The Independent Associations of Anti-Müllerian Hormone and Estradiol Levels over the Menopause Transition with Lipids/lipoproteins: The Study of Women’s Health Across the Nation

Abstract

Background:

The menopause transition (MT) is linked to adverse changes in lipids/lipoproteins. However, the related contributions of Anti-Müllerian hormone (AMH) and estradiol (E2) are not clear.

Objective:

To evaluate the independent associations of premenopausal AMH and E2 levels and their changes with lipids/lipoproteins levels [total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), apolipoprotein B (apoB) and apolipoprotein A-1 (apoA-1)] over the MT.

Methods:

SWAN participants who transitioned to menopause without exogenous hormone use, hysterectomy, or bilateral oophorectomy with data available on both exposure and outcomes when they were premenopausal until the 1st visit postmenopausal were studied.

Results:

The study included 1,440 women (baseline-age:mean±SD=47.4±2.6) with data available from up to 9 visits (1997–2013). Lower premenopausal levels and greater declines in AMH were independently associated with greater TC and HDL-C, whereas lower premenopausal levels and greater declines in E2 were independently associated with greater TG and apo B and lower HDL-C. Greater declines in AMH were independently associated with greater apoA-1, and greater declines in E2 were independently associated with greater TC and LDL-C.

Conclusions:

AMH and E2 and their changes over the MT relate differently to lipids/lipoproteins profile in women during midlife. Lower premenopausal and/or greater declines in E2 over the MT were associated with an atherogenic lipid/lipoprotein profile. On the other hand, lower premenopausal AMH and/or greater declines in AMH over the MT were linked to higher apo A-1 and HDL-C; the later found previously to be related to a greater atherosclerotic risk after menopause.

INTRODUCTION

The menopause transition (MT) is a critical stage for cardiovascular disease (CVD) risk acceleration. ¹ As women traverse the MT, they experience a reduction in ovarian reserve and adverse changes in multiple CVD risk factors such as lipids and lipoproteins. Within 1 year of the final menstrual period (FMP), low density lipoprotein-cholesterol (LDL-C) and apolipoprotein B (apo B) rise sharply. ^2,3 This non-linear association between lipids/lipoproteins and time relative to FMP supports a critical contribution of ovarian aging to pro-atherogenic lipid/lipoprotein changes during the MT.

The significant decline in estradiol (E2) over the MT may contribute to the adverse changes in LDL-C and apo B. Many clinical trials of hormone therapy use found that estrogens lowered LDL-C and raised high density lipoprotein-cholesterol (HDL-C). ⁴ However, observational studies linking endogenous estradiol level with lipid/lipoprotein levels have not shown similar associations. ^5–11 A key limitation is that all previous observational studies were cross-sectional in nature, and thus did not capture dynamic changes in E2 on lipids/lipoproteins over the MT. Another possibility is that E2 may not be the only MT-related hormone contributing to adverse changes in lipids/lipoproteins.

Anti-Müllerian hormone (AMH), which declines progressively with age, is a widely used clinical marker of ovarian aging. Using a new, ultrasensitive AMH assay (picoAMH ELISA, Ansh LABS, Webster, TX), AMH can predict the FMP within a window of 12 to 24 months in late-reproductive aged women.¹² Compared to E2, AMH levels measured through a full menstrual cycle did not show consistent fluctuation patterns ¹³, making it a cycle-independent marker for ovarian reserve. AMH level has been linked to lipids/lipoproteins in either reproductive-aged women or women with comorbidities. ^14–21 However, results from these studies were not consistent. Notably, none of these studies assessed changes in AMH as related to lipid/lipoprotein levels in women transitioning through menopause.

Considering the changes in AMH, E2, and lipids/lipoproteins over the MT, we aimed to evaluate the associations of premenopausal AMH and E2 levels and their changes since baseline with lipids (total cholesterol (TC), triglyceride (TG), LDL-C, and HDL-C) and apolipoprotein (apo B and apo A-1) levels during midlife. Moreover, we assessed the independent associations of AMH and E2 with lipids/lipoproteins. We hypothesized that AMH level and AMH changes since baseline would be associated with levels of lipids/lipoproteins independently of level and changes in E2 in midlife women traversing the MT.

