ORIGINAL
E n d o c r i n e
ARTICLE
R e s e a r c h
Reproductive and Lifestyle Determinants of AntiMüllerian Hormone in a Large Population-based
Study
M. Dólleman, W. M. M. Verschuren, M. J. C. Eijkemans, M. E. T. Dollé,
E. H. J. M. Jansen, F. J. M. Broekmans, and Y. T. van der Schouw
Department of Reproductive Medicine and Gynaecology (M.D., F.J.M.B.); Julius Center for Health
Sciences and Primary Care (M.D., M.J.C.E., Y.T.v.d.S.), University Medical Center Utrecht, 3508 GA
Utrecht, The Netherlands; and National Institute for Public Health and the Environment (W.M.M.V.,
M.E.T.D., E.H.J.M.J.), 3720 BA Bilthoven, The Netherlands
Context: Anti-müllerian hormone (AMH) is an ovarian reserve marker that is increasingly applied in
clinical practice as a prognostic and diagnostic tool. Despite increased use of AMH in clinical practice,
large-scale studies addressing the influence of possible determinants on AMH levels are scarce.
Objective: We aimed to address the role of reproductive and lifestyle determinants of AMH in a
large population-based cohort of women.
Design: In this cross-sectional study, age-specific AMH percentiles were calculated using general
linear modeling with CG-LMS (Cole and Green, Lambda, Mu, and Sigma model, an established
method to calculate growth curves for children).
Setting: Women from the general community participating in the Doetinchem Cohort study were
assessed.
Participants: Two thousand three hundred twenty premenopausal women were included.
Main Outcome Measure: The effect of female reproductive and lifestyle factors on shifts in agespecific AMH percentiles was studied.
Results: In comparison to women with a regular menstrual cycle, current oral contraceptive (OC) users,
women with menstrual cycle irregularity, and pregnant women had significantly lower age-specific
AMH percentiles (for OC use, 11 percentiles lower; for cycle irregularity, 11 percentiles lower; and for
pregnancy, 17 percentiles lower [P value for all ⬍.0001]). Age at menarche and age at first childbirth
were not associated with the age-specific AMH percentile. Higher parity was associated with 2 percentiles higher age-specific AMH (P ⫽ .02). Of the lifestyle factors investigated, current smoking was
associated with 4 percentiles lower age-specific AMH percentiles (P ⫽ .02), irrespective of the smoking
dose. Body mass index, waist circumference, alcohol consumption, physical exercise, and socioeconomic status were not significantly associated with age-specific AMH percentiles.
Conclusions: This study demonstrates that several reproductive and lifestyle factors are associated
with age-specific AMH levels. The lower AMH levels associated with OC use and smoking seem
reversible, as effects were confined to current use of OC or cigarettes. It is important to give careful
consideration to the effect of such determinants when interpreting AMH in a clinical setting and
basing patient management on AMH. (J Clin Endocrinol Metab 98: 2106 –2115, 2013)
ISSN Print 0021-972X ISSN Online 1945-7197
Printed in U.S.A.
Copyright © 2013 by The Endocrine Society
Received November 26, 2012. Accepted February 27, 2013.
First Published Online March 26, 2013
2106
jcem.endojournals.org
Abbreviations: AMH, anti-müllerian hormone; BMI, body mass index; CG-LMS model, Cole
and Green Lambda Mu Sigma model; DCS, Doetinchem Cohort Study; OC, oral contraceptive; PCOS, polycystic ovary syndrome; SES, socioeconomic status.
J Clin Endocrinol Metab, May 2013, 98(5):2106 –2115
doi: 10.1210/jc.2012-3995
doi: 10.1210/jc.2012-3995
nti-müllerian hormone (AMH) is a marker of female
“ovarian reserve,” or the number of follicles remaining in a woman’s ovaries (1). AMH is increasingly
used in clinical practice for both prognostic purposes, such
as prediction of ovarian response during in vitro fertilization treatment, prediction of age at menopause, and diagnostic purposes, such as the identification of women
with premature ovarian insufficiency or polycystic ovary
syndrome (2– 6).
Several studies have presented nomograms for clinical
interpretation of AMH levels in different populations (7–
12). However, there are suggestions that AMH varies according to oral contraceptive (OC) use, smoking, and
body mass index (BMI), casting doubt on the clinical utility of these nomograms (13–17).
