Associations of Endogenous Sex Hormones with Carotid Plaque Burden and Characteristics in Midlife Women

Abstract

Context

Endogenous sex hormones may be involved in the pathogenesis of cardiovascular disease (CVD) in women. Carotid plaque characteristics, such as echogenicity, an ultrasound measure that reflects plaque composition, may identify unstable plaques that are more likely to rupture, precipitating a CVD event. However, few studies have considered sex steroids in relation to carotid plaque and its characteristics.

Objective

To evaluate estrone (E1), estradiol (E2), testosterone (T), sex hormone binding globulin (SHBG), and free T (FT) in relation to carotid plaque in women.

Design, Setting, and Participants

In MsHeart, a cross-sectional study of 304 women aged 40 to 60 years, participants underwent a carotid artery ultrasound assessment. The current analysis included MsHeart participants with carotid plaque (n = 141, 46%). E1, E2, and T were assayed using liquid chromatography–tandem mass spectrometry; FT was estimated using ensemble allostery models. Regression models were adjusted for sociodemographic characteristics and CVD risk factors.

Main Outcomes

Carotid plaque burden (number of plaques, total plaque area [TPA]) and characteristics (calcification, echogenicity) were determined using semi-automated software.

Results

SHBG was inversely related to TPA (odds ratio [OR] 0.39; 95% confidence interval [CI] 0.21, 0.74; multivariable) and higher FTs were associated with greater TPA (OR 2.89; 95% CI 1.31, 6.37; multivariable). Higher E1 was related to echogenicity (OR 2.31; 95% CI 1.26, 4.33; multivariable), characteristic of more stable plaque.

Conclusions

SHBG and FT are related to TPA while E1 is related to plaque echogenicity, suggesting these hormones have different roles in the development of carotid plaque. Our findings highlight the importance of sex hormones in the development of carotid plaque in midlife women.

Cardiovascular disease (CVD) is the leading cause of death in women across all racial/ethnic groups (1). Female sex hormones, particularly estrogens, may help protect women from CVD until their postmenopausal years (2). Notably, the menopause transition is accompanied by variability of sex hormone secretion, which may adversely affect CVD risk factors (ie, body composition, lipoprotein particle size) (3, 4). It has been postulated that the cardioprotective effects of estrogen may explain why women develop CVD a decade later than men, largely during the postmenopausal years (5). Estrogens have been shown to have metabolic actions on the liver, pancreas, and adipose tissue (6). Studies have shown that estradiol (E2), the major form of ovarian estrogen, plays an active role in endothelial function, including maintenance of vascular tone (7). However, less is known about estrone (E1), produced in peripheral tissues, particularly adipose tissue (8). While there is a large body of literature regarding the impact of exogenous hormones on midlife women’s cardiovascular health (9), there is limited research on a range of endogenous sex hormones and their impact on women’s CVD risk.

In addition to estrogen, recent studies suggest that sex hormone binding globulin (SHBG) and testosterone (T) may be associated with future risk of CVD (10–12). Among postmenopausal women, low levels of SHBG have been associated with greater carotid intima–media thickness (13) and incident carotid plaque (14), although adiposity may in part account for these associations. T is an essential precursor for estradiol with a high affinity for SHBG. T exerts physiological effects throughout the body including increasing renin activity (15, 16) and vasoconstriction (17). The evidence on T and atherosclerosis is conflicting (18, 19). While higher T has been associated with more favorable subclinical CVD risk markers in older men and women (18, 20), studies among healthy postmenopausal women have found that higher levels of T are associated with elevated risk of CVD (13, 19).

A limitation of many existing studies is that they have used standard direct immunoassays for the measurement of sex hormones, which may have limited accuracy and utility for postmenopausal women (21, 22). Liquid chromatography–tandem mass spectrometry (LC-MS/MS) has emerged as an accurate and reproducible measurement of serum sex steroids that can quantify the low levels of these hormones observed in postmenopausal women (21, 23, 24). Few studies have used LC-MS/MS when examining associations between endogenous sex hormones and women’s cardiovascular health (25).

Noninvasive measures of subclinical atherosclerosis, including carotid intima–media thickness and plaque assessed by ultrasonography, are inexpensive and reliable predictors of coronary heart disease (CHD) and future CVD events (26, 27). Emerging evidence suggests that carotid plaque may be a better marker of CVD risk than intima–media thickness (28, 29). Several large cohort studies have shown that the presence of carotid plaque is predictive of cerebral infarction and myocardial infarction in asymptomatic individuals (30, 31), particularly women (32). It has been suggested that total carotid plaque area and morphology more accurately predict CVD events than plaque presence alone (33, 34). Carotid plaque characteristics, such as plaque echogenicity, an ultrasound measure that reflects plaque content, may identify unstable plaques that are more likely to rupture, precipitating a major cardiovascular event (35). Higher echogenicity of carotid artery plaque is consistent with a greater content of fibrous tissue and calcification (36,37), while lower echogenicity of carotid plaque has been associated with increased lipid content and necrotic core presence. Measures of carotid plaque burden and characteristics (eg, number of plaques, total plaque area, calcification, and echogenicity) have been proposed to evaluate response to therapy and better understand the mechanisms underlying the progression of atherosclerosis (38,39). Thus, measures of carotid plaque burden and characteristics may be useful to understand the development of CVD in midlife women, as well as potential preventive strategies.