MATERIALS AND METHODS

Participants

The Study of Women’s Health Across the Nation (SWAN) is an ongoing multi-racial/ethnic longitudinal study of the MT across 7 study sites in the U.S. (Boston, MA; Detroit, MI; Oakland, CA; Los Angeles, CA; Pittsburgh, PA; Chicago, IL; or Newark, NJ). SWAN eligibility criteria included: 1) having a uterus and at least one intact ovary, 2) having a menstrual period within the past 3 months of enrollment, 3) not being pregnant, and 4) not having taken hormone therapy in the last 3 months of enrollment. Between 1996 and 1997, 3302 women aged 42–52 years old from the 7 sites completed the baseline visit (v00). Since baseline, SWAN participants have completed up to 16 follow-up visits (spanning from 1997 to 2018). The SWAN AMH ancillary study measured AMH level over a maximum of 10 time points between v00 and v13 (1997–2013) among 1,536 SWAN participants who had their final menstrual period (FMP) without having a hysterectomy, bilateral ovariectomy, or taking hormone therapy; and had at least 1 blood sample available while pre- (had menstrual bleeding in the previous 3 months with no change in cycle predictability in the past year) or early perimenopausal (had menstrual bleeding in the previous 3 months with a decrease in cycle regularity in the past year).

For the current analysis, 1,530 SWAN AMH participants (6679 observations) with at least one lipid or apolipoprotein measurement assessed concurrently with the AMH measures were considered. Women whose menopause status at the first available time point was not pre- or early perimenopause (n=18), those who were on lipid-lowering medication during the study (n=27), and women with missing data or extreme values on study covariates (n=45) were excluded, while observations at which women were taking hormone therapy were dropped, resulting in 1,440 women (5,863 observations) in the final analytical sample. The included women had a better overall health condition as indicated by more favorable levels of multiple clinical measures including lower body mass index, fewer comorbidities, and better lipids compared to the excluded women.

Research protocols were approved by the institutional review board at each study site and all the participants provided a written informed consent prior to enrollment.

Study Measures

At each study visit, a fasting blood draw was scheduled early in the morning on menstrual cycle days 2 through 5 to measure AMH, E2, and lipids/lipoproteins. If a timed blood sample could not be obtained at any visit, typically due to menstrual cycle irregularity accompanying progress through the MT, a fasting blood sample was collected within 90 days of that visit. Accordingly, cycle day of blood draw was reported as either days 2–5 or outside that period. Serum was processed promptly, frozen, and stored at −80°C until thawed for measurements.

Lipids and apolipoproteins

At visits 00, 01, and 03–07, lipids and apolipoproteins were measured at the Medical Research Laboratory, Lexington KY. At visit 12 and 13, lipids were measured at the University of Michigan Pathology Laboratory, Ann Arbor, MI, and apolipoproteins were measured at the University of Pittsburgh Heinz Laboratory, Pittsburgh, PA. HDL-C was isolated based upon the method of Izawa et al.²² TC and TG were determined by coupled enzymatic methods. LDL-C was calculated by the Friedewald equation when triglycerides were <400 mg/dL ²³ or set to missing when triglycerides were >400 mg/dL. Apo A-1 and apo B were measured by immunonephelometry.

Because different laboratories performed the lipids/lipoproteins assays over time, calibration analyses were conducted to convert the University of Michigan and University of Pittsburgh results to values comparable to those from the Medical Research Laboratory. In brief, a random sample (340 samples for TC, LDL-C and TG; 100 samples for apo A-1 and apo B) was drawn across the full range of values, with checks for the distribution of the selected sample by menopause status, race/ethnicity, and study visit to assure adequate representation of the full cohort. Based on these analyses, correction factors were applied.