Although it is generally believed that smoking is deleterious to the ovarian reserve and AMH values (13, 18,
19), not all studies agree (8, 20) and most studies have not
been designed to look at dose response and the influence
of quitting smoking. Thus far, studies have remained inconclusive about the effect of other potential determinants, such as cycle regularity, current pregnancy, OC use,
parity, age at menarche, age at first childbirth, BMI, waist
circumference, alcohol consumption, physical exercise, or
socioeconomic status (13–16, 21–24). These are factors
that have been suggested to affect either female endocrinology, age at menarche, or age at menopause (25–27).
The aim of this study is therefore to address the possible
role of both reproductive and lifestyle determinants of
AMH in a large cohort of women from the general
population.
A
Materials and Methods
Participants
The Doetinchem Cohort Study (DCS) is an ongoing multipurpose prospective study, initially carried out in a random general population sample of men and women aged 20 to 59 years
(1987–1991). Over 95% of the DCS population are Caucasian
women (28). The aim of the DCS was to study the impact of
(changes in) lifestyle factors and biological risk factors on various aspects of health, such as the incidence of chronic diseases,
physical and cognitive functioning, and quality of life. The cohort is re-examined every 5 years with questionnaires and a physical examination at the local health service. Three follow-up
examination rounds were completed during 1993–1997, 1998 –
2002, and 2003–2007. All participants gave written informed
consent, and the study was approved according to the guidelines of the Helsinki Declaration by the Medical Ethics Committee of the Netherlands Organization of Applied Scientific
Research. Details on the DCS have been extensively described
elsewhere (29).
For the present study we used a cross-sectional design among
women participating in the second examination round of the
jcem.endojournals.org
2107
DCS. Women were excluded from the current study if they were
postmenopausal (n ⫽ 1700), if they were taking hormone replacement therapy (n ⫽ 23), if they had a history of surgery to the
uterus or ovaries (n ⫽ 26), or if the information on their reproductive status was missing (n ⫽ 59). After exclusion, 2 320 from
the original 4 128 women remained eligible for the current study.
Potential AMH determinants
The reproductive determinants included were age at menarche, OC use, cycle regularity status, pregnancy, parity, and age
at first childbirth. The lifestyle determinants encompassed BMI,
waist circumference, smoking, alcohol consumption, physical
exercise, and socioeconomic status.
Reproductive history was assessed via extensive questionnaires. Women were asked at what age they had their first menstrual period, if their period was regular, how long their menstrual cycle was in number of days, and how many menstruations
they had in the last 12 months. Furthermore women were asked
whether they were currently pregnant, if they had ever been pregnant, if so, how many pregnancies and how many children they
had, as well as when the children were born. In addition, women
were asked if they had ever used OC or hormone replacement
therapy, how old they were when they started using this medication, if they were using it at the time of blood sampling for
AMH measurements, and how many years they had used it in
total. It was also inventoried whether and at what age women
had undergone a hysterectomy and/or if they had ever been operated on one or both ovaries and at what age that occurred. A
regular cycle was defined as having a regular cycle with a mean
cycle length of 24 to 36 days. OC use was categorized as previous
OC use, current OC use, and never used OC. The number of
years of OC use was divided into ⬍1, 1–5, 5–10, 10 –15, 15–20,
and ⬎ 20 years. Reproductive status at time of blood sampling
was categorized into mutually exclusive categories of regularly
cycling, irregularly cycling, currently pregnant, or currently taking OC. Parity was assessed as a continuous variable.
Body weight and height were measured by trained staff. Body
weight was measured to the nearest 100 g on calibrated scales
with participants wearing light indoor clothing without shoes,
with emptied pockets (29). BMI was categorized as normal (⬍25
kg/m2), overweight (25–30 kg/m2), and obese (ⱖ30 kg/m2).
Waist circumference was measured to the nearest 0.5 cm according to World Health Organization instructions and categorized
into normal (⬍80 cm), moderately increased (80 – 88 cm), and
abdominal obesity (⬎88 cm) (30). Ever smokers were identified
based on the question “did you ever smoke regularly.” For ever
smokers, information on age at which the respondent started
smoking, as well as the total number of years of smoking and
average number of cigarettes smoked, was assessed, followed by
a question on current smoking (“do you smoke at present”).
