Associations Between Anti-Mullerian Hormone and Cardiometabolic Health in Reproductive Age Women Are Explained by Body Mass Index

Abstract

Context

The relationship between reproductive and cardiometabolic aging is unclear. It is unknown if the relationship differs across different clinical populations.

Objective

To determine whether markers of ovarian reserve are associated with cardiometabolic risk in reproductive aged women with unexplained infertility (UI), polycystic ovary syndrome (PCOS), and regularly cycling women (OVA).

Design and setting

Cross-sectional data from 8 US-based academic centers.

Participants

Women aged 25–40 from 3 clinical populations: 870 with UI, 640 with PCOS, and 921 community-based OVA.

Main Outcome Measures

Multivariable linear regression models were used to relate anti-mullerian hormone (AMH) and antral follicle count with cardiometabolic parameters including body mass index (BMI), waist circumference (WC), fasting glucose and insulin, homeostasis model assessment-insulin resistance (HOMA-IR), lipids, and C-reactive protein.

Results

In age and study site-adjusted models, AMH inversely related to BMI in the UI and OVA groups (P = 0.02 and P < 0.001). Among women with PCOS, AMH inversely related to BMI (P < 0.001), and also to WC (P < 0.001), fasting insulin (P < 0.01), HOMA-IR (P < 0.01), triglycerides (P = 0.04), and C-reactive protein (P < 0.001) and directly related to higher total (P = 0.02), low-density lipoprotein (P < 0.01), and high-density lipoprotein cholesterol (P < 0.01). In OVA, AMH also varied inversely with WC (P < 0.001), fasting insulin (P = 0.02), and HOMA-IR (P = 0.02). Adjustment for BMI eliminated associations in the OVA group but in PCOS, the relationship of AMH to total (P = 0.03) and low-density lipoprotein cholesterol (P = 0.003) remained.

Conclusion

Associations observed between AMH and cardiometabolic indices are largely explained by BMI in women with and without PCOS. (J Clin Endocrinol Metab XX: 0-0, 2019)

Menopause has been associated with deterioration of cardiovascular status as well as alterations in cardiometabolic risk factors including dyslipidemia and dysglycemia irrespective of age at menopause (1). The Study of Women's Health Across the Nation confirmed this association, demonstrating worsening lipid status with progress through the stages of the menopausal transition that was independent of age (2). There is also evidence that the particular hormone pattern that is produced during the menopausal transition may influence the appearance and progression of subclinical cardiovascular disease (3, 4). Other studies have demonstrated a link between the age at menopause and cardiovascular mortality and heart failure (5–8).

The inevitable decline of the ovarian follicle pool is a key determinant of time of menopause, and serum anti-mullerian hormone (AMH) has emerged as a promising single marker of ovarian reserve (9, 10). AMH has been shown to reflect the gradual decline in reproductive capacity with increasing age (11). As such, investigators have questioned whether decreased AMH, independent of age, may also serve as an early indicator of worsening cardiometabolic status. Women with primary ovarian insufficiency have increased risk for cardiovascular disease (12) and have low AMH (13), which supports the notion that AMH may serve as an early marker for cardiovascular disease. Yet, the relationship between AMH and markers of cardiovascular health does not appear to be straightforward. A number of large and small studies in regular cycling women with and without infertility have reported associations between AMH and various cardiometabolic parameters, but there is a lack of clarity around potential confounding effects of age and BMI (9, 14–17). Thus, there remains uncertainty over the degree to which AMH may reflect underlying cardiometabolic health in regularly cycling women (OVA).

A gap also exists regarding whether AMH might relate to cardiometabolic markers differently in unique clinical populations, including infertile women and those with polycystic ovary syndrome (PCOS). AMH levels have been reported to be 2–3 times higher in women with PCOS compared with those without this diagnosis (18–21). It has been well established through cross-sectional data that women with PCOS have increased risks for cardiometabolic disease compared with women without PCOS (22). However, the relationship between AMH level and cardiometabolic risk factors in a PCOS population remains unclear and may differ in its biological basis from the relationship between AMH and cardiovascular risk in eumenorrheic populations. Several cohort and case-control studies have demonstrated no association with AMH and markers of cardiometabolic risk in women with PCOS (17, 23, 24); however, other small studies have reported both positive (15, 25, 26) and negative (27) correlations with lipids, homeostatic model assessment of insulin resistance (HOMA-IR), and/or 2-hour oral glucose tolerance test. Feldman et al. recently reported that young PCOS women with low AMH have increased risk for metabolic syndrome (28).

