Associations between low sex hormone concentrations and depression in older women: An observational study

Abstract

Objective:

We investigated whether low sex hormone concentrations are associated with depression in older women.

Study design:

This was a cross-sectional study of Australian women, aged at least 70 years, not taking medications modulating sex hormone levels. Associations between hormones, measured by liquid chromatography-tandem mass spectrometry, and depression were examined by logistic regression adjusted for potential confounders.

Main outcome measures:

The primary outcome was a Center for Epidemiologic Studies Depression score >10, designated as ‘depression’, with an expanded definition that included anti-depressant use as a secondary outcome.

Results:

For the 5535 participants in the analysis, median age 74.0 years (interquartile range 71.7–77.7), depression prevalence was 5.8 % (95 % CI 5.2–6.4 %). In the adjusted models, a statistically significantly greater likelihood of depression was seen for women with testosterone and oestrone blood concentrations in quartile 1 compared with quartiles 2–4 (odds ratio 1.33, 95 % CI 1.04 to 1.70, p = 0.022; and 1.37, 95 % CI 1.06 to 1.78, p = 0.017, respectively). For the expanded definition, the odds ratios for the lowest testosterone and oestrone quartile compared with other quartiles were 1.47 (95 % CI 1.24 to 1.75, p < 0.001) and 1.31 (95 % CI 1.09 to 1.58, p < 0.001), respectively. A significant association for low DHEA was seen only for the expanded definition of depression (1.36, 95 % CI 1.13 to 1.64, p = 0.001). Receiver operating characteristic curves showed that the contribution of each sex hormone to the likelihood of depression was small.

Conclusions:

Amongst older women not taking medications that influence sex hormone concentrations, low testosterone and oestrone levels are associated with a greater likelihood of depression, but the effects are small.

Trial registration:

International Standard Randomized Controlled Trial Number Register (ISRCTN83772183) and clinicaltrials.gov (NCT01038583).

1. Introduction

Depression is the most common mental illness world-wide, and has been reported as more prevalent in women than in men [1]. It has been suggested that relative sex hormone insufficiency, notably lower testosterone concentrations in postmenopausal women compared with men [2], may contribute to a greater likelihood of depression. A systematic review of depression in the elderly indicated the point prevalence of depression to be between 4.0 % and 10.3 % for women over 75 years [3]. Women of this age are approximately 50 % more likely to be affected than men [3].

We have shown that blood testosterone concentrations in women decline from the age of 18 years, with no acute change at natural menopause [4]. Following menopause, both testosterone and oestrone, the predominant postmenopausal oestrogen, are primarily metabolites of dehydroepiandrosterone (DHEA) produced by the adrenal glands. After reaching a nadir at approximately 62–63 years of age, blood testosterone concentrations increase and approximate those of young premenopausal women by the age of 70 years. We recently found that women aged 70 years or greater with blood testosterone or DHEA concentrations in the lowest quartile have a two-fold greater risk of an ischaemic cardiovascular event [5]. As depression is a major cause of morbidity in older women, whether low testosterone is associated with a greater risk of depression in this age group merits investigation.

Our systematic review of association between testosterone and depressive symptoms in older postmenopausal women has yielded inconsistent results with some reporting an inverse association and others no association [6]. Most of these studies measured testosterone by immunoassays that lack precision at the low testosterone concentrations seen in women, rendering the findings unreliable. Few studies have used liquid chromatography-tandem mass spectrometry (LCMS) which provides precision for the simultaneous measurement of several sex hormones at low concentrations. A retrospective analysis of women aged 20 to 80 years found no significant association between testosterone measured by LCMS and depressive symptoms [7]. Our systematic review did not find any association between DHEA and depressive symptoms in postmenopausal women, however most of the included studies were limited by risk of bias [8]. Few studies have taken into consideration the influence of medications commonly used by older women that influence sex hormone blood levels, such as anti-oestrogens, anti-androgens and glucocorticosteroids, further limiting their interpretation. In postmenopausal women the main circulating oestrogen is oestrone, with oestradiol being below the limit of detection when measured either by immunoassay [9,10] or LCMS [11] in the majority of postmenopausal women. However, data for oestrone and depressive symptoms in older women are lacking.