Smoking was categorized as current smoker, ex-smoker, and
never smoker. The amount of smoking was categorized into the
following 5 levels: ⬍1 cigarette per month, ⬍1 cigarette per day,
1–9 cigarettes per day, 10 –19 cigarettes per day, and ⬎20 cigarettes per day. Pack-years of smoking was divided into 7 strata
with 5-year intervals. Alcohol intake was assessed in the questionnaire as number of glasses consumed per week, and subsequently operationalized as number of beverages consumed daily.
Physical activity was assessed using the validated (31) European
Prospective Investigation into Cancer and Nutrition questionnaire and categorized according to the Cambridge Physical
2108
Dólleman et al
Determinants of AMH
J Clin Endocrinol Metab, May 2013, 98(5):2106 –2115
linear regression and the extent of smoothing can be
modified by altering the degrees of freedom (33).
The fitted model was used to predict the age-specific
percentile of AMH of each woman in the database.
Data were analyzed with SPSS version 20.0 (SPSS
Inc, Chicago, Illinois) and with R version 2.13
(http://www.r-project.org/). The “gamlss” package,
which is a collection of functions to fit Generalized
Additive Models for Location Scale and Shape, was
used to perform the CG-LMS method in R
(www.gamlss.org) (34).
Determinants of AMH
Figure 1. Flowchart of included women.
Activity Index into (moderately) active and (moderately) inactive
(32). Socioeconomic status (SES) was classified into 4 categories
according to the highest level of education that a woman had
reached: primary school (SES level 1), lower secondary or vocational school (SES level 2), intermediate vocational or higher
secondary school (SES level 3), and higher vocational or university (SES level 4) (29).
AMH assay
Blood samples were collected on a random day of the menstrual or OC cycle. Serum was frozen on the day of vena cubiti
puncture and stored in liquid nitrogen. Before the AMH measurement, each sample went through one thaw-freeze cycle on ice
for 4 hours; intermitted storage of samples until AMH measurement was at ⫺80°C for a maximum of 4 weeks. Serum AMH was
measured with the AMH Gen II ELISA (Beckman-Coulter,
Sinsheim, Germany) in a single laboratory, by the same experienced laboratory technician. The precision of assay results was
validated with linearity-of-dilution assessment. The limit of detection for this assay is 0.08 ng/mL, and the limit of quantification is 0.16 ng/mL. The interassay and intra-assay coefficients of
variation were 3.35% and 4.0%, respectively.
Statistical analysis
Baseline characteristics were described for the group as a
whole with mean values and corresponding standard deviation
for continuous variables and frequencies and percentages for
categorical data.
Age-specific reference values of AMH
We age-standardized the distribution of AMH using the CGLMS method. This CG-LMS method, originally described by
Cole and Green, is an established method to estimate growth
curves for children. The LMS method summarizes the distribution of age-specific AMH by 3 smoothing splines, which represent the skewness (L), the median (M), and the coefficient of
variation (S). These 3 splines are simultaneously fitted by non-
The relationships between potential determinants
and age-specific AMH percentiles were assessed in the
cohort using univariable and multivariable regression
analysis. Determinants were investigated for both
linear and nonlinear associations with AMH by categorizing determinants into the clinically relevant
categories as described above. The magnitude of the
effect of the determinants was calculated as the number of percentiles of change. As percentiles of agestandardized AMH were used, no further correction
for age in a multivariable analysis was needed. The
change in percentiles is displayed with the standard error and the
corresponding P value. Individual P values for categorical values
in comparison to the reference value were calculated and the
overall P value was measured with ANOVA between groups
(p-ANOVA). For reproductive status at time of blood sampling,
regularly cycling women were chosen as the reference category
for the 3 other mutually exclusive categories of women. A P value
of ⱕ.05 was considered to be statistically significant. The effect
can thus be interpreted for dichotomous variables as “in the
presence of determinant x, age-specific AMH is y percentiles
lower/higher” and for continuous variables as “with a one unit
higher level of determinant x, the AMH level is y percentiles
higher/lower.”
