Higher serum total testosterone levels correlate with increased risk of depressive symptoms in Caucasian women through the entire menopausal transition

Psychoneuroendocrinology 62 (2015) 107–113

Contents lists available at ScienceDirect

Psychoneuroendocrinology journal homepage: www.elsevier.com/locate/psyneuen

Higher serum total testosterone levels correlate with increased risk of depressive symptoms in Caucasian women through the entire menopausal transition Lauren W. Milman a , Mary D. Sammel b , Kurt T. Barnhart a , Ellen W. Freeman a , Anuja Dokras a,∗ a b

Department of Obstetrics and Gynecology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, United States Center for Clinical Epidemiology and Biostatistics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104, United States

a r t i c l e

i n f o

Article history: Received 29 April 2015 Received in revised form 10 July 2015 Accepted 28 July 2015 Keywords: Testosterone Androgens Depressive symptoms Menopause Menopausal transition

a b s t r a c t Background: Despite the high prevalence of depressive symptoms in women, the precise role of sex hormones in mood changes during the menopausal transition is unclear. Previous studies have been inconsistent with regard to identifying the association of androgens, namely total testosterone, with depressive symptoms. Objective: The objectives of this study were to evaluate changes in serum total testosterone levels and depressive symptoms during the entire menopausal transition, and examine the impact of covariates on the association between concurrent serum total testosterone levels and depressive symptoms during this time period. Methods: A longitudinal cohort study (428 women at baseline with 3634 repeated measures) using data from the Penn Ovarian Aging Study, a population-based cohort of late reproductive-aged women, followed through the menopausal transition. Serum hormone parameters and depression scores using the Center for Epidemiological Studies of Depression scale (CES-D) were measured at each annual visit over a 14-year period. General linear (for testosterone) and a generalized negative-binomial model (for depressive symptoms) for repeated measures were used for analysis. Results: Serum total testosterone levels increased progressively over the study period and were significantly associated with older age and with current smoking (p < 0.001, respectively). In the post menopause total testosterone levels were significantly higher in African Americans compared to Caucasians (p = 0.012). The proportion of women with CES-D ≥16 significantly decreased with increasing age and in the post-menopausal period, and were higher in women with a history of depression and hot flashes (p < 0.001). The association between concurrent testosterone levels and high depressive symptoms (CES-D ≥16) differed by race (p = 0.008). In Caucasians, but not African Americans, higher serum testosterone levels were associated with increased depressive symptoms after controlling for several variables including age, obesity status, hot flashes and menopausal status (RR 1.09, 95% CI 1.00–1.17, p = 0.042). Conclusion: In our cohort, testosterone levels were low but progressively increased from premenopause through post menopause. In addition to age and history of depression, we identified race to have a significant interaction between the association of testosterone levels and depressive symptoms. This study further supports the associations between sex hormones and increased risk of having depressive symptoms, although the precise underlying mechanisms for this association remain unclear. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction

∗ Corresponding author at: 3701 Market Street, Suite 800, Philadelphia, PA 19104, United States. Fax: +1 215 349 5512. E-mail address: [email protected] (A. Dokras). http://dx.doi.org/10.1016/j.psyneuen.2015.07.612 0306-4530/© 2015 Elsevier Ltd. All rights reserved.

Depression is a significant medical problem from an individual and a public health perspective, given both the emotional and physical impairment, expense of treatment and loss of productivity (Beck et al., 2011). It is the most common psychiatric disorder glob-

108

L.W. Milman et al. / Psychoneuroendocrinology 62 (2015) 107–113

ally, with a 1.5–3 times higher risk in women than men (Weissman et al., 1996; Angst et al., 2002). In the United States, there is a greater than 20% lifetime prevalence of major depression among women (Kessler et al., 1994) with the highest prevalence seen in women between 40 and 59 years, encompassing the late reproductive years through the menopausal transition (NCHS, 2010). Despite conflicting evidence in cross-sectional population based studies (Dennerstein et al., 1994; McKinlay et al., 1987; Hunter et al., 1986) some longitudinal studies show a trend towards increasing depressive symptoms during the progression through the menopause transition compared to the premenopause. Specifically, data from the Study of Women’s Health Across the Nation (SWAN, (Bromberger et al., 2007; Bromberger et al., 2010)) at 5 and 8-year follow-up time points has shown an increased risk of depressive symptoms during the menopausal transition compared to premenopause. In a prior analysis of the Penn Ovarian Aging Study (POAS), we also reported an increase in depressive symptoms during the transition phase compared to premenopause (Freeman et al., 2004). However, recently in the POAS cohort followed for 14 years the final menstrual period appeared to be pivotal in determining the risk of depression such that there was a higher risk for depressive symptoms before the final menstrual period, and a lower risk following this period into the post menopause (Freeman et al., 2014). The association between the menopausal period and risk of depressive symptoms suggests a potential link between changes in reproductive hormones such as estradiol, follicle stimulating hormone, and luteinizing hormone, and mood during this time period. However several large cohort studies have reported conflicting results (Bromberger et al., 2010; Freeman et al., 2006; Ryan et al., 2009). The association between depressive symptoms and androgens of both adrenal and ovarian origin has also been studied. Two cross-sectional studies showed an inverse relationship between the serum levels of dihydroepiandrosterone sulfate (DHEAS) and depression (Schmidt et al., 2002; Morsink et al., 2007), while longitudinal studies found either no significant association (Bromberger et al., 2010) or a positive association between DHEAS levels and depressive symptoms during the menopause transition (Morrison et al., 2011). The association between testosterone levels and depressive symptoms is also controversial across the menopausal transition. In the 10-year longitudinal Seattle Midlife Women’s Health Study (SMWHS) there was no association between urinary testosterone levels and depressive symptoms (Woods et al., 2008). Interestingly in the early SWAN cohort, which only included premenopausal and early perimenopausal women, there was an inverse association between testosterone and high depressive symptoms (Santoro et al., 2005), while after 8 years of follow-up, increasing levels of testosterone were associated with depressive symptoms (Bromberger et al., 2010). In the Multiethnic Study of Atherosclerosis (MeSA), a significant inverse relationship between free testosterone levels and high depressive symptoms were reported in women who were within the first ten years following menopause (Colangelo et al., 2012). Race can influence both testosterone levels in post-menopausal women (Ouyang et al., 2009) and risk of depression (Bromberger et al., 2010; Freeman et al., 2004) but the association between race and testosterone on depressive symptoms in the perimenopause is unclear (Breslau et al., 2006; Brown et al., 2014). There is evidence to support a positive association between elevated testosterone levels and depressive symptoms in other subgroups of women, such as premenopausal women with major depression (Baischer et al., 1995), those in the peripartum period (Hohlagschwandtner et al., 2001) or those with polycystic ovary syndrome (Dokras et al., 2011). The aims of our study were two fold. First, to describe changes in serum total testosterone levels over the menopause transition, and secondly to evaluate associations between depressive

