Abstract
Objective:
Studies suggest a reversal in the protective association of high-density lipoprotein cholesterol (HDL-C) and cardiovascular disease (CVD) in women traversing menopause. Decreasing estrogen levels during the transition, as well as inflammation, may explain this reversal. We tested whether either estradiol or C-reactive protein (CRP) concentrations modified the association of HDL-C with aortic (AC) or coronary artery calcification (CAC).
Methods:
478 participants between ages 46–59 from the SWAN Heart baseline visit were included. AC and CAC presence were defined as Agatston score ≥100 and ≥10, respectively. Logistic regression was used for analysis.
Results:
112 (23.53%) participants had AC ≥ 100 and 104 (21.76%) had CAC ≥ 10. In unadjusted models, a 1-mg/dL higher in HDL-C was associated with 3% lower odds of AC (95% CI: 0.95 to 0.99) and 4% lower odds of CAC (95% CI: 0.95 to 0.98). In adjusted models, a significant interaction between HDL-C and estradiol with respect to AC but not CAC was detected, such that higher HDL-C level was protective at the highest estradiol quartile (OR: 0.91, 95% CI: 0.84 to 0.99 per 1 mg/dL higher HDL-C, P=0.03) but tended to associate with greater risk at the lowest quartile (OR: 1.04, 95% CI: 0.98 to 1.10 per 1 mg/dL higher HDL-C, P=0.16). CRP did not modify any association.
Conclusions:
The protective cardiovascular association of higher HDL-C levels on AC was modified by estradiol but not CRP concentrations. The pathways through which estradiol might influence this association should be further investigated.
Keywords: cardiovascular disease, menopause, high density lipoprotein cholesterol, estradiol
1. Introduction
Cardiovascular disease (CVD) is the leading cause of death worldwide [1]. As women approach midlife, risk of CVD increases [2]. Growing evidence suggests that menopause may play a role in this increased risk [3].
The menopause transition is characterized by both hormonal changes and menstrual cycle irregularities [4]. During this transition, estradiol levels decrease, which may contribute to CVD risk. Findings from both the Women’s Ischemia Syndrome Evaluation (WISE) study and the Study of Women’s Health Across the Nation (SWAN) have suggested that estradiol concentration was associated with coronary artery disease and development of atherosclerosis [5, 6].
Interestingly, as women transition through menopause, they experience dramatic increases in low-density lipoprotein cholesterol (LDL-C), which may contribute to an increased risk of CVD [6, 7]. However, the trajectory of high-density lipoprotein cholesterol (HDL-C) over the menopause transition has not been consistent across studies [8]. In the SWAN cohort, a slight increase in HDL-C levels was reported around the final menstrual period [7]. The study included 863 postmenopausal SWAN participants who showed an annual mean change in HDL-C levels of 0.57 mg/dL around the time of the final menstrual period, with smaller increases in the year prior (0.40 mg/dL) and the year following (0.21 mg/dL) the final menstrual period [7]. The reported increase seems to be related to estradiol level as suggested by another SWAN analysis [9]. The average HDL-C level among participants in the lowest estradiol quartile (E2 < 21.45) was 57.9 mg/dL and among participants in the highest estradiol quartile (E2 > 78.62) was 59.0 mg/dl [9].
Not only the direction of changes in HDL-C over the menopausal transition that has not been clear, but also the role of HDL-C in CVD risk prediction among women transitioning through menopause. Whether high levels of HDL-C may further contribute to the increase in risk of CVD during midlife is a current open research question [10–13].
Recent work from the Multi-Ethnic Study of Atherosclerosis (MESA) and SWAN suggest that increasing HDL-C levels are associated with an increased risk of subclinical CVD in women transitioning through menopause and in older women [10–13]. SWAN research has reported a reversal in the protective association between HDL-C and subclinical CVD among midlife women [10, 11]. As women transition through menopause, SWAN has observed a positive association in HDL-C level and subclinical CVD, particularly in AC presence and carotid intima-media thickness [10, 11]. Results from the MESA study are consistent with results from SWAN, reporting that higher HDL-C levels were associated with an increased risk of carotid plaque in postmenopausal women [13]. These prior studies have suggested that high levels of HDL-C may contribute to an increased risk of CVD in midlife women. However, the mechanism of this association remains unclear.
Estradiol may impact the size and number of HDL particles and the ability of HDL particles to promote cholesterol efflux from macrophages [12]. Since estradiol is considered an anti-inflammatory agent in many cases, its decrease may lead to an increase in inflammation [14, 15]. Therefore, inflammation may modify the association of HDL-C with CVD in the same way as estradiol.
We aimed to build off prior findings showing a reversal in the protective association of HDL-C and CVD and examine whether estradiol concentration or inflammation (C-reactive protein, or CRP) modifies the association of HDL-C and CVD with respect to two subclinical measures of CVD: AC and coronary artery calcification (CAC). We hypothesized that lower estradiol and/or higher CRP levels modify the association of HDL-C and subclinical measures of CVD, such that HDL-C would be positively associated with calcification at low estradiol levels or higher CRP levels, which would explain the reversal seen in previous research.
2. Methods
2.1. Study Participants
The Study of Women’s Health Across the Nation (SWAN) is a longitudinal, multi-site study which recruited over 3,000 women between the ages of 42 and 52 to understand biological and psychological factors which impact the health of women around the menopause transition. Details about SWAN study design have been published previously [16]. Briefly, SWAN focused on recruitment of geographically and racially diverse groups at seven different sites across the United States. Eligibility criteria included having at least one ovary and an intact uterus, having menstruated during the past 3 months but not having taken either hormone therapy or oral contraceptive in the previous three months, and not being pregnant or breast feeding.
Several ancillary studies have arisen from the parent study, including SWAN Heart, which focused on progression of subclinical measures of CVD at midlife. Recruitment for SWAN Heart occurred between annual SWAN visits four through seven, from 2001 to 2003 at the Pittsburgh and Chicago sites, with 608 eligible participants recruited for the study. Study protocol was approved by the institutional review boards at each site, and participants submitted written informed consent before enrollment.