Results
The study cohort comprised 2 320 women, as depicted
in the flowchart in Figure 1, of whom 1 090 were regularly
cycling, 908 were taking OC for family planning purposes,
49 were pregnant, and 273 had an irregular unpredictable
cycle. The baseline characteristics of the cohort are shown
in Table 1. Age-specific percentiles of AMH as estimated
using the CG-LMS model are depicted in Supplemental
Figure 1, published on The Endocrine Society’s Journals
Online web site at http://jcem.endojournals.org.
Reproductive determinants
There was no association between age at menarche and
AMH (␤ ⫽ .7, P ⫽ .09). In comparison to women with a
regular menstrual cycle, current OC use, current pregnancy, and having an irregular menstrual cycle were all
reproductive factors that were associated with having significantly lower age-specific AMH (Table 2). Current OC
doi: 10.1210/jc.2012-3995
jcem.endojournals.org
Table 1. Baseline Characteristics of the Cohort
Mean (SD) or % (n)
Lifestyle factors
Age, y
BMI
Waist circumference
Smoking
Current smokers, %
Ex-smokers, %
Never smoked, %
Number of cigarettes packs per
day among smokers
Number of pack years among
smokers
Alcohol
Drink alcohol daily, %
Amount of alcoholic beverages
per day among drinkers
Physical activity
CPAI 1 (inactive)
CPAI 2 (moderately inactive)
CPAI 3 (moderately active)
CPAI 4 (active)
Socioeconomic status
SES 1
SES 2
SES 3
SES 4
Reproductive factors
Age at menarche
Ever taken oral contraceptives
Parity
Age at birth of first child
Reproductive status at AMH
measurement
Regularly cycle
Irregular cycle
Oral contraceptives
Pregnant
2109
Table 2. Number of Percentiles that AMH Shifts in the
Presence of a Specific Reproductive Determinant
Linear Regression
37.3 (9.2)
24.3 (3.9)
84.6 (10.6)
Determinants
Percentiles
of Change
Standard
Error
P Value
Reference
⫺11.0
⫺11.0
⫺16.6
2.2
1.3
4.1
⬍.0001
⬍.0001
⬍.0001
Reference
⫺0.3
⫺9.0
2.3
2.3
.88
⬍.0001
Reference
2.6
⫺2.2
⫺2.6
⫺3.9
⫺8.8
0.7
2.1
⫺0.2
3.1
3.0
3.2
3.5
4.2
0.4
0.9
0.2
.39
.46
.42
.26
.04
.09
.02
.39
5.2% (120)
47.4% (1099)
31.7% (736)
15.3% (356)
Current reproductive
status
Regular cycle
Irregular cycle
Current OC use
Current pregnancy
History of OC use
Never
Previous
Current
Years of OC use in
current or past
users (n ⫽ 1996)
⬍1 y
1–5 y
5–10 y
10 –15 y
15–20 y
⬎20 y
Age at menarche
Parity
Age at birth of first child
13.2 (1.4)
90.6% (2102)
2.1 (0.8)
25.4 (4.0)
The association between reproductive determinants and age-specific
AMH is quantified as the number of percentiles that AMH shifts in the
presence of a specific determinant. Corresponding standard error and
P value are shown.
47% (1090)
11.8% (273)
39.1% (908)
2.1% (49)
Having given birth to a higher number of children was
associated with higher age-specific AMH levels (per additional child, ␤ ⫽ 2.1, P ⫽ .02). There was no association
between the age at first childbirth and AMH (␤ ⫽ ⫺.2, P ⫽
.39). Results are displayed in Table 2.
32.8% (760)
29.2% (678)
38% (882)
12.6 (7.5)
10.2 (9.1)
46% (1068)
1.1 (0.9)
2.8% (62)
22.4% (464)
31.8% (622)
43% (872)
Abbreviations: CPAI, Cambridge Physical Activity Index; SES,
socioeconomic status. Means and standard deviations (SD) are shown
for continuous variables. Percentage and numbers are shown for
categorical variables.
users had an age-specific AMH of 11 percentiles lower
than women with a regular cycle (␤ ⫽ 11; P ⬍ .0001).