symptoms and total testosterone during the menopausal transition and to examine the impact of important covariates on the association between concurrent serum testosterone levels and depressive symptoms during this time period. We hypothesized that higher total testosterone levels will correlate with higher depressive symptoms during the menopause transition. 2. Methods 2.1. Study cohort The PENN Ovarian Aging Study (POAS) is a population-based cohort identified via random digit dialing of households in Philadelphia County, Pennsylvania and stratified to include equal numbers of Caucasians and African Americans. Of the 436 women enrolled in the cohort, 428 participants who were included in this study had at least one testosterone value and one rating of depression on the Center for Epidemiologic Studies Depression (CES-D) scale (Radloff, 1977). The details of enrollment have been previously described (Hollander et al., 2001). Briefly, women between 35–47 years with regular menstrual cycles (22–35 days) for the previous 3 months were eligible for the study. Exclusion criteria included current use of psychotropic or hormonal medications, history of a chronic health problem that could compromise ovarian function (e.g. liver disease, breast or endometrial cancer), pregnancy, breast feeding, alcohol or other drug abuse within the past year. The Institutional Review Board of the University of Pennsylvania approved this study, and all enrolled women provided informed written consent. 2.2. Study design In the present study, 428 women contributed 3634 repeated measures over 14 assessment periods from 1996 to 2010. Each assessment consisted of 2 visits per year that were scheduled in the early follicular phase (cycle day 1–6) of 2 consecutive menstrual cycles. After they reached menopause, the assessments were conducted approximately 1 month apart. At each visit a trained research interviewer obtained anthropomorphic measurements, blood samples, and conducted a structured interview to collect general health information, behaviors and demographic information. Participants completed the CES-D at the first visit in each assessment period. 2.3. Study variables 2.3.1. Menopausal status During the assessment visit the current and two preceding menstrual cycle dates were recorded and used to assign menstrual status. Menstrual status was based on the consensus statements of the Stages of Reproductive Aging Workshop (Soules et al., 2001; Harlow et al., 2012). Menopausal status was defined as (1) premenopause, with regular menstrual cycles and less than 7 days cycle length variation; (2) menopause transition, comprising all assessment periods with any cycle length variation beginning with a change in cycle length of 7 or more days from the subject’s baseline for at least one cycle through 11 months of amenorrhea; and (3) postmenopause, defined as greater than or equal to 12 months of amenorrhea. 2.3.2. Hormone levels Non-fasting blood samples were obtained in the early follicular phase in cycling women (days 2–6) at each of the two visits one month apart during each of the 14 assessment periods. Samples were obtained approximately one month apart in non-cycling women. The samples were centrifuged and frozen in aliquots at

