailable testosterone parallels the increase in sex hormone
TABLE 1 -- Age-specific mean ( SD) sex hormone levels and Beck Depression Inventory (BDI) scores in 856 men, The Rancho Bernardo Study, 1984-87

50-59 n = 152 60-69 n = 221 70-79 n = 328 80-89 n = 155 P for trend
Testosterone (nmol/L) 10.52 (3.07) 10.41 (3.12) 10.76 (3.76) 10.34 (4.40) 0.95
Bioavailable testosterone 4.27 (1.09) 3.65 (0.97) 3.18 (0.99) 2.61 (1.07) 0.00
Estradiol (pmol/L) 75.62 (21.65) 71.22 (21.65) 74.89 (26.43) 72.68 (27.53) 0.73
Bioavailable estradiol (pmol/L) 54.33 (15.42) 48.46 (15.05) 47.36 (16.52) 42.21 (17.25) 0.00
DHT (nmol/L) 1.47 (0.58) 1.44 (0.55) 1.54 (0.65) 1.55 (0.72) 0.08
BDI score 3.3 (3.2) 4.1 (3.4) 4.9 (3.4) 5.6 (3.3) 0.00


575

TABLE 2 -- Age and other potentially confounding variables by quartile of Beck Depression Inventory (BDI) scores in 856 men, The Rancho Bernardo Study, 1984-87

Quartile of BDI scores
1 (<2.33) n = 196 2 (2.33-3.50) n = 226 3 (3.52-5.83) n = 221 4 (>5.83) n = 213 P for trend
Age 66.9 70.2 73.0 73.4 0.00
BMI 25.9 26.0 25.7 25.7 0.46
Waist-to-hip ratio 0.91 0.91 0.92 0.92 0.17
Weight change (Kg) (visit 4 minus visit 1) 0.15 0.57 -0.57 -1.96 0.00
Current regular exercise (%) 36.0 28.0 22.0 20.0 0.00
Current smoker (%) 13.0 11.5 10.0 10.0 0.36
Alcohol gm/past wk (among drinkers) 104.1 97.0 97.2 104.0 0.98


TABLE 3 -- Adjusted sex hormone mean by quartile of Beck Depression Inventory

Quartile of BDI scores
1 (<2.33) n = 196 2 (2.33-3.50) n = 226 3 (3.52-5.83) n = 211 4 (>5.83) n = 213 P trend
Testosterone (nmol/L)




  age-adjusted 10.92 10.80 9.91 10.62 0.13
  age, reg exercise, wt change 10.87 10.87 10.03 10.45 0.07
Bioavailable testosterone (nmol/L)




  age-adjusted 3.51 3.44 3.39 3.25 0.01
  age, reg exercise, wt change 3.71 3.44 3.29 3.11 0.007
Estradiol (pmol/L)




  age-adjusted 74.15 74.89 71.95 73.79 0.57
  age, reg exercise, wt change 74.15 74.89 71.95 73.79 0.66
Bioavailable Estradiol (pmol/L)




  age-adjusted 48.09 48.46 48.09 46.99 0.58
  age, reg exercise, wt change 48.09 48.09 48.46 47.36 0.76
DHT (nmol/L)




  age-adjusted 1.55 1.55 1.43 1.49 0.13
  age, reg exercise, wt change 1.55 1.56 1.45 1.48 0.03

binding globulin with age, a change that accompanies increasing body fat independent of weight change [
20] . The prevalence of depressive symptoms increases with age in community-dwelling older adults [3] [4] . In the Rancho Bernardo cohort, the average BDI scores for each decade were relatively low ( Table 1 ) compared with other community-based studies [21] [22] . The relatively low BDI scores in this cohort likely reflect the inverse association between socioeconomic status and depression [23] ; more than 80% of the Rancho Bernardo men were middle or upper-middle class, based on a summary score of occupation and education [24] . On average, hormone levels in Rancho Bernardo were somewhat lower than levels reported in Massachusetts [25] , possibly reflecting the older age of our cohort or the higher sensitivity of our assay.

Despite the low BDI scores, there was a graded stepwise decrease in bioavailable testosterone with increasing level of depressed mood. This association was independent of age, weight change, and physical activity, the only three significant confounding covariates of the association between sex hormone levels and BDI scores. The testosterone-BDI association was present throughout the whole range of BDI scores; it was not explained by the 25 men who had categorically defined depression and who had bioavailable testosterone levels that were, on average, 0.58 nmol/L lower than all other men. To our knowledge, no other large studies have examined the bioavailable testosterone-depressed mood association in community-dwelling older men.