When the history of OC use was assessed in the entire
cohort, current OC users had significantly lower AMH
values than women who had never used OC (␤ ⫽ ⫺9; P
value ⬍.0001) but previous OC use was not associated
with lower AMH values (␤ ⫽ ⫺.3; P ⫽ .9). There was no
clear dose-response relationship of duration of OC use
with age-specific AMH levels; only women who used OC
⬎20 years had a 8.8 ⫾ 4.2 percentiles lower age-specific
AMH compared to women who had used OC for a year or
less. Being pregnant was associated with age-specific
AMH values of almost 17 percentiles lower (␤ ⫽ 16.6; P ⬍
.0001) and having an irregular cycle with age-specific
AMH values of 11 percentiles lower (␤ ⫽ 11, P ⬍ .0001).
A significant trend was seen between parity and AMH.
Lifestyle determinants
There was no significant association between BMI (␤ ⫽
⫺.2; P ⫽ .16), waist circumference (␤ ⫽ ⫺.09; P ⫽ .12),
alcohol consumption (␤ ⫽ ⫺.4; P ⫽ .74), physical activity
(p-ANOVA ⫽ 0.62), socioeconomic status (p-ANOVA ⫽
0.62), and age-specific AMH. When determinants were
modeled in clinically relevant categories instead of linear
variables, there was still no significant association with
age-specific AMH (Table 3). Smoking, however, was associated with lower AMH percentiles: Current smoking
was associated with 3.6 percentiles lower age-specific
AMH values (P ⫽ .02) in comparison to never smokers.
Ex-smoking was not associated with lower AMH values
(␤ ⫽ ⫺.6. P ⫽ .67). Daily smoking was associated with
significant decreases in age-specific AMH. However, the
number of cigarettes smoked daily and AMH does not
seem to be related to AMH, as the number of percentiles
difference of age-specific AMH with nonsmokers was similar for women who smoked 1–9 cigarettes per day (␤ ⫽
2110
Dólleman et al
Determinants of AMH
J Clin Endocrinol Metab, May 2013, 98(5):2106 –2115
Table 3. Number of Percentiles that AMH Shifts in the
Presence of a Specific Lifestyle Determinant
Table 4. Multivariable Analysis of Determinants
Linear Regression
Linear Regression
Determinants
Anthropometrics
BMI
⬍25 kg/m2 (normal)
25–30 kg/m2
(overweight)
⬎30 kg/m2 (obese)
Waist circumference
⬍80 cm (normal)
80 – 88 cm (increased)
⬎88 cm (abdominal
obesity)
Smoking
Never smoked
Previous smoker
Current smoker
Number of cigarettes
smoked per day in
current smokers
Smoking frequency in
current smokers
⬍1 per month
⬍1 per day
1–9 per day
10 –19 per day
ⱖ20 per day
Number of pack-years in
current/previous
smokers
Pack-years in categories
0 –5
5–10
10 –15
15–20
20 –25
ⱖ25
Alcohol
Daily alcohol intake
(yes/no)
Number of alcoholic
beverages daily
Physical Exercise
CPAI 1 (inactive)
CPAI 2 (moderately
inactive)
CPAI 3 (moderately
active)
CPAI 4 (active)
Social economic status
SES 1 (lowest
education level)
SES 2
SES 3
SES 4 (highest
education level)
Percentiles
of Change
Standard
Error
P
Value
⫺0.2
Reference
⫺2.5
0.2
.16
1.4
.08
2.4
0.06
.52
.12
1.49
1.5
.19
.12
Reference
⫺0.6
⫺3.6
⫺0.3
1.5
1.5
0.1
.67
.02
.06
Reference
1.3
⫺12.8
⫺12.3
⫺11.8
⫺0.3
7.2
6.5
6.2
6.3
0.09
.86
.05
.05
.06
.001
⫺3.1
⫺7.0
⫺8.5
⫺6.6
⫺6.0
2.1
2.3
2.6
3.3
2.9
.15
.003
.001
.05
.04
⫺0.4
1.3
.74
1.6
1.0
.12
⫺5.1
4.2
.23
⫺2.8
4.2
.50
⫺4.4
Reference
4.1
.29
3.2
4.0
3.4
3.0
3.0
3.2
.28
.19
.29
⫺1.6
⫺0.09
Reference
⫺1.9
⫺2.3
Reference
Reference
The association between lifestyle determinants and age-specific AMH
is quantified as the number of percentiles that AMH shifts in the
presence of a specific determinant. Corresponding standard error and
P value are shown.
Determinants
Irