Association of serum androgens and coronary artery calcium scores in women

Abstract

OBJECTIVE:

To determine the association between serum androgens measured by high resolution liquid chromatography-mass spectrometry (LC-MS) and coronary artery calcium (CAC) scores.

DESIGN:

Cross sectional study

SETTING:

Academic institution

PATIENTS:

239 women, ages of 40–75, with CAC testing and complete CVD risk evaluation. Total testosterone, DHEA, and androstenedione were measured using high resolution LC-MS, while estradiol and sex hormone-binding globulin were measured using commercial assays.

INTERVENTIONS:

NONE

MAIN OUTCOME MEASURES:

Independent associations between CAC scores and sex steroids.

RESULTS:

Overall 164 subjects had a CAC score <10, 48 had a CAC score between 10–100 and 27 had a score >100. There were no differences in sex hormone levels between women with CAC scores >10 versus CAC scores ≤10. In multivariable models adjusting for age, BMI and LDL-C, a higher testosterone to estradiol ratio was associated with an elevated CAC score, with an unadjusted OR associated with 1SD change in log transformed testosterone to estradiol of 1.38 (1.01–1.89) p<0.046 and adjusted OR 1.02 (1.002–1.04), p<0.028. Total testosterone, DHEA, androstenedione, SHBG, and estradiol levels were not associated with increased CAC. CONCLUSIONS In the general population, there are mixed reports regarding the relationship between serum androgens and risk factors for CVD, and limited information on the relationship between androgens and subclinical atherosclerosis. Our study shows that increased androgens relative to estrogens may have a weak but independent association with subclinical atherosclerosis, as measured by CAC scores.

Introduction

The leading cause of mortality in women in the United States and worldwide is cardiovascular disease (CVD)(1,2). Despite many advancements in our understanding of CVD, its pathophysiology and clinical determinants, particularly in women, have not been fully recognized. This is in part due to previous studies primarily including men and to decreased utilization of guideline-directed medical therapies in women (3). Recently, an increase in the awareness of CVD and implementation of newer therapies in women has resulted in decreased mortality. However, the annual mortality rate stratified by age continues to be greater for women compared to men (4). Moreover, the etiology for the higher risk of mortality in younger women after myocardial infarction (MI) is unclear and likely multifactorial.

Some of the key differences in CVD manifestations between genders point to a role for endogenous sex hormones in the development of coronary artery disease (CAD) (5,6). Several population-based studies have suggested a relationship between estrogen and CAD in women based on a lower prevalence of MI in premenopausal women than men of similar age and narrowing of this difference after menopause (7). Further, women with premature ovarian insufficiency, a state characterized by low serum estrogen levels, are more likely to have early onset high risk cardiovascular profiles (8). The protective effects of estrogen may be related to its effects on the vascular system including release of nitric oxide leading to vasodilation, regulation of prostaglandin production, and inhibition of smooth muscle proliferation (5). The association between androgens and CVD risk factors has also been examined in large population studies as well as in select groups, such as women with polycystic ovary syndrome (PCOS) and postmenopausal women with diabetes. Reproductive-age women with PCOS, an endocrine disorder characterized by hyperandrogenism, are more likely to have obesity, dyslipidemia, diabetes, and elevated serum biomarkers for CVD compared to women without PCOS (9,10). This association is more evident in the hyperandrogenic PCOS phenotype, suggesting an association between serum androgens and CVD risk (11). However, studies examining the risk of cardiovascular (CV) events in this population show mixed results (12). In postmenopausal women, serum testosterone levels positively associate with type 2 diabetes, suggesting that endogenous sex hormones may differentially modulate glycemic control (13). In the multiethnic Study of Women’s Health Across the Nation (SWAN) free androgen index correlated with biomarkers of CV risk including lipids, glucose, and inflammation(14). In a subsequent publication analyzing the same population, women with the highest tertile of testosterone levels had increased odds of metabolic syndrome adjusted for age, body mass index (BMI), smoking, and ethnicity (15). However, other studies show no significant associations between serum androgens and CV risk markers such as C-reactive protein (CRP) and lipids (16).