The Sex Hormones in Older Women (SHOW) study documented age-specific reference ranges for each of the main sex hormones, measured by LCMS, in community-dwelling women aged ≥70 years [11], who were also screened for depression by a validated questionnaire. We now report the associations between low blood concentrations of testosterone and oestrone, and their precursor hormone DHEA, and depressive symptoms in this large sample of older women, taking into account other factors likely to contribute to the pathogenesis of depression. For the purpose of this study, we defined “depression” as a symptom score above the threshold for the Center for Epidemiological Studies Depression Scale (CESD-10) that has been shown to correlate strongly with clinical depression.

2. Materials and methods

2.1. Participants and study design

The SHOW study was a prospective sub-study of the ASPirin in Reducing Events in the Elderly Study (ASPREE) [11]. Australian participants in ASPREE were recruited between March 10, 2010 and December 31, 2014 through partnerships with over 2500 primary care practitioners. Eligible women were aged 70 years and older at recruitment, living in the Australian states of Victoria, South Australia, New South Wales, and Tasmania, or the Australian Capital Territory. ASPREE study exclusions included a life expectancy of <5 years, impaired or inability to perform any of the 6 basic activities of daily living, dementia or documented cognitive impairment, a prior major ischaemic vascular event, a high risk of bleeding or other contraindication to aspirin, significant anaemia, or blood pressure of 180/105 mmHg or more [11].

Biobank samples were available for the measurement of sex hormones for 6358 women. For the present study, we excluded women taking sex hormone therapy (oestrogens, androgens and progestogens), anti-oestrogens (tamoxifen, aromatase inhibitors, and gonadotrophin releasing hormone analogues), anti-androgens (cyproterone acetate, spironolactone) or systemic glucocorticosteroids.

2.2. Clinical variables

All demographic data, including smoking and alcohol consumption, and all medication use were documented at randomization. Antidepressants were defined as medications included in the NO6A code of the WHO ATC/DDD Index 2022 [12]. Clinical measurements included weight and height.

2.3. Measurement of sex hormones

Study baseline non-fasting blood samples were stored under nitrogen vapour. Testosterone, DHEA, oestradiol and oestrone were measured simultaneously in plasma samples by LCMS at the ANZAC Research Institute, University of Sydney, Australia as previously described [11]. Deuterated isotopes were d3-testosterone, d2-DHEA, d4-oestradiol and d4-oestrone. All hormone standards and internal standards were from the National Measurement Institute (NMI, Sydney, Australia), except for d4-oestrone (Steraloids, Newport RI, USA) and oestradiol and oestrone (Cerilliant, Round Rock, Texas, USA). The LCMS was performed on an API-5000 triple-quadrupole mass spectrometer (Applied Biosystems/MDS SCIEX, Foster City, CA/Concord, Ontario, Canada). The limits of detection (LOD), limits of quantification, and within-run and between-run range of coefficients of variation (CV) over 3 quality control concentrations were for testosterone 0.035 nmol/L, 0·09 nmol/L, 2·0 %, 3·9–6·5 %, for DHEA 0·07 nmol/L, 0·17 nmol/L, 3–6 %, 8–12 %, for oestradiol 11 pmol/L, 18 pmol/L, 6·6 %, 4·8–8·6 %, and for oestrone 3·7 pmol/L, 11 pmol/L, 4·7 %, 4·6–7·5 %, respectively [11].

2.4. Outcomes

Depression was a prespecified secondary outcome of ASPREE [13]. Participants completed the CESD-10 at study baseline [13]. The CESD-10 is a 10-item questionnaire that identifies symptoms aligned to the diagnostic criteria for depression including depressed mood, hopelessness, and sleep disruption. It has a sensitivity of 97 % and specificity of 84 % for clinically diagnosed late life depression [14]. We defined prevalent depression as a CESD-10 cut-off score of ≥10, irrespective of antidepressant medication use, as scores of 10 or more as this correlates with clinical depression [14]. However, this would miss women with treated depression whose hormones may have contributed to the depression etiology. Therefore, an expanded definition of depression, that included prevalent depression combined with current use of an antidepressant, was examined as a secondary outcome, cognisant that these medications may be prescribed for conditions other than depression.

2.5. Statistical analysis

A sample size of 5535 women provides for a prevalence estimate for depression of 5.8 % with a 95 % confidence interval (CI) from 5.2 % to 6.4 %. The primary outcome was the association between low sex hormones levels and prevalent depression, with the expanded definition as secondary outcome.