In a prior analysis we tested LC-MS/MS-assessed E1, E2, and T as well as free T (FT) and SHBG in relation to several subclinical CVD measures (25). While this analysis found that endogenous E1 levels were related to endothelial function and E2 to vascular remodeling, higher SHBG and lower FT were associated with carotid plaque presence. Among late postmenopausal women in the Rotterdam Study, higher E2 and lower total T were associated with measures of carotid plaque composition (40), but these associations have not been examined in healthy peri- and early postmenopausal (<10 years since final menstrual period). A more detailed ultrasound characterization of plaques may provide higher diagnostic accuracy for the prediction of future CVD events (41). Thus, in the current study, we evaluated LC-MS/MS-assessed endogenous sex hormones (E1, E2, T) as well as FT and SHBG in relation to measures of carotid plaque burden and characteristics in a community sample of nonsmoking, CVD-free midlife women. We hypothesized that higher E1 and T levels would be related to a more adverse carotid plaque profile (ie, greater number of plaques and total plaque area, lower plaque calcification and echogenicity), while E2, SHBG, and FT levels would be inversely related to these measures of plaque.

Patients and Methods

Design and study sample

The MsHeart Study was a cross-sectional study of 304 perimenopausal and postmenopausal women aged 40 to 60 years who were nonsmoking and free of clinical CVD. The study was designed to examine associations between hot flashes and CVD risk. By design, half of the women reported hot flashes and half reported no hot flashes. Details of the study have been previously published (42). Briefly, women were recruited from the community using advertisements, mailings, and message boards. Exclusion criteria included hysterectomy and/or bilateral oophorectomy; history of heart disease, stroke, arrhythmia, ovarian/gynecological cancer, pheochromocytoma, pancreatic tumor, kidney failure, seizures, Parkinson’s disease, Raynaud’s phenomenon; current pregnancy; or use of the following medications in the past 3 months: oral/transdermal estrogen or progesterone, selective estrogen receptor modulators, selective serotonin reuptake inhibitors, serotonin norepinephrine reuptake inhibitors, gabapentin, insulin, beta-blockers, calcium channel blockers, alpha-2 adrenergic agonists, or other antiarrhythmic agents. Women who had undergone endometrial ablation, endarterectomy, or lymph node removal or who were undergoing dialysis or chemotherapy were also excluded. Since the focus of the current analysis is the association between sex hormones and carotid plaque characteristics, women without carotid plaque (n = 161; 53% of original cohort) or with missing/poor quality plaque data (n = 2) were excluded, resulting in 141 women in the current sample.

Participants underwent physical measurements, psychosocial and medical history assessments, ambulatory hot flash monitoring, phlebotomy for assessment of lipids, glucose, insulin, and sex hormone concentrations, and a carotid artery ultrasound. Procedures were approved by the University of Pittsburgh Institutional Review Board. Participants provided written, informed consent.

Sex hormones

Blood draw was performed after a 12-hour overnight fast. Endogenous sex hormones (E1, E2, T) were measured from serum samples using LC-MS/MS, the gold standard for steroid sex hormones. E1 and E2 were assessed at the University of Pittsburgh’s Small Molecule Biomarker Core and T was assessed at the Brigham and Women’s Hospital Research Assay Core Laboratory, certified by the Center for Disease Control’s Hormone Assay Standardization Program for Testosterone (43). SHBG was measured at the University of Pittsburgh Chemistry and Nutrition Laboratory via the enzyme-linked immunosorbent assay (ALPCO, Salem, NH). As described in an earlier report (40), the lower limit of detection and of quantification for estrogens (E1 and E2) were 1.0 pg/mL and 2.5 pg/mL, respectively. Intraday statistics showed errors below 8.1% and interday statistics showed errors below 5.0%. For T, the lower limit of quantitation was 1.0 ng/dL, with intra-assay variation of <4% and interassay variation of <5%. Sensitivity of SHBG was 0.1 nmol/L, with intra-assay and interassay coefficients of variation of 1.3% and 4.9%, respectively. FT was calculated from total T and SHBG, using the ensemble allostery model as previously described (44).

