SHBG, Sex Hormones, and Inflammatory Markers in Older Women

In older women SHBG and estradiol are independently associated with inflammatory markers, suggesting a protective role of SHBG and the negative role of estradiol in cardiovascular disease.

Abstract

Context:

In premenopausal and older women, high testosterone and estradiol (E2) and low SHBG levels are associated with insulin resistance and diabetes, conditions characterized by low-grade inflammation.

Objective:

The aim of the study was to examine the relationship between SHBG, total testosterone, total E2, and inflammatory markers in older women.

Design and Patients:

We conducted a retrospective cross-sectional study of 433 women at least 65 yr old from the InCHIANTI Study, Italy, who were not on hormone replacement therapy or recently hospitalized and who had complete data on SHBG, testosterone, E2, C-reactive protein (CRP), IL-6, soluble IL-6 receptor (sIL-6r), and TNF-α. Relationships between sex hormones and inflammatory markers were examined by multivariate linear regression analyses adjusted for age, body mass index, smoking, insulin, physical activity, and chronic disease.

Results:

In fully adjusted analyses, SHBG was negatively associated with CRP (P = 0.007), IL-6 (P = 0.008), and sIL-6r (P = 0.02). In addition, testosterone was positively associated with CRP (P = 0.006), IL-6 (P = 0.001), and TNF-α (P = 0.0002). The negative relationship between testosterone and sIL-6r in an age-adjusted model (P = 0.02) was no longer significant in a fully adjusted model (P = 0.12). E2 was positively associated with CRP (P = 0.002) but not with IL-6 in fully adjusted models. In a final model including E2, testosterone, and SHBG, and all the confounders previously considered, SHBG (0.23 ± 0.08; P = 0.006) and E2 (0.21 ± 0.08; P = 0.007), but not testosterone (P = 0.21), were still significantly associated with CRP.

Conclusion:

In late postmenopausal women not on hormone replacement therapy, SHBG and E2 are, respectively, negative and positive, independent and significant correlates of a proinflammatory state.

Aging is characterized by a low-grade inflammatory status. Serum levels of inflammatory markers increase with age in both sexes, and the level of inflammatory markers is a strong and independent risk factor for frailty, disability, and cardiovascular events (1–3). It has been suggested that the mild proinflammatory state that is often detected in older persons is connected with the hormonal milieu occurring with aging (2). In women, the abrupt decline in circulating estradiol (E2) levels during menopause together with the smaller decline in testosterone levels leads to a dramatic increase in testosterone to estrogen ratio (4). Hormonal changes and their potential effects on inflammation may also explain the sharp increase in risk of cardiovascular disease occurring in women after menopause (5). The association between sex hormones and atherosclerosis and cardiovascular events occurring in older women has been widely studied. Most of those studies specifically targeted estrogens (5–10). We recently showed that higher E2 levels are associated with higher risk of all-cause mortality in late postmenopausal women independent of testosterone, supporting the timing hypothesis of hormone replacement therapy (HRT) in women (11, 12).

Estrogens and hyperandrogenemia have also been associated with an adverse coronary heart disease risk profile. This hormonal pattern, a hallmark of polycystic ovarian syndrome, supports the view that androgen excess may affect cardiovascular risk profile in women (13). In older women, testosterone levels have also been associated with insulin resistance, metabolic syndrome, and prevalent cardiovascular disease (14). Interestingly, the balance between testosterone and E2 is regulated by SHBG, which is a glycoprotein mainly synthesized by hepatocytes, for which the concentration tends to increase with age. In adult women, SHBG levels seem to mirror insulin and body mass index (BMI) trajectories (15), supporting its role as an independent predictor of metabolic syndrome and type 2 diabetes (16–18) conditions, both characterized by low-grade inflammatory status (19). Although several lines of research suggest a possible relationship between sex hormones and a proinflammatory state, especially in postmenopausal women, such a relationship has been only partially explained in large populations and with conflicting findings (9, 20–25).

Using data from the InCHIANTI Study, we hypothesized a negative relationship between SHBG and a positive relationship between sex hormones and inflammatory markers in late postmenopausal women.