Hormone assays

Estradiol (E2) was measured at the Clinical Ligand Assay Satellite Services Central Laboratory, University of Michigan using a modified, off-line Automated Chemiluminescence System: ACS-180 (E2–6) (Bayer Diagnostics Corp). The lower limit of detection (LLD) was between 1 and 7 pg/mL. Only 11 observations (0.2% of all included samples) were below LLD. A random number between 0 and LLD was used to fill in missing for those 11 observations. The inter- and intra-assay coefficients of variation were 10.6 and 6.4%, respectively, at an E2 level of 50 pg/mL. ²⁴

AMH was measured at Ansh Labs under the joint supervision of the Ansh Laboratory Director and the Director of Special Chemistry, Clinical Pathology Core Laboratory at Massachusetts General Hospital using picoAMH ELISA, a quantitative three-step sandwich type immunoassay with an analytical measurable range of 2.0–11,000 pg/mL and LLD of 1 pg/mL. The assay employs recombinant pro-mature human AMH (BA 047, Ansh labs, USA) as a calibrator.

Study covariates

Self-reported race/ethnicity and education level (categorized as 1) college or post college degree vs. 2) less than high school or high school or some college for this analysis) were collected at the SWAN screening interview. Menopause status was categorized based on bleeding status into: 1) premenopausal: bleeding in the past 3 months without menstrual cycle changes during the past year; 2) early perimenopause: bleeding in the past 3 months with some menstrual cycle changes during the past year; 3) late perimenopause: bleeding at least once between 3 to 12 months prior to the current visit; 4) natural postmenopause: no bleeding in the last 12 months prior to the current visit. Age was the difference between date of birth and the interview completion date of each visit. Smokers were defined as women who currently smoked cigarettes or who had ever smoked a total of at least 20 packs of cigarettes over the life span or at least one cigarette per day for at least 1 year (never vs. current or past smoking). Systolic blood pressure (SBP) was the average of two continuous blood pressure measures assessed with at least a 2-minute interval. Body mass index (BMI) was calculated by dividing weight (kg) by height squared (m²), measured after removing participants’ shoes. Physical activity score was calculated as the sum of active living index (incorporating time spent in watching television and daily walking), household/childcare activity index (incorporating time spent in housework and childcare), and sports index (incorporating intensity of and time spent in sport and exercise). ²⁵ Anti-hypertensive medications included alpha blockers, beta blockers, calcium channel blockers, ace inhibitors, angiotensin receptor blockers, and diuretics. Anti-diabetic medications included metformin, sulfonylureas, meglitinide, thiazolidinediones, DPP-IV inhibitors, incretins, insulins, and other hypoglycemics. SWAN participants were considered diabetic by meeting any of the following criteria: 1) used ani-diabetic medication at any visit, 2) had a fasting glucose≥126 mg/dL at 2 consecutive visits or 3 attended visits, or 3) had at 2 visits with self-report diabetes and at least 2 visits with fasting glucose≥126 mg/dL. Cardiovascular events included self-reported heart attack or stroke. All included covariates were collected concurrently at the same time lipids/lipoproteins and hormones were measured.

Data Analysis

Normality of each continuous variable was examined. We have reported median, 25^th, and 75^th percentiles for non- normally distributed variables. AMH, E2, and TG were natural log transformed in all models. Changes in AMH (AMH_c) and E2 (E2_c) since baseline (the first available visit when women were either pre- or early perimenopausal) were calculated as the difference between log transformed AMH and E2 at each subsequent visit and log transformed AMH (AMH₀) and E2 (E2₀) at baseline. T tests and Chi square tests were applied to compare distributions of continous and categorical variables measured at baseline between included and excluded women.

Linear mixed effect modeling with random intercept was used to assesss associations of premenopausal (AMH₀, E2₀) and time-varying change since baseline in AMH (AMH_c) and E2 (E2_c) with each lipid/lipoprotein measures over time. Hormone values at the first available visit (baseline) and changes in hormone values since baseline (changes in AMH and E2 over the MT) were tested in relation to repeated measures of each lipid/lipoproteins over time separately and jointly (AMH and E2 measures in the same models) to assess independent associations of AMH vs. E2 baseline/changes with lipids/lipoproteins. Covariates were selected in an a priori fashion and included study site, race/ethnicity, education level, and time varying: age, menopause status, BMI, physical activity score, smoking status, and cycle day of blood draw (included in both univariable and multivariable models that included E2) and ever reported a cardiovascular event (stroke or heart attack). Since lipids/lipoproteins were quantified at different laboratories, we conducted sensitivity analysis by assessing the above associations after removing records measured at visit 12 and 13. Statistical analyses were carried out using SAS software 9.4. All P-values are two sided, and p<0.05 was considered to be statistically significant.