Carotid ultrasound

Bilateral carotid images were obtained by trained and certified vascular sonographers at the University of Pittsburgh’s Ultrasound Research Laboratory using a Sonoline Antares (Siemens, Malvern, PA) high-resolution duplex scanner equipped with a VF10-5 transducer. Consistent with the Mannheim Consensus Statement (45), a carotid artery plaque was defined as a focal area protruding into the vessel lumen that was at least 50% thicker than the adjacent intima–media thickness. Sonographers evaluated the presence and extent of plaque in each of five segments of the left and right carotid artery (distal and proximal CCA, carotid bulb, and proximal internal and external carotid arteries). Plaque presence was defined as having plaque in any arterial segment (distal and proximal CCA, carotid bulb, and proximal internal and external carotid arteries). Total number of plaques was defined as the sum of all plaques across the 5 segments. Total plaque area (TPA) and gray-scale median (GSM) were determined by a Ultrasound Research Laboratory-certified reader using the semi-automated Carotid Analyzer software from the Vascular Research Tools 5 Suite (Medical Imaging Applications LLS, Coralville, IA) (46). TPA (mm²) was defined as the sum of plaque area across all segments. GSM, a continuous measure of plaque echogenicity, was calculated by the software using the frequency distribution of the gray scale pixels within the plaque, which was normalized by the reader first selecting a region of blood from the lumen (black) and a region of adventitia (white) on the image. The normalized gray scale ranged from 0 (blood) to 256 (adventitia). TPA and GSM were categorized into 3 groups based on tertile of the distribution. TPA ranged from low (tertile 1) to severe (tertile 3). Low GSM (tertile 1) was considered a more unstable echolucent plaque with thrombus and a lipid-rich core, while high GSM (tertile 3) was considered highly echogenic plaque, more consistent with fibrous tissue and calcification (36). Within-reader intraclass correlation coefficients for TPA and GSM were 0.95 and 0.99, respectively (46). Carotid plaque calcification (any versus none) was assessed subjectively based on whether the plaque consisted of highly echogenic areas and acoustic shadowing.

Covariates

Sociodemographic characteristics including race/ethnicity, education, and financial strain (ie, difficulty paying for basics) were assessed by self-report at baseline. Height and weight were obtained using a fixed stadiometer and a calibrated balance beam scale, and body mass index (BMI kg/m²) calculated. Seated blood pressure was measured via a Dinamap device (GE Medical Systems Information Technologies Inc., Milwaukee, WI) after a 10-minute rest (blood pressure was the average of the 2nd and 3rd of 3 seated measurements). Medical, reproductive, and psychosocial history was assessed by standard instruments. Menopause status was determined from reported menstrual bleeding patterns and classified according to STRAW +10 criteria (47). Medication use (eg, antihypertensives, lipid-lowering medications, glucose control) was assessed by self-report. Leisure time physical activity was evaluated using the International Physical Activity Questionnaire (48). Glucose, total cholesterol, high-density lipoprotein cholesterol (HDL-C), and triglycerides were measured enzymatically (Vital Diagnostics, Lincoln, RI). Low-density lipoprotein cholesterol (LDL-C) was calculated using the Friedewald equation (49). Insulin was measured via radioimmunoassay. The HOMA-IR index, an indicator of insulin resistance, was calculated from fasting insulin and glucose as (insulin (mU/L) × glucose (mmol/L/22.5) (49).

Statistical analysis

Analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC). Variables were examined for outliers and normality. E1, E2, T, FT, and SHBG were log-transformed for analyses. Associations between each hormone and each outcome were tested using logistic (TPA tertiles, GSM tertiles, any calcification) and Poisson regression (plaque number). Primary models considered each hormone separately. Models adjusted for covariates associated with sex hormones and carotid plaque measures at P < .10, including age, race/ethnicity, education, BMI, systolic blood pressure, insulin, HOMA-IR, LDL-C, and triglycerides. Due to the high correlation between insulin and HOMA-IR (r = 0.93), and the stronger correlation between HOMA-IR and the outcomes of interest (TPA: HOMA-IR r = 0.19 vs insulin = –0.02), we adjusted for HOMA-IR instead of insulin in all models. Additional models adjusted for blood pressure, diabetes, and lipid-lowering medications. Because increased testosterone levels (>52 ng/dL) may be a sign on an underlying pathology such as polycystic ovarian syndrome, we performed sensitivity analyses excluding high T levels (>52 ng/dL, n = 3). As in previous analyses (25), sensitivity analyses were performed excluding E2 values below the limit of quantitation (<2.0 pg/mL, n = 1) and high FT (>2.0 ng/dL, n = 0) values. Sensitivity analyses were also performed excluding 5 women with menstruation-related issues (eg, premature ovarian insufficiency, polycystic ovarian syndrome). Interactions between hormones and BMI using both a continuous and a dichotomous variable (BMI < 30) were tested. Menopause status was considered as an additional covariate. We also examined interactions between sex hormones and menopausal status. Models were repeated excluding perimenopausal women and women with reported statin use. Residual analysis and diagnostic plots were conducted to verify model assumptions. All models were 2-sided at α = .05.

