Anti-Mullerian Hormone and Inhibin B in the Definition of Ovarian Aging and the Menopause Transition

Abstract

Context/Objective: The objective of the study was to determine whether anti-Mullerian hormone (AMH) and inhibin B are viable endocrine biomarkers for framing the menopause transition from initiation to the final menstrual period (FMP).

Design: We assayed AMH, inhibin B, and FSH in 300 archival follicular phase specimens from 50 women with six consecutive annual visits commencing in 1993 when all women were in the pre- and perimenopausal menopause stages. Subsequently each woman had a documented FMP. The assay results were fitted as individual-woman profiles and then related to time to FMP and age at FMP as outcomes.

Results: Based on annual values from six time points prior to the FMP, _logAMH longitudinal profiles declined and were highly associated with a time point 5 yr prior to FMP [including both observed and values below detection (P < 0.0001 and P = 0.0001, respectively)]. Baseline AMH profiles were also associated with age at FMP (P = 0.035). Models of declining _loginhibin B profiles (including both observed and values below detection) were associated with time to FMP (P < 0.0001 and P = 0.0003, respectively). There was no significant association of _loginhibin B profiles with age at FMP.

Conclusions: AMH, an endocrine marker that reflects the transition of resting primordial follicles to growing follicles, declined to a time point 5 yr prior to the FMP; this may represent a critical biological juncture in the menopause transition. Low and nondetectable levels inhibin B levels also were observed 4–5 yr prior to the FMP but were less predictive of time to FMP or age at FMP.

A 15-year longitudinal study of premenopausal women followed to their final menstrual period finds that longitudinal anti-Mu" llerian hormone profiles predicted age of and time to final menstrual period. Declines to low anti-Mu" llerian hormone levels may represent a critical biological juncture in the menopausal transition.

Biomarkers with well-understood biological mechanisms and metrics for assay interpretation are needed to provide an ovarian frame for the onset and end of the menopause transition, indicate the proximity to the final menstrual period (FMP), and contribute to clinical decision making. Current biomarker candidates include anti-Mullerian hormone (AMH), inhibin B, and FSH. In females, AMH, also known as Müllerian-inhibiting substance, is produced in the granulosa cells of ovarian follicles (1) and is thought to be one of the earliest markers of ovarian aging because it reflects the transition of resting primordial follicles into growing follicles (2). This then leads to the subsequent recruitment of FSH-sensitive follicles in the early antral stage (3). Because AMH is produced solely in the growing ovarian follicles, serum levels are regarded as a marker for ovarian reserve, representing the quantity of the ovarian follicle pool (4).

Inhibin B is produced by granulosa and theca cells and its activities include suppression of FSH production (5,6,7). Inhibin B has been described as a marker for the growth of the small antral follicle cohort (8). As the size of the follicle pool declines, inhibin B levels fall and FSH levels rise, the latter reflecting a loss of inhibin restraint. Inhibin B levels change within the menstrual cycle, peaking in the early to midfollicular phase, with a secondary peak at the time of ovulation (9). To provide maximum interpretable information and detection, levels must be assessed in the early follicular phase of the menstrual cycle.

There are recent reports of AMH as an accurate and noninvasive method for determining ovarian follicle reserve and reproductive aging (10,11,12,13,14,15). Likewise, studies of early follicular serum inhibin B are thought to be an indicator of ovarian aging, but a single-time inhibin B value is more difficult to interpret because it may be influenced by cycle phase as well as body mass and perhaps ethnicity (16,17,18,19). However, there are fewer studies of AMH and inhibin B to ascertain their utility as ovarian markers of the menopause transition, especially using a longitudinal design. Therefore, the goal of this research was to evaluate whether AMH and inhibin B levels and their changes over time were associated with time to the FMP and age at FMP, thereby acting as menopause transition landmarks. As depicted in Fig. 1, we hypothesized that characterizing a marker reflecting the size of the recruitable follicle pool would benchmark an important step in the menopause transition. We further hypothesized that low or very low levels of AMH were the critical signal of the progressive diminution of the recruitable follicle pool and reflective of minimal capacity to recruit FSH-sensitive antral follicles. By extension, the ensuing cascade of events would then include declining levels of inhibin B that, in turn, would be reflected with rising FSH levels. Within this framework, levels of AMH, inhibin B, and FSH could provide an endocrine definition of key phases during the menopause transition. We evaluated this concept in a longitudinal study of 50 premenopausal women with six annual measures of AMH, inhibin B, and FSH before their transition to the postmenopause.

Figure 1.