L.W. Milman et al. / Psychoneuroendocrinology 62 (2015) 107–113

−80 ◦ C. The assays for total testosterone, total estradiol (E2), and DHEAS were measured using radioimmunoassay (Coat-A-Count kits, Siemens, Deerfield, IL), and conducted in batched samples from 20 women. Each batch consisted of 4 samples from 4 consecutive time points per subject. The assays were performed in duplicate and were repeated if the duplicate values differed by more than 15%. The interassay and intraassay coefficients of variation were consistently less than 5%. The mean value of the two visits was used to represent the hormone level for that assessment period. 2.3.3. Outcome variable The primary outcome measure was the CES-D score (Radloff, 1977). The CES-D is a validated 20-item self-administered questionnaire that screens for depression by measuring the presence of depressive symptoms in the past week. The outcome was dichotomized, using the standard cut-off score of 16 or greater to define ‘high depressive symptoms’ (Boyd et al., 1982; Weissman et al., 1977). The CES-D has been shown to be both a reliable and valid screening tool for depression in African American and Caucasian populations (Roberts, 1980). Potential risk factors were identified based on prior evidence of their association with depressive symptoms (Freeman et al., 2006; Harlow et al., 1999; Roberts et al., 2000). At each assessment period, data were gathered on current smoking (yes, no), currently employed (yes, no), hot flashes (yes, no) and marital status (married, living with partner vs all other). During the initial assessment period, education status (high school or less vs more than high school) and race (African American or Caucasian) were assessed. BMI was calculated at each assessment and dichotomized to obese (>30 kg/m2 ) and non- obese (≤30 kg/m2 ). We used similar criteria as in our previous studies to classify subjects as having a history of depression; based on the participant’s medical history form, or diagnosis on the Primary Care Evaluation form (PRIMEMD), or a CES-D score ≥16 all completed at baseline (Freeman et al., 2014). Current medications (yes, no) included anti-depressants and anxiolytic medications reported at follow up assessments. These measurements were exclusions at enrollment. 2.4. Statistical analysis Descriptive statistics of the cohort at baseline were analyzed using the Chi-square test for categorical variables, and means with standard deviations and the Student’s t-test for continuous variables. Hormone levels were converted to the natural logarithmic scale to reduce the influence of extreme values and meet the assumptions for statistical modeling. Power calculations were conducted using STATA and assumed type I, alpha error of 5%, 80% power, prevalence of depressive symptoms of 30%, on average 10 repeated measures per woman, and within woman correlation among the replicates of 0.6. We also assumed a comparison of the women in the lowest quartile of total testosterone (at baseline) to the highest quartile. Given these assumptions this study had 80% power to detect a 55% increase in CESD symptoms between these groups. To describe changes in serum total testosterone levels over the menopause transition, we employed a general linear regression model (GLM) for repeated measures with generalized estimating equations (GEE) (Zeger et al., 1988) to compare natural-log transformed total testosterone levels over the 14 assessment periods. We report Geometric mean (GM) ratios to quantify these associations (Table 2). For the second objective, associations, prevalence/risk ratios (RR), between concurrent CES-D scores (<16 or ≥16) and total testosterone levels were estimated using a generalized negative-binomial model (Zou, 2004) given the high prevalence of CES-D symptoms. This model also adjusted for repeated measures per woman using (GEE) as above and was

109

Table 1 Characteristics of the POAS cohort at the baseline and final assessments. Covariate

Assessment 1 (n = 428)

Assessment 14 (n = 293)

Age mean years (SD) Menopausal stage, n (%): Premenopause Menopausal transition Postmenopause Body Mass Index (mean BMI kg/m2 (SD)) BMI group, n (%) <30 kg/m2 ≥30 kg/m2 Current Smokers, n (%) African Americans, n (%)

41.4 (3.5)

53.5 (3.5)

428 (100.0)

29.1 (7.8)

1 (0.4) 114 (39.0) 175 (60.8) 32.1 (8.5)

263 (62.6) 157 (37.4) 165 (38.6) 214 (50.0)

124 (47.5) 137 (52.5) 86 (29.4) 139 (47.4)

initially used to perform bivariate analyses with each covariate thought to influence the main exposure and outcome variables, including age, race, marital and employment status, education level, obesity status, menopausal status, history of depression, antidepressant or anti-anxiolytic medication, hot flashes and smoking status. DHEAS and estradiol levels were modestly correlated with total testosterone values (correlation coefficient = 0.26 and 0.12, respectively; p = <0.001) and were included concurrently in building the final model. A priori individual covariates were selected and tested for effect modification with total testosterone level. Race was determined to be an effect modifier of the association of total testosterone levels and high CES-D. All covariates found to be associated with high CES-D with p < 0.2 were considered in multivariable modeling. A backwards step-wise elimination approach was used to select covariates for the final multivariable model. That is, the variable with the largest p-value is dropped from the model, and the association between total testosterone and high CES-D from this model is compared to the previous model. If the association parameter for total testosterone, the log prevalence ratio, changes by 15% or greater, the dropped covariate is considered as a confounder, and retained in the model. If, however, there is no change or only a small change in the association parameter for total testosterone, the covariate is dropped. The covariate with the largest p-value from this new model is evaluated similarly until all remaining covariates are statistically significant, or have been identified as confounders. Obesity status was retained as a covariate in multivariable models as it was a confounder of the race/total testosterone interaction association with high CES-D. The Stata statistical software package, version 12.1 (Stata Corporation, College Station, TX), was used for all analyses. Statistical tests were two-tailed and considered significant with a p ≤ 0.05.

3. Results 3.1. Baseline characteristics Characteristics of the cohort at baseline (n = 428) and at the final assessment period (visit 14, n = 293) are shown in Table 1. Based on our study design at the time of recruitment, fifty percent of the subjects were Caucasian. There was no evidence of differential loss to follow up by race through to the last study visit (p = 0.99). All subjects were premenopausal at baseline, and 60.6% were post-menopausal and 39.0% were in the menopausal transition at the final assessment in year 14. At baseline, 56.1% of the study population had education beyond high school, 57.2% were married, 84.1% had at least one child, and 81.5% were employed. In the entire cohort, 31.6% (1330) observations occurred during the premenopausal time period, 52.2% (2200) during the menopausal transition and 16.2% (683) post menopause.

110

L.W. Milman et al. / Psychoneuroendocrinology 62 (2015) 107–113

Fig. 1. Trend of total testostrone by menupausal stage.

Table 2 Multivariable model for associations of Total Testosterone and covariates during the entire study period. Covariate

GMa Ratio

95% CI

p value

Age BMI ≥30 kg/m2 Current smoking Race group × menopausal stage interaction Race effect stratified by Menopausal stage: Caucasian as reference -Pre-menopause-African American -Transition-African American -Post-menopause-African American

1.03 0.93 1.25

1.02–1.04 0.85–1.02 1.14–1.38

<0.001 0.110 <0.001 0.008

a

0.95 1.00 1.27

0.81–1.12 0.88–1.15 1.05–1.53

0.565 0.961 0.012

GM = geometric mean.