Subjects and Methods

Study sample

The study population included 556 women who participated in the Invecchiare nel CHIANTI (InCHIANTI) study, randomly selected from all female residents 65 yr and older in the CHIANTI catchment area (Tuscany, Italy). Women selected for this study had complete data on SHBG, total testosterone, total E2, IL-6, soluble IL-6 receptor (sIL-6r), TNF-α, and C-reactive protein (CRP).

Exclusion criteria

After exclusion of 43 women who were using oral HRT (n = 33) or had been recently hospitalized (n = 10), 513 women from the original subset of 556 participants 65 yr and older were used in this analysis. Of these, 450 had natural menopause, and 63 had surgical menopause. None of the participants were taking androgens.

The Italian National Institute of Research and Care on Aging Institutional Review Board ratified the study protocol. Participants consented to participate and to have their blood samples analyzed for scientific purposes (26).

Biological samples

Blood samples were obtained from participants after a 12-h fast and after a 15-min rest. Aliquots of serum were stored at −80 C and were not thawed until analyzed.

Laboratory measures

Hormone assays

Total E2 was measured in the Laboratory of the University of Parma using ultrasensitive RIA (DSL-4800, Chematil, Angri, Italy) with a minimum detectable concentration (MDC) of 2.2 pg/ml and intra- and interassay coefficients of variation (CVs) of 8 and 10%, respectively. Total testosterone was assayed using commercial radioimmunological kits (Diagnostic Systems Laboratories, Webster, TX). The MDC was 0.08 nmol/liter; intraassay and interassay CVs for three different concentrations were 9.6, 8.1, and 7.8%, and 8.6, 9.1, and 8.4%, respectively. SHBG was measured using immunoradiometric assay (Diagnostic Products, Los Angeles, CA) with an MDC of 3.00 nmol/liter, and inter- and intraassay CV concentrations for three different concentrations were 3.7, 1.1, and 3.4% and 11.5, 10.3, and 8.7%, respectively. Dehydroepiandrosterone sulfate (DHEAS) was assayed using RIA commercial kits (Diagnostic Systems Laboratories). For DHEAS, the minimum detection limit was 1.7 μg/dl; intraassay CVs for three different concentrations (low, medium, high) ranged between 4.1 and 5.3%, and interassay CVs ranged between 4.6 and 7.0%. Plasma insulin level was determined using a double-antibody, solid-phase RIA (interassay and intraassay CVs, 3.1 and 0.3%, respectively; Sorin Biomedica, Milan, Italy). Cross-reactivity with human proinsulin was 0.3%.

Cytokine assays

Serum levels of IL-6, sIL-6r, and TNF-α were measured in duplicate by high-sensitivity ELISAs (BioSource International, Camarillo, CA). The lower detectable limit was 0.1 pg/ml for IL-6, 8.00 pg/ml for sIL-6r, and 0.09 pg/ml for TNF-α. For TNF-α, intraassay CVs ranged from 1.4 to 7.9%, whereas the interassay CV was less than 21%; for sIL-6r, intraassay and interassay CVs were less than 6.0% and less than 10.0%, respectively. For IL-6, the interassay and intraassay CVs were 7%. CRP was measured with a particle-enhanced- immunonephelometric assay and a Dade Behring BN11 nephelometer (Dade Behring, Inc., Deerfield, IL). The MDC was 0.16 mg/liter. Intraassay and interassay CVs ranged from 2.3 to 4.4% and from 2.1 to 5.7%, respectively.

Health behaviors

Smoking history was determined from self-report and dichotomized in the analysis as “current smoking” vs. “ever smoked” or “never smoked.” Education was assessed as years of schooling.

Physical activity in the year before the interview was coded as: 1) sedentary, completely inactive or light-intensity activity less than 1 h/wk; 2) light physical activity, light-intensity activity 2–4 h/wk; and 3) moderate to high physical activity, light activity at least 5 h/wk or more or moderate activity at least 1–2 h/wk.

Body size and composition

Weight and height were measured using standard techniques. BMI was calculated as weight (in kilograms) divided by the square of height (in meters). The waist circumference was measured at the midpoint between the lower rib margin and the iliac crest (normally umbilical level) and hip circumference at the level of the trochanters. Both were measured to the nearest 0.5 cm with an elastic plastic tape measure to calculate the waist-to-hip ratio. These measures are generally considered an acceptable alternative to more complex imaging-based assessment.