Hypothesized ovarian markers of the menopausal transition. As the recruitable and FSH-insensitive follicle cohort shrinks (values on the right vertical axis), AMH, solely produced in the growing ovarian follicles, will fall to a critical point, which we hypothesize occurs on average 5 yr before FMP. This, in turn, impacts the recruited, FSH-sensitive follicle pool (values shown on the left vertical axis) as reflected in the declining inhibin B levels. Furthermore, the decline of inhibin B to increasingly lower levels is permissive for an accelerated rise in FSH levels. We therefore hypothesize that the low levels of AMH and inhibin B observed, on average, 5 yr before the FMP frame the endocrine definition of the initiation of the perimenopause.

Subjects and Methods

Population

This report is based on assay of 300 follicular phase specimens for AMH, inhibin B, and FSH from the repository of the longitudinal Michigan Bone Health and Metabolism Study (MBHMS) as previously described (20,21). The 300 specimens came from 50 MBHMS women who had six consecutive annual visits and repository specimens collected during the time period that the women were pre- and perimenopausal and before their subsequent documented FMP.

Since 1992 the MBHMS cohort, comprised of 664 women identified from two sampling sources, has participated in annual health evaluations. MBHMS participants are the female offspring of the Tecumseh (Michigan) Community Health Study enrollees who were premenopausal and between the ages of 20 and 40 yr in 1988. In 1992 the MBHMS cohort of 543 women was made population based by incorporating additional Tecumseh women whose parents had not participated in the Tecumseh (Michigan) Community Health Study. Using a sampling frame that included age, name, address, and telephone numbers (Kohl’s Directory), 135 additional age-eligible premenopausal women were identified and 121 (87%) were enrolled. Written informed consent has been obtained from all participants with protocol approval by the University of Michigan Institutional Review Board.

Annual cohort health evaluations have incorporated phlebotomy for metabolic measures, physical measures of bone, body composition, and physical functioning and interviews about health status, menstrual bleeding patterns and health-related behaviors. We selected MBHMS repository specimens to correspond to six consecutive measures, all beginning in 1993, when eligible women were in the pre- and early menopause stages and before their subsequent FMP.

Inclusion criteria for specimen selection were an FSH value in 1992–1993 and 1993–1994 less than 14 mIU/ml from a specimen collected in d 2–7 of the follicular phase of the menstrual cycle; regular menstrual cycling with at least nine menstrual cycles per year; documentation of 12 months of amenorrhea to identify a FMP during the follow-up period; and age at FMP older than 41 yr. Exclusion criteria for specimen selection were an FSH value in 1992–1993 or 1993–1994 of 14 mIU/ml or greater; absence of a FMP; an age at FMP less than 41 yr; hormone therapy use; and hysterectomy or oophorectomy.

The 50 women selected for study were, on average, 4 yr older than women not selected. The mean age in 1993 of those women selected was 42 yr (sd 2.7), compared with 38 yr (sd 5.0) in women not selected. Body mass index (BMI) and smoking frequency was not significantly different between the selected and nonselected groups of women. Mean estradiol (E2) values from the 1992–1993 visit were not statistically different between selected and nonselected women.

Specimens and assays

Specimens were collected fasting in d 2–7 of the follicular phase of the menstrual cycle. When women became amenorrheic, specimens were collected on the anniversary of study enrollment ± 15 d. Biological samples were aliquoted and stored at −80 C without thaw until assay.

A commercially available ELISA [Diagnostic Systems Laboratories (Beckman Coulter), Webster, TX)] were used for the in vitro quantitative measurement of Müllerian-inhibiting substance/AMH in human serum. This ELISA is a direct competitive immunoassay without sample extraction or hydrolysis. Sample wells were coated with a primary antibody. The detection system consisted of a biotinylated secondary antibody and streptavidin-labeled horseradish peroxidase. Samples were assayed in duplicate. There is no detectable cross-reactivity with closely related compounds. The assay measured AMH concentrations from 0.017 to 10 ng/ml with an assay range (standard curve) of 0.05–10 ng/ml. Manufacturer-specified interassay coefficients of variation (CVs) were 8.0% at 0.15 ng/ml, 4.8% at 0.85 ng/ml, and 6.7% at 4.28 ng/ml (mean = 6.5%); intraassay CVs were 4.6% at 0.14 ng/ml, 2.4% at 0.84 ng/ml, and 3.3% at 4.41 ng/ml (mean 4.0%).