Fewer studies have examined the association between androgens and markers of subclinical atherosclerosis. These studies show variable associations between androgens and carotid intima-medial thickness (CIMT) (17,18,19,20). Even fewer studies have examined the association between coronary artery calcium (CAC) scores, a better predictor of CVD (21), and serum androgen levels. Two large population based studies (MESA and CARDIA) did not demonstrate a significant association between serum androgen levels and CAC scores (22,23). The use of immunoassays versus liquid chromatography-mass spectrometry (LC-MS) to measure low concentrations of androgens, as seen in post-menopausal women, may have been a limiting factor in these studies (24). The aim of the present study was to determine the association between sex hormones, as measured by high resolution LC-MS, and CAC scores in a well-characterized group of community-based women.

Subjects and Methods

Study Subjects

Participants were selected from IRB-approved community-based studies that measured CAC scores in subjects with varying numbers of metabolic syndrome criteria without clinical evidence of CVD conducted at the University of Pennsylvania (25). Subjects had previously provided consent for use of de-identified data and stored serum in future studies. IRB approval was not required for this study as it was determined non-human subjects research. Subjects were included if they were female, had a documented CAC score, had sufficient residual serum sample for sex hormone assays, and complete data on risk factors for CVD: age at visit, BMI, waist circumference, blood pressure, serum glucose and lipid profile, and data on history of hypertension, diabetes, smoking and hormone replacement therapy (HRT) status. Subjects taking HRT were excluded from this analysis.

Coronary artery calcium (CAC) scores

Global CAC scores were determined with customized software used according to the method of Agatston (26) from 40 continuous 3-mm-thick computed tomograms collected on an electron beam tomography (EBT) scanner (Imatron). Reproducibility studies with the use of these techniques suggest that artifact in EBT estimation of CAC accounts for a small proportion of CAC variability.

Sample preparation and Girard P derivatization for androgen measurements.

Blood was drawn in the morning after a 12-hour fast and serum was frozen in −70 C. Total testosterone (TT), dehydroepiandrosterone (DHEA), and androstenedione (AD) were measured using liquid chromatography high resolution mass spectrometry (LC-MS/HRMS) (26). All plasma samples were centrifuged at 5300g at 4° C for 10 min. A 50μL aliquot was transferred to a glass tube, and 20 μL of internal standards mixture solution was spiked followed by the addition of 0.1 mL water, 0.1 mL 0.5% L-ascorbic acid, and 0.2 mL sodium acetate buffer (200mM, pH 5.0). Stable isotope-labeled androgens ([¹³C₃]-T, [¹³C₃]-AD, and [²H₅]-DHEA) were obtained from Cambridge Isotope Laboratories (Andover, MA). Samples underwent liquid-liquid extraction (LLE) with MTBE by vortex-mixing. The top organic layer was evaporated to dryness under nitrogen, then re-suspended in 200 μL of 10% acetic acid in methanol for derivatization as previously described²⁷. To evaluate linearity of standard curves, calibration standards were prepared at concentrations of 0, 15.6, 31.2, 62.5, 125, 250, 500, and 1000pg/mL of each androgen in charcoal stripped human serum, except for DHEA, which was prepared at 5 times higher concentration. 20 μL of internal standard mixture solution was added to each sample. Calibration curves were generated by plotting the area ratios of the analyte to internal standard peak using linear regression with 1/x weighting. The lower limit of quantification was defined as the lowest calibration level that could be fitted to the calibration curve with a residual of less than 10% and peak area ratio deviating less than 25%. Satisfactory linearity was observed over 100-fold concentration with linear regression correlation coefficients all better than 0.99. Plasma concentrations of all androgens were calculated using Xcalibur software (version 2.6) from Thermo Fisher Scientific.

LC-MS/HRMS analysis of Girard P derivatized androgens.