Results

The sample of 141 women (72% non-Hispanic White) were on average 54 years old, overweight, and 82% were postmenopausal with final menstrual period on average 4.8 (standard deviation ± 4.5) years ago (Table 1). Consistent with a sample of primarily early postmenopausal women, the median (interquartile range [IQR]) of sex hormone levels were estrone 27.5 pg/mL (IQR 17.3–37.6), estradiol 4.8 pg/mL (IQR 2–11.3), testosterone 25.3 ng/dL (IQR 20.2–31.6), FT 0.3 (IQR 0.2–0.44), and SHBG 86.1 nmol/L (IQR 48.4–140.8). Approximately one-third of women had evidence of calcified carotid plaque.

Table 1.

Sample characteristics

Variable, N	N = 141
Age, y, mean (SD)	54.2 (4.2)
Race/ethnicity, n (%)
Non-Hispanic White	101 (71.6)
Black/othera	40 (28.4)
Education, n (%)
High school/some college/vocational	60 (42.6)
College or higher	81 (57.4)
Hard to pay for basics, n (%) (N = 138)
Not hard	91 (65.9)
Somewhat/very hard	47 (34.1)
Menopause status, n (%)
Perimenopausal	26 (18.4)
Postmenopausal	115 (81.6)
Years since FMP, M (SD)	4.8 (4.5)
Parity, number of live births, median (IQR)	2.0 (1.0–3.0)
BMI, mean (SD)	28.39 (6.14)
SBP, mm Hg, mean (SD)	121.55 (14.94)
DBP, mm Hg, mean (SD)	70.17 (9.07)
LDL-C, mg/dL, mean (SD)	133.14 (35.62)
HDL-C, mg/dL, mean (SD)	63.20 (14.62)
Triglycerides, mg/dL, median (IQR)	98.00 (75.0–142.5)
Insulin, median (IQR)	10.5 (8.2, 13.3)
HOMA-IR, median (IQR)	2.35 (1.69–3.25)
Medications, n (%)
Blood pressure lowering	34 (24.1)
Antidiabetic	8 (5.7)
Lipid lowering	23 (16.3)
Physical activity, leisure time, median (IQR)	480 (24.75–1188)
E1, pg/mL, median (IQR)	27.50 (17.30–37.60)
E2, pg/mL, median (IQR)	4.80 (2.00–11.30)
T, ng/dL, median (IQR)	25.30 (20.20–31.60)
SHBG, nmol/L, median (IQR)	0.30 (0.20–0.44)
FT, median (IQR)	86.10 (48.40–140.80)
Total no. of plaques, median (IQR)	2.00 (1.00–2.50)
Total plaque area (mm2), median (IQR)	17.10 (9.85–26.34)
Gray-scale median <32	21 (14.9)
Average gray-scale median of total plaques, median (IQR)	59.78 (47.74–72.93)
Calcification	46 (32.6)

Sample sizes for all variables N = 141, unless otherwise noted.

Abbreviations: BMI, body mass index; CRP, C-reactive protein; DB, diastolic blood pressure; FMP, final menstrual period; FT, free testosterone; IQR, interquartile range; SBP, systolic blood pressure. Of those on lipid-lowering medication nearly half reported statin use (n, 11).

Table 2 presents Pearson correlations between endogenous sex hormones, CVD risk factors, and carotid plaque measures in our total sample. We found that SHBG was inversely correlated with BMI, insulin, HOMA-IR, and C-reactive protein in the total sample (Table 2) and when excluding perimenopausal women. There was a strong negative correlation between SHBG and FT in the total sample (r = –0.78) and when excluding perimenopausal women (r = –0.77). TPA and plaque number were positively correlated with age, BMI, LDL-C, and HOMA-IR. E1 was negatively correlated with GSM.

Table 2.

Pearson correlation matrix for continuous measures of carotid plaque, sex hormones, and cardiovascular risk factors in the total sample (n = 141)

	TPA	Gray-scale median	Plaque number	Log E1	Log E2	Log T	Log SHBG	Log FT
Age	0.19b	0.06	0.31d	–0.13	–0.27c	0.13	0.0001	0.07
BMI	0.14a	–0.03	0.15a	0.34d	0.31c	0.20b	–0.38d	0.45d
SBP	0.08	–0.07	0.14a	0.16a	0.09	0.19b	–0.10	0.19b
DBP	–0.006	–0.07	0.02	0.05	0.05	0.14a	0.14a	–0.03
HDL-C	–0.07	–0.09	–0.01	–0.16a	–0.20b	–0.005	0.32d	–0.26c
Triglyceride	0.17a	0.06	0.14	0.03	0.06	0.008	–0.28c	0.23c
LDL-C	0.25c	0.10	0.18b	0.06	0.13	0.11	–0.15a	0.20b
HOMA-IR	0.19b	–0.11	0.12	0.16a	0.18b	0.16a	–0.52d	0.52d
Insulin	–0.02	–0.10	–0.08	0.03	0.16	0.03	–0.50d	0.44d
CRP	0.07	0.02	0.15a	0.18b	0.22c	–0.02	–0.43d	0.34d
TPA	1.00	–0.04	0.62d	0.08	–0.001	0.12	0.07	0.007
Gray-scale median	–0.04	1.00	–0.06	–0.23b	–0.13	0.05	–0.01	0.03
Plaque number	0.62d	–0.06	1.00	–0.05	–0.10	0.12	0.05	0.02
Log E1	0.08	–0.23b	–0.05	1.00	0.67d	0.32c	–0.11	0.29c
Log E2	–0.001	–0.13	–0.10	0.67d	1.00	0.31c	–0.16a	0.32c
Log T	0.12	0.05	0.12	0.32c	0.31c	1.00	0.13	0.52d
Log SHBG	0.07	–0.01	0.05	–0.11	–0.16a	0.13	1.00	–0.78d
Log FT	0.007	0.03	0.02	0.29c	0.32c	0.52d	–0.78d	1.00