Diseases

Diseases were ascertained by an experienced clinician according to preestablished criteria that combine information from self-reported physician diagnoses, current pharmacological treatment, medical records, clinical examinations, and blood tests. Diseases included in the current analysis were coronary heart disease (including angina and myocardial infarction), congestive heart failure (CHF), stroke, diabetes, hypertension, Parkinson's disease, peripheral artery disease, asthma, cancer, and chronic obstructive pulmonary disease (COPD).

Statistical analysis

Variables with symmetric distribution were reported as means ± sd values. Variables with asymmetric distribution were summarized as medians and interquartile ranges and were log-transformed in regression analyses. Because of skewed distributions, log-transformed values for E2, testosterone, SHBG, IL-6, sIL-6r, TNF-α, CRP, and insulin were used in regression analyses and back-transformed for data presentation.

Multivariate regression analyses were used to test the relationship between SHBG, E2, testosterone, and inflammatory markers. The regression analysis was initially adjusted for age (model 1); BMI, smoking, physical activity, log (insulin), COPD, CHF, and stroke (model 2); and then finally for the other sex hormones and SHBG. All the analyses were performed by the SAS statistical package, version 9.1 (SAS Institute Inc., Cary, NC).

Results

Table 1 shows the main characteristics of the study population. Mean age of the studied population was 76.3 ± 7.8 yr, with an age range between 65.2 and 102 yr; 8.6% of the population were current smokers, 4.4% were affected by COPD, 2.4% were affected by stroke, and 2.9% were affected by CHF.

Table 1.

Demographic and clinical characteristics of the study population (n = 513)

Age (yr)	76.3 ± 7.8
BMI (kg/m2)	27.7 ± 4.7
Waist to hip ratio	0.89 (0.84–0.93)
Smoking (current)	44 (8.6)
Inflammatory markers
IL-6 (pg/ml)	1.38 (0.82–2.09)
sIL-6r (ng/liter)	94.47 (69.72–124.60)
TNF-α (pg/ml)	0.12 (0.00–1.45)
CRP (mg/liter)	2.71 (1.38–5.71)
SHBG (nmol/liter)	117.84 (85.84–172.94)
Total E2 (pg/ml)	5.61 (3.92–7.89)
Total testosterone (ng/ml)	0.61 (0.44–0.80)
Fasting insulin (mIU/liter)	9.88 (7.20–14.26)
Physical activity (%)
Sedentary/light	27.2
Moderate	70.2
High	2.56
COPD	23 (4.4)
Stroke	13 (2.4)
CHF	15 (2.9)

Data are expressed as median (interquartile range), except for age and BMI, expressed as means ± sd; smoking, COPD, stroke, and CHF, expressed as number (percentage); and physical activity, expressed as percentage.

SHBG and inflammatory markers

As shown in Table 2, in age-adjusted analyses log (SHBG) was negatively associated with log (CRP) (β ± se = −0.47 ± 0.08; P < 0.001), log (IL-6) (β ± se = −0.26 ± 0.06; P < 0.001), and log (sIL-6r) (P = 0.03). There was an inverse nonsignificant relationship between log (SHBG) and TNF-α (P = 0.09). After adjustment for multiple confounders, the strength of the association between log (SHBG) and log (CRP) was attenuated by almost 40% (β ± se = −0.28 ± 0.08) but remained statistically significant (P < 0.001). The strengths of the relationship between log (SHBG) and log (IL-6) (β ± se = −0.23 ± 0.07; P = 0.007) and between log (SHBG) and log (sIL-6r) (β ± se = −0.10 ± 0.04; P = 0.02) were unaffected by further adjustment for BMI, physical activity, smoking, fasting insulin, COPD, stroke, and CHF (Table 2).

Table 2.