The majority of studies that have examined the relationship between ovarian reserve and cardiometabolic health have focused mainly on AMH (9). Serum AMH levels are positively associated with the number of antral follicles seen on ultrasound or antral follicle count (AFC). Therefore, examining both AMH and AFC as markers of ovarian reserve is important to understanding the relationship between ovarian reserve and cardiometabolic heath (29).

The purpose of this study is to investigate the relationship between markers of ovarian reserve, including AMH and AFC, to markers of cardiovascular and metabolic health in 3 distinct populations: ovulatory women with unexplained infertility, infertile women with PCOS, and regularly cycling reproductive aged women. We hypothesized that increased AMH and AFC levels would associate with healthier cardiometabolic risk profiles across all three populations.

Materials and Methods

Design

This was a cross-sectional analysis of 870 women with unexplained infertility, 640 women with PCOS, and 921 regular cycling women from a community-based cohort. We restricted the study cohort to women 25–40 years old from each population. All participants provided written informed consent and institutional review board approval was obtained at each participating study site.

Participants

Unexplained infertility cohort. 

This study population included 870 female subjects with unexplained infertility (UI) who participated in the Assessment of Multiple Intrauterine Gestations from Ovarian Stimulation study. This was a multicenter, prospective, partially blinded clinical trial that compared gonadotropins, clomiphene citrate, and aromatase inhibitors for the treatment of UI. Inclusion criteria for the UI study cohort have been previously described (30–32). Participants were recruited between 2010 and 2013 at 12 participating Reproductive Medicine Network clinical sites throughout the United States. Women were included if they had 1 or more years of infertility, desired conception, and were regularly ovulating, with a normal uterine cavity and at least 1 patent fallopian tube. Regular ovulation was considered as 9 or more menses per year. To rigorously define UI, cycle day 3 follicle-stimulating hormone level had to be less than 12 IU/L within 1 year of study initiation. Normal thyroid-stimulating hormone and prolactin were required within 1 year of study initiation and the male partner needed as least 5 million sperm per milliliter in the ejaculate.

Exclusion criteria for this study were previously reported (31, 32) and included contraindications to study medications, history of self-reported endometriosis (moderate to severe), PCOS, type 1/2 diabetes or use of antidiabetic medications, known heart disease (New York Heart Association class II or higher), liver disease (aspartate transaminase or alanine transaminase >2 times normal or bilirubin >2.5 mg/dL), renal disease (blood urea nitrogen > 30 mg/dL or creatinine > 1.4 mg/dL), significant anemia (hemoglobin < 10 g/dL), history of thromboembolic events, Cushing disease, 21-hydroxylase deficiency, and known or suspected androgen-secreting tumor. Subjects using donor sperm or history of sterilization with reversal were also excluded. A washout period of 2 months was required for women taking oral contraceptives, depo-progestins, or hormonal implants before initial baseline screening assessment.

PCOS cohort. 

The PCOS cohort included 640 female subjects with PCOS [defined as menstrual interval at least 45 days and/or ≤8 menses/year, combined with either hyperandrogenism (hirsutism or hyperandrogenemia) or polycystic-appearing ovaries by ultrasound according to the Rotterdam criteria] who participated in the Pregnancy in Polycystic Ovary Syndrome II trial, a multicenter, double-blind, randomized clinical trial (33, 34) comparing live-birth rate in response to treatment with escalating doses of clomiphene citrate or letrozole for up to a total of 5 cycles. The study design, methods, inclusion, and exclusion criteria have been described in detail elsewhere (33–35). In brief, women with PCOS with anovulation (36) who were actively seeking pregnancy were enrolled. All had normal thyroid-stimulating hormone and prolactin levels. Documentation of tubal patency and a normal uterine cavity was required, as well as a semen analysis with at least 14 million sperm per milliliter for the male partners.