Previous studies of sex hormones and depression in men, including clinical trials, have yielded mixed results, possibly reflecting the different patient populations and androgen assays. In one of the earliest such studies, Sachar et al. [26] studied 15 severely depressed men who improved on electroshock therapy or chloral hydrate, but they found no change in plasma testosterone after treatment. In the first study with a control group, Vogel et al. [27] reported that both total and free testosterone were significantly lower in 27 men with primary unipolar (neurotic) depression compared with 13 age-matched controls. Rubin et al. [28] also reported lower nocturnal testosterone levels in 9 depressed men compared with 6 nondepressed men, but the differences were not statistically significant. In 1985, Yesavage [29] found that plasma testosterone levels correlated with the severity of depression in 18 men who had major depression. In a randomized, placebo-controlled clinical trial conducted in 56 healthy older men who were randomly assigned to testosterone patches or placebo patches designed to restore physiologic levels of testosterone, Janowsky and colleagues [30] found no effect on mood as measured by the Profile of Mood States, a standardized self-rating scale [31] . A study of male body builders who were self-administering androgen [32] found that men who had the highest serum testosterone levels had the lowest BDI scores. However, a double-blind, randomized trial [33] reported no effect of supraphysiologic doses (600 mg) of testosterone enanthate in healthy young men on an unspecified mood inventory.


576


Figure 1. Mean BDI score adjusted for age, regular exercise, and change of body weight by quartile of sex hormone.


TABLE 4 -- Linear regression adjusted association between sex hormones and BDI scores among 856 men, The Rancho Bernardo Study
Sex hormone B Adjusted a SE B P
Testosterone -0.038 0.03 0.24
Bioavailable testosterone -0.302 0.11 0.007
Estradiol -0.003 0.01 0.57
Bioavailable estradiol -0.006 0.01 0.42
DHT -0.363 0.18 0.048
a Adjusted for age, weight change between visit 1 and visit 4, and regular exercise.




The present study has several limitations. Because this is a cross-sectional study, we cannot exclude the possibility that depressed mood lowers testosterone directly or indirectly through weight loss or inactivity. This study is based on a single hormone assay; stronger associations might be observed if hormonal status included more samples obtained on a single day or over multiple days. Although Bremner et al. [34] reported no circadian variation in testosterone levels among older men, all plasma samples for the present study were obtained in the morning after a 12-h fast. Although there was an 82% participation rate, selective nonparticipation by men who were clinically depressed is plausible. This would be expected to reduce the observed association, because the high end of depressed mood scores would be truncated.

This study is based on a single hormone assay, and the classification of depression or depressed mood is based only on one standardized instrument. However, testosterone levels appear to be highly repeatable [35] , and BDI score has been shown to be remarkably valid in studies of older adults [13] . In this cohort a higher BDI score was indirectly validated by its association with weight loss; depressed men, in contrast to depressed women, are likely to lose weight [36] . The association of depression with lack of physical exercise has been reported previously [37] ; it is unclear whether physical activity prevents depression or whether depressed people become sedentary.

It is biologically plausible that testosterone has an effect on depression. As reviewed elsewhere [38] , androgen and estrogen binding sites overlap in the brain. Testosterone in the brain is extensively converted to estradiol and DHT, which binds to androgen receptors [39] . The complete absence of an association between circulating estradiol and mood scores makes the thesis that the testosterone mood association is mediated by aromatization to estradiol unattractive but does not entirely exclude it. Clinical trials are needed to further assess this association.


577

References


1. Davidson JM, Chen JJ, Crapo L, Gray GD, Greenleaf WJ, Catania JA. 1983 Hormonal changes and sexual function in aging men. J Clin Endocrinol Metab. 57:71-77.   Abstract

2. Vermeulen A. 1991 The aging male. J Clin Endocrinol Metab. 73:221-224.   Citation

3. Murrell SA, Himmelfarb S, Wright K. 1983 Prevalence of depression and its correlates in older adults. Am J Epidemiol. 117:173-185.   Abstract

4. Blazer D, Burchett B, Service C, George L. 1991 The association of age and depression among the elderly: an epidemiological exploration. J Gerontol. 46:M210-M215.   Abstract

5. Burris AS, Banks SM, Carter CS, Davidson JM, Sherins RJ. 1992 A long-term, prospective study of the physiologic and behavioral effects of hormone replacement in untreated hypogonadal men. J Androl. 13:297-304.   Abstract

6. Wang C, Alexander G, Berman N, et al. 1995 Testosterone replacement therapy improves mood in hypogonadal men - a clinical research center study. J Clin Endocrinol Metab. 81:3578-3583.  

7. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. 1961 An inventory for measuring depression. Arch Gen Psychiatry. 4:561-571.  

8. Frette C, Barrett-Connor E, Clausen JL. 1996 Effect of active and passive smoking on ventilatory function in elderly men and women. Am J Epidemiol. 143:757-765.   Abstract

9. Barrett-Connor E, Suarez L. 1982 A community study of alcohol and other factors associated with the distribution of high density lipoprotein cholesterol in older vs. younger men. Am J Epidemiol. 115:888-893.   Abstract

10. Reaven PD, McPhillips JB, Barrett-Connor E, Criqui MH. 1990 Leisure time exercise and lipid and lipoprotein levels in older population. J Am Geriatr Soc. 38:847-854.   Abstract

11. Shrout PE, Yager TJ. 1989 Reliability and validity of screening scales: effect of reducing scale length. J Clin Epidemiol. 42:69-78.   Abstract