Association between log (SHBG) and log (IL-6), log (sIL-6r), log (TNF-α), and log (CRP)

	Log (SHBG) (n = 513)
	β	se	P
Model 1
Log (CRP)	−0.47	0.08	<0.001
Log (IL-6)	−0.26	0.06	<0.0001
Log (sIL-6r)	−0.09	0.04	0.03
Log (TNF-α)	−0.29	0.18	0.09
Model 2
Log (CRP)	−0.28	0.08	0.0008
Log (IL-6)	−0.23	0.07	0.0007
Log (sIL-6r)	−0.10	0.04	0.02
Log (TNF-α)	−0.28	0.19	0.14

Each line refers to the results of a separate model adjusted for indicated covariates. Model 1 was adjusted for age. Model 2 was adjusted for age, BMI, physical activity, smoking, fasting insulin, COPD, stroke, and CHF.

Testosterone and inflammatory markers

Log (testosterone) was positively and significantly associated with log (CRP) (P = 0.001), log (IL-6) (P = 0.02), and log (TNF-α) (P = 0.0004) in age-adjusted analyses (Table 3). The strengths of the associations between testosterone and CRP, IL-6, and TNF-α remained substantially unchanged and still statistically significant in the multivariate analyses. Log (testosterone) was inversely and significantly associated with log (sIL-6r) (P = 0.02) in the age-adjusted model, but the association was no longer significant in the fully adjusted model (P = 0.12).

Table 3.

Association between log (total testosterone) and log (IL-6), log (sIL-6r), log (TNF-α), and log (CRP)

	Log (testosterone) (n = 513)
	β	se	P
Model 1
Log (CRP)	0.19	0.06	0.001
Log (IL-6)	0.09	0.05	0.02
Log (sIL-6r)	−0.07	0.03	0.02
Log (TNF-α)	0.41	0.11	0.0004
Model 2
Log (CRP)	0.16	0.06	0.006
Log (IL-6)	0.07	0.05	0.001
Log (sIL-6r)	−0.05	0.03	0.12
Log (TNF-α)	0.43	0.11	0.0002

Each line refers to the results of a separate model adjusted for indicated covariates. Model 1 was adjusted for age. Model 2 was adjusted for age, BMI, smoking, physical activity, fasting insulin, COPD, stroke, and CHF.

E2 and inflammatory markers

As summarized in Table 4, log (E2) was positively associated with log (CRP) in age-adjusted analysis (β ± se = 0.37 ± 0.08; P < 0.0001), and the strength of the association was attenuated but still statistically significant after further adjustment for confounders (β ± se = 0.24 ± 0.08; P = 0.001). Log (E2) was positively associated with log (IL-6) in the age-adjusted (P = 0.02) but not in the fully adjusted model (P = 0.10). There was no significant relationship between log (E2) and sIL-6r and TNF-α in both age-adjusted and fully adjusted analyses.

Table 4.

Relationship of total E2 and log (IL-6), log (sIL-6r), log (TNF-α), and log (CRP)

	Log (E2) (n = 513)
	β	se	P
Model 1
Log (CRP)	0.37	0.08	<0.0001
Log (IL-6)	0.13	0.06	0.02
Log (sIL-6r)	0.02	0.04	0.56
Log (TNF-α)	−0.03	0.15	0.84
Model 2
Log CRP	0.24	0.08	0.001
Log (IL-6)	0.10	0.06	0.10
Log (sIL-6r)	0.23	0.04	0.57
Log (TNF-α)	−0.02	0.16	0.87

Each line refers to the results of a separate model adjusted for indicated covariates. Model 1 was adjusted for age. Model 2 was adjusted for age, BMI, smoking, log (insulin), COPD, stroke, and CHF.

Testosterone, E2, SHBG, and CRP

Because SHBG, total testosterone, and E2 were all significantly associated with CRP, we created a final model including SHBG, total testosterone, total E2, and all the confounders previously considered. In the final model, shown in Table 5, log (SHBG) (β ± se = −0.23 ± 0.08; P = 0.006) was negatively associated, and log (E2) (β ± se = 0.21 ± 0.08; P = 0.007) was positively and significantly associated with log (CRP), with no change in the strength of the associations in this analysis compared with the analyses when SHBG, testosterone, and E2 were analyzed separately. Log (testosterone) was no longer significantly associated with CRP (P = 0.12). When we introduced DHEAS in the final model, the relationship between log (SHBG), log (E2), and log (CRP) was unaffected (data not shown), with P values of 0.002 and 0.009, respectively. Interestingly, also adjusting for surgical menopause (n = 63; 12.8%), the strength and the significance of the association were substantially unaffected (data not shown).