RESULTS

At the first available visit, women on average were 47.41(2.3) years old, pre- or early perimenopause, with 44.1% White, 28.9% Black, 11.4% Japanese, 9.9% Chinese and 5.6% Hispanic. None of the study participants were on lipid lowering medication. Summary statistics of clinical factors are presented in Table 1. As expected, both AMH and E2 declined since baseline with median (Q1, Q3) of −56.4(−362.1, 0) pg/ml and 0(−26.05, 7.05) pg/ml, respectively].

Table 1.

Participants Characteristics at baseline (first available visit with AMH measure)

Characteristics	n=1440
Age, mean (SD), year	47.4 (2.6)
Race, n (%)
Black	417 (29.0%)
White	635 (44.1%)
Chinese	143 (10.0%)
Japanese	164 (11.4%)
Hispanic	81 (5.6%)
Education, n (%)
<= High school/some college	791 (54.9%)
College degree/post college	649 (45.1%)
Menopause status, n (%)
Premenopausal	448 (31.1%)
Early perimenopausal	991 (68.9%)
BMI, mean (SD), kg/m2	28.2 (7.3)
Physical activity score, mean (SD)	7.7 (1.8)
Diabetes status, n (%)
No	1357 (94.2%)
Yes	83 (5.8%)
Smoking Status, n (%)
Never	830 (57.7%)
Past Only	399 (27.7%)
Current	210 (14.6%)
Current use of antihypertensive medications, n (%)	213 (14.8%)
Current use of antidiabetic medications, n (%)	44 (3.1%)
Ever had stroke or heart attack, n (%)	7 (0.5%)
E2, pg/mL, median (Q1, Q3) a	49.9 (29.7, 88.7)
AMH, pg/mL, median (Q1, Q3) a	136.4 (15.3, 470.9)
Total cholesterol, mean (SD), mg/dL	195.9 (34.7)
Triglycerides, median (Q1, Q3), mg/dL	94.0 (69.0, 134.0)
LDL-C, mean (SD), mg/dL	114.8 (31.3)
apo B, mean (SD), mg/dL	106.0 (27.7)
HDL-C, mean (SD), mg/dL	58.2 (14.8)
apo A-1, mean (SD), mg/dL	152.0 (28.2)

^aTo convert E2 from pg/mL to pmol/L, multiply by 3.671; AMH from pg/mL to pmol/L, multiply by 0.0071

Abbreviations: AMH: Anti-Müllerian hormone; apo A1: apolipoprotein A1; apo B: apolipoprotein B; BMI: body mass index; E2: estradiol; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol

Associations of baseline AMH (AMH₀) and change in AMH since baseline (AMH_c) with lipids/lipoproteins- E2 measures not included in models

In unadjusted models, negative associations were reported between each of AMH₀ and AMH_C and levels of TC, TG, LDL-C, apo B, HDL-C and apo A-1 over time (Table 2).

Table 2.

Associations ^a of premenopausal AMH and E2 levels ^b and their changes over the menopause transition with lipids/lipoproteins