Exclusion criteria for this study that were previously reported (33–35) included contraindications to study medications, bariatric surgery <12 months from screening visit, uncontrolled diabetes type 1/2 (hemoglobin A1c > 7.0%) or use of antidiabetic medications that would interfere with study medications, known heart disease, liver disease (aspartate transaminase or alanine transaminase > 2 times normal or bilirubin > 2.5 mg/dL), renal disease (blood urea nitrogen > 30 mg/dL or creatinine > 1.4 mg/dL), significant anemia (Hemoglobin < 10 g/dL), history of thromboembolic events, uncontrolled hypertension (systolic > 160 mm Hg or diastolic > 100 mm Hg), Cushing disease, uncorrected thyroid disease (thyroid-stimulating hormone <0.2 or >5.5 mIU/mL), hyperprolactinemia (prolactin > 30 ng/mL), 21-hydroxylase deficiency, and known or suspected androgen-secreting tumor. Subjects using donor sperm or history of sterilization with reversal were also excluded. A washout period of 2 months was required for women taking any medications to affect reproductive function or metabolism.

OVA cohort. 

The OVA cohort included 921 ovulatory women not seeking fertility treatment. Subjects were a part of a community-based cohort, the Ovarian Aging study (16), a prospective, observational study designed to investigate reproductive aging. OVA participants were evaluated at University of California, San Francisco. Participants were recruited between 2006 and 2010 from a sampling frame of all age-eligible female members of the Kaiser Permanente of Northern California Health Plan, within a reasonable travel distance to the research center. Kaiser Permanente is an integrated healthcare delivery system covering 30% of the regional population. Inclusion criteria for the OVA population included regular menses at 22- to 35-day intervals and self-identification in 1 of 4 racial/ethnic groups categorized as Caucasian, Asian, Hispanic, or African American. Women reporting multiethnic origin were not enrolled.

Exclusion criteria for the OVA cohort included estrogen- or progestin-containing medication use in the 3 months before enrollment, history of endometriosis, or any history of uterine or ovarian surgery. Further information regarding the study design and methodology for OVA have been previously published (16, 37).

Ultrasound, anthropomorphic measurements, and serum laboratory testing

UI and PCOS.

Transvaginal ultrasonography for AFC was performed (38) between cycle days 3 and 5 (for cycling women) during an index visit at 1 of the participating sites using standard clinical ultrasound machines. Anthropometric measurements were obtained within 6 months before the start of the trial. Serum was collected for AMH, fasting glucose, insulin, lipids, and high-sensitivity C-reactive protein (hsCRP) before ovulation induction. All serum testing was completed at a central core laboratory (University of Virginia) as previously described (31, 35). Serum AMH for subjects from UI and PCOS cohorts was performed using a commercially available enzyme-linked immunosorbent assay (Ansh Laboratories, Webster, TX); lower limit of detection was 0.25 ng/mL, with an intra-assay coefficient of variation (CV) of 3.0% and inter-assay CV of 7.0%.

OVA cohort.

Anthropomorphic measurements and transvaginal ultrasounds for AFC women (38) were performed at an index visit between cycle days 2 and 4. Serum was collected at the baseline visit and frozen for future use. Laboratory assays were performed at a single commercial laboratory. AMH values were analyzed on the 2 commercially available enzyme-linked immunosorbent assays from Beckman Coulter (Marseille, France). The Immunotech assay was used for the 84% of OVA samples until this assay was retired and the second-generation assay (Gen II) was used for the remainder of the samples. Forty-four women had serum tested with both assays and a regression analyses showed excellent correspondence (R² = 0.94). AMH values based on the Immunotech assay were adjusted using the equation of the line with Immunotech predicting Gen II. The Gen II assay sensitivity was 0.16 ng/mL, the intra-assay CV was 1.4%, and the inter-assay CV was 12.5%. Because the OVA cohort's AMH samples were analyzed on a different assay than UI and PCOS samples, a subset of 244 OVA participants had frozen serum also analyzed with Ansh assay at the Central Ligand Assay and Analysis Core at the University of Virginia. Assays for triglycerides, high-density lipoprotein (HDL), fasting glucose, and insulin were performed by Quest Diagnostics (San Jose, CA) (39). Lipids were assayed using enzymatic methods, fasting glucose was assayed by the glucose oxidase method, and insulin was assayed using the Siemens Immulite (Tarrytown, NY) immunochemiluminometric assay.