LC-MS/HRMS analysis on the Girard P derivatives of TT, AD, and DHEA were conducted on an Ultimate 3000 quaternary UPLC equipped with a refrigerated autosampler (6° C) and a column heater (60° C) coupled to a Thermo Scientific™ Q Ex active™ plus HRMS. LC separations were conducted on a Phenomenex Kinetex biphenyl column (2.6 μm, 100 A, 100 × 2.1 mm). A multistep gradient at 0.2 mL/min flow with solvent A (water 1% acetic acid) and solvent B (acetonitrile 1% acetic acid) was as follows: 20% B from 0 to 1 minute, increasing to 25% B from 1 to 5 minutes, increasing to 100% B from 5 to 8 minutes then holding 100% B until 12 minutes, then the column was returned to starting conditions and re-equilibrated at 20% B from 13 to 17 minutes. The mass spectrometer was operated in positive ion mode alternating between full scan (200–800 m/z) at a resolution of 70,000 and parallel reaction monitoring at 17,500 resolutions with a precursor isolation window of 0.7 m/z. Molecular (MH⁺) precursor and product ions (m/z) were as follows: TT and DHEA (422.2802 to 343.2380), AD (420.2646 to 341.2224), [¹³C₃]-TT (425.2903 to 346.2481), [¹³C₃]-AD (423.2746 to 344.2324), and [²H₅]-DHEA (428.3179 to 349.2757).

The samples were run as a single batch for estradiol using a chemiluminescent competitive immunoassay and sex hormone-binding globin (SHBG) using a chemiluminescent immunometric (sandwich) assay (Siemens Medical Solutions USA, Inc., Norwood MA, USA). The manufacturer reports a CV < 10% for both assays. The free androgen index (FAI) was calculated FAI = TT in nmol/L / SHBG in nmol/L X 100.

Statistical Analysis

Demographic characteristics and outcome variables were compared between women with a clinically significant CAC scores >10 versus CAC scores ≤10 using Wilcoxon rank sum tests and Pearson Chi-squared and Fisher exact tests for categorical variables, as appropriate. Sex hormone concentrations were natural log transformed to decrease skewness of high values. Multivariable logistic regression models were used to determine the relationship between sex hormones and CAC score >10, adjusting for confounders as appropriate. Age, BMI, smoking, low-density lipoprotein cholesterol (LDL-C), history of diabetes and hypertension were considered as potential confounders in this association. Each covariate was fit one at a time, and variables were entered into the model if the associated P-value was < 0.2 in adjusted analysis. A backwards model selection strategy was used to evaluate confounding, and remove non-significant variables. Our final model adjusted only for participant age, BMI and LDL-cholesterol. Odds ratios were reported per standard deviation change in order to place them on the same scale. A sub-analysis was performed including all women with estradiol levels <50pg/ml, indicating that assessments were made at the early follicular phase of the menstrual cycle for subjects that were pre-menopausal. A second sub-analysis was performed examining the relationship between highest quartile values of sex hormone concentrations and a CAC score >10. A two-sided P-value of < 0.05 was considered statistically significant. Statistical analysis was performed using Stata version 13.0 (StataCorp, College Station, TX).

Results

Demographic characteristics of 239 study participants ages 40–75years are outlined in Table 1. There were no significant demographic differences between women with clinically significant CAC score >10 and women with CAC score ≤10. Most of the subjects were overweight and white. A larger proportion of women with CAC score >10 reported a history of smoking (44% versus 35%), however, this difference was not statistically significant (P = 0.18). The mean CAC score was 37.8±126.4, 164 subjects had a CAC score <10, 48 had a CAC score 10–100 and 27 had a score >100.

Table 1:

Baseline characteristics of the study cohort.

	CAC score ≤10 (n= 167)	CAC score >10 (n=72)	P valuea
Mean age in years (SD)	55.8 (5.6)	56.5 (7.2)	0.22
Mean BMI in kg/m2 (SD)	29.8 (6.5)	28.1 (5.9)	0.11
Race n(%)			0.66
White	151 (90.4%)	66 (91.7%)
Black	11 (6.6%)	5 (6.9%)
White Hispanic	2 (1.2%)	0 (0%)
Asian	2 (1.2%)	0 (0%)
Mixed	0 (0%)	1 (1.4%)
Unknown	1 (0.6%)	0 (0%)
Ever smoker n (%)	59 (35.3%)	32 (44.4%)	0.18
H/o Diabetes mellitus n (%)	4 (2.5%)	1 (1.4%)	0.5
H/o Hypertension n(%)	52 (32.9%)	19 (26.4%)	0.32
Metabolic syndrome n (%)	33 (26%)	12 (23.5%)	0.73

SD, standard deviation.