Abbreviations: BMI, body mass index; CRP, C-reactive protein; DBP, diastolic blood pressure; E1, estrone; E2, estradiol; FT, free testosterone; GSM, gray-scale median; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; IL6, interleukin-6; LDL, low-density lipoprotein; SBP, systolic blood pressure; SHBG, sex hormone binding globulin; T, testosterone; TPA, total plaque area.

^a P ≤ .10.

^b P < .05.

^c P < .01.

^d P < .0001.

In adjusted models evaluating associations between sex hormones and measures of carotid plaque burden (ie, total number of plaques and TPA), higher SHBG levels were associated with risk of a greater number of carotid plaques, but lower TPA (Table 3). These findings were unchanged when excluding perimenopausal women (Table 3). In models adjusting for all covariates, higher FT was related to a greater TPA tertile. Levels of E1, E2, and T were not associated with carotid plaque burden.

Table 3.

Associations between sex hormones and measures of carotid plaque burden

	Total no. of plaques			Tertile TPA
	RR (95% CI)			OR (95% CI)
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
E1	0.95 (0.80, 1.14)	0.94 (0.78, 1.14)	0.89 (0.71, 1.13)	0.85 (0.51, 1.39)	0.87 (0.49, 1.55)	0.94 (0.48, 1.85)
E2	0.96 (0.89, 1.03)	0.95 (0.87, 1.03)	0.91 (0.82, 1.01)a	1.04 (0.85, 1.27)	1.08 (0.85, 1.37)	1.04 (0.85, 1.27)
T	1.28 (0.91, 1.18)	1.09 (0.76, 1.58)	1.11 (0.75, 1.63)	0.79 (0.30, 2.12)	1.01 (0.34, 2.95)	0.78 (0.25, 2.47)
SHBG	1.05 (0.89, 1.25)	1.26 (1.02, 1.55)b	1.29 (1.04, 1.61)b	0.61 (0.38, 0.98)b	0.39 (0.21, 0.74)c	0.28 (0.13, 0.57)c
FT	1.03 (0.83, 1.27)	0.81 (0.62, 1.05)	0.80 (0.61, 1.05)	1.67 (0.92, 3.01)	2.89 (1.31, 6.37)c	3.58 (1.53, 8.42)c

Abbreviations: E1, estrone; E2, estradiol; T, testosterone; FT, free testosterone; SHBG, sex hormone binding globulin; TPA, total plaque area. Hormones log-transformed. RR determined by Poisson distribution. Tertile TPA assessed using ordinal logistic regression (ref, lowest tertile). Model 1: unadjusted. Model 2: age; race/ethnicity; education; BMI; SBP; HOMA-IR; LDL-C; triglycerides. Model 3: Sensitivity analysis excluding perimenopausal women (n, 115).

^a P ≤ .10.

^b P < .05.

^c P < .01.

^d P < .0001.

Tables 4 and 5 present associations between sex hormones and carotid plaque characteristics (ie, calcification, echogenicity), which correlate with the histological composition of the plaque. For example, echolucent plaques (lower GSM) represent rupture-prone plaques which are lipid-rich with a large necrotic core, whereas echogenic plaques (greater GSM) represent more stable plaques with a greater content of fibrous tissue and calcification (49). Higher E1 was associated with greater GSM tertile, indicating more stable plaques appearing echogenic with fibrous or calcific content; the magnitude of the effect was greater (OR 3.16, 95% CI 1.09, 9.18) when excluding women who reported statin use (n = 11). T levels were inversely associated with the presence of calcified carotid plaque, but the association did not persist after adjusting for CVD risk factors and covariates. E2, FT, and SHBG were not associated with carotid plaque characteristics.

Table 4.