	Total Cholesterol mg/dL		Log Triglyceridec		LDL-C mg/dL		Apo B mg/dL		HDL-C mg/dL		Apo A-1 mg/dL
	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value
AMH Only
Unadjusted model
AMH0	−5.76(0.81)	<.0001	−0.04(0.01	0.02	−3.91(0.75)	<.0001	−3.3(0.66	<.0001	−0.95(0.36)	0.009	−1.56(0.62)	0.01
AMHc	−4.80(0.29)	<.0001	−0.01(0.004)	0.0006	−3.49(0.25)	<.0001	−1.8(0.23)	<.0001	−0.90(0.10)	<.0001	−3.87(0.31)	<.0001
Adjusted model d
AMH0	−3.73(0.84)	<.0001	−0.003(0.01)	0.75	−2.09(0.77)	0.007	−1.61(0.67)	0.01	−1.39(0.34)	<.0001	−0.11(0.63)	0.86
AMHc	−1.49(0.35)	<.0001	0.01(0.01)	0.26	−1.23(0.31)	<.0001	−0.50(0.27)	0.06	−0.22(0.13)	0.09	−1.11(0.35)	0.001
E2 Only
Unadjusted model e
E20	−2.72(0.80)	0.0007	−0.05(0.01)	<.0001	−2.27(0.74)	0.002	−2.66(0.65)	<.0001	1.10(0.36)	0.002	1.09(0.62)	0.08
E2c	−2.67(0.25)	<.0001	−0.01(0.004)	0.004	−2.47(0.22)	<.0001	−1.73(0.19)	<.0001	0.12(0.09)	0.22	−0.48(0.25)	0.06
Adjusted model d
E20	−1.86(0.80)	0.02	−0.03(0.01)	0.002	−1.59(0.74)	0.03	−1.89(0.64)	0.003	0.88(0.32)	0.007	1.27(0.59)	0.03
E2c	−1.73(0.25)	<.0001	−0.01(0.004)	0.08	−1.84(0.22)	<.0001	−1.36(0.19)	<.0001	0.34(0.09)	0.0003	0.28(0.25)	0.27

^aCoefficients indicated changes in lipids/apolipoproteins per one SD increase in baseline AMH₀ and E2₀ as well as per one SD increase in the change in AMH and E2 since baseline.

^bAMH and E2 were log transformed

^cLog (triglyceride level in mg/dL)

^dAdjusted model: Adjusted for study site, race, education, and time-varying age, menopausal status, BMI, physical activity score, smoking status, cycle day (for models including E2), ever had stroke or heart attack.

^eUnadjusted model adjusted for cycle day.

Abbreviations: AMH: Anti-Müllerian hormone; apo A1: apolipoprotein A1; apo B: apolipoprotein B; BMI: body mass index; E2: estradiol; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol; TC: total cholesterol; TG: triglyceride; AMH₀: baseline AMH; AMH_c: change in AMH since its premenopausal baseline level; E2₀: baseline E2; E2_c: change in E2 since its premenopausal baseline level

In adjusted models, the negative associations of AMH₀ and AMH_c remained significant for TC, and LDL-C, but not TG. AMH₀, but not AMH_C, remained negatively associated with apo B and HDL-C, while AMH_C, but not AMH₀, remained negatively associated with apo A-1 (Table 2 and Figures 1, 2, 3).

Figure 1. Average changes in cholesterol and triglycerides over the menopause transition as related to premenopausal levels and changes in AMH/E2.

Abbreviation: AMH₀: baseline AMH; AMH_c: change in AMH since its premenopausal baseline level; E2₀: baseline E2; E2_c: change in E2 since its premenopausal baseline level; TC: total cholesterol, TG: triglycerides. AMH and E2 were log transformed Presented estimates are changes in total cholesterol (A) and triglycerides (B) per one SD increase in baseline AMH and E2 as well as per one SD increase in AMH and E2 change since baseline, adjusting for study site, race, education, and time-varying age, menopausal status, BMI, physical activity score, smoking status, cycle day (for models including E2), ever had stroke or heart attack.

* P value < 0.05

** P value < 0.001

Figure 2. Average changes in LDL-C and apo B over the menopause transition as related to premenopausal levels and changes in AMH/E2.

Abbreviation: AMH₀: baseline AMH; AMH_c: change in AMH since its premenopausal baseline level; E2₀: baseline E2; E2_c: change in E2 since its premenopausal baseline level; LDL-C: low-density lipoprotein cholesterol; apoB: apolipoprotein B. AMH and E2 were log transformed Presented estimates are changes in LDL cholesterol (A) and apolipoprotein B (B) per one SD increase in baseline AMH and E2 as well as per one SD increase in the change in AMH and E2 since baseline, adjusting for study site, race, education, and time-varying age, menopausal status, BMI, physical activity score, smoking status, cycle day (for models including E2), ever had stroke or heart attack.

* P value < 0.05

** P value < 0.001

Figure 3. Average changes in HDL-C and apo A-1 over the menopause transition as related to premenopausal levels and changes in AMH/E2.