3.2. Total testosterone levels during the menopausal transition The median total testosterone levels at assessment 1 were 10.3 ng/dl (IQR 3.7–17.9). The Geometric mean total testosterone increased by 35% from premenopause through the transition period (p < 0.001), and by 22% from the transition period through post menopause (p < 0.001, Fig. 1). Total testosterone levels were significantly positively associated with age and smoking (Table 2). In the multivariable model adjusting for obesity status, age and smoking status, the association between total testosterone and race was significantly modified by menopausal stage (p = 0.008, Table 2). In the premenopausal and transition stages, the difference in total testosterone levels between races was not significant, however in the post-menopausal period, total testosterone was 27% higher in African Americans than in Caucasians (GM ratio, 1.27; 95% CI, 1.05–1.53; p = 0.012, Table 2). Obesity status did not modify this association further, as a three-way interaction between race, menopausal stage and obesity status was not statistically significant. 3.3. CES-D scores during the menopausal transition Overall, the mean CES-D score decreased from 15.2 ± 10.8 during the first assessment to 9.1 ± 8.7 during the fourteenth assessment. The percent of the study cohort who had a CES-D score ≥16 decreased from 41.1% (172/418) during the first assessment to 29.4% (68/231) in the last assessment. Table 3 describes the

Table 3 Risk of high depressive symptoms (CESD≥16) over the late reproductive years through the menopause transition (unadjusted). Covariate

Risk Ratio (RR)

95% CI

p value

Age Menopausal status: Pre menopause Transition Post menopause Hot Flashes (yes/no) History of depression Current anti-depressants or anti-anxiolytics BMI > 30 kg/m2 Current smoking Married Education > high school Race: Caucasian African American Race group × mean log testosterone interaction, Per 1 sd increasea in log testosterone: Caucasian subgroup African American subgroup Log Estradiol, per 1 SD increasea Log DHEAS, per 1 SD increasea

0.93

0.93-0.95

<0.001

1.00 0.79 0.47 1.04 3.53 1.23

– 0.72-0.86 0.39–0.58 0.96–1.14 2.94–4.24 1.05–1.46

– <0.001 <0.001 0.351 <0.001 0.013

0.89 1.34 0.82 0.64

0.79–1.01 1.19–1.52 0.69–0.97 0.54–0.76

0.068 <0.001 0.024 <0.001

1.00 1.36

– 1.14–1.61

– 0.001 0.009

1.03 0.90 1.05 1.09

0.94–1.11 0.86–0.95 1.01-1.10 1.01-1.17

0.497 <0.001 0.011 0.032

a

sd(log-testosterone) = 0.86; sd(log-estradiol) = 0.60; sd(log-dheas) = 0.65.

associations between high CES-D scores and important covariates of interest. The prevalence of high CES-D scores also declined across the menopause transition, 44.3% during premenopause, 34.8% during the transition period and 21.1% in post menopause (p < 0.001). Advancing age and menopause status both had a significant inverse relationship with depressive symptoms (p < 0.001, Table 3) whereas women with a prior history of depression and current smoking were more likely to have CES-D ≥16 (p < 0.001). Compared to Caucasians, African Americans had a higher risk of depression (RR = 1.36, 95%CI: 1.14–1.61, p = 0.001). 3.4. Association between total testosterone and depressive symptoms We next examined an a priori-defined interaction between race group and total testosterone on CES-D ≥16 (Table 3, p = 0.009). In

L.W. Milman et al. / Psychoneuroendocrinology 62 (2015) 107–113 Table 4 Multivariable model for risk of high depressive symptoms (CESD ≥16) over the late reproductive years through the menopause transition. Covariate

Risk ratio

95% CI

p value

Age Menopausal status: Pre menopause Transition Post menopause History of depression BMI > 30 kg/m2 Current anti-depressant or anti-anxiolytic Hot Flashes Race group × mean log testosterone interaction, Per 1 sd increasea in log testosterone: Caucasian subgroup: African American subgroup:

0.95

0.94–0.97

<0.001

1.00 0.97 0.78 3.15 1.01 1.52

– 0.87, 1.07 0.60,0.99 2.59-3.82 0.88-1.1 1.26, 1.84

Ref 0.523 0.048 <0.001 0.891 <0.001

1.20

1.08, 1.32

0.001 0.008

1.09 0.95

1.00-1.17 0.90-1.01

0.042 0.085

a

sd(log-testosterone) = 0.86.

this unadjusted model, CES-D ≥16 was not significantly associated with total testosterone in Caucasian women (RR = 1.03, 95% CI, 0.94–1.11, p = 0.497), while in African American women the risk of a CES-D ≥16 significantly decreased by 10% with each increasing unit SD log total testosterone (RR = 0.90, 95% CI, 0.86–0.95, p < 0.001). There was also a significant positive association between each unit increase in estradiol and DHEAS (p = 0.011 and p = 0.032, respectively), but no significant interaction with race. In multivariable models, history of depression, anti-depressant or anti-anxiolytic medication, post-menopausal stage, hot flashes and age remained significant independent risk factors, while smoking did not retain its significance (Table 4). Obesity status was retained in the multivariable model as a confounder of the race/CES-D association. After adjusting for all other variables in the model, the interaction between race and total testosterone remained significant (p=0.008) such that the risk of a CES-D ≥16 increased by 9% with each increasing unit SD log total testosterone in Caucasian women (RR = 1.09, 95% CI, 1.00–1.17, p = 0.042) (Table 4). There was however, no significant association between total testosterone and depressive symptoms in African Americans (RR = 0.95, 95% CI, 0.90–1.01, p = 0.085). Both DHEAS and estradiol did not retain their significance with CESD ≥16 in this model.