Serum inhibin B concentrations were measured in duplicate with a α-βB dimeric ELISA and referenced to a standard of human inhibin B preparation isolated from human follicular fluid provided by Nigel Groome (Oxford Brookes University, Oxford, UK). The assay sensitivity is 4 ng/liter. The within- and between-assay variations were 11.7 and 15.6%, respectively. Assays for both AMH and inhibin B were measured in a single time period and kits came from single lots.

FSH concentrations were measured with a two-site chemiluminometric immunoassay directed to different regions on the β-subunit with intra- and interassay CV of 12.0 and 6.0% and a lower limit of detection of 1.05 mIU. Laboratory quality control preparations at the levels of 8.3 and 13.7 mIU/ml (n = 415 and 462) gave CVs of 9.4 and 7.2%, respectively. Serum E2 concentrations were measured with a modified, off-line Automated Chemiluminometric System 180 (E₂-6) immunoassay to increase sensitivity. The assay has a lower limit of detection of 1.0 pg/ml and intra- and inter CVs of 10.6 and 6.4%, respectively.

Other measures

Height (centimeters) and weight (kilograms) were measured at annual study visits with a calibrated stadiometer and balance-beam scale and used to calculate BMI as weight (kilograms) divided by height squared (meters). Smoking status at study entry was included as an ever vs. never dichotomous variable.

Statistical analysis

Variable distributions were examined for normality, the presence of nonplausible outliers and/or changing variability over time with contingency tables, scatter plots, and box plots. Transformations were applied to outcome measures for satisfying model assumptions such as normality and constant variance.

A two-stage modeling approach was used to relate biomarker (_logAMH, _loginhibin B, and _logFSH levels) to time to FMP and age at FMP as the outcomes of interest (see additional details in Appendix I published as supplimental data on The Endocrine Society’s Journals Online website at http://jcem.endojournals.org). In the first stage, the six annual biomarker values from each woman were decomposed into subject-specific intercept and slope random effects with a flag variable to incorporate information that a value was below the assay detection level. In the second stage, these were incorporated as random independent variables and related to the outcome measures. When values below assay detection were present as for the biomarkers AMH and inhibin B, first-stage models were implemented using the nonlinear mixed model procedure, Proc NLMixed (in SAS; SAS Institute, Cary, NC), to address those below detection values, as reported by Thiebaut et al. (22). Because the random effects were replaced by their estimates from the first stage, modeling in the second stage was undertaken linear models using generalized estimating equation implemented in Proc GENMOD. The results were confirmed by maximum likelihood estimation.

The appropriateness of model fitting was assessed both graphically and using residual analyses. We examined the difference in the predictive power between AMH and inhibin B by putting the AMH intercept and slope in a model, then putting the inhibin B intercept and slope in the model and applying Wald tests that coefficients were zero for the inhibin B intercept and slope. Then hormone order was reversed and the magnitude of the χ² compared. In fitting models and using Cook’s D and maximum likelihood distance influence statistics, we identified that three individual values of inhibin B were strikingly out of range and were excluded from model fitting. Table 1 shows the characteristics of the sample with and without excluding these values.

Table 1.

Hormone characteristics from 50 pre- and perimenopausal women at six points across time (MBHMS)

	1993–1994	1994–1995	1995–1996	1997–1998	1998–1999	1999–2000
	Mean ± sd	Mean ± sd	Mean ± sd	Mean ± sd	Mean ± sd	Mean ± sd
AMH (ng/ml)	0.626 ± 0.51	0.462 ± 0.37	0.335 ± 0.30	0.104 ± 0.10	0.112 ± 0.10	0.086 ± 0.08
% ↓ detection	3 (6%)	5 (10%)	8 (16%)	26 (52%)	27 (54%)	32 (64%)
Inhibin B (pg/ml)	69.8 ± 45.2	54.3 ± 43.6	50.4 ± 37.2	28.0 ± 21.5	30.0 ± 55.5	23.0 ± 24.9
Inhibin Ba	65.3 ± 32.1	50.8 ± 36.1	50.4 ± 37.2	28.0 ± 21.4	22.6 ± 20.0	23.0 ± 24.9
% ↓ detection	2 (4%)	2 (4%)	9 (18%)	18 (36%)	23 (46%)	23 (46%)
FSH (mIU/ml)	8.0 ± 2.4	7.6 ± 3.7	7.7 ± 4.9	16.4 ± 15.7	18.3 ± 15.5	21.5 ± 18.3
E2 (pg/ml)	58.9 ± 24.5	72.0 ± 36.0	65.1 ± 44.3	66.1 ± 59.6	57.1 ± 48.8	57.4 ± 51.9

^aExcluding three values greater than 200 pg/ml. 