Statistical analysis

A multivariable linear regression model was performed to assess the relationship between AMH and markers of cardiometabolic risk parameters [waist circumference, fasting glucose, HOMA-IR, total cholesterol, HDL cholesterol, low-density lipoprotein (LDL) cholesterol, and hsCRP; outcomes], controlling for age and study site. Models were also adjusted for body mass index (BMI) to determine the relationship of AMH cardiometabolic risk indices independent of BMI. AFC was also compared with markers of cardiometabolic risk as a secondary measure of ovarian reserve. Data analyses were performed using STATA 14.2 (STATA Corp).

Results

Subject demographics, anthropomorphic measurements, and serum laboratory values are shown in Table 1 for all cohorts.

Table 1.

Subject baseline characteristics

Characteristic	UI (n = 870)	PCOS (n = 640)	OVA (n = 921)
Age, y	32.5 (3.9)	30.0 (3.6)	33.0 (4.3)
BMI, kg/m2	26.9 (6.6)	35.1 (9.3)	26.7 (7.0)
Waist circumference, cm	86.5 (15.8)	105.7 (20.5)	83.5 (15.4)
AMH, ng/mL	2.6 (2.0)	7.9 (6.8)	4.1 (3.1)
AFC	20.4 (11.8)	46.4 (28.3)	17.2 (9.4)
AMH/AFC	0.14 (0.14)	0.18 (0.14)	0.24 (0.14)
Fasting glucose mg/dL	85.2 (11.8)	86.0 (13.0)	86.7 (12.5)
Fasting insulin mg/dL	9.6 (15.3)	18.7 (27.7)	5.3 (5.9)
HOMA-IR	2.1 (3.6)	4.3 (9.5)	1.2 (1.7)
Total cholesterol, mg/dL	169.2 (31.5)	180.1 (36.6)	172.3 (30.4)
HDL, mg/dL	46.1 (11.2)	38.1 (10.7)	59.8 (15.5)
LDL, mg/dL	109.0 (33.6)	122.7 (33.4)	95.0 (27.3)
Triglycerides, mg/dL	94.1 (52.0)	118.2 (57.4)	88.1 (57.9)
hsCRP mg/L	3.2 (4.6)	6.3 (6.8)	2.7 (3.8)

Values are in mean (standard deviation). Abbreviations: AFC, antral follicle count; AMH, anti-mullerian hormone; BMI, body mass index; HOMA-IR, homeostatic model assessment of insulin resistance, hsCRP, high-sensitivity C-reactive protein; PCOS, polycystic ovary syndrome; UI, unexplained infertility.

Anti-mullerian hormone

UI cohort.

Lower AMH was associated with higher BMI in this cohort. Other cardiometabolic indices were not associated with AMH in this population in unadjusted or adjusted multivariable regression models (Tables 2 and 3).

Table 2.

Linear regression models for AMH and cardiometabolic risk markers, adjusted for age and study site

	AMH: coefficient (P value)
	UI	PCOS	OVA
BMI, kg/m2	-0.27 (0.02)	-0.31 (<0.001)	-0.42 (<0.001)
Waist circumference, cm	-0.31 (0.27)	-0.68 (<0.001)	-0.98 (<0.001)
Fasting glucose mg/dL	-0.39 (0.06)	-0.11 (0.14)	-0.14 (0.35)
Fasting insulin, mg/dL	-0.20 (0.45)	-0.46 (<0.01)	-0.17 (0.02)
HOMA-IR	-0.06 (0.33)	-0.12 (0.04)	-0.05 (0.02)
Total cholesterol mg/dL	0.07 (0.90)	0.49 (0.02)	0.39 (0.28)
HDL mg/dL	0.11 (0.57)	0.17 (<0.01)	0.38 (0.05)
LDL, mg/dL	-0.43 (0.45)	0.55 (<0.01)	0.14 (0.68)
Triglycerides, mg/dL	-0.15 (0.90)	-0.72 (0.04)	-0.21 (0.77)
hsCRP	-0.15 (0.06)	-0.17 (<0.001)	-0.25 (0.07)

Values are in mean (standard deviation). Abbreviations: AMH, anti-mullerian hormone; BMI, body mass index; HOMA-IR, homeostatic model assessment of insulin resistance; hsCRP, high-sensitivity C-reactive protein; PCOS, polycystic ovary syndrome; UI, unexplained infertility. P values <0.05 were considered statistically significant and are in boldface type.

Table 3.