Table 5.

Association between CRP and sex hormones

	Log (CRP) (n = 513)
	β	se	P
Model fully adjusted
Log (SHBG)	−0.23	0.08	0.006
Log (Total testosterone)	0.09	0.06	0.12
Log (E2)	0.21	0.08	0.008
Age	0.01	0.007	0.006
COPD	0.34	0.10	0.0009
BMI	0.05	0.009	<0.0.0001

Values were adjusted for age, BMI, smoking, log (insulin), COPD, stroke, CHF, fasting insulin, and physical activity.

Discussion

In the cohort of late postmenopausal women not on HRT, SHBG is negatively associated and E2 is positively and independently associated with inflammatory markers. To the best of our knowledge, this is the first study aimed at investigating the relationship between SHBG, sex hormones, and inflammatory markers in late postmenopausal women.

Our study provides support for the recent suggestion that age-related changes in sex hormones contribute to the development of a proinflammatory state. Research previously conducted in this area has been scant. Most of these studies tested the association between sex hormones and CRP in early postmenopausal healthy women without providing any information on inflammatory cytokines and diseases and limiting their observation to the first years after menopause.

Our data are consistent with the results of Störk et al. (24); in a large, healthy, nonmedicated sample of 889 early postmenopausal women, estrogenic and androgenic sex hormones were independently associated with increased levels of CRP after adjustment for age, traditional cardiovascular risk factors, and markers of body composition. However, this analysis was limited to early postmenopausal women and evaluated only CRP (24). Similarly, Joffe et al. (9) found a negative and independent correlation between SHBG and CRP in a population of menopausal women, including HRT users and nonusers who subsequently developed clinical cardiovascular disease. However, no significant relationship was found between E2 and CRP in HRT nonusers in the age- and BMI-adjusted analysis. Surprisingly, a negative correlation between testosterone and CRP was also found, but this relationship was not present in cardiovascular disease-free participants, although 20% of the study population were smokers and half of those were HRT users. In addition, a positive association between CRP and free androgen index was found when the entire group of women not using hormonal therapy was analyzed (9). The discrepancy of these results could also be explained by the limited number of confounders used in that analysis.

Crandall et al. (23), in 623 postmenopausal women aged 45–64 yr, found that bioavailable testosterone was positively associated and SHBG was negatively associated with CRP. In contrast with our study, no association was found between E2 and CRP and between SHBG, testosterone, and IL-6 (23). However, the younger age of participants and the lower levels of IL-6 of that population can explain such a difference.

Our data are consistent with the results of a cross-sectional study by Maturana et al. (20) conducted in 53 recent postmenopausal women where CRP was negatively associated with SHBG and positively associated with bioavailable testosterone after adjusting for age, BMI, physical activity, alcohol consumption, and tobacco use. Interestingly, this association was independent of insulin levels, suggesting that higher testosterone in women may be part of a proatherogenic profile (20).

Finally, analyzing a subsample of the Atherosclerosis Risk in Communities study (n = 57), Folsom et al. (22) showed that, after adjusting for age, race, and case-control status, mean CRP was 2-fold lower across quartiles of SHBG. However, probably because of the small sample size, not all of these associations reached the statistical significance (22).

We tested in a final model including all the hormones the relationship between sex hormones and CRP. Interestingly, log (E2) was positively and significantly associated with log (CRP), whereas the relationship between testosterone and CRP was no longer significant. Because 80% of circulating E2 levels in postmenopausal women are derived from aromatization from testosterone, especially in adipose tissue, we hypothesized that testosterone could mediate the relationship between E2 and inflammatory markers. However, after adjusting for baseline testosterone levels, the strength of the relationship between total E2 levels and mortality was unchanged, suggesting that the proinflammatory role of E2 in older postmenopausal women is independent of its precursor testosterone. Although the production of androgens by the postmenopausal ovary remains controversial (27, 28), the history of surgical or natural menopause was accounted for, and the relationship between E2 level and CRP was unaffected. This is consistent with our previous observation that higher E2 levels are a significant and independent risk factor for all-cause mortality, independent of testosterone levels (11). Interestingly, the relationship between E2 and CRP was also independent of DHEAS, which is an important precursor of E2 levels.