12. Reynolds WM, Gould JW. 1981 A psychometric investigation of the standard and short form Beck Depression Inventory. J Consult Clin Psychol. 49:306-307.   Citation

13. Gallagher D, Nies G, Thompson LW. 1982 Reliability of the Beck Depression Inventory with older adults. J Consult Clin Psychol. 50:152-153.   Citation

14. Tremblay RR, Dube JY. 1974 Plasma concentrations of free and non-TeBG bound testosterone in women on oral contraceptives. Contraception. 10:599-605.   Citation

15. SPSS Inc. 1996 SPSS Base 7 for Windows user's guide. Chicago, IL.  

16. SAS Institute Inc. 1989 SAS/STAT user's guide, version 6, ed 4. Cary, NC:SAS Institute.  

17. Ferrini RL, Barrett-Connor E. 1998 Sex hormones and age: a cross-sectional study of testosterone, estradiol and their bioavailable fractions in community-dwelling men. Am J Epidemiol. 147:750-754.   Abstract

18. Vermeulen A, Kaufman JM. 1995 Aging of the hypothalamo-pituitary-testicular axis in men. Horm Res. 43:25-28.   Abstract

19. Gray A, Feldman HA, McKinlay JB, Longcope C. 1991 Age, disease and changing sex hormones levels in middle-aged men: results of the Massachusetts Male Aging Study. J Clin Endocrinol Metab. 73:1016-1025.   Abstract

20. Longcope C, Goldfield SRW, Brambilla DJ, McKinlay J. 1990 Androgens, estrogens, and sex hormone binding globulin in middle-aged men. J Clin Endocrinol Metab. 71:1442-1446.   Abstract

21. Paolo AM, Troster AI, Ryan JJ. 1997 California Verbal Learning Test: normative data for the elderly. J Clin Exp Neuropsychol. 19:220-234.   Abstract

22. Tandberg E, Larsen JP, Aarsland D, Cummings JL. 1996 The occurrence of depression in Parkinson's disease: a community-based study. Arch Neurol. 53:175-179.   Abstract

23. Rabbitt P, Donlan C, Watson P, McInnes L, Bent N. 1995 Unique and interactive effects of depression, age, socioeconomic advantage, and gender on cognitive performance of normal healthy older people. Psychol Aging. 10:307-313.   Abstract

24. Hollingshead AB. 1957 Two factor index of social position. Privately published.  

25. Gray A, Jackson DN, McKinlay JB. 1991b The relation between dominance, anger, and hormones in normally aging men: results from the Massachusetts male aging study. Psychosom Med. 53:375-385.  

26. Sachar EJ. 1973 Plasma and urinary testosterone levels in depressed men. Arch En Psyche. 28:15-18.  

27. Vogel W, Clabber EL, Broverman DM. 1978 Roles of the gonadal steroids hormones in psychiatric depression in men and women. Prog Neuropsychopathol. 2:487-503.  

28. Rubin RT. 1981 Sex steroid hormone dynamics in endogenous depression: a review. J Ment Health. 10:43-59.  

29. Yesavage JA, Davidson J, Widrow L, Berger PA. 1985 Plasma testosterone levels, depression, sexuality, and age. Biol Psychiatry. 20:222-225.   Citation

30. Janowsky J, Oviatt SK, Orwoll ES. 1994 Testosterone influences spatial cognition in older men. Behav Neurol. 108:325-332.  

31. McNair DM, Lorr M, Droppleman LF. 1971 Profile of mood states manual. San Diego, CA: Educational and Industrial Testing Services.  

32. Dickerman RD, Mcconathy WJ. 1997 Testosterone, vasopressin and depression. Prog Neuropsychol. 21:247-248.  

33. Bhasin S, Storer TW, Berman N, et al. 1996 The effects of supraphysiologic doses of testosterone on muscle size and strength in normal men. N Engl J Med. 335:1-7.   Abstract

34. Bremner WJ, Vitiello MV, Prinz PN. 1983 Loss of circadian rhythmicity in blood testosterone levels with aging in normal men. J Clin Endocrinol Metab. 56:1278-1281.   Abstract

35. Dai WS, Kuller LH, La Porte RE, Gutai JP, Falvo-Gerard L, Caggiula A. 1981 The epidemiology of plasma testosterone levels in middle-aged men. Am J Epidemiol. 114:804-816.   Abstract

36. Frank E, Carpenter LL, Kupfer DJ. 1988 Sex differences in recurrent depression: are there any that are significant? Am J Psychiatry. 145:41-45.   Abstract

37. Palinkas LA, Wingard DH, Barrett-Connor E. 1990 Chronic illness and depressive symptoms in the elderly: a population-based study. J Clin Epidemiol. 43:1131-1141.   Abstract

38. McEwen BS. 1991 Our changing ideas about steroids effects on an ever changing brain. Semin Neurosci. 3:497-507.  

39. McEwen BS. 1981 Neural gonadal steroid actions. Science. 211:1303-1311.   Abstract