Linear regression models for AMH and cardiometabolic risk markers, adjusted for age, study site, and BMI

	AMH: coefficient (P value)
	UI	PCOS	OVA
Waist circumference, cm	0.23 (0.11)	0.01 (0.17)	-0.13 (0.07)
Fasting glucose, mg/dL	-0.34 (0.08)	-0.05 (0.54)	0.01 (0.93)
Fasting insulin mg/dL	0.01 (0.97)	-0.22 (0.19)	0.03 (0.68)
HOMA-IR	-0.02 (0.71)	-0.06 (0.23)	0.003 (0.88)
Total cholesterol, mg/dL	-0.36 (0.50)	0.50 (0.03)	0.63 (0.09)
HDL, mg/dL	0.14 (0.42)	0.04 (0.40)	-0.01 (0.95)
LDL, mg/dL	-0.67 (0.24)	0.60 (0.003)	0.57 (0.08)
Triglycerides, mg/dL	0.47 (0.58)	-0.25 (0.46)	0.76 (0.27)
hsCRP	-0.09 (0.20)	-0.05 (0.15)	-0.21 (0.10)

Values are in mean (standard deviation). Abbreviations: AFC, antral follicle count; AMH, anti-mullerian hormone; BMI, body mass index; HOMA-IR, homeostatic model assessment of insulin resistance; hsCRP, high-sensitivity C-reactive protein; PCOS, polycystic ovary syndrome; UI, unexplained infertility. P values <0.05 were considered statistically significant and in boldface type.

PCOS cohort.

Increased BMI, waist circumference, fasting insulin, HOMA-IR, triglycerides, and hsCRP were seen with lower levels of AMH in the age- and site-adjusted model (Table 2). In addition, increased total, HDL, and LDL cholesterol levels were found to associate with higher levels of AMH (Table 2). After adjusting for BMI, only higher levels of total cholesterol and LDL cholesterol were found to independently associate with higher AMH, whereas the significance of all other metabolic associations was lost (Table 3). To further explore the association between AMH and LDL and total cholesterol, sensitivity analyses were performed adding total testosterone to our models. These analyses revealed persistence of the AMH association with LDL and slight attenuation of the association with total cholesterol (coefficient for LDL 0.51, P = 0.02; coefficient for total cholesterol 0.40, P = 0.08). To explore potential effect modification by BMI, we further completed age, site, BMI, and testosterone adjusted analyses in the PCOS group, stratified into obese (BMI ≥ 30 kg/m²) and nonobese (<30 kg/m²) groups. Here, we found that the positive association between AMH and LDL and total cholesterol was restricted to the obese cohort (coefficient 1.14; P = 0.0001; coefficient 1.0; P = 0.003, respectively). In the nonobese cohort, AMH had a negative association with triglyceride levels of borderline significance (coefficient = -0.78, P = 0.06).

Ovarian aging cohort.

In our age adjusted model, higher BMI, waist circumference, fasting insulin, and HOMA-IR were found in subjects with lower AMH (Table 2). After controlling for BMI, no associations were noted with AMH with any of the cardiometabolic indices (Table 3). In the subset of 244 OVA participants with AMH values completed at the Ligand Assay and Analysis Core, similar results were seen as for the entire cohort. In our age-adjusted model, higher BMI, waist circumference, fasting insulin, and HOMA-IR were noted with a lower AMH, and a high HDL was seen with higher AMH (data not shown). None of these associations persisted after adjustment for BMI (data not shown).

Antral follicle count

AFC was only found to be negatively associated with HOMA-IR in the PCOS cohort (coefficient = -0.013, P = 0.04), but was not significantly associated after controlling for BMI (coefficient = -0.01, P = 0.43). AFC did not have any significant associations with cardiometabolic markers in either the UI or OVA populations in either model (Table 4).

Table 4.