Notably, the negative association between SHBG and inflammatory markers including CRP was the most robust finding of our study. Abundance of SHBG in postmenopausal women and the precision of measurement compared with that of total testosterone may account for the persistence of SHBG as an independent predictor of CRP in joint models.

Both SHBG and inflammatory markers are altered during insulin resistance (16, 18), which is a well-known modulator of testosterone and SHBG. However, after adjusting for fasting insulin, the relationship between SHBG and CRP did not change. We cannot exclude that SHBG might also exert direct effects on CRP and inflammatory markers. SHBG has been recently reported to act by a novel steroid-signaling system that is independent of the classic intracellular steroid receptor (29, 30).

Given the cross-sectional nature of the analysis, the alternative hypothesis that inflammation down-regulates SHBG and up-regulates E2 and testosterone cannot be neglected. It is known that CRP and inflammatory cytokines are stimulating factors of aromatase, the enzyme converting testosterone into E2 (31). The best source of this enzyme is the adipose tissue, and its serum levels tend to be strongly related to BMI. However, the introduction of BMI in the multivariate analysis did not affect the relationship between E2 and serum levels of inflammatory markers. Another possible interpretation of our results is that changes in sex hormone levels and SHBG may merely represent surrogates of underlying disease and systemic associated inflammation. This may also explain the peculiarity of our study vs. the previous studies performed in healthier and younger populations. We tried to account for the interference of acute diseases by excluding participants who were hospitalized in the last 10 d before the evaluation. We also adjusted the multivariate analysis for chronic diseases such as stroke, COPD, and CHF. However, the presence of residual confounding is always possible and cannot be definitively excluded.

The relationship between sex hormones and sIL-6r was tested here for the first time in an older female population. SHBG was inversely associated with sIL-6r in both age-adjusted and fully adjusted analysis. Concomitantly, we found no relationship between E2 and sIL-6r.

Similarly to what we found in the older men (32) and for the first time in the literature, we found a negative relationship between testosterone and sIL-6r, although in the fully adjusted model the relationship was no longer statistically significant. sIL-6r was the only marker negatively associated with testosterone. The receptor for IL-6 in the cell membranes consists of two subunits, a membrane-bound glycoprotein 130 and an extramembrane domain (IL-6r), which also exists in a soluble form (sIL-6r). sIL-6r enhances the IL-6 activity by making cells that have a glycoprotein 130 membrane domain sensitive to IL-6 signaling. Interestingly, testosterone was positively and not negatively associated with IL-6. The different association between testosterone and different components of the IL-6 system deserves further investigation.

Limitations

We acknowledge that our study has limitations. The cross-sectional nature of the study does not allow establishing any causality in this relationship. No information on estrone (the principal estrogen in postmenopausal women) and androstenedione was available for analysis.

The assay method used for testosterone is not considered the absolute gold standard, especially in postmenopausal women who have very low levels. We also acknowledge that there is considerable variability in the measurement of TNF-α. We cannot exclude that by using another methodology, different results could have been possible. However, the findings concerning the relationship between each hormone (including SHBG) and inflammatory markers were in the same direction as those observed for TNF-α.

Strengths of the study

The limitations are offset by important strengths. The consistent sample size was carefully selected by excluding factors interfering with inflammatory status and hormonal milieu. We excluded participants who were on HRT and androgen therapy and those with recent hospitalization in the last 10 d preceding the time visit. The mean age of the population was significantly older than that of all previous studies. The analysis included a pool of the most important inflammatory markers, giving the opportunity to investigate the relationship between sex hormones and inflammatory markers (i.e. sIL-6r) never investigated before. Finally, information on a wide range of potential confounders was available for inclusion in the statistical analysis.

Conclusion and perspective

In late menopausal women who were not taking HRT and androgens and were not recently hospitalized, SHBG and E2 were independently associated with inflammatory markers.

These data could explain the protective role of SHBG and the negative role of E2 in cardiovascular disease observed in older women. Longitudinal studies are needed to establish the clinical relevance of the association between SHBG, E2, and inflammatory markers in late postmenopausal women.