Abbreviation: AMH₀: baseline AMH; AMH_c: change in AMH since its premenopausal baseline level; E2₀: baseline E2; E2_c: change in E2 since its premenopausal baseline level; HDL-C: high-density lipoprotein cholesterol; apoA-1: apolipoprotein A-1. AMH and E2 were log transformed Presented estimates are changes in HDL cholesterol (A) and apolipoprotein A1 (B) per one SD increase in baseline AMH and E2 as well as per one SD increase in the change in AMH and E2 since baseline, adjusting for study site, race, education, and time-varying age, menopausal status, BMI, physical activity score, smoking status, cycle day (for models including E2), ever had stroke or heart attack.

* P value < 0.05

** P value < 0.001

Associations of baseline E2 (E2₀) and change in E2 since baseline (E2_c) with lipids/lipoproteins- AMH measures not included in models

In unadjusted models, negative associations were reported between each of E2₀ and E2_c and levels of TC, TG, LDL-C, and apo B over time. E2₀, but not E2_C, positively associated with HDL-C, and both E2₀ and E2_C were not related to apo A-1. Compared to E2c, per 1 SD unit increase, E2₀ showed stronger associations with TC, TG, apo B, HDL-C, but not LDL-C, which showed a stronger negative association with E2_c (Table 2).

In adjusted models, E2₀ and E2_c remained negatively associated with TC, LDL-C, apo B, and positively associated with HDL-C. Only E2₀ remained negatively associated with TG and positively associated with apo A-1(Table 2 and Figures 1, 2, 3).

Independent associations of AMH and E2 measures with lipids/lipoproteins- AMH and E2 measures included in the same models

In unadjusted models that includes both baseline AMH and E2, and their changes since baseline, negative independent associations were reported for AMH and E2 measures with TC, LDL-C and apo B. Negative independent associations were reported for AMH measures with HDL-C, while positive independent associations were reported for E2 measures with HDL-C. Negative independent association was reported for AMH_C with apo A-1, while positive independent association was reported for E2₀ with apo A-1. Higher E2 measures but not AMH measures independently associated with lower TG (Table 3).

Table 3.

Independent associations ^a of premenopausal AMH and E2 levels ^b and their changes over the menopause transition with lipids/lipoproteins

	Total Cholesterol mg/dL		Log Triglyceride c		LDL-C mg/dL		Apo B mg/dL		HDL-C mg/dL		Apo A-1 mg/dL
	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value	β(SE)	P-value
Unadjusted model d
AMH0	−4.06(0.83)	<.0001	−0.02(0.01)	0.09	−2.61(0.77)	0.0007	−2.29(0.68)	0.0007	−0.96(0.37)	0.009	−1.16(0.64)	0.07
AMHc	−3.14(0.32)	<.0001	−0.004(0.005)	0.43	−2.18(0.27)	<.0001	−0.96(0.25)	0.0001	−0.76(0.12)	<.001	−3.42(0.33)	<.0001
E20	−1.96(0.81)	0.02	−0.05(0.01)	<.0001	−1.79(0.75)	0.03	−2.27(0.66)	0.0006	1.29(0.36)	0.0003	1.38(0.62)	0.03
E2c	−1.95(0.26)	<.0001	−0.01(0.003)	0.01	−1.97(0.23)	<.0001	−1.51(0.20)	<.0001	0.30(0.10)	0.002	0.21(0.26)	0.43
Adjusted model e
AMH0	−2.94(0.85)	0.0005	0.005(0.01)	0.66	−1.44(0.79)	0.07	−1.07(0.68)	0.01	−1.52(0.34)	<.0001	−0.19(0.63)	0.76
AMHc	−0.79(0.36)	0.03	0.01(0.01)	0.06	−0.59(0.32)	0.06	−0.04(0.28)	0.89	−0.27(0.13)	0.04	−1.12(0.36)	0.002
E20	−1.45(0.81)	0.07	−0.03(0.01)	0.002	−1.39(0.75)	0.06	−1.75(0.64)	0.007	1.09(0.33)	0.0008	1.31(0.59)	0.03
E2c	−1.59(0.26)	<.0001	−0.01(0.002)	0.03	−1.74(0.23)	<.0001	−1.35(0.20)	<.0001	0.40(0.10)	<.0001	0.45(0.26)	0.08

^aCoefficients indicated changes in lipids/apolipoproteins per one SD increase in baseline AMH₀ and E2₀ as well as per one SD increase in the change in AMH and E2 since baseline.