The median and interquartile range (IQR) (for hormone levels, inter-decile range 10th and 90th percentiles) were used to summarise skewed data. Categorical data were reported using the frequency and percentage. Differences in depression prevalence by age group and antidepressant use were assessed using the Chi-squared test. There was no missing data for the CESD-10 questionnaire and <1 % missing data for sex hormone levels. Serum oestradiol was not examined in our analysis as 66 % of the sample had oestradiol concentrations below the assay’s LOD [11].

Based on our prior findings [5], a low sex hormone blood concentration was designated as a concentration in the lowest quartile (Q). An omnibus test was performed for each hormone to determine whether the outcomes for each of Q2, Q3 and Q4 differed from each other. Where this was not the case, the outcomes for women with low sex hormone concentrations (Q1) were compared with those with not low concentrations (Q2–4).

To assess the association between low sex hormones and depression, we used univariable and multivariable logistic regression models. Potential risk factors for depression included in the multivariable regression models were age, body mass index (BMI), smoking status, alcohol consumption, years of education and living status. Receiver operating characteristic (ROC) curves were generated and the area under the curve (AUC) for each was used as an indication as to the degree to which the model discriminated between depressed and non-depressed individuals. Although our a priori analysis was to examine quartiles, in a sensitivity analysis a similar analytical approach was used for an exploratory analysis for tertiles and quintiles of each hormone.

All statistical tests were two-sided, and a p value of <0.05 was considered statistically significant. All statistical analyses were performed using Stata 16.0 (Stata Corporation, College Station, TX, USA). This manuscript was written in accordance with the STROBE statement for observational studies [15].

3. Results

SHOW study participants comprised 6392 (69·6 %) of the 9180 Australian women recruited to the ASPREE trial who provided biobank samples. Sufficient serum for baseline serum concentrations of sex hormones measurement was available for 6358 (99·5 %) participants (Fig. 1). After excluding 823 women taking medications that would affect their sex hormone levels, 5535 women, median (IQR) age at recruitment 74.0 (71.7, 77.7) years, provided data for analysis. Most (99 %) were of European ancestry, over half (52 %) had <12 years of education, and 31 % had obesity (Table 1). The prevalence of depression was 5.8 % (95 % CI: 5.2 to 6.4) and was similar amongst those aged 70 to 79 years and 80 years and older (5.7 % [95 % CI: 5.7 to 6.4] vs 5.9 % [95 % CI: 4.5 to 7.5]; p = 0.877). Overall, 7.9 % of the women were current antidepressant users and women with depression were more likely to be taking an antidepressant than non-depressed women (17.0 % versus 7.4 %, p < 0.001).

Fig. 1.

Inclusion of participants in the analysis.

Table 1.

Baseline characteristics of the study participants.

Clinical characteristics	Overall (N = 5535)
Age in years, median (IQR)	74.0 (71.7–77.7)
Age group, n (%)
70–74	2781 (50)
75–79	1745 (32)
80–84	760 (14)
≥85	249 (4)
Weight (kg), median (IQR)a	69.8 (61.9–79.2)
Height (cm), median (IQR)b	159.0 (155.0–163)
Body mass index (kg/m2)c, n (%)
<18.5	44 (0.8)
18.5 to <25	1560 (28.3)
25 to <30	2187 (39.7)
≥30	1715 (31.2)
Baseline living statusb
Alone	2272 (41)
Not alone	3249 (59)
Ethnicity, n (%)
European ancestry	5470 (98.8)
Other	65 (1.2)
Years of educationb, n (%)
< 12 years	2868 (52)
12 years and above	2653 (48)
Smoking status, n (%)
Current	156 (2.8)
Former	1739 (31.4)
Never	3640 (65.8)
Alcohol consumption, n (%)
Current	4130 (74.6)
Former	216 (3.9)
Never	1189 (21.5)
Baseline antidepressants usee, n (%)
Yes	423 (7.9)

Number with non-missing data: ^an = 5517

^bn = 5521

^cn = 5506

^dn = 5382

^en = 5326.

Other include 32 Asian and 7 Aboriginal/Torres Strait Islander, IQR = interquartile range.