Associations between sex hormones and measures of carotid plaque characteristics

	Gray-scale median ≥ 32			Tertile GSM
	OR (95% CI)			OR (95% CI)
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
E1	1.10 (0.54, 2.21)	1.22 (0.54, 2.74)	0.52 (0.08, 3.60)	2.08 (1.23, 3.52)c	2.31 (1.26, 4.22)c	1.78 (0.89, 3.58)a
E2	0.85 (0.63, 1.14)	0.88 (0.63, 1.22)	0.91 (0.53, 1.58)	1.22 (0.99, 1.50)	1.27 (0.99, 1.63)	1.22 (0.99, 1.50)
T	0.62 (0.15, 2.56)	0.46 (0.10, 2.11)	0.30 (0.02, 6.15)	0.81 (0.30, 2.17)	0.61 (0.21, 1.77)	0.66 (0.21, 2.05)
SHBG	1.67 (0.83, 3.39)	1.41 (0.59, 3.35)	1.32 (0.26, 6.62)	1.20 (0.76, 1.91)	1.24 (0.69, 2.22)	1.56 (0.82, 2.97)
FT	0.50 (0.21, 1.21)	0.53 (0.19, 1.50)	0.50 (08, 3.12)	0.81 (0.745, 1.44)	0.65 (0.31, 1.37)	0.53 (0.724, 1.18)

Abbreviations: E1, estrone; E2, estradiol; T, testosterone; FT, free testosterone; SHBG, sex hormone binding globulin; GSM, gray-scale median. Hormones log-transformed. RR determined by Poisson distribution. Tertile TPA assessed using ordinal logistic regression (ref, lowest tertile). Model 1: unadjusted. Model 2: age; race/ethnicity; education; BMI; SBP; HOMA-IR; LDL-C; triglycerides. Model 3: Sensitivity analysis excluding perimenopausal women (n, 115).

^a P ≤ .10.

^b P < .05.

^c P < .01.

^d P < .0001.

Table 5.

Associations between sex hormones and carotid plaque calcification

	Presence of calcification
	OR (95% CI)
	Model 1	Model 2	Model 3
E1	0.76 (0.44, 1.30)	0.77 (0.39, 1.52)	0.57 (0.25, 1.31)
E2	1.04 (0.84, 1.28)	1.01 (0.76, 1.34)	0.97 (0.70, 1.35)
T	0.33 (0.11, 1.01)	0.40 (0.10, 1.58)	0.33 (0.08, 1.37)
SHBG	1.06 (0.63, 1.78)	0.67 (0.31, 1.43)	0.56 (0.25, 1.25)
FT	0.58 (0.30, 1.12)	0.84 (0.34, 2.11)	0.91 (0.36, 2.28)

Abbreviations: E1, estrone; E2, estradiol; T, testosterone; FT, free testosterone; SHBG, sex hormone binding globulin; TPA, total plaque area. Hormones log-transformed. RR determined by Poisson distribution. Model 1: unadjusted. Tertile TPA assessed using ordinal logistic regression (ref, lowest tertile). Model 1: age; race/ethnicity; education; BMI; SBP; HOMA-IR; LDL-C; triglycerides (postmenopausal women only). Model 2, Model 1 adjusting for family history of CVD. Model 3, Model 2 adjusting for gestational hypertension/preeclampsia; and gestational diabetes.

^a P ≤ .10.

^b P < .05.

^c P < .01.

We performed several sensitivity analyses to examine the relationship between endogenous sex hormones and carotid plaque. We excluded perimenopausal women (n = 26), which did not alter findings. We additionally adjusted for medications that have been shown to alter plaque (ie, antihypertensives, antidiabetics, and lipid-lowering), but results were unchanged; when we excluded women on any of these medications, findings remained unchanged. We also repeated analyses excluding women with high T levels (>52 ng/dL, n = 3) or FT (>2.0 ng/dL, n = 0) values, or low E2 values (<2.0 pg/mL, n = 1); findings were unchanged. We additionally tested for interactions between sex hormones and BMI and found no significant interactions.

Discussion

This cross-sectional study extends earlier work in MsHeart by examining the association between endogenous sex hormones and measures of carotid plaque burden and characteristics in nonsmoking midlife women free of clinical CVD. Furthermore, while prior analyses have shown differential associations between endogenous sex hormones and plaque composition in men with coronary artery disease (20), we assess this relationship in peri- and postmenopausal women using highly accurate LC-MS/MS assays for E1, E2, and T. We found that after adjusting for a range of CVD risk factors, higher SHBG was related to a greater number of plaques yet lower total carotid plaque area. Higher FT was related to greater carotid plaque area, whereas higher E1 was associated with indicators of more stable plaque (greater GSM) (50). Taken together, these findings suggest that higher SHBG levels and FT are related to greater plaque size, while higher E1 is associated with more stable plaque (echogenicity). Carotid plaque size and echogenicity have been shown to predict atherosclerotic events, such as stroke and myocardial infarction (27, 51).