For this analysis, women were excluded if they had a history of heart attack, stroke, or angina (n=11), were surgically menopausal (n=16), were missing CAC and AC scores (n=40), or were missing data on estradiol (n=39). Women who had unknown menopausal status due to hormone therapy (HT) use in the past three months were also excluded from this analysis (n=24). Our final sample included 478 participants for the analysis of effect modification by estradiol levels. For the analysis of effect modification by CRP level, an additional 136 women who did not have CRP values measured at the Chicago site were excluded. Although SWAN Heart is a longitudinal study, this analysis was cross-sectional and used data from the SWAN Heart baseline visit.
2.2. Aortic and Coronary Artery Calcium Scores
Aortic (AC) and coronary artery calcium (CAC) scores were measured via non-contrast CT scan on an Imatron C-150 Ultrafast CT Scanner at SWAN Heart baseline. Two passes were made, gathering 30–40 contiguous 3-mm images on the second pass. Calcium presence was defined as having 3 contiguous pixels of greater than 130 Hounsfield units (HU). Results from CT scans were recorded and sent to Pittsburgh for scoring using a DICOM station. An Agatston score was calculated by multiplying the density score (130 to199 HU = 1, 200 to299 HU = 2, 300to 399 = 3, 400 + = 4) by the total area of observed calcification, in millimeters [17]. AC score was the total Agatston score for the aorta. CAC score was the sum of the Agatston scores from the left main, left anterior descending, left circumflex, and right coronary arteries. These scores report relatively high reproducibility ratings, with low interobserver variability (0.98 for AC and 0.99 for CAC) [17].
2.3. Study Covariates
Waist circumference, systolic blood pressure (SBP), HDL-C, LDL-C, triglycerides, glucose, CRP, and estradiol were measured at SWAN Heart baseline. Waist circumference was recorded as the smallest circumference from the ribs and the top of the hip. Blood pressure was recorded as an average of two readings. Blood panels were performed after an overnight fast of at least 10 hours and on days 2 through5 of a menstrual cycle that coincided with an annual visit. If a sample could not be obtained during this timeframe, a random sample was collected within 90 days. Cycle day of blood draw was recorded as days 2–5 of the menstrual cycle or unknown and was adjusted for in all analyses which included estradiol. All assays were performed at the Medical Research Lab in Lexington, KY. Plasma for HDL-C measures was treated with heparin and manganese chloride. Total cholesterol was measured in plasma with a variation of 0.8%. Triglycerides were measured in plasma in an enzymatic assay with a variation of 1.2%. LDL-C was calculated by subtracting 1/5 triglycerides as well as HDL-C from total cholesterol (LDL-C = Total cholesterol – 1/5 triglycerides – HDL-C). Serum glucose was measured using a Hitachi 747–200 analyzer. CRP was measured on plasma, using different assay methods over time. A calibration equation was developed and applied to convert all values from different assays to a high sensitivity assay (ELISA), a plate assay which employs the quantitative sandwich enzyme immunoassay technique using a monoclonal antibody specific for CRP which has been pre-coated onto a microplate. The assay has a range of 0.78 – 125.0 pg/mL with a sensitivity of 0.10 pg/mL. The inter-assay CV: 6.0% at 4.84 pg/mL, 7.0% at 8.66 pg/mL, 6.6% at 17.7 pg/mL; the intra-assay CV: 4.4% at 4.79 pg/mL, 3.8% at 8.66 pg/mL, 8.3% at 18.9 pg/mL. Estradiol was measured via E2–6 immunoassay, which has a lower limit of detection of 7.0 pg/mL. Estradiol was measured twice and was recorded as the mean of two values. Participants who had estradiol levels under 7.0 pg/mL were assigned a random number between 1 and 7 to represent their estradiol concentrations.
Self-reported covariates used in this analysis include age, race, smoking status, household income, education level, and alcohol consumption. Race was reported as either White or Black. Age was calculated at the time of CT scan based on birth date given at SWAN enrollment. Smoking status was reported as either currently smoking or not currently smoking. Annual income was dichotomized as less than $50,000 or at least $50,000 annually. Education was categorized as holding less than a bachelor’s degree or holding a bachelor’s degree or higher. Alcohol consumption was categorized into less than two drinks per week or greater than or equal to two drinks per week.
Menopausal status was determined by self-reported answers to questions about bleeding patterns, reproductive surgeries, and hormone use. Participants who had not noted a change in bleeding patterns within 12 months were classified as premenopausal. Those who noted a change in regularity in the prior 3 months were considered early perimenopausal. Those who had no bleeding for 3 months or more, but less than 12 months were considered late perimenopausal. Finally, women who had no bleeding for 12 months or more were considered postmenopausal. Due to small sample sizes in some groups, menopausal status was dichotomized into pre-/early perimenopausal and late peri-/postmenopausal, with the expectation that hormone profiles will be similar within these groups.
2.4. Statistical Analysis
AC and CAC are known to be skewed with many 0 values. These values were thus dichotomized into “low” and “high” categories. Similar to previous work [18, 19], CAC presence was defined as ≥ 10 Agatston units, and AC presence was defined as ≥100 Agatston units.
Continuous variables were assessed for normality, and those which were not normally distributed were log-transformed. Glucose, estradiol, triglycerides, and CRP were all log-transformed for this analysis. All other variables were approximately normally distributed and were analyzed as continuous variables. Characteristics were compared among participants who were included and excluded using chi-square tests for categorical variables and t-tests or Wilcoxon rank-sum tests for continuous variables. Means and standard deviations were compared for normally distributed variables, whereas medians and inter quartile ranges were compared for non-normally distributed or skewed variables. Additionally, because waist circumference and CRP were moderately correlated, models which include both covariates were analyzed using the waist circumference residuals from a linear regression model between waist circumference and CRP to reduce collinearity in the models.
Univariate logistic regression was performed to assess the association of the defined study exposure variables with AC and CAC as outcome variables. Multivariable adjusted models were created for each outcome measure, using forward stepwise variable selection methods. Age, race, study site, and hormone therapy usage were included as covariates in all final models, regardless of their level of significance, per SWAN protocol. The models containing all additional variables that were associated with the outcome at the p < 0.05 level were considered the final main effect models. Effect modification by estradiol and by CRP was then tested by including the cross-product term for each of these two independent variables with HDL-C in separate models. The interaction was assessed in unadjusted and fully adjusted models.
After defining the final main effect models for AC and CAC, models were stratified by estradiol quartile (Q1 < 16.15 pg/mL, Q2: 16.15 to < 29.67 pg/mL, Q3: 29.67 to < 74.15 pg/mL, and Q4: 74.15 and above). Forest plots were produced from the stratified models in order to visualize the association of HDL-C on AC and CAC at various estradiol levels. We also performed a sensitivity analysis, excluding participants whose estradiol levels were below the lower limit of detection (n=18).
All statistical procedures were performed using SAS version 9.4.
3. Results
478 women were included in the main analysis. Average age of included participants was 50.9 years, and the sample was 37% Black (Table 1). Women who were excluded from these analyses differed from those included in that a smaller proportion of the excluded women were from the Chicago site compared to the Pittsburgh site (43% excluded participants were Chicago based, whereas 61% of included women were Chicago participants). More included women reported moderate alcohol consumption (more than one drink per month up to one drink per week) compared to those who were excluded (p=0.005). However, when alcohol consumption was dichotomized into at most one drink per week versus more than one drink per week, this difference between included and excluded groups was not significant, with 32% of the excluded women and 24% of the included women reporting drinking more than 2 drinks per week (p=0.07). Finally, women who did not have CRP measures were excluded from the analysis in models containing CRP as a covariate. It should be noted that the characteristics of the excluded women were not significantly different from those who were included.
Table 1:
Baseline characteristics of SWAN Heart participants who were included versus excluded participants
| Demographic characteristics | Included (n=478) | Excluded (n=130) | P value |
|---|---|---|---|
| Age, years, mean (SD) | 50.9 (2.91) | 50.8 (2.72) | 0.79a |
| Black, n (%) | 176 (36.82%) | 50 (38.46%) | 0.73b |
| Current smoker, n (%) | 79 (16.53%) | 22 (16.92%) | 0.91b |
| Site = Chicago, n (%) | 293 (61.30%) | 56 (43.08%) | 0.002b |
| Education < college degree, n (%) | 228 (47.70%) | 62 (47.69%) | 0.99b |
| Annual income < $50k, n (%) | 141 (29.50%) | 43 (33.08%) | 0.43b |
| Alcohol consumption > 2x/week, n (%) | 117 (24.48%) | 42 (32.31%) | 0.0057b |
| HT use, n (%) | 29 (6.07%) | 46 (35.38%) | <0.0001b |
| Menopausal status | |||
| Premenopausal, n (%) | 47 (9.83%) | 10 (12.20%) | 0.79b |
| Early peri-menopausal, n (%) | 233 (48.74%) | 38 (46.34%) | |
| Late peri-menopausal, n (%) | 53 (11.09%) | 7 (8.54%) | |
| Post-menopausal, n (%) | 145 (30.33%) | 27 (32.93%) | |
| Body measurements | |||
| Waist circumference, cm, mean (SD) | 89.4 (14.5) | 87.8 (13.3) | 0.26a |
| SBP, mm Hg, mean (SD) | 119.8 (16.7) | 118.6 (16.5) | 0.45a |
| HDL-C, mg/dL, mean (SD) | 57.4 (14.4) | 58.2 (14.6) | 0.57a |
| Total cholesterol, mg/dL, mean (SD) | 199.9 (37.7) | 206.7 (34.2) | 0.06a |
| LDL-C, mg/dL, mean (SD) | 118.7 (32.7) | 124.2 (32.2) | 0.09a |
| Triglycerides, mg/dL, median (Q1, Q3) | 98.0 (74, 137) | 106.5 (86.0, 147.0) | 0.06c |
| Glucose, mg/dL, median (Q1, Q3) | 88 (83, 96) | 89 (82.0, 95.0) | 0.78c |
| Estradiol, pg/mL, median (Q1, Q3) | 29.7 (16.2, 74.2) | 32.8 (15.9, 68.7) | 0.99c |
| Estradiol (quartiles) | |||
| Q1 (E2 < 16.15 pg/mL) | 121 (25.31%) | 24 (26.97%) | 0.65b |
| Q2 (16.15pg/mL ≤ E2 <29.67 pg/mL) | 122 (25.52%) | 18 (20.22%) | |
| Q3 (29.67 ≤ E2 <74.15 pg/mL) | 119 (24.90%) | 26 (29.21%) | |
| Q4 (E2 ≥ 74.15 pg/mL) | 120 (25.10%) | 21 (23.60%) | |
| CRP, mg/dL, median (Q1, Q3) | 1.98 (0.64, 5.39) | 1.60 (0.76, 5.66) | 0.35c |
| Outcome measures | |||
| AC, median (Q1, Q3) | 16 (0, 84) | 9 (0, 49) | 0.17c |
| CAC, median (Q1, Q3) | 0 (0, 6.52) | 1.03 (0, 7.9) | 0.41c |
| AC ≥ 100, n (%) | 112 (23.53%) | 13 (15.66%) | 0.12b |
| CAC ≥ 10, n (%) | 104 (21.76%) | 18 (21.69%) | 0.87b |
AC = aortic calcification; CAC = coronary artery calcification; CRP = C-reactive protein; E2 = estradiol; HDL-C = high-density lipoprotein cholesterol; HT = hormone therapy; LDL-C = low-density lipoprotein cholesterol; Q = quartile; SBP = systolic blood pressure
Indicates T-test was performed
Indicates chi-square test was performed
Indicates Wilcoxon rank-sum test was performed
AC and CAC were skewed with many 0 values. Median AC score was 16.5, and median CAC score was 0 among included participants. 112 women (23.53%) had AC ≥ 100, and 104 (21.76%) had CAC ≥10, with 55 (53%) of all women having both high AC and high CAC.