An omnibus test did not demonstrate any significant differences between quartiles 2, 3 and 4 in the any of the adjusted models for any of the hormones. Therefore, these quartiles were combined as the reference against which concentrations in the lowest quartile (Q1) were compared. The unadjusted odds ratios (ORs) for prevalent depression for Q1 of each sex hormone compared with the higher quartiles (Q2–4), and adjusted ORs allowing for age, BMI, smoking, alcohol consumption, education, and living circumstances are provided in Table 2. A statistically significantly greater likelihood of depression was seen for testosterone and oestrone blood concentrations in Q1 in comparison with Q2–4 (OR 1.33 (95 % CI, 1.04 to1.70) p = 0.022 and 1.37 (95 % CI, 1.06 to 1.78) p = 0.017, respectively).

Table 2.

Associations between sex hormones quartiles and depression.

Concentrations of sex steroids by quartile, median (10th, 90th percentile)	Depression (CESD-10 score ≥10)		Depression (CESD-10 score ≥10 and/or antidepressants use)
	Unadjusteda OR (95 % CI) p value	Adjustedb OR (95 % CI) p value	Unadjusteda OR (95 % CI) p value	Adjustedb OR (95 % CI) p value
Testosterone
Quartile 2 to 4 combined (ref) 0.45 (0.28–1.07) nmol/Ld n = 3894, 70.4 %	1.00	1.00	1.00	1.00
Quartile 1 0.17 (0.10–0.24) nmol/Ld n = 1641, 29.6 %	1.33 (1.05–1.69)	1.33 (1.05–1.69)	1.46 (1.24–1.73)	1.47 (1.24–1.74)
	0.018	0.019	<0.001	<0.001
AUC	0.531	0.583	0.542	0.589
Oestrone
Quartile 2 to 4 combined (ref) 214.52 (144.24–366.16) pmol/Lc n = 4130, 74.6 %	1.00	1.00	1.00	1.00
Quartile 1 96.16 (48.08–122.05) pmol/Lc n = 1405, 25.4 %	1.33 (1.04–1.70)	1.42 (1.11–1.83)	1.28 (1.08–1.53)	1.35 (1.12–1.61)
	0.022	0.006	0.006	0.001
AUC	0.529	0.590	0.525	0.584
Dehydroepiandrosterone
Quartile 2 to 4 combined (ref) 3.22 (1.98–6.73) nmol/Le n = 4109, 74.2 %	1.00	1.00	1.00	1.00
Quartile 1 1.14 (0.55–1.59) nmol/Le n = 1426, 25.8 %	1.18 (0.92–1.52)	1.16 (0.91–1.50)	1.37 (1.15–1.63)	1.36 (1.14–1.62)
	0.184	0.248	<0.001	0.001
AUC	0.519	0.578	0.532	0.583

Quartiles did not include exactly 25 % of the observations for each hormone as for some hormones, especially testosterone and dehydroepiandrosterone, a number of women shared the same value.

OR = odds ratio; CI = confidence interval; AUC = area under the curve.

^aSeparate univariable models fitted for each hormone.

^bModels for each hormone were all adjusted for age, BMI, smoking status, alcohol consumption, years of education, and living status.

^cTo convert to pg/mL divide by 3.699.

^dTo convert to ng/dL divide by 0.0347.

^eTo convert to mg/L divide by 3.467.

In order to better understand the contribution of each hormone to depression, we developed a separate model without any hormone variable but including the potential risk factors. We then explored the effect of adding each hormone to the model. The AUC for the adjusted model with no hormone included was 0.574. The addition of each hormone had marginal effects; the AUC for the separate models that included testosterone, oestrone or DHEA were 0.583, 0.590, and 0.578, respectively (data not shown).

When prevalent depression was combined with current antidepressant use, the prevalence of the expanded definition of depression was 12.4 % (95 % CI: 11.6 to 13.3). Application of the expanded definition did not change the findings for testosterone and oestrone (ORs for Q1 vs Q2–4 were 1.47 (95 % CI, 1.24 to 1.75) p < 0.001 and 1.31 (95 % CI, 1.09 to 1.58) p < 0.001, respectively). A statistically significant association for low DHEA was also seen for the expanded definition of depression (OR 1.36 (95 % CI, 1.13 to 1.64) p = 0.001). In the expanded definition model, the AUC excluding hormones was 0.569, and with the inclusion of testosterone, oestrone or DHEA, increased to 0.589, 0.584 and 0.583, respectively (data not shown).