Increasing evidence suggests that lower levels of SHBG, a major carrier of androgens, is associated with a proatherogenic risk factor profile in postmenopausal women, including higher waist circumference, insulin, glucose, inflammatory markers, and adverse lipids (52, 53). However, findings of the association between SHBG and atherosclerosis have been inconsistent. While some studies show that lower SHBG is associated with presence of coronary artery calcification (11), others suggest higher levels of SHBG are associated with greater coronary calcium independent of adiposity (13). Similarly, an analysis of midlife women in the Coronary Artery Risk Development in Young Adults (CARDIA) study showed an inverse relationship between SHBG and carotid intima–media thickness (11), but our earlier analysis in MsHeart found that higher SHBG is related to greater carotid intima–media thickness (25). Of note, participants in CARDIA were mostly premenopausal (62%) and SHBG levels were on average lower than MsHeart participants; studies have postulated a threshold effect for SHBG that may be related to metabolic and cardiovascular outcomes during midlife in women (54). Though carotid plaque is a direct measure of carotid atherosclerosis, few studies have examined the relationship between SHBG and carotid plaque. We show that in a sample of mostly postmenopausal women, higher SHBG is related to greater number of carotid plaques, but lower plaque area.

It is possible that lower levels of bioavailable E2, which are associated with higher SHBG (55), could be associated with a greater number of carotid plaques, while insulin sensitivity and lower inflammation reduce plaque area. We found higher levels of SHBG correlated with lower bioavailable E2 and insulin sensitivity (Table 2). Though low trajectories of E2 during the menopause transition have been associated with greater prevalence of carotid plaque (56), the current analysis was cross-sectional and could not assess these changes in E2 over time. Insulin sensitivity and lower inflammation, observed with increasing levels of SHBG during the menopause transition (57,58) have been associated with lower plaque area (39,59). Importantly, our findings were similar when we excluded perimenopausal women, suggesting that our findings were not linked to menopause status. In addition, findings were similar when we adjusted for HOMA-IR or insulin, but there was a stronger correlation between SHBG and HOMA-IR. Our findings extend prior work, including our own, that found an association between higher levels of SHBG and carotid plaque presence (14,25). Whether the menopause-related reduction in E2 (60) contributes to the higher risk of CVD in women after menopause (61,62) has been a matter of debate. Longitudinal analyses are related to carotid plaque burden and characteristics. To further elucidate whether these associations persist over time, future work should consider progression of carotid plaque over time.

While prior findings in MsHeart showed that higher FT levels are associated with lower carotid plaque presence (25), here we found that higher FT is associated with greater TPA, which mirrors our findings with SHBG. Our results underscore the importance of assessing additional measures of carotid plaque burden to better determine the association between FT and carotid plaque. These findings are consistent with the Multi-Ethnic Study of Atherosclerosis, which reported a positive association between FT and carotid intima–media thickness (13). Studies in midlife women have shown that higher FT levels are related to the development of metabolic syndrome (63) and increases in visceral fat (64). Thus, potential mechanisms driving associations between FT and total carotid plaque area may include elevated blood glucose, insulin resistance, and lipoprotein-related inflammation (65). BMI is particularly important to consider in these analyses given its association with T and SHBG and its role as a strong risk factor for CVD; our findings were independent of adiposity. A prior analysis among postmenopausal women found an inverse association between FT and cardiovascular mortality in diabetic women, but no such association in women without diabetes (66). Mixed findings may in part be due to the use of older androgen assays in earlier studies, which have poorer accuracy and precision.

Our work highlights the important role of E1 in carotid atherosclerosis. While lower levels of E1 have been associated with increased all-cause mortality among postmenopausal women (67), less is known about carotid artery disease. The Estrogen in the Prevention of Atherosclerosis Trial found an inverse association between E1 and carotid intima–media thickness progression, controlling for age and BMI (68). Similarly, our prior work in this cohort found that higher E1 levels were associated with more favorable endothelial function (25). Carotid plaque characteristics, such as plaque echogenicity, may identify “vulnerable” or unstable plaques that are more likely to rupture, precipitating a major atherosclerotic event (ie, myocardial and cerebral infarction). Vulnerable/unstable plaques are characterized by a large lipid-rich core, a thin fibrous cap, and a greater number of inflammatory cells; stable plaques, on the other hand, are characterized by a smaller lipid content, more collagen and calcium, and lower concentration of inflammatory cells (36,69). In the current study we showed that higher E1 is related to greater echogenicity, characteristic of a more stable (low-risk) plaque. Echolucency of carotid plaques has been associated with low HDL-C and increased BMI (70). E1 is produced mostly by the conversion of androgens in peripheral tissues, particularly body fat. Thus, BMI may play a critical role in the observed associations between E1 and GSM; notably, our findings persisted adjusting for BMI. Though our findings are consistent with prior analyses suggesting the possible protective effects of higher E1 levels in postmenopausal women (67), there is a need for longitudinal analyses assessing patterns of E1 across the menopause transition. Few studies have examined patterns of E1 concentration changes across the menopause transition (71) and none has examined whether these patterns differentially impact carotid plaque development and progression.