^aP-values calculated using Wilcoxon Rank sum and Pearson chi-squared test of Association

Traditional CVD risk factors were compared between groups; total cholesterol and LDL-C levels were higher in women with CAC scores >10 compared to women with CAC scores ≤10 (p < 0.05, Table 2). There were no significant differences in mean levels of sex hormones between the two groups (Table 2). adjusted for age and BMI, DHEA correlated with LDL-C (β coefficient 6.29 [0.53–12.04], P < 0.04), androstenedione correlated with triglycerides (TG) (β coefficient - 0.70 [−1.35 - −0.06], P < 0.04), and high-density lipoprotein cholesterol (HDL-C) (β coefficient 2.48 [0.20 – 4.76], P < 0.04), while SHBG correlated with total cholesterol (β coefficient 0.09 [0.003–0.19] P <0.04, TG (β coefficient −0.08 [−0.14- −0.01]) and HDL-C (β coefficient 0.5 [0.28 – 0.72]).

Table 2:

Differences in serum sex hormones and cardiometabolic risk factors.

	CAC score ≤10	CAC score>10	P-valuea
Mean Sex Steroid (SD)
Testosterone pg/mL	165.3 (292.1)	278.6 (582.5)	0.4
DHEA pg/mL	966.7 (1450.6)	1159.5 (1618)	0.97
Androstenedione pg/mL	483.4 (286)	452.5 (261.4)	0.45
SHBG nmol/L	40.2 (22.9)	53.7 (39.0)	0.06
Estradiol pg/mL	42.1 (52.4)	45.6 (84.0)	0.78
Testosterone/Estradiol Ratio	6.1 (9.5)	11.6 (25)	0.36
Free Androgen Index	2.69 (4.85)	10.08 (53.11)	0.11
CVD risk factors Mean (SD)
Glucose mg/dL	70.2 (16.7)	71.6 (10.9)	0.68
Systolic blood pressure mmHg	125.5 (17.4)	121.9 (16.2)	0.13
Diastolic blood pressure mmHg	75.8 (9.9)	73.3 (9.4)	0.07
Total cholesterol	204.5 (34.0)	221.0 (43.1)	0.01
Low-density lipoprotein mg/dl	120.1 (29.1)	134.2 (39.3)	0.02
High-density lipoprotein mg/dl	60.3 (16.2)	61.6 (16.8)	0.55
Triglycerides mg/dl	104.5 (54.7)	111.2 (53.8)	0.24

SD, standard deviation.

^aAll p-values calculated using Wilcoxon Ranksum tests

In univariate models, a higher testosterone to estradiol ratio was associated with an elevated CAC score, with an OR associated with 1SD change in log transformed testosterone to estradiol of 1.38 [CI] 1.01–1.89, p<0.048 (Table 3). In multivariate models including age, BMI and LDL-C this association was weaker 1.02 (1.002–1.04, p<0.028). Total testosterone levels were not associated with increased CAC scores on adjusted analysis (OR 1.35 (0.91–2.0), p=0.18). Estradiol, DHEA, androstenedione, SHBG, and free-androgen index were also not associated with CAC scores. We performed a sub-analysis including subjects with an estradiol level <50pg/ml and found similar associations between 1SD change in steroid hormone levels and elevated CAC scores (Table 4).