When exploratory analyses of the hormone concentrations were conducted using tertiles and quintiles the odds ratios for the lowest tertile compared with tertiles 2 and 3 combined and the lowest quintile compared with quintiles 2 to 5 combined were similar to those seen for the primary analysis (Supplementary Tables 1 and 2).

4. Discussion

This study indicates that older women with low testosterone and oestrone have over a 30 percent greater likelihood of prevalent depression than women with higher levels of these hormones, in addition to other adverse physical and psychological health factors and socio-cultural influences. However, the overall contribution to the likelihood of depression explained by the blood concentrations of these hormones was very small. This was not unanticipated in view of the complexities of both sex hormone physiology in postmenopausal women and depression. Thus, while our findings indicate that low testosterone and oestrone may be detrimental to mood, their measurement does not provide useful information about depression risk in individual older women.

When the definition of depression was expanded to include all antidepressant users, the ORs for low testosterone and low oestrone were minimally changed, and low DHEA became significantly different in both the unadjusted and adjusted models. The apparently smaller AUC for the expanded definition of depression suggests that the same model discriminated less effectively in this larger group. This may be due to inclusion of women with effectively treated depression and women taking antidepressants for other conditions. Irrespective of the definition of depression used, the omnibus test demonstrated that the findings for each of the sex hormones did not differ between the quartiles above Q1, in either the unadjusted or adjusted models. Together this suggests that relative ‘insufficiency’ of each of testosterone and oestrone is associated with a greater likelihood of depression, but that having concentrations of either hormone above the second quartile conveys no additional benefit.

Further interpretation of our findings requires consideration of sex hormone physiology. During the female reproductive years, circulating testosterone is derived from the ovaries and adrenals, and oestradiol from the ovaries, is the main circulating oestrogen. Following menopause, testosterone and oestrone, the main postmenopausal oestrogen, are produced from their adrenal precursors, DHEA and androstenedione, in peripheral tissues such as the brain, bone, adipose and endothelial cells. The blood concentrations of testosterone, oestrone and oestradiol are not determined by blood levels of the adrenal hormones, which are the most abundant sex hormones in women, but by the amounts and activity of the enzymes required for their biosynthesis in peripheral tissues, and the extent to which these hormones escape intracellular metabolism and spill over into the circulation.

Our primary outcome was the association between low sex hormone concentrations and prevalent depression. Although depression in the elderly has been described as under diagnosed and under treated [16], antidepressant use is greatest in Australian women over the age of 75 years [17] and 7.9 % of the women in our study were taking an anti-depressant. Hence, exclusion of antidepressant users without prevalent depression from our analysis meant excluding a large number of effectively treated women, whose sex hormone milieu was potentially contributing to their depression. As we were not able to identify the women prescribed antidepressants for conditions other than depression, we do not know if their inclusion meaningfully diluted the pool of women with diagnosed depression. Nonetheless, the findings for the expanded definition were consistent with those for prevalent depression, supporting an association between low testosterone and oestrone and depression in older women.

The picture that low testosterone was associated with a greater likelihood of depression is in line with studies of hypogonadal men [18]. It has been hypothesised that testosterone may be protective against depression through its effects in the brain, particularly in the hippocampus, a structure linked to depression and anxiety [2]. Testosterone exerts androgen receptor-mediated neuroprotective and anti-inflammatory actions in the brain, enhances neuroplasticity and neurogenesis, and protects against oxidative stress in animal models [19]. While testosterone therapy may be mood elevating, it remains to be established that low testosterone adversely effects neural networks influencing mood.

Testosterone and oestrone are both immediate precursors for oestradiol production, with oestradiol in postmenopausal women primarily produced and metabolised within target tissues. Thus, over 98 % of our study sample had measurable testosterone and oestrone, with median concentrations similar to those of premenopausal women, while two-thirds had oestradiol levels below the LOD [11]. Oestradiol may have direct effects on neural networks involved in affective mood regulation in the brain, such that low oestradiol may contribute to amplification of response to stressful life events and processing and memory for negative emotional experiences.

As most older women have extremely low oestradiol, it is not surprising that two previous studies did not find an association between blood oestradiol concentrations and depression in older women, despite the use of LCMS [7,20]. Neither study reported the number of women with oestradiol levels below the LOD [7,20]. We have shown that blood oestrone concentrations are a reliable proxy for oestradiol in older postmenopausal women, independent of age, BMI and testosterone levels [21]. Therefore, in our study higher oestrone being protective against depression may reflect a concurrent protective effect of oestradiol. Alternatively, oestrone may exert direct effects on neural networks important for mood. Traditionally oestrone has been considered to be of little biological importance as its potency has been estimated as only one-tenth that of oestradiol [22]. However, blood oestrone concentrations in our sample, on average, were at least 10-fold greater than oestradiol [11]. In postmenopausal women, oestrone concentrations are positively associated with higher bone mineral density [23] and greater breast cancer risk, [24] and protective against bowel cancer [25]. Therefore, it is biologically plausible that oestrone exerts direct brain effects that influence mood in older women.

Mixed findings have been reported for DHEA and depression in older women [8]. To our knowledge, prior studies have not measured DHEA by LCMS and have not excluded, or taken into account, the use of systemic glucocorticosteroid therapy which suppresses adrenal hormone production and elevates mood. The association between low DHEA and depression in our study was limited to the expanded definition. DHEA and its sulphate are primarily precursors for testosterone and oestrogen production, and circulate in sufficient concentrations for this throughout life. Hence the lack of an association between DHEA and prevalent depression may reflect DHEA primarily exerting indirect biological effects through its metabolites.

While our findings support potential mood modulating effects of sex hormones in older women, they should not be interpreted as indicating a role for sex hormone therapy in this population. The AUC of the model that excluded any hormone variable was only very modestly increased with the addition of any of the sex hormones, indicating none of these meaningfully increased the likelihood of depression. Oestrogen therapy does not improve depressive symptoms in small studies of older postmenopausal women [26,27]. Similarly, our systematic review and meta-analysis of clinical trials of testosterone therapy for postmenopausal women showed no benefit of testosterone for depressive symptoms or psychological and general wellbeing [28], with the former reaffirmed by a more recent clinical trial in treatment-resistant depressed women [29]. DHEA did not improve quality of life or wellbeing in 2 clinical trials involving older women [30,31].

Strengths of the study include the large sample size, comprehensive characterisation of the participants, measurement of sex hormones by LCMS, use of a validated depression questionnaire and exclusion of women using exogenous sex hormones, glucocorticosteroids, anti-oestrogen and anti-androgen therapy. Our findings could be due to reverse causation, with depression causing low testosterone and oestrone through impaired adrenal function and reduced adrenal precursor DHEA. We consider this unlikely as DHEA concentrations were more than adequate for oestrone and testosterone production, and DHEA was not associated with prevalent depression. As the majority of study participants were of European ancestry our findings cannot be extrapolated to all women. Furthermore, as ASPREE study eligibility included not having had a prior major cardiovascular event, our sample was relatively healthy compared with the general population. However, this is also a study strength, as having a relatively healthy sample potentially eliminated some health conditions that may have been confounders in our analysis.

The study did not include a structured clinical assessment of depression but used a validated questionnaire with a conservative cut-off. Thus, more fine-grained fluctuations in mood may have been missed. Although our analysis was based on a single blood sample, not necessarily drawn on the same day of questionnaire completion, we have shown stability of sex hormones measured longitudinally in this cohort [32]. While sufficient data for oestradiol was lacking due to the very low levels in older postmenopausal women, oestrone provides a robust surrogate for overall oestrogen exposure in older postmenopausal women [21]. Despite adjusting for several potential confounders, the possibility of residual confounding effect cannot be excluded. Perhaps, most importantly, whether blood concentrations are a true reflection of either testosterone or oestrone’s neurobiological effects is uncertain.

5. Conclusions

In conclusion, we have observed statistically significant associations between both low testosterone and oestrone and the likelihood of depression in older postmenopausal women. However, as the contribution to the likelihood of depression was extremely small our findings do not confirm these hormones make a direct biological contribution to mental health. Therefore, our findings should not be interpreted as supporting the use of sex hormones for the treatment of depression in older women.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.maturitas.2023.107822.