Carotid plaque assessment by ultrasound is a noninvasive and cost-effective measure of subclinical atherosclerosis that is associated with future CVD events (41, 72). Because atherosclerotic plaques are not static lesions, plaque presence alone may not provide sufficient information on plaque progression and vulnerability, which is an important to risk factor for CVD events (39, 72). In an earlier analysis in MsHeart, SHBG and FT levels were positively and negatively associated with carotid plaque presence, respectively. This study found that SHBG and FT are related to TPA while E1 is related to plaque echogenicity, highlighting different aspects of carotid atherosclerosis development. Although FT was related to carotid plaque measures, serum T was not related to any parameter of carotid plaque. The high correlation between SHBG and FT, but not total T, suggests that changes in SHBG may be the primary determinant of FT levels, which may affect carotid health. Our findings differ from the Rotterdam Study which found that higher E2 and lower total T are related to a more adverse carotid plaque composition (ie, intraplaque hemorrhage) in postmenopausal women (73). However, the Rotterdam Study assessed plaque composition using magnetic resonance imaging, which is able to delineate plaque morphology with higher resolution. Future studies using magnetic resonance imaging of carotid plaque may provide additional information on the association between endogenous sex hormones and carotid plaque burden in midlife women.

The results of this study should be viewed within the context of its limitations and strengths. First, this is a cross-sectional analysis and causality cannot be determined. Moreover, we could not account for several factors related to endogenous sex hormones, including genetic variability (eg, estrogen receptor subtypes), enzymatic activities of aromatase, stress hormones, and body fat mass. Another limitation was the lack of information on pregnancy history. Additional pregnancy-related data is warranted given accumulating evidence relating pregnancy complications to subclinical CVD (74). Similarly, data on family history of CVD was only available for a subsample of women (n = 65). Analyses considering these important factors are necessary to determine the physiologic mechanisms linking sex hormones to carotid plaque and/or potential moderators of these associations. Future research should consider an assessment of the change in carotid plaque size and echogenicity. In addition, plaque composition (or morphology), such as calcification, lipid core, and intraplaque hemorrhage, is more accurately visualized using magnetic resonance imaging rather than ultrasound. It is important to note that FT was calculated using the ensemble allostery model, which provides excellent conformity with values measured by equilibrium dialysis; yet FT was not measured directly. Though our sample size of 141 women with carotid plaque is smaller than epidemiologic studies with similar plaque data (ie, the Multi-Ethnic Study of Atherosclerosis), prior studies did not measure sex steroids via LC-MS/MS, a robust measure that can more accurately detect low levels of steroid sex hormones. Furthermore, consistent with the demographics of the study’s geographic area, few Hispanic/Latina or Asian women were included. Notably, 28% of the sample identified as African-American/non-Hispanic Black; earlier studies suggest prevalence of carotid plaque is greater among White women compared to Black men and women (75–77). Future studies in a racially/ethnically diverse cohort are necessary.

Strengths of the present study include the use of LC-MS/MS, a gold standard measure for low levels of steroid sex hormones typically present in postmenopausal women. Moreover, we considered a wide range of endogenous sex hormones, including E1, an understudied hormone in relation to CVD. We assessed several measures of carotid plaque burden and characteristics, including total carotid plaque area and GSM, predictive of a CVD event. We considered multiple confounders and potential explanatory factors, including a range of CVD risk factors; associations persisted with adjustment for these CVD risk factors.

Conclusions

This study suggests that endogenous sex hormone levels may play an important role in the development of carotid atherosclerosis. In addition, our findings underscore the complexities of the effect of endogenous sex hormones on cardiovascular health. We found that SHBG and FT were related to carotid plaque burden and E1 was related to plaque characteristics. Peri- and postmenopause are critical periods of dynamic vascular changes. Our findings support the potential importance of endogenous sex hormones in the development of carotid atherosclerosis as women age.

Glossary

Abbreviations

BMI

body mass index

CHD

coronary heart disease

CI

confidence interval

CVD

cardiovascular disease

E1

estrone

E2

estradiol

GSM

gray-scale median

HDL-C

high-density lipoprotein cholesterol

LC-MS/MS

liquid chromatography–tandem mass spectrometry

LDL-C

low-density lipoprotein cholesterol

OR

odds ratio

SHBG

sex hormone binding globulin

T

testosterone

TPA

total plaque area

Additional Information

Disclosure Summary: R.C.T. consults for Astellas, Pfizer, and Procter and Gamble and has received research grants from the National Institutes of Health (NIH). S.B. has received research grants from NIH, National Institute on Aging (NIA), the National Institute of Nursing Research, the Patient Centered Outcomes Research Institute, the Foundation for the NIH, AbbVie, Alivegen, MIB, Althea Biosciences, and Transition Therapeutics; these research grants are managed by his institution. S.B. has received consulting fees from AbbVie, Novartis, and Opko. S.B. holds an equity interest in FPT, LLC. N.S. consults for Astellas, serves on the scientific advisory board for Menogenix, and owns stock options in Menogenix. The remaining authors have nothing to disclose.

Data Availability: Restrictions apply to the availability of data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.