4. Discussion In our 14-year longitudinal cohort starting in the premenopause we found serum total testosterone levels increased progressively over the menopause transition, with higher levels in African Americans compared to Caucasian women only in the post menopause. Overall depressive symptoms decreased over the menopause transition into the post menopause. We evaluated race specific effects between total testosterone levels and depressive symptoms and report for the first time a significant interaction of race in this association in women during the perimenopause transition, with the direction of the association being positive in Caucasians, but negative in African Americans. The statistical significance of these associations does appear to be confounded by obesity status so that the conclusions differ between the unadjusted and adjusted models. In the multivariable model, higher serum total testosterone levels were significantly associated with increased depressive symptoms in Caucasian women, while the negative association between higher serum total testosterone and CESD ≥16 was not statistically significant for African Americans at the nominal 0.05 level. These models were adjusted for several known variables including age, obesity status, and menopausal status.

111

We comprehensively examined the change in serum total testosterone levels and report low but progressively increasing levels associated with the change in menopausal stage in the entire cohort. Our overall findings are consistent with those of the Michigan Bone Health and Metabolism Study that reported a 44% rise in serum testosterone during a 20-year perimenopausal span (p < 0.0001) (Sowers et al., 2009). In a Norwegian prospective study there was an increase in the free androgen index by 80% over six years associated with decrease in SHBG (Judd and Yen, 1973). On the contrary, the SWAN study showed minimal to no significant change in serum testosterone levels during the menopause transition although the changes were not examined by menopause stage (Bromberger et al., 2010; Burger et al., 2000). Other studies, potentially limited by their smaller sample size, have reported an overall decrease or no change in serum testosterone over the menopausal transition (Rannevik et al., 1995; Overlie et al., 1999). These differences in total testosterone trends might be explained by variations in methodology and timing of sample collection. In order to overcome some of these known limitations in total testosterone measurements, we used the average of two blood draws one-month apart during each assessment period and batched samples from women of the same age. In addition, in regularly cycling women, early follicular phase samples were used to mitigate differences that might be observed due to phase of menstrual cycle (Judd and Yen, 1973). The validity of our findings is further increased as we accounted for several factors known to affect total testosterone levels including age, obesity status, race, menopausal stage and smoking in the presented analysis (Sowers et al., 2001; Longcope and Johnston, 1988; Randolph et al., 2003). Next, we examined change in mean CES-D scores over the menopause transition and noted a decrease in scores and percent women with CES-D ≥16. In the adjusted model however, only post- menopausal stage had significantly lower depressive symptom scores compared to the premenopausal stage. These findings are in agreement with our recent study showing a marked decrease in depressive symptoms in the same cohort after the final menstrual period irrespective of a prior history of depression (Freeman et al., 2014). A significant percentage of women experienced symptoms of depression through the menopause. In our cohort we found associations between previously described risk factors such as age, history of depression, menopausal status and high CES-D scores. In examining the relationship between serum total testosterone and depressive symptoms we found that race played a significant role in the relationship between total testosterone and CES- D ≥16 during this time period. In the multivariable analysis including age, menopausal stage, and other risk factors, Caucasian women with higher total testosterone levels had an increased likelihood of reporting depressive symptoms, compared to women with same covariates but lower total testosterone levels. Our findings are similar to the SWAN study where higher testosterone levels were associated with high CES-D scores although, there was no race interaction reported (Bromberger et al., 2010). Other studies that have shown either no association or an inverse relationship between testosterone and depression have been limited by a small sample size or the cohort was confined to one menopausal stage. For example, in the SMWHS cohort which had 10 years of follow-up there was no association between urinary testosterone measurements and depressive symptoms however, only a subset of subjects provided urine samples (n = 170) (Woods et al., 2008). In the Multiethnic Study of Atherosclerosis (MeSA), a significant inverse relationship between free testosterone levels and high depressive symptoms were reported however, all women in this cohort were menopausal for at least ten years (Colangelo et al., 2012). As described in the Introduction, studies have reported conflicting associations between other steroid hormones and depressive