Table 3:

Adjusted* Odds Ratio per 1SD change in risk factor for associations with Coronary Artery Calcium Score of Greater than 10

	Unadjusted OR	P value	Adjusted ORa	P value	Adjusted ORb	P value
Testosterone	1.21 (0.92–1.6)	0.18	1.23 (0.92–1.63)	0.16	1.35 (0.91–2.0)	0.14
DHEA	1.03 (0.78–1.36)	0.83	1.06 (0.79–1.42)	0.69	0.99 (0.73–1.34)	0.93
Androstenedione	0.88 (0.67–1.16)	0.36	0.88 (0.66–1.18)	0.4	0.82 (0.50–1.34)	0.43
SHBG	1.22 (0.9–1.66)	0.21	1.17 (0.84–1.62)	0.35	1.12 (0.80–1.58)	0.50
E2	0.97 (0.73–1.29)	0.82	1.01 (0.75–1.37)	0.94	1.07 (0.69–1.65)	0.78
Testosterone/estradiol ratio	1.38 (1.01–1.89)	0.046	1.4 (1.02–1.9)	0.035	1.02 (1.002–1.04)	0.028
FAI	2.06 (0.62–6.88)	0.24	2.21 (0.65–7.51)	0.21	1.03 (0.99–1.07)	0.19

^aMultivariable model adjusted for age and BMI

^bMultivariable model adjusted for age, BMI and LDL

Table 4:

Adjusted Odds Ratio per 1SD change in risk factor for associations with CAC Score >10 in women with E2<50pg/ml

	Unadjusted OR	P value	Adjusted ORa	P value	Adjusted ORb	P value
Testosterone	1.27 (0.95–1.69)	0.10	1.40 (0.94–2.10)	0.097	1.41 (0.94–2.13)	0.098
DHEA	0.99 (0.73–1.35)	0.96	0.99 (0.72–1.36)	0.96	0.88 (0.63–1.23)	0.46
Androstenedione	0.94 (0.70–1.28)	0.71	0.91 (0.55–1.51)	0.72	0.90 (0.53–1.54)	0.71
SHBG	1.28 (0.91–1.80)	0.16	1.22 (0.84–1.77)	0.29	1.14 (0.78–1.65)	0.50
E2	1.52 (0.50–4.62)	0.46	2.43 (0.49–12.1)	0.28	2.46 (0.45–13.3)	0.30
Testosterone/ estradiol ratio	1.37 (0.99–1.89)	0.054	1.02 (1.001–1.04)	0.037	1.02 (1.002–1.04)	0.028
FAI	2.21 (0.64–7.69)	0.21	1.03 (0.99–1.08)	0.19	1.03 (0.99–1.08)	0.15

^aMultivariable model adjusted for age and BMI

^bMultivariable model adjusted for age, BMI and LDL

Next, a sub-analysis was performed to evaluate the relationship between the highest quartile values of sex hormone concentrations and a CAC score > 10. After adjustment for age, BMI, and LDL-C there were no significant associations between the highest quartile of steroid hormone values and an elevated CAC score.

Discussion

In a large cohort of well-characterized women, we demonstrated a weak but independent association between sex-hormones, namely testosterone to estradiol ratio, and subclinical atherosclerosis, as measured by CAC scores. Our findings support the notion that increased androgens relative to estradiol concentrations, in peri- and postmenopausal women may be associated with an increased risk of CVD.

Several studies have previously shown a positive association between serum androgen levels and serum markers of CVD risk in pre- and postmenopausal women (14). Similarly, we found significant correlations between some androgens and lipids in our cohort. However, fewer studies have examined the independent association of sex hormones with markers of subclinical atherosclerosis as measured by CAC scores (22,23). In the prospective CARDIA study, primarily including premenopausal women (37–52 years), no association was reported between CAC scores and total and free testosterone levels and the highest quartile of testosterone at year 20 of follow-up.(23) In the Multi-Ethnic Atherosclerosis Study (MESA), which included older women (mean age 60 years), no association was reported between serum androgens (total testosterone and DHEA) and CAC scores.(22) In a sub-analysis of women with only positive CAC scores in the MESA study, lower free testosterone levels were associated with significantly greater CAC scores. Neither of these studies examined the relationship between testosterone to estradiol ratio and CAC scores and both used commercially available immune assays to measure testosterone levels, which were typically at the lower end of the normal range in these cohorts.

Contrary to the above two studies, we noted a weak positive association between testosterone to estradiol ratio and CAC scores. Across the menopausal transition studies have demonstrated a substantial decrease in circulating estradiol levels, whereas testosterone levels do not change significantly (28,29). This suggests that the testosterone to estradiol ratio increases after menopause and, consistent with the results of our study, higher testosterone to estradiol ratios may contribute to increased risk of atherosclerosis in the post-menopausal period. The concept of a synergistic effect between estradiol and testosterone with regards to CVD has been suggested previously (30). The mechanisms by which estrogens and androgens alter CVD risk vary. In ovariectomized primates, coronary artery atherosclerosis, as measured by plaque size, was increased by exogenous long-term androgen administration; this effect was abrogated by administration of estrogens (31). In another study including postmenopausal women, a lower testosterone to estradiol ratio was associated with a favorable lipid profile (32). An in vitro study showed that an optimal testosterone to estradiol ratio (1:5) protects against CRP-induced apoptosis in human umbilical vein endothelial cells (33). These hormones injected in a similar ratio in mice with early-stage atherosclerosis were found to be more cardioprotective (assessed by reduced lipid lesions, foam cells, and endothelial injury) than estrogen or testosterone administration alone. The effect of androgens in the pathogenesis of atherosclerosis in women may therefore be modulated by estradiol.

The impact of hyperandrogenism in young women is best illustrated in PCOS, a disorder characterized by androgen excess and high prevalence of cardiometabolic risk factors independent of obesity. An association between free androgen index and several CVD risk factors have been described independent of age, ethnicity and smoking (34). Also, young premenopausal women with PCOS have reduced flow-mediated dilatation independent of BMI (35) and increased prevalence of high CIMT compared to controls (36). Further, women with PCOS have an increased prevalence of higher CAC scores compared to controls (37, 38, 39). In a sub-analysis of the CARDIA study, women with PCOS (defined as irregular menses and hyperandrogenism) had increased odds for having CAC (OR 2.69 95% CI [1.37–5.25])⁴⁰. It is not clear that these observations result in increased CV events and mortality in PCOS. Larger prospective studies consisting of well-phenotyped women with hyperandrogenism prior to menopause are needed to clarify the clinical significance of these findings.

In women in the general population, androgens have been shown to have mixed associations with different CVD outcomes. In some studies, higher total testosterone levels associated with an increased odds of CHD (41,42), others found that lower testosterone levels associated with all-cause mortality (43, 44) while another study did not demonstrate significant associations (45). A recent study analyzed the MESA population and reported an increased risk of CVD and CHD associated with increased testosterone to estradiol ratio (46). Most of these studies used commercially available immunoassays to measure endogenous steroids. One of the first studies to use LCMS to measure endogenous steroids in elderly women (47) did not report a significant correlation between estradiol to testosterone ratio and prevalent CVD. However, only 8 of 72 women reported CVD in this study. Another study using LCMS to measure androgens did not find significant associations between serum androgens and CIMT (48).

The strengths of our study include subjects recruited from the community, similar prevalence of well-characterized traditional CVD risk factors in both groups, exclusion of women on HRT, measurement of androgens by high resolution LC-MS (a gold standard for androgen measurements) (49), and assessment of subclinical atherosclerosis by measurement of CAC. Limitations include a relatively small sample size, lack of data on menopausal status, use of a marker for subclinical atherosclerosis, and lack of longitudinal data to more fully elucidate the temporal relationship between CVD and sex hormones in women. Longitudinal studies that measure sex hormones in the premenopaual period and then follow women for 2–3 decades to capture sufficient CV events are difficult to conduct. As noted above, the few studies that have examined CV events have looked at different CV adverse outcomes (i.e. CHD, all-cause mortality) making adequate comparisons between these studies difficult. Therefore, another approach is to examine surrogate outcomes such as CAC scores (as reported in CARDIA and MESA studies) and measure sex hormones in a slightly younger cohort.

Based on highly sensitive LC-MS measurements of androgens and valid measures of subclinical atherosclerosis, our study found weak positive associations between the testosterone to estradiol ratio, and calcification of the coronary arteries. Understanding the interplay between androgens, estrogen and CVD in women is critical to counselling women especially those with PCOS or those prescribed androgen therapy. Due to the limitations in the currently published studies, longitudinal studies are needed to examine this association both prior to menopause and after menopause.