Linear regression models for AFC and cardio-metabolic risk markers

	AFC: coefficient (P value), age-adjusted model			AFC: coefficient (P value), age- and BMI-adjusted model
	UI	PCOS	OVA	UI	PCOS	OVA
BMI kg/m2	-0.005 (0.81)	-0.27 (0.65)	-0.009 (0.76)	NA	NA	NA
Waist circumference, cm	-0.02 (0.57)	-0.58 (0.07)	-0.002 (0.97)	0.04 (0.14)	-0.005 (0.74)	-0.02 (0.43)
Fasting glucose, mg/dL	-0.05 (0.11)	-0.01 (0.50)	-0.19 (0.72)	-0.06 (0.11)	-0.007 (0.70)	-0.02 (0.67)
Fasting insulin, mg/dL	-0.04 (0.42)	-0.036 (0.43)	0.005 (0.83)	-0.04 (0.43)	-0.01 (0.07)	0.001 (0.98)
HOMA-IR	-0.01 (0.42)	-0.013 (0.04)	0.0005 (0.94)	-0.01 (0.43)	-0.01 (0.58)	-0.001 (0.91)
Total cholesterol, mg/dL	-0.002 (0.98)	0.09 (0.10)	0.01 (0.96)	-0.002 (0.98)	0.09 (0.11)	0.002 (0.99)
HDL, mg/dL	-0.05 (0.13)	0.04 (0.02)	-0.43 (0.53)	-0.05 (0.12)	0.03 (0.06)	-0.03 (0.59)
LDL, mg/dL	-0.60 (0.57)	0.07 (0.89)	0.02 (0.88)	-0.06 (0.56)	0.77 (0.16)	0.004 (0.97)
Triglycerides, mg/dL	0.20 (0.23)	-0.01 (0.89)	0.21 (0.38)	0.19 (0.22)	0.03 (0.47)	0.19 (0.41)
hsCRP	-0.02 (0.12)	-0.02 (0.18)	-0.11 (0.09)	-0.02 (0.10)	-0.01 (0.62)	-0.10 (0.09)

Values are in mean (standard deviation). Abbreviations: AMH, anti-mullerian hormone; BMI, body mass index; HOMA-IR, homeostatic model assessment of insulin resistance; hsCRP, high-sensitivity C-reactive protein; NA, not available; PCOS, polycystic ovary syndrome; UI, unexplained infertility. P values <0.05 were considered statistically significant and are in boldface type.

Discussion

We sought to determine if markers of ovarian reserve, AMH and AFC, would be associated with increased cardiovascular risk during the reproductive period in 3 distinct large populations. We found that low AMH levels were associated with several adverse cardiometabolic indices, including waist circumference, fasting insulin, and HOMA-IR, in both the PCOS and OVA populations. However, importantly, we also found that these relationships were entirely explained by the strong inverse relationship between AMH and BMI. Furthermore, contrary to our hypothesis, higher AMH associated with higher total and LDL cholesterol in the PCOS sample. We found that AFC had no independent associations with cardiometabolic indices.

Because menopause has been associated with a worsening of cardiometabolic status, the question of whether markers of ovarian reserve might predict cardiovascular health has been a recent question in the literature. For eumenorrheic women, studies have shown an association between AMH and various cardiometabolic parameters but a clear relationship has not been established because of confounding by age and BMI. Our study found that in OVA without infertility, low AMH was associated with poorer cardiometabolic profile by several parameters; however, these associations were mainly accounted for by BMI, similar to prior studies that controlled for BMI. In contrast to our study, Tehrani et al. studied 1015 OVA ages 20–50 in a longitudinal study assessing the relationship between age-specific AMH quartiles and lipid parameters and found that women whose AMH was in the lowest age-specific quartile had poorer lipid profiles (14). Interpretation of these findings is limited by the fact that this study did not control for BMI.

We were also interested to understand if associations between AMH and cardiometabolic health would differ in an infertile population. Similar to the findings of our study, Cui et al. reported in 1896 infertile eumenorrheic women only an inverse relationship between AMH and BMI (17) In contrast, Nardo et al. described no associations between AMH and markers of insulin resistance or BMI after controlling for age in 183 regularly cycling infertile women (15). However, Nardo's sample had overall lower BMI compared with our study and thus may not have had enough of a range of BMIs to discern the association.