112

L.W. Milman et al. / Psychoneuroendocrinology 62 (2015) 107–113

symptoms during the menopause transition. We found an association between estradiol and DHEAS levels and CES-D ≥16 in our unadjusted analysis, but in the multivariate model, only total testosterone continued to show a significant positive association modified by race. Using data from 11 assessment periods of the POAS study, we have previously reported that DHEAS levels positively associated with CES-D scores used as a continuous variable, with no significant interaction between DHEAS and race (Morrison et al., 2011). In our current study with a longer period of assessment and use of CES-D ≥16 as a dichotomous outcome we had adequate power to identify the association between testosterone and depressive symptoms and the influence of race. We have previously also reported a weak association between estradiol levels and CES-D scores (Freeman et al., 2004) and a positive association between estradiol variability and high CES-D scores (Freeman et al., 2006). In our current study, in the multivariable model, estradiol was not significant after adjustment for important confounders such as age or menopausal status in the model. The Penn Ovarian Aging Cohort offers unique contributions to understanding the relationship between hormonal changes, psychological or physical symptoms through the entire menopausal transition period. The study period spanned 14 years, and is one of the longest time periods available for study of menopause and depressive symptoms, with a baseline entirely in the premenopausal stage for the entire cohort and the majority of subjects included in this study observed to reach menopause. The ascertainment of each of the key study variables at every assessment was important in minimizing recall bias, and allowing the examination of within-subject change across the entire transition period. Because the population was equally stratified with Caucasians and African Americans, it was possible to compare the impact of race on total testosterone levels, depressive symptoms and their association. Despite numerous strengths, our study has limitations. Total testosterone was measured via direct radioimmunoassay, which may have limited accuracy in the low ranges observed in our study (Vesper and Botelho, 2010; Rosner et al., 2007). Although evidence suggests that radioimmunoassay is comparable to liquid chromatography mass spectrometry (LC/MS), this evidence was obtained from reproductive age women (Legro et al., 2010). As mentioned earlier we applied several strategies to reduce the variability of total testosterone values, such as averaging two hormone levels from the same time period, and optimized the sensitivity of the radioimmunoassay in the lowest ranges of testosterone by performing assays using women of the same age group. Our study examined concurrent levels of total testosterone and CES-D scores in a longitudinal cohort and was not designed to determine causation or determine a temporal relationship. Women with severe depression have been reported to have significantly elevated DHEAS and testosterone levels associated with high cortisol levels (McEwen, 2010). These authors speculated that over activity of the hypothalamic pituitary adrenal axis may result in an elevation of adrenal androgens in women with severe depression. Other investigators have described a U-shaped relationship between gonadal steroid hormones and depression such that higher levels of depressive symptoms are associated with both higher levels of testosterone as seen in reproductive age women with PCOS and lower levels of testosterone as seen in hypogonadal males (Mueller et al., 2014). The serum levels of androgens in women with PCOS are significantly lower than men, suggesting that the absolute levels of testosterone may not be relevant to the postulated underlying mechanisms. The impact of estradiol on depressive symptoms in the perimenopause has been variable (Bromberger et al., 2010; Freeman et al., 2006) suggesting that testosterone may have a direct effect, mediated via its receptor rather than through aromatization to estradiol. There is emerging data on the role of sex

steroids on the structural plasticity as well as regulation of signaling pathways that directly affect neuronal excitability, metabolism and survival (McEwen, 2010). Our findings indicate that the effects of testosterone should be considered in future studies designed to understand the effects of sex hormones in the menopausal transition. Conflict of interest L.W.M., K.T.B, and A.D. have nothing to declare. Acknowledgement M.D.S is a consultant for Swiss Precision Diagnostics, GmbH. E.W.F. received grant support from Forest Laboratories, Inc. This work was supported by National Institutes of Health Grants RO1AG-12745 (to E.W.F., Principal Investigator) and RR024134 (to Clinical and Translational Research Center, Perelman School of Medicine). References Angst, J., Gamma, A., Gastpar, M., Lepine, J.P., Mendlewicz, J., Tylee, A., Depression, R., esearch, in, E., uropean, S., ociety, S., 2002. Gender differences in depression. Epidemiological findings from the European DEPRES I and II studies. Eur. Arch. Psychiatry Clin. Neurosci. 252, 201–209. Baischer, W., Koinig, G., Hartmann, B., Huber, J., Langer, G., 1995. Hypothalamic–pituitary–gonadal axis in depressed premenopausal women: elevated blood testosterone concentrations compared to normal controls. Psychoneuroendocrinology 20, 553–559. Beck, A., Crain, A.L., Solberg, L.I., Unutzer, J., Glasgow, R.E., Maciosek, M.V., Whitebird, R., 2011. Severity of depression and magnitude of productivity loss. Ann. Fam. Med. 9, 305–311. Boyd, J.H., Weissman, M.M., Thompson, W.D., Myers, J.K., 1982. Screening for depression in a community sample. Understanding the discrepancies between depression symptom and diagnostic scales. Arch. Gen. Psychiatry 39, 1195–1200. Breslau, J., Aguilar-Gaxiola, S., Kendler, K.S., Su, M., Williams, D., Kessler, R.C., 2006. Specifying race-ethnic differences in risk for psychiatric disorder in a USA national sample. Psychol. Med. 36, 57–68. Bromberger, J.T., Matthews, K.A., Schott, L.L., Brockwell, S., Avis, N.E., Kravitz, H.M., Everson-Rose, S.A., Gold, E.B., Sowers, M., Randolph Jr., J.F., 2007. Depressive symptoms during the menopausal transition: the Study of Women’s Health Across the Nation (SWAN). J. Affect. Disord. 103, 267–272. Bromberger, J.T., Schott, L.L., Kravitz, H.M., Sowers, M., Avis, N.E., Gold, E.B., Randolph Jr., J.F., Matthews, K.A., 2010. Longitudinal change in reproductive hormones and depressive symptoms across the menopausal transition: results from the Study of Women’s Health Across the Nation (SWAN). Arch. Gen. Psychiatry 67, 598–607. Brown, C., Bromberger, J.T., Schott, L.L., Crawford, S., Matthews, K.A., 2014. Persistence of depression in African American and Caucasian women at midlife: findings from the Study of Women Across the Nation (SWAN). Archives of Women’s Mental Health. Burger, H.G., Dudley, E.C., Cui, J., Dennerstein, L., Hopper, J.L., 2000. A prospective longitudinal study of serum testosterone, dehydroepiandrosterone sulfate, and sex hormone-binding globulin levels through the menopause transition. J. Clin. Endocrinol. Metabol. 85, 2832–2838. Colangelo, L.A., Craft, L.L., Ouyang, P., Liu, K., Schreiner, P.J., Michos, E.D., Gapstur, S.M., 2012. Association of sex hormones and sex hormone-binding globulin with depressive symptoms in postmenopausal women: the multiethnic study of atherosclerosis. Menopause 19, 877–885. Dennerstein, L., Smith, A.M., Morse, C., 1994. Psychological well- being, mid-life and the menopause. Maturitas 20 (1), 1–11. Dokras, A., Clifton, S., Futterweit, W., Wild, R., 2011. Increased risk for abnormal depression scores in women with polycystic ovary syndrome a systematic review and meta-analysis. Obstet. Gynecol. 117, 145–152. Freeman, E.W., Sammel, M.D., Liu, L., Gracia, C.R., Nelson, D.B., Hollander, L., 2004. Hormones and menopausal status as predictors of depression in women in transition to menopause. Arch. Gen. Psychiatry 61, 62–70. Freeman, E.W.S.M., Lin, H., Nelson, D.B., 2006. Associations of hormones and menopausal status with depressed mood in women with no history of depression. Arch. Gen. Psychiatry 63, 375–382. Freeman, E.W., Sammel, M.D., Boorman, D.W., Zhang, R., 2014. Longitudinal pattern of depressive symptoms around natural menopause. JAMA Psychiatry 71, 36–43. Harlow, B.L., Cohen, L.S., Otto, M.W., Spiegelman, D., Cramer, D.W., 1999. Prevalence and predictors of depressive symptoms in older premenopausal women: the Harvard Study of Moods and Cycles. Arch. Gen. Psychiatry 56, 418–424.