^bAMH and E2 were log transformed

^cLog (triglyceride level in mg/dL)

^dUnadjusted model adjusted for cycle day.

^eAdjusted model: Adjusted for study site, race, education, and time-varying age, menopausal status, BMI, physical activity score, smoking status, cycle day, ever had stroke or heart attack.

Abbreviations: AMH: Anti-Müllerian hormone; apo A1: apolipoprotein A1; apo B: apolipoprotein B; BMI: body mass index; E2: estradiol; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol; TC: total cholesterol; TG: triglyceride; AMH₀: baseline AMH; AMH_c: change in AMH since its premenopausal baseline level; E2₀: baseline E2; E2_c: change in E2 since its premenopausal baseline level

In adjusted models, only the negative independent associations of AMH measures with TC and HDLC remained significant. higher AMH_c, but not AMH₀ independently associated with lower apo A-1. On the other hand, higher E2 measures remained independently associated with lower TG and apo B, and higher HDL-C. Higher E2_C remained independently associated with lower TC, LDL-C, while higher E2₀ remained independently associated with higher apo A-1 (Table 3 and Figures 1, 2, 3).

Sensitivity analysis

Compared with results listed in Table 1, after removing outcomes measured at visit 12 and 13, the direction, magnitude, and significance of associations of premenopausal levels of AMH and E2 and their changes since baseline with lipids and apolipoproteins did not change meaningfully (Table S1).

DISCUSSION

Using comprehensive longitudinal data from the SWAN study, we showed that premenopausal levels of AMH and E2 and their changes over the MT may contribute differently to lipids/lipoproteins profile in women during midlife. Independent of premenopausal levels of E2 and their changes since baseline, lower levels of premenopausal AMH and greater declines of AMH over the MT associated with higher TC, HDL-C, and apo A-1 (only with greater decline of AMH). On the other hand, independent of premenopausal AMH levels and their changes since baseline, lower levels of premenopausal E2 and greater declines of E2 over the MT associated with higher TG, and apo B, and lower HDL-C, and apo A-1 (only with lower premenopausal E2). Additionally, greater declines in E2 levels over the MT associated with higher TC and LDL-C. The reported findings suggest that both AMH and E2 play independent role in lipids/lipoproteins in women traversing the menopause. An atherogenic profile of higher TG, apo B, TC and LDL-C seem to be more driven by premenopausal levels of E2 and/or their changes over the MT, while higher levels of HDL-C and apo A-1 seem to be more driven by premenopausal levels of AMH and their decline over the MT.

Limited studies assessed associations of AMH and lipids/lipoproteins in women. ^14–21 All previous studies focused on young reproductive-aged women, and none assessed changes in AMH over time, included women transitioning through menopause, or adjusted for level of and changes in E2. Results from previous studies were not consistent, which could be due to differences in the study population (healthy participants vs. participants with diabetes or suffer from polycystic ovary syndrome), methods to measure AMH, or analytical approaches and/or study designs. ^14–21 Among healthy reproductive-aged women, lower AMH was associated with lower HDL-C ^14–16 and/or higher TC ¹⁷. Interestingly, in studies that included young women with a health condition such as type 1 diabetes or polycystic ovary syndrome, lower AMH was associated with higher HDL-C ¹⁸, with one study even reporting a U-shaped relationship between AMH and HDL-C ¹⁹. Findings from previous studies in young women (healthy or unhealthy) were inconsistent with our findings of a link of AMH with specific lipids/lipoproteins including TC and HDL-C.