SAS 9.1 and Macro facilities (SAS Institute) were used to perform the statistical analyses and plot the findings.

Results

The hormone characteristics of specimens from the 50 women at initial evaluation were as follows: mean AMH was 0.626 ± 0.51 ng/ml; mean inhibin B was 69.8 ± 45.2 pg/ml (vs. 65.3 ± 31.3 without outliers); mean FSH was 7.6 ± 2.4 mIU/ml; and the mean E2 level was 58.9 ± 24.5 pg/ml (see Table 1 for values across the six annual visits assayed in specimens collected in d 2–7 of the follicular phase of the menstrual cycle). At initial examination, the mean age of the women was 42 yr (sd = 2.7); mean BMI was 26.5 ± 5.3 kg/m² and 18% smoked. The mean age at FMP of the 50 women was 50.5 yr.

Figure 2 shows the pattern of _logAMH change in relation to the time to FMP based on fitted models shown in Table 2. The best-fitting regression model for _logAMH was a two-piece pattern with a linear decline of _logAMH values from −10 yr to a node located at 5 yr before the FMP. After the node 5 yr to the time of the FMP, a single line around the level of the assay detection level (0.05 ng/ml) was the best fit. Figure 2 shows the curvilinear patterns of _loginhibin B decline and _logFSH rise in relation to the time to FMP, based on regression models shown in Table 2. Inhibin B values declined to below the limits of the assay detection (10.0 pg/ml) 5 yr before the FMP at which time FSH levels had doubled to approximately 15 mIU/ml.

Figure 2.

Serum AMH declines to very low and nondetectable (*) levels 5 yr before the FMP (A), whereas inhibin B declines to very low and nondetectable (*) levels 4 yr before the FMP, accompanied by increased FSH levels in the absence of restraint by inhibin B (B). *, Lower levels of detection: AMH about 0.05 ng/ml; inhibin B about 10 pg/ml.

Table 2.

Association of AMH, inhibin B, and FSH in relation to time to FMP with and without dummy variables for values below assay detection (constrained to 11 yr prior to and 2 yr after FMP)

Covariates	β-Coefficient (se)	P value
AMH
	AMH model 1
logAMHa	−1.86 ± 0.13	<0.0001
	AMH model 2
logAMH (observed data)	−1.75 ± 0.14	<0.0001
AMH below detection (dummy variable)	6.57 ± 0.31	<0.0001
Inhibin B
	Inhibin B model 1b
logInhibin Ba	−1.76 ± 0.21	<0.0001
	Inhibin B model 2b
logInhibin B (observed data)	−1.84 ± 0.23	<0.0001
Inhibin B below detection (dummy variable)	−3.39 ± 0.92	0.0003
FSH
	FSH model 1c
logFSH	−1.54 ± 1.05	0.14
logFSH X logFSH (quadratic term)	0.71 ± 0.20	0.0006

^aModels exclude all observations less than lower limit of detection. 

^bInhibin models exclude n = 3 observations greater than 200 pg/ml. 

^cThere is no second model because no values are below detection. 

Table 3 shows that baseline _logAMH level, but not _logAMH change, was associated with age at FMP (P = 0.035). As shown in Table 3, there was no association of inhibin B baseline or profile over time values with age at FMP.

Table 3.

The association of age at FMP with _logAMH and _loginhibin B profiles

	β ± se	P value
logAMH intercept	0.83 ± 0.38	0.035
logAMH slope	0.75 ± 3.52	0.83
logInhibin B intercept	1.83 ± 1.77	0.31
logInhibin B slope	−0.07 ± 3.52	0.98

In direct comparisons in which statistical tests were used to compare AMH and inhibin B, inhibin B was not significant (P < 0.47), whereas AMH tended to be significant (P < 0.09). Therefore, we concluded that _logAMH was likely a more informative marker than inhibin B based direct comparisons using orthogonal contrasts and its significant associations with both age and time to FMP.

Discussion

These data indicate a linear decline in _logAMH profiles to values below detection at a time 5 yr before the FMP. This coupled with an understanding of the biological role of AMH in oocyte recruitment indicates that observation of these very low AMH levels could reflect the endocrine onset of the menopause transition in individual women. The concordance of observations between AMH and inhibin B tends to reinforce the soundness of the findings. The absence of detectable inhibin B occurring coincident with the time point 5 yr before the FMP is reflective of the diminishing FSH-sensitive antral follicle pool; furthermore, the inverse relationship between declining inhibin B and rising FSH is consistent with loss of FSH suppression by inhibin B (8).