L.W. Milman et al. / Psychoneuroendocrinology 62 (2015) 107–113 Harlow, S.D., Gass, M., Hall, J.E., Lobo, R., Maki, P., Rebar, R.W., Sherman, S., Sluss, P.M., de Villiers, T.J., 2012. Straw +10 collaborative group. executive summary of the stages of reproductive aging workshop +10: addressing the unfinished agenda of staging reproductive aging. J. Clin. Endocrinol. Metab. 97, 1159–1168. Hohlagschwandtner, M., Husslein, P., Klier, C., Ulm, B., 2001. Correlation between serum testosterone levels and peripartal mood states. Acta Obstet. Gynecol. Scand. 80, 326–330. Hollander, L.E.F.E., Sammel, M.D., Berlin, J.A., Grisso, J.A., Battistini, M., 2001. Sleep quality, estradiol levels, and behavioral factors in late reproductive age women. Obstet. Gynecol. 98, 391–397. Hunter, M., Battersby, R., Whitehead, M., 1986. Relationships between psychological symptoms, somatic complaints and menopausal status. Maturitas 8, 217–228. Judd, H.L., Yen, S.S., 1973. Serum androstenedione and testosterone levels during the menstrual cycle. J. Clin. Endocrinol. Metab. 36, 475–481. Kessler, R.C., McGonagle, K.A., Zhao, S., Nelson, C.B., Hughes, M., Eshleman, S., Wittchen, H.U., Kendler, K.S., 1994. Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States. Results from the National Comorbidity Survey. Arch. Gen. Psychiatry 51, 8–19. Legro, R.S., Schlaff, W.D., Diamond, M.P., Coutifaris, C., Casson, P.R., Brzyski, R.G., Christman, G.M., Trussell, J.C., Krawetz, S.A., Snyder, P.J., Ohl, D., Carson, S.A., Steinkampf, M.P., Carr, B.R., McGovern, P.G., Cataldo, N.A., Gosman, G.G., Nestler, J.E., Myers, E.R., Santoro, N., Eisenberg, E., Zhang, M., Zhang, H., 2010. Total testosterone assays in women with polycystic ovary syndrome: precision and correlation with hirsutism. J. Clin. Endocrinol. Metab. 95, 5305–5313. Longcope, C., Johnston Jr., C.C., 1988. Androgen and estrogen dynamics in pre- and postmenopausal women: a comparison between smokers and nonsmokers. J. Clin. Endocrinol. Metab. 67, 379–383. McEwen, B.S., 2010. Stress, sex, and neural adaptation to a changing environment: mechanisms of neuronal remodeling. Ann. N. Y. Acad. Sci. 1204, E38–E59, Suppl. McKinlay, J.B., McKinlay, S.M., Brambilla, D., 1987. The relative contributions of endocrine changes and social circumstances to depression in mid-aged women. J. Health Soc. Behav. 28, 345–363. Morrison, M.F., Freeman, E.W., Lin, H., Sammel, M.D., 2011. Higher DHEA-S (dehydroepiandrosterone sulfate) levels are associated with depressive symptoms during the menopausal transition: results from the PENN Ovarian aging study. Arch. Women’s Mental Health 14, 375–382. Morsink, L.F., Vogelzangs, N., Nicklas, B.J., Beekman, A.T., Satterfield, S., Rubin, S.M., Yaffe, K., Simonsick, E., Newman, A.B., Kritchevsky, S.B., Penninx, B.W., 2007. Health ABCs Associations between sex steroid hormone levels and depressive symptoms in elderly men and women: results from the Health ABC study. Psychoneuroendocrinology 32, 874–883. Mueller, S.C., Grissom, E.M., Dohanich, G.P., 2014. Assessing gonadal hormone contributions to affective psychopathologies across humans and animal models. Psychoneuroendocrinology 46, 114–128. NCHS, 2010. CfDCaPCNCfHS National Health and Nutrition Examination Survey Data 2007–2010. U. S. Department of Health and Human Services. Centers Dis. Control Prev. 60, 1747. Ouyang, P., Vaidya, D., Dobs, A., Golden, S.H., Szklo, M., Heckbert, S.R., Kopp, P., Gapstur, S.M., 2009. Sex hormone levels and subclinical atherosclerosis in postmenopausal women: the multi-ethnic study of atherosclerosis. Atherosclerosis 204, 255–261. Overlie, I., Moen, M.H., Morkrid, L., Skjaeraasen, J.S., Holte, A., 1999. The endocrine transition around menopause-a five years prospective study with profiles of gonadotropines, estrogens, androgens and SHBG among healthy women. Acta Obstet. Gynecol. Scand. 78, 642–647.