Interestingly, in young healthy women, lower ovarian reserve was associated with lower HDL-C, while in our study of women traversing the menopause, lower premenopausal levels of AMH and greater declines in AMH over the MT were associated with greater HDL-C and apo A-1. These findings suggest a change in the association between AMH and HDL-C in young vs. midlife women. Accumulating lines of evidence show that higher HDL-C in women traversing the MT may not necessarily be cardioprotective, and may be a marker of HDL dysfunctionality. ^26,27 Elevated postmenopausal HDL-C and larger increases in HDL-C and apo A-1 since the final menstrual period were associated with atherogenesis in previous work. ^28,29 Interestingly, large HDL particles, which known to be highly correlated with HDL-C, was associated with a higher carotid intima-media thickness (cIMT) close to menopause but with lower cIMT later in life, suggesting that the quality of large HDL particles may be reduced over the MT. ²⁹ The ability of large HDL particles to promote cholesterol efflux capacity was found to be lower after as compared to before menopause. ²⁶ Our findings of associations between lower levels and decline in AMH with higher HDL-C during midlife suggest a unique contribution of AMH to the cardioprotective features of HDL in women beyond E2. Future studies should assess whether declines in AMH from reproductive years to midlife year are associated with dysfunctional HDL.

In our study, associations of AMH and E2 with HDL-C and apo A-1 were in the opposite direction, such that lower AMH measures associated with higher HDL-C and apo A-1 independent of E2 measures, while lower E2 measures associated with lower HDL-C and apo A-1 independent of AMH measures. The reported findings suggest that levels of HDL-C in midlife women depend on the opposing contributions of AMH and E2, which may explain the inconsistent results across studies of the patterns of HDL-C over the MT. In other work, oral estrogens associate with increases in HDL-C and this increase seems to be blunted by the addition of progestin. ⁴ This suggests that progestin may counteract the effects of exogenous estrogens or AMH on HDL-C. The US Preventive Services Task Force recommends against the use of conjugated equine estrogen combined with progestin for primary prevention of cardiovascular disease among postmenopausal women based on findings from the Women’s Health Initiative, Heart and Estrogen/progestin Replacement Study, and Estrogen Replacement and Atherosclerosis studies. ^30–35 The implication of our findings on the effect of hormone therapy on lipids/lipoproteins should be assessed. Whether the increase in HDL-C achieved by oral estrogens augmented with HDL-C increases that naturally accompanying AMH declines contribute to the greater CVD risk related to hormone therapy use in midlife women ⁴ is an open question that remain to be addressed.

Unlike AMH measures, measures of endogenous E2 in midlife women showed independent and significant negative associations with TG, LDL-C, and apo B in our study. Results from multiple randomized clinical trials indicated a similar effect of exogenous estrogens on lipids/lipoproteins, such that use of hormone therapy was associated with decreases in LDL-C and TC among postmenopausal women. ^{31,32,36–43}

Our study has several strengths including the large sample size of midlife women with repeated measures of AMH and E2 from pre- to 1 year postmenopausal; the use of a new, ultrasensitive AMH assay (picoAMH ELISA, Ansh LABS, Webster, TX) with a very low limit of detection enabling better characterization of AMH decline over the MT; and the well characterized SWAN cohort. The current study is limited by chemiluminescence assays utilization for E2 compromising our ability to detect lower levels of E2 in postmenopausal women. However, we do not expect this limitation to bias our results, since only 11 women (11 observations) had E2 below the LLD in the current study and a sensitivity analysis excluding this data did not change our overall findings.

Using a strong study design and repeated measures of AMH, E2 and lipids/lipoproteins in a relatively large sample size of women traversing the menopause, the current study provided unique contributions to the literature on lipids/lipoproteins profile over the MT. AMH and E2 contributed differently to lipids/lipoproteins profiles during midlife. Atherogenic changes in TC, TG, apo B and LDL-C were more driven by lower premenopausal and/or greater declines in E2 over the MT, while greater increases in HDL-C and apo A-1, which may be viewed as markers of HDL dysfunctionality ²⁸, were more driven by lower premenopausal AMH and/or greater declines in AMH over the MT.

Highlights.

• AMH and E2 changes relate differently to lipid/lipoprotein profile in midlife women

• Lower midlife E2 may be related to atherogenic lipid/lipoprotein profile in women

• Lower midlife AMH may be related to HDL dysfunctionality in women