The rationale for the predictive value of AMH for the time to FMP is predicated on the age-related decline in follicle number (23) and those processes whereby resting primordial follicles transition into growing follicles (2). The essence of the ovarian aging concept is that the timing of reproductive events, including the transition from menstrual cycle regularity to irregularity, and finally menopause (24), is dictated by the decreasing quantity and quality of the resting follicle pool. During the years before the FMP, oocytes undergo an accelerated rate of loss until eventually the stock of oocytes is mostly depleted. Faddy and colleagues (25,26) reported that this accelerated loss is initiated when the total number of oocytes reaches approximately 25,000 and that, on average, women reach this threshold at approximately 37–38 yr of age. At menopause, the number of oocytes has decreased to some hundreds (24,26,27,28). This 10- to 15-yr period between accelerated loss and the FMP leaves open the potential for other critical thresholds related to follicle recruitment and development. Because of its role in the regulation of primordial follicle recruitment and that it is produced at all follicle development stages until FSH dependency is achieved, AMH is a viable candidate to mark these critical thresholds (2). Our study suggests that AMH levels are indeed a marker of one those critical thresholds and this threshold occurs at 5 yr before the FMP.

The decrease in AMH that we observed with advancing age may be present before changes in other ovarian aging-related markers, and this evidence is largely supported with the limited number of longitudinal studies. Based on evaluation of 41 women studied twice over an interval ranging from 1.1 to 7.3 yr, one study reported a mean AMH reduction of 38%, although the number of antral follicles and the levels of FSH and inhibin B did not change (23). We identified a 20% loss/yr in _logAMH levels. Recently a report of 81 women (mean age of 39.6 yr) studied prospectively for 4 yr found that antral follicle count did not change over time, whereas AMH, FSH, and inhibin B did change significantly (29).

It has been argued that inhibin B reaches a threshold level in which the follicle pool is too limited for successful recruitment or that follicles are not responsive to the increasingly and persistently elevated FSH levels (30). Our data suggest that this constellation of endocrine events occurs 4–5 yr before the FMP. The actual timing is more difficult to identify with inhibin B values, compared with AMH, because the _loginhibin B values were decelerating in a curvilinear manner and the _logFSH values were accelerating in a curvilinear manner; in contrast, _logAMH values had a significant linear decline. In an Australian cohort, it was identified that the late phase of the menopause transition can be identified in menstrual cycling patterns commencing within 2.5 yr of the FMP (31), and Burger et al. (32) reported that in that same cohort, a decline in inhibin B was more correlated with menstrual irregularity than FSH values. AMH was not measured.

We identified three women who had single-time individual inhibin B values that were strikingly out of range (values >200 pg/ml), and Table 1 shows how their exclusion affects the mean and sd. However, E2 values, menstrual bleeding patterns, and laboratory duplication of measures suggested that these were biologically relevant values and that these may represent individual excursions related to incomplete follicle advancement beyond the intermediate stage as women experienced ovarian aging (33). It has been reported that polycystic ovarian syndrome is characterized by increased inhibin levels due to the persistence of a cohort of small follicles with incomplete follicular growth that contribute to the pool of circulating inhibins. It is further suggested that these elevated levels are associated with a relative FSH deficit (34).

This report includes strengths and limitations. The data are derived from appropriately collected specimens obtained from women in a population-based cohort participating in a study that has documented their natural progression through stages of the menopause transition to the postmenopause. The statistical analysis methodology appropriately acknowledged when values were below assay detection. The population is Caucasian, so findings may not be generalizable to women of other race/ethnic groups. Furthermore, samples were collected annually, and whereas these were specimens timed to menstrual bleeding, they may not adequately capture influential circadian or menstrual cycle variation. Additionally, counts of antral follicles were not available to substantiate the inferences.

In summary, we identified that the low levels and decline in AMH was associated with time to FMP and age at FMP. Furthermore, the majority of AMH values were not measurable, with precision, 5 yr before the FMP, yet this also predicted FMP. There was an expected inverse relation between inhibin B and FSH levels. The increasing curvilinear FSH values and the decreasing curvilinear values of inhibin B were related to time to FMP but had a less direct time to FMP interpretation than AMH and were not related to age at FMP. These data, coupled with the belief that there may be minimal menstrual cycle variability in AMH values, suggest that AMH may act as an endocrine marker of follicle depletion, may reflect minimal ovarian reserve, and may signal the onset of the late stage of the menopause transition.