113

Radloff, L.S., 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Meas. 1, 385–401. Randolph Jr., J.F., Sowers, M., Gold, E.B., Mohr, B.A., Luborsky, J., Santoro, N., McConnell, D.S., Finkelstein, J.S., Korenman, S.G., Matthews, K.A., Sternfeld, B., Lasley, B.L., 2003. Reproductive hormones in the early menopausal transition: relationship to ethnicity, body size, and menopausal status. J. Clin. Endocrinol. Metab. 88, 1516–1522. Rannevik, G., Jeppsson, S., Johnell, O., Bjerre, B., Laurell-Borulf, Y., Svanberg, L., 1995. A longitudinal study of the perimenopausal transition: altered profiles of steroid and pituitary hormones, SHBG and bone mineral density. Maturitas 21, 103–113. Roberts, R.E., Kaplan, G.A., Shema, S.J., Strawbridge, W.J., 2000. Are the obese at greater risk for depression? Am. J. Epidemiol. 152, 163–170. Roberts, R.E., 1980. Reliability of the CES-D Scale in different ethnic contexts. Psychiatry Res. 2, 125–134. Rosner, W., Auchus, R.J., Azziz, R., Sluss, P.M., Raff, H.P., osition statement, 2007. Utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement. J. Clin. Endocrinol. Metabol. 92, 405–413. Ryan, J., Burger, H.G., Szoeke, C., Lehert, P., Ancelin, M.L., Henderson, V.W., Dennerstein, L., 2009. A prospective study of the association between endogenous hormones and depressive symptoms in postmenopausal women. Menopause 16, 509–517. Santoro, N., Torrens, J., Crawford, S., Allsworth, J.E., Finkelstein, J.S., Gold, E.B., Korenman, S., Lasley, W.L., Luborsky, J.L., McConnell, D., Sowers, M.F., Weiss, G., 2005. Correlates of circulating androgens in mid-life women: the study of women’s health across the nation. J. Clin. Endocrinol. Metabol. 90, 4836–4845. Schmidt, P.J., Murphy, J.H., Haq, N., Danaceau, M.A., St, C., lair, L., 2002. Basal plasma hormone levels in depressed perimenopausal women. Psychoneuroendocrinology 27, 907–920. Soules, M.R., Sherman, S., Parrott, E., Rebar, R., Santoro, N., Utian, W., Woods, N., 2001. Executive summary: stages of reproductive aging workshop (STRAW). Fertil. Steril. 76, 874–878. Sowers, M.F., Beebe, J.L., McConnell, D., Randolph, J., Jannausch, M., 2001. Testosterone concentrations in women aged 25-50 years: associations with lifestyle, body composition, and ovarian status. Am. J. Epidemiol. 153, 256–264. Sowers, M.F., Zheng, H., McConnell, D., Nan, B., Karvonen-Gutierrez, C.A., Randolph Jr., J.F., 2009. Testosterone, sex hormone-binding globulin and free androgen index among adult women: chronological and ovarian aging. Hum. Reprod. 24, 2276–2285. Vesper, H.W., Botelho, J.C., 2010. Standardization of testosterone measurements in humans. J. Steroid Biochem. Mol. Biol. 121, 513–519. Weissman, M.M., Sholomskas, D., Pottenger, M., Prusoff, B.A., Locke, B.Z., 1977. Assessing depressive symptoms in five psychiatric populations: a validation study. Am. J. Epidemiol. 106, 203–214. Weissman, M.M., Bland, R.C., Canino, G.J., Faravelli, C., Greenwald, S., Hwu, H.G., Joyce, P.R., Karam, E.G., Lee, C.K., Lellouch, J., Lepine, J.P., Newman, S.C., RubioStipec, M., Wells, J.E., Wickramaratne, P.J., Wittchen, H., Yeh, E.K., 1996. Cross-national epidemiology of major depression and bipolar disorder. JAMA 276, 293–299. Woods, N.F., Smith-DiJulio, K., Percival, D.B., Tao, E.Y., Mariella, A., Mitchell, S., 2008. Depressed mood during the menopausal transition and early postmenopause: observations from the Seattle midlife women’s health study. Menopause 15, 223–232. Zeger, S.L., Liang, K.Y., Albert, P.S., 1988. Models for longitudinal data: a generalized estimating equation approach. Biometrics 44, 1049–1060. Zou, G., 2004. A modified poisson regression approach to prospective studies with binary data. Am. J. Epidemiol. 159, 702–706.