The relationship between AMH and cardiometabolic health in PCOS has also been an unresolved question. AMH levels are increased in women with PCOS, likely resulting from either an increased number of pre-antral and antral follicles (40) and/or a higher per follicle production of AMH (41). Higher AMH values have been linked to a more severe PCOS reproductive phenotype, with increased menstrual dysfunction and resistance to ovulation induction in the PCOS population (42–45), but whether increased AMH also correlates with a more severe metabolic phenotype in PCOS has been unclear. Similar to our study, Cui et al. reported in 304 PCOS women that AMH was only negatively related to BMI and age but found no association with HOMA-IR or serum lipid parameters (17). A number of small cohort studies also reported no association between AMH and parameters associated with insulin resistance (9). In contrast, Feldman et al. reported increased odds of metabolic syndrome in PCOS women with lower AMH (28), but this study did not control for BMI. Our results show that a lower AMH associates with a poorer cardiometabolic risk profile; however, this effect was attenuated once we controlled for BMI. Similar to another study by Skalba et al. (46) and contrary to our hypotheses, we also found that higher AMH was related to higher total and LDL cholesterol in PCOS, even after controlling for both age and BMI. A stratified analysis demonstrated that this relationship was restricted to the obese PCOS population. The etiology behind this relationship is unclear. AMH may be a marker for increased production of other steroid hormones, including androgens, which may directly contribute to dyslipidemia through effects on the liver. However, addition of testosterone to our models only slightly attenuated the relationship between AMH and total and LDL cholesterol. Triglyceride did not share the same relationship with AMH as total and LDL cholesterol. This dichotomous finding may suggest triglyceride elevations in PCOS result from a different pathway than LDL and cholesterol elevations. Further research is needed into the interplay between lipids, sex steroid hormones, and adiposity to elucidate the pathophysiology and clinical consequence behind these findings.

Our findings regarding the association between BMI and AMH has been shown in most (27, 39, 47), but not all (15, 48) prior studies that have examined this relationship. In our 3 distinct populations, we clearly demonstrated a relationship between BMI and AMH. Given that cardiovascular disease risk is increased in women with higher BMI (49–51), we controlled for this key covariate and the previous associations between AMH and cardiovascular risk factors were no longer found. The etiology of the inverse association between AMH and BMI is unknown, but may be related to a suppressive effect of obesity on follicular AMH production or to a dilutional effect on serum AMH concentrations. Interestingly, our findings demonstrate that waist circumference has an inverse association of greater magnitude with AMH compared with BMI, lending support to the concept that adiposity may have a toxic effect on granulosa cell function. These findings highlight a need for further research on the specific effects of adverse metabolic factors on ovarian health.

Serum AMH levels have been shown to positively correlate with the number of antral follicles by ultrasound and with the size of the primordial follicle pool (29); therefore, we expected to find similar relationships of AMH and AFC with cardiometabolic indices. However, only a lower AFC was associated with higher HOMA-IR in the PCOS population, and this finding was not maintained after controlling for BMI. The discrepancy in findings between AMH and AFC further supports the theory that obesity selectively affects AMH levels and thus granulosa cell function, but not ovarian reserve itself.

Our study has several strengths, including: 1) reporting on both AMH and AFC as ovarian reserve markers; 2) assessing multiple cardiometabolic indices; and 3) describing these associations in 3 well-phenotype large populations of reproductive aged women. Our study is limited by laboratory assays run in the same laboratory for only 2 of the 3 populations preventing direct comparison across the 3 groups. Also, the unexplained infertility group excluded those with a follicle-stimulating hormone level > 12 IU/mL and therefore may have also excluded women with lower AMH and AFC influencing results seen in this particular population. Further studies that include infertile women with a broader range of ovarian reserve are needed to further validate these findings.

In summary, to our knowledge, this is 1 of the largest studies evaluating the relationship between ovarian reserve markers and cardiometabolic risk parameters. Contrary to our hypothesis, we did not find that low ovarian reserve markers were associated poorer cardiometabolic state in well-phenotyped women with and without PCOS. Rather, we found BMI was a primary driver of a worse cardiometabolic status. Additionally, BMI has an impact on AMH but not AFC in reproductive aged women with and without PCOS. Further investigation is needed to determine why and how BMI affects AMH, and longitudinal cohort studies are needed to follow trends in markers of ovarian reserve with the development of cardiovascular disease and metabolic syndrome as reproductive age women transition to menopause.

Additional Information

Disclosure Summary: EAG, MGP, MIC, FS, RDR, GC, HZ, and HGH have nothing to disclose. JSR received lecture fees from AbbVie. RSL is consultant for Ferring, Fractyl, AbbVie, and Bayer and receives research funds from Guerbet. NS serves on the Scientific Advisory Board for Astellas/Ogeda and Menogenix, Inc, and has equity interests in Menogenix, Inc. MPD serves on the board of directors for Advanced Reproductive Care, has equity interest in ObsEva, AbbVie, and Bayer, and has clinical trial contracts with ObsEva, AbbVie, and Bayer through Augusta University.

Data Availability: The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.