In unadjusted regression models, each 1-mg/dL higher in HDL-C level was associated with a 3% reduction in odds of AC presence (95% CI: 0.95 to 0.99) and a 4.0% reduction in odds of CAC presence (95% CI: 0.95 to 0.98) (Table 2). Estradiol was negatively associated with CAC but not AC presence.
Table 2:
Univariable association of AC and CAC with study variables
| Variable | AC | CAC | ||
|---|---|---|---|---|
| OR (95% CI)a | P value | OR (95% CI)a | P value | |
| Age, years | 1.10 (1.02 to 1.18) | 0.01 | 1.19 (1.10 to 1.29) | <.0001 |
| Race=Black (ref: White) | 1.21 (0.79 to 1.87) | 0.39 | 1.45 (0.92 to 2.27) | 0.11 |
| Study site=Pittsburgh (ref: Chicago) | 0.75 (0.49 to 1.15) | 0.18 | 0.99 (0.63 to 1.56) | 0.96 |
| Menopausal status=Late peri-/post- (ref: pre-/early peri) | 1.63 (1.06 to 2.50) | 0.03 | 1.76 (1.12 to 2.74) | 0.01 |
| Family History of CVD=Yes (ref: no) | 1.23 (0.77 to 1.96) | 0.38 | 1.00 (0.62 to 1.60) | 0.99 |
| Income < $50k/year (ref: ≥$50k/year) | 1.25 (0.79 to 1.97) | 0.34 | 1.00 (0.61 to 1.62) | 0.99 |
| Education < college (ref: ≥ some college) | 1.92 (1.25 to 2.96) | 0.003 | 1.42 (0.91 to 2.21) | 0.13 |
| Smoking=Yes (ref: non-smokers) | 3.13 (1.88 to 5.21) | <.0001 | 0.72 (0.38 to 1.37) | 0.32 |
| Alcohol consumption < 2 drinks per wk (ref: ≥ 2 drinks/ wk) | 1.31 (0.78 to 2.20) | 0.31 | 2.82 (1.48 to 5.37) | 0.002 |
| Waist circumference, cm | 1.06 (1.04 to 1.08) | <.0001 | 1.10 (1.08 to 1.12) | <.0001 |
| SBP, mm Hg | 1.02 (1.01 to 1.03) | 0.001 | 1.04 (1.02 to 1.05) | <.0001 |
| Total cholesterol, mg/dL | 1.01 (1.00 to 1.01) | 0.002 | 1.01 (1.00 to 1.01) | 0.01 |
| HDL-C, mg/dL | 0.97 (0.95 to 0.98) | 0.0002 | 0.96 (0.95 to 0.98) | 0.0002 |
| LDL-C, mg/dL | 1.01 (1.00 to 1.01) | 0.02 | 1.01 (1.00 to 1.01) | 0.02 |
| Log (glucose)b | 6.42 (1.92 to 21.53) | 0.003 | 9.91 (2.91 to 33.76) | 0.0002 |
| Log (triglycerides)b | 3.20 (2.06 to 4.99) | <.0001 | 2.52 (1.61 to 3.93) | <.0001 |
| Hormone therapy use = Yes (ref: no) | 0.89 (0.35 to 2.25) | 0.80 | 0.78 (0.29 to 2.09) | 0.62 |
| Log (CRP)a | 1.50 (1.23 to 1.81) | <.0001 | 1.49 (1.22 to 1.81) | <.0001 |
| Estradiol (Quartile) (ref: E2 ≥ 74.15 pg/mL) | ||||
| Q1 (E2 < 16.15 pg/mL) | 0.97 (0.51 to 1.85) | 0.92 | 2.36 (1.17 to 4.76) | 0.016 |
| Q2 (16.15pg/mL ≤ E2 <29.67pg/mL) | 1.86 (1.02 to 3.37) | 0.042 | 2.88 (1.45 to 5.72) | 0.003 |
| Q3 (29.67 ≤ E2 <74.15 pg/mL) | 1.41 (0.76 to 2.61) | 0.28 | 1.90 (0.93 to 3.87) | 0.08 |
| Log (E2)b | 0.89 (0.73 to 1.09) | 0.27 | 0.76 (0.62 to 0.94) | 0.01 |
AC = aortic calcification; CAC = coronary artery calcification; CRP = C-reactive protein; E2 = estradiol; HDL-C = high-density lipoprotein cholesterol; HT = hormone therapy; LDL-C = low-density lipoprotein cholesterol; Q = quartile; SBP = systolic blood pressure
Odds ratios of AC/CAC presence (AC ≥ 100, CAC ≥ 10) compared to no AC/CAC (AC < 100, CAC < 10) per 1-unit increase for continuous variables, or compared to the referent group for categorical variables.
Variables were found to violate normality assumptions. Therefore, these variables were log-transformed. Models were created using the transformed variables.
Without the interaction term, the multivariable adjusted models show no significant association of HDL-C with either AC or CAC (Table 3). However, the interaction between HDL-C and estradiol was significant in the unadjusted and the multivariable adjusted models for AC. Adjusting for cycle day, hormone use, age, race, site, waist circumference, smoking status, triglycerides, SBP, glucose, LDL-C, menopausal status, education, and alcohol consumption, the interaction between HDL-C and estradiol was significant (Table 3). The addition of CRP as a covariate in the model did not change the significance of the interaction term. To illustrate this interaction with respect to AC, a stratified analysis by estradiol quartiles was performed unadjusted (Figure 1-A) and then adjusted for CRP (Figure 1-B). Higher HDL-C was significantly associated with lower risk of AC presence at the highest estradiol quartile and tended to be associated with greater risk at the lowest estradiol quartile in the final model that did not adjust for CRP (Figure 1-A). Interestingly, after an additional adjustment for CRP, the positive association between HDL-C and risk of AC presence at the lowest estradiol quartile became significant (Figure 1-B).
Table 3:
Multivariable logistic regression of effect modification of estradiol on the association between HDL-C and calcification measures, per 1 mg/dL higher HDL-C in SWAN Heart participants
| Model 0 | Model 1 | Model 2 | Model 3 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | P | OR (95% CI) | P | OR (95% CI) | P | OR (95% CI) | P | ||
| AC ≥ 100 | HDL-C | 0.97 (0.95 to 0.99) | 0.0003 | 1.04 (0.98 to 1.101) | 0.17 | 1.00 (1.01 to 1.16) | 0.04 | 1.13 (1.03 to 1.25) | 0.02 |
| Log E2a | 0.901 (0.734 to 1.11) | 0.35 | 3.12 (1.145 to 8.49) | 0.03 | 3.51 (1.11 to 11.14) | 0.03 | 6.68 (1.36 to 32.87) | 0.02 | |
| HDL-C* log E2 | ------ | ----- | 0.98 (0.96 to 1.00)a | 0.02 | 0.98 (0.96 to 1.00)a | 0.03 | 0.97 (0.94 to 0.99) | 0.02 | |
| Log CRPa | ------ | ----- | ------ | ----- | ------ | ----- | 1.56 (1.21 to 2.01) | 0.001 | |
| CAC ≥ 10 | HDL-C | 0.97(0.95 to 0.98) | 0.0001 | 1.00 (0.94 to 1.06) | 0.94 | 1.04 (0.95 to 1.12) | 0.40 | 1.02 (0.92 to 1.13) | 0.68 |
| Log E2a | 0.77 (0.62 to 0.96) | 0.02 | 1.32 (0.47 to 3.69) | 0.60 | 1.61 (0.42 to 6.12) | 0.49 | 1.38 (0.27 to 7.05) | 0.70 | |
| HDL-C* log E2 | ------ | ----- | 0.99 (0.97 to 1.01) | 0.32 | 0.99 (0.97 to 1.01) | 0.33 | 0.99 (0.96, to 1.02) | 0.54 | |
| Log CRPa | ------ | ---- | ------ | ---- | ------ | ----- | 1.45 (1.12 to 1.89) | 0.01 | |
AC = aortic calcification; CAC = coronary artery calcification; CRP = C-reactive protein; E2 = estradiol; HDL-C = high-density lipoprotein cholesterol
Indicates an upper CI that was less than 1. However, due to rounding, these numbers are now listed as 1.00.
Model 0: unadjusted model
Model 1: main effects with interaction term, adjusted for cycle day.
Model 2: main effects with interaction term, adjusted for cycle day, hormone use, age, race, site, waist circumference, smoking, log triglycerides, SBP, log glucose, LDL-C, menopausal status, education, and alcohol consumption
Model 3: main effects with interaction term, adjusted for cycle day, hormone use, age, race, site, waist circumference (residuals), smoking, log triglycerides, SBP, log glucose, LDL-C, menopausal status, education, alcohol consumption, and CRP. Model 3 used a smaller sample size of 342 due to missing CRP data.
Figure 1-Aa, Bb:
Odds ratio of AC presence per 1 mg/dL higher HDL-C, by estradiol quartiles
a Model is adjusted for cycle day, hormone use, age, race, site, waist circumference, smoking status, triglycerides, SBP, glucose, LDL-C, menopausal status, education, and alcohol consumption.
b Model is adjusted for cycle day, hormone use, age, race, site, waist circumference (residuals), smoking status, triglycerides, SBP, glucose, LDL-C, menopausal status, education, alcohol consumption, and CRP.
The interaction between HDL-C and estradiol was not significant for CAC in any model (Table 3). The stratified analysis showed no association between HDL-C and the odds of CAC at any estradiol level before (Figure 2-A) or after (Figure 2-B) adjusting for CRP, although the CAC results show a pattern similar to the AC analysis.
Figure 2-Aa, Bb:
Odds ratio of CAC presence per 1 mg/dL higher HDL-C, by estradiol quartiles
a Model is adjusted for cycle day, hormone use, age, race, site, waist circumference, SBP, triglycerides, glucose, alcohol consumption, menopausal status, LDL-C, education, smoking status.
b Model is adjusted for cycle day, hormone use, age, race, site, waist circumference (residuals), smoking status, triglycerides, SBP, glucose, LDL-C, menopausal status, education, alcohol consumption, and CRP.
Similar methods were used to assess the interaction between HDL-C and CRP with respect to AC and CAC. None of these interactions were significant. These results are presented as in Supplemental Table 1.
Results from the sensitivity analyses are consistent with those from the main analysis (Supplemental Table 2). With all groups examined, the interaction between HDL-C and estradiol was significant for AC, but not for CAC.
4. Discussion
In a large sample of midlife women, HDL-C was negatively associated with presence of AC and CAC. The associations between HDL-C and the risk of AC presence differed by estradiol concentration, controlling for confounders such as age, race, study site, waist circumference, and systolic blood pressure. In contrast, estradiol level did not impact the association between HDL-C and CAC presence. The protective association of HDL-C on AC was only evident at higher levels of estradiol, as suggested by our stratification analyses showing higher HDL-C to be associated with lower AC risk at the highest estradiol quartile but with higher risk at the lowest estradiol quartile after accounting for CRP. Our findings of a stronger protective association of HDL-C on AC at higher levels of estradiol was also evident in sensitivity analyses with samples defined by different inclusion criteria. The consistency of these results suggests that estradiol level may contribute to the association of HDL-C on AC presence.
The lack of interaction between CRP and HDL-C for both AC and CAC suggests that inflammation does not influence the association of HDL-C on coronary outcomes, although inflammation shows a strong independent association on AC and CAC.
The current results are in agreement with previous research from SWAN suggesting a reversal in the association of HDL-C with subclinical CVD over the menopause transition, in that estradiol concentration, which is related to menopause, modified the association between HDL-C and AC presence [10, 11]. A cross-sectional analysis investigating effect modification of menopausal status on the association between HDL-C and calcification found that, among SWAN Heart participants, menopausal status significantly modified the association between HDL-C and AC, but not CAC [10]. In particular, pre-/early perimenopausal women showed lower odds of AC with high HDL-C, but this association was reversed in late peri-/postmenopausal women [10]. Since estradiol concentrations fall during the menopause transition, our findings of different associations of higher HDL-C with AC presence by estradiol level support past SWAN work and gives new evidence of one potential pathway through which HDL-C might show different associations with aortic calcium score. The combination of past SWAN work as well as the present study gives some evidence that estradiol could contribute to the association of HDL-C on cardiovascular disease. However, due to the nature of this research, we cannot determine whether the reversal of this association is permanent, or if it is only present during the time surrounding the menopause transition. To answer that question would require a longitudinal analysis, which is beyond the scope of the research we have performed.
Our findings could potentially be explained by reduced cholesterol efflux as a result of changes in HDL particle size, which is associated with low estradiol concentrations [13, 19, 20]. Previous research suggests that low estradiol concentrations may impact HDL particles in various ways. A 2014 clinical trial reported that, under high estradiol concentrations, apolipoprotein A (apoA) levels, a major component of HDL particles, increased by 8%, despite similar levels of HDL-C [20]. ApoA reduces oxidation and assists in reverse cholesterol transport [19]. It is possible that lower apoA levels as a result of a decrease in estradiol could therefore affect HDL’s cholesterol-carrying ability and lead to an increased risk of atherosclerosis [20, 21].
Finding effect modification by estradiol level on the association of HDL-C with AC but not CAC may be explained by differences in pathology of calcification. Calcification may occur in the aorta before the coronary artery [22]. In women over 50, half of women showed distal aortic calcification, whereas 31% showed coronary artery calcification, and increasing age was associated with higher odds of calcification in the distal and proximal aorta compared to calcification in the coronary artery [22]. Therefore, calcification in the aorta may develop earlier than in the coronary artery, or atherosclerosis may occur faster in the aorta than the coronary artery, which may explain why an interaction was observed between HDL-C and estradiol for AC but not for CAC. However, almost equal proportion of our study participants had AC and CAC presence in this analysis, suggesting other potential contributors. Additional research by these same authors indicates that differences in variables associated with AC versus variables associated with CAC could partially explain some of the differences between these measures in our study. For example, researchers reported that a history of hypertension was more strongly associated with CAC than AC in women, with point estimates of 17.5 and 4.7, respectively, and asserted that age was more strongly associated with AC than CAC [23]. These results suggested underlying differences in the pathologies associated with calcification. Interestingly, there is inconclusive evidence of the association between cholesterol efflux and calcification measures [24, 25]. A decrease in cholesterol efflux may more negatively impact AC than CAC, which could explain difference in the observed effect modification in our analysis. Future research may give insights into this conundrum.
While our analysis only included a portion of SWAN Heart participants, we observed few differences in characteristics of included and excluded participants. One difference of note is the significant difference in the proportion of women who reported alcohol consumption in the included and excluded groups. Alcohol consumption is thought to be negatively associated with risk of CVD and positively associated with HDL-C [26]. Our sample included a lower proportion of women who reported drinking two or more drinks per week compared to the women who were excluded from the analysis. However, the removal of the alcohol consumption variable from our models did not seem to impact the association of estradiol and HDL-C. Despite the difference, we believe that these results are generalizable to midlife women with similar characteristics. Additionally, the sensitivity analyses performed alongside the main analysis indicate that these results are robust and have high internal validity. This study is limited, however, in that it is cross-sectional, which prevents us from assessing temporality and causality. Interestingly, with a smaller sample size and controlling for CRP, HDL-C was negatively associated with risk of AC presence at high estradiol levels and positively associated with risk of AC presence at low estradiol levels. However, these results were based on a smaller sample size, so replication of these results would be necessary. While we observed effect modification of estradiol on the association of HDL-C with AC, residual confounding may influence the degree of this observed association. Additionally, the use of immunoassay for measurement of estradiol concentration is less than ideal for women with low estradiol concentrations. The use of mass spectrometry is considered the gold standard, and SWAN researchers are working on calibrating immunoassay measurements to the gold standard. Current research suggests that the immunoassay used by SWAN shows strong correlation for estradiol measurements over 15 pg/mL [27]. Our results indicate that the association between HDL-C and CVD risk is weaker when estradiol levels are above this threshold, thus falling within the range that is shown to be strongly correlated with mass spectrometry measures.
The reported results should be used as guidance for future studies, as this work is hypothesis-generating. To better understand this effect modification of estradiol on the relationship between HDL-C and subclinical CVD, future research could address this question in a longitudinal setting. Additionally, understanding how estradiol impacts novel metrics of HDL, such as HDL particle concentration or size as well as particle proteomics and lipidomics, may shed light on the biological mechanism for the association we observed.
Current clinical guidelines recommend risk estimation for individuals between the ages of 40 and 75 using the Pooled Cohort Equation, which predicts lower 10-year risk of atherosclerotic cardiovascular disease (ASCVD) with increasing HDL-C [28]. Our results indicate that this equation may underestimate that risk for women who have transitioned through menopause, even if this reversal is only temporary. Recommendations to raise HDL-C concentration to decrease risk of CVD may have the opposite of the desired effect in women who have recently reached postmenopause with low estradiol concentrations, which could be interesting to test in a clinical trial. Current recommendations suggest additional risk assessment for individuals with intermediate (7.5% to 20%) 10-year risk of ASCVD, such as the use of CT scans to measure CAC [28]. Future studies could give insight into whether these recommendations should be refined to assess CAC in more situations. It should be noted that these guidelines only mention CAC screening for ASCVD risk, not AC screening, which we found more relevant in our study to women at midlife. Future studies should also assess effect modification of exogenous estradiol on associations between AC and HDL-C in women. While this association was not observed with regards to CAC, more research could help understand differences in calcification patterns in midlife women.
5. Conclusions
HDL-C may not be cardioprotective in midlife women with respect to aortic calcification. The degree to which HDL-C is negatively associated with risk of CVD may vary based on estradiol concentration in women. With higher levels of estradiol, HDL-C was associated with lower risk of aortic calcification. At lower estradiol levels, however, higher HDL-C was associated with a higher risk in this study when adjusting for CRP. These results confirm the reversal in protective association of HDL-C demonstrated in previous literature [10–13]. While this association was only seen with respect to aortic calcification, the implications may span to other subclinical markers of CVD as well. This work underlines the limited information about conventional HDL-C measures and highlights the need for improvement in novel metrics of HDL which may serve as better predictors of CVD risk in certain populations, including midlife women.
Supplementary Material
Supplemental Table 1: Multivariate logistic regression of effect modification of CRP level on association between HDL-C and calcification measures, per 1 mg/dL higher HDL-C in SWAN Heart participants.
Supplemental Table 2: Sensitivity analyses based on exclusion of observations with estradiol levels less than the lower limit of detection.
Acknowledgments:
Clinical Centers: University of Michigan, Ann Arbor - Siobán Harlow, PI 2011 - present, MaryFran Sowers, PI 1994-2011; Massachusetts General Hospital, Boston, MA - Joel Finkelstein, PI 1999 - present; Robert Neer, PI 1994 - 1999; Rush University, Rush University Medical Center, Chicago, IL - Howard Kravitz, PI 2009 - present; Lynda Powell, PI 1994 - 2009; University of California, Davis/Kaiser - Ellen Gold, PI; University of California, Los Angeles - Gail Greendale, PI; Albert Einstein College of Medicine, Bronx, NY - Carol Derby, PI 2011 - present, Rachel Wildman, PI 2010 - 2011; Nanette Santoro, PI 2004 - 2010; University of Medicine and Dentistry - New Jersey Medical School, Newark - Gerson Weiss, PI 1994 - 2004; and the University of Pittsburgh, Pittsburgh, PA - Karen Matthews, PI.
NIH Program Office: National Institute on Aging, Bethesda, MD - Chhanda Dutta 2016- present; Winifred Rossi 2012-2016; Sherry Sherman 1994 - 2012; Marcia Ory 1994 - 2001; National Institute of Nursing Research, Bethesda, MD - Program Officers.
Central Laboratory: University of Michigan, Ann Arbor - Daniel McConnell (Central Ligand Assay Satellite Services).
Coordinating Center: University of Pittsburgh, Pittsburgh, PA - Maria Mori Brooks, PI 2012 - present; Kim Sutton-Tyrrell, PI 2001 - 2012; New England Research Institutes, Watertown, MA - Sonja McKinlay, PI 1995 - 2001.
Steering Committee: Susan Johnson, Current Chair; Chris Gallagher, Former Chair
We thank the study staff at each site and all the women who participated in SWAN.
Funding:
The Study of Women’s Health Across the Nation (SWAN) has grant support from the National Institutes of Health (NIH), DHHS, through the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR) and the NIH Office of Research on Women’s Health (ORWH) (Grants U01NR004061; U01AG012505, U01AG012535, U01AG012531, U01AG012539, U01AG012546, U01AG012553, U01AG012554, U01AG012495). SWAN Heart was supported by the National Heart, Lung, and Blood Institute (grants HL065581, HL065591). The Study of Women’s Health Across the Nation (SWAN) HDL ancillary study has grant support from National Institute on Aging (NIA) AG058690. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the NIA, NINR, ORWH or the NIH.
Footnotes
Conflicts of Interest:
The authors have no conflicts of interest.
References
- 1.World Health Organization (WHO). The top 10 causes of death. Available online: https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death (accessed on 10 January 2020)
- 2.Centers for Disease Control and Prevention, National Center for Health Statistics. Underlying Causes of Death 1996–2016 on CDC WONDER Online Database, released December 2017. Available online: https://wonder.cdc.gov/wonder/help/ucd.html (accessed on 10 January 2020)
- 3.El Khoudary SR, Thurston RC. Cardiovascular implications of the menopause transition: Endogenous sex hormones and vasomotor symptoms. Obstet Gynecol Clin North Am. 2018, 45(4): 641–661. DOI: 10.1016/j.ogc.2018.07.006. [DOI] [PubMed] [Google Scholar]
- 4.El Khoudary SR, Greendale G, Crawford SL, et al. The menopause transition and women’s health at midlife: a progress report from the Study of Women’s Health Across the Nation (SWAN). Menopause. 2019. 26(10): 1213–1227. DOI: 10.1097/GME.0000000000001424 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Merz CNB, Johnson BD, Sharaf BL, et al. Hypoestrogenemia of hypothalamic origin and coronary artery disease in premonpausal women: a report from the NHLBI-sponsored WISE study. J. of the Am. Coll. Cardiol 2003. 41(3): 413–419. DOI: 10.1016/S0735-1097(02)02763-8. [DOI] [PubMed] [Google Scholar]
- 6.El Khoudary SR, Santoro N, Chen HY, et al. Trajectories of estradiol and follicle-stimulating hormone over the menopause transition and early markers of atherosclerosis after menopause. Eur J Prev Cardiol. 2016. 23(7):694–703. DOI: 10.1177/2047487315607044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Matthews KA, El Khoudary SR, Brooks MM, et al. Lipid changes around the final menstrual period predict carotid subclinical disease in postmenopausal women. Stroke. 2017. 48(1): 70–76. DOI: 10.1161.STROKEAHA.116.014743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.El Khoudary SR. HDL and the menopause. Curr Opin Lipidol. 2017. 28(4): 328–336. DOI: 10.1097/MOL.0000000000000432. [DOI] [PubMed] [Google Scholar]
- 9.Derby CA, Crawford SL, Pasternak RC, Sowers MF, Sternfeld B, Matthews KA. Lipid changes during the menopause transition in relation to age and weight: The Study of Women’s Health Across the Nation. Am J Epidemiol. 2009. 169(11):1352–1361. DOI: 10.1093/aje/kwp043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Woodard GA, Brooks MM, Barinas-Mitchell E, Mackey RH, Matthews KA, Sutton-Tyrrell K. Lipids, menopause, and early atherosclerosis in SWAN Heart Women. Menopause. 2011. 18 (4): 376–384. DOI: 10.1097/gme.0b013e3181f6480e. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.El Khoudary SR, Wang L, Brooks MM, Thurston RC, Derby CA, Matthews KA. Increase in HDL-C level over the menopausal transition is associated with greater atherosclerotic progression. J Clin Lipidol. 2016. 10(4): 962–969. DOI: 10.1016/j.jacl.2016.04.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.El Khoudary SR, Hutchins PM, Matthews KA, et al. Cholesterol efflux capacity and subclasses of HDL particles in healthy women transitioning through menopause. J Clin Endocrinol Metab. 2016. 101(9): 3419–3428. DOI: 10.1210/jc.2016-2144. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.El Khoudary SR, Ceponiene I, Samargandy S, et al. HDL (High-Density Lipoprotein) metrics and atherosclerotic risk in women. Arter. Thromb. Vasc. Biol 2018. 38 (9): 2236–2244. DOI: 10.1161/ATVBAHA.118.311017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Koka S, Petro TM, Reinhardt RA. Estrogen inhibits interleukin 1-beta-induced interleukin-6 production by human osteoblast-like cells. J Interferon Cytokine Res. 1998. 18 (7): 479–483. DOI: 10.1089/jir.1998.18.479 [DOI] [PubMed] [Google Scholar]
- 15.Galien R, Garcia T. Estrogen receptor impairs interleukin-6 expression by preventing protein binding on the NF-kappaB site. Nucleic Acids Res. 1997. 25 (12): 2424–2429. DOI: 10.1093/nar/25.12.2424 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Sowers MF, Sternfeld B, Morganstein D, et al. Design, survey, sampling and recruitment methods of SWAN: a multi-center, multi-ethnic, community based cohort study of women and the menopausal transition. Menopause: biology and pathobiology. 2000. 175–188. [Google Scholar]
- 17.Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viatmonte M, Detrano R. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol. 1990. 15: 827–832. DOI: 10.1016/0735-1097(90)90282-t [DOI] [PubMed] [Google Scholar]
- 18.Janssen I, Powell LH, Matthews KA, et al. Relation of persistent depressive symptoms to coronary artery calcification in women aged 46 to 59 years. Am J Cardiol. 2016. 117(12): 1884–1889. DOI: 10.1016/j.amjcard.2016.03.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Agatisa PK, Matthews KA, Bromberger JT, Edmundowicz D, Chang YF, Sutton-Tyrrell K. Coronary and aortic calcification in women with a history of major depression. Arch Intern Med. 2005. 165(11): 1229–36. DOI: 10.1001/archinte.165.11.1229 [DOI] [PubMed] [Google Scholar]
- 20.Persson L, Henriksson P, Westerlund E, Hovatta O, Angelin B, Rudling M. Endogenous estrogens lower plasma PCSK9 and LDL cholesterol but not Lp(a) or bile acid synthesis in women. Arterioscler Thromb Vasc Biol. 2012; 32: 810–814. DOI: 10.1161/ATVBAHA.111.242461. [DOI] [PubMed] [Google Scholar]
- 21.Rader DJ, Hovingh GK. HDL and cardiovascular disease. The Lancet. 2014. 384: 618–625. DOI: 10.1016/S1040-6736(14)61217-4. [DOI] [PubMed] [Google Scholar]
- 22.Criqui MH, Kamineni A, Allison MA, et al. Risk factor differences for aortic vs. coronary calcified atherosclerosis: The Multi-Ethnic Study of Atherosclerosis. Arterioscler Thromb Vasc Biol. 2010. 30(11):2289–2296. DOI: 10.1161/ATVBAHA.110.208.181 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Allison MA, Criqui MH, Wright CM. Patterns and risk factors for systemic calcified atherosclerosis. Arterioscler Thromb Vasc Biol. 2004. 24:331–336) [DOI] [PubMed] [Google Scholar]
- 24.Mody P, Joshi PH, Khera A, Ayers CR, Rohatgi A. Beyond coronary calcification, family history, and C-reactive protein: Cholesterol efflux capacity and cardiovascular risk prediction. J Am Coll Cardiol. 2016. 67(21): 2480–2487. DOI: 10.1016/j.jacc.2016.03.0538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zimetti F, Freitas WM, Campos AM, et al. Colesterol efflux capacity does not associate with coronary calcium, plaque vulnerability, and telomere length in healthy octogenarians. J Lipid Research. 2018. 59: 714–721. DOI: 10.1194/jlr.P079525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hansel B, Thomas F, Pannier B, et al. Relationship between alcohol intake, health and social status, and cardiovascular risk factors in the urban Paris-Ile-de-France Cohort: is the cardioprotective action of alcohol a myth? Eur. J. Clin. Nutr 2010. 64(6): 561–568. DOI: 10.1038/ejcn.2010.61 [DOI] [PubMed] [Google Scholar]
- 27.Santoro N, Auchus R, Greendale G, et al. SAT-027 Comparison of Estradiol by Mass Spectrometry Versus Immunoassay in Women Undergoing Menopause: Study of Womens Health Across the Nation (SWAN). Journal of the Endocrine Society. 2020. 4(S_1): SAT–027. DOI: 10.1210/jendso/bvaa046.446 [DOI] [Google Scholar]
- 28.Arnett DK, Blumenthal RS, Albert MA, et al. 2019. ACC/AHA guideline on the primary prevention of cardiovascular disease: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. J Am Coll Cardiol. 2019 74:e177–232 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Table 1: Multivariate logistic regression of effect modification of CRP level on association between HDL-C and calcification measures, per 1 mg/dL higher HDL-C in SWAN Heart participants.
Supplemental Table 2: Sensitivity analyses based on exclusion of observations with estradiol levels less than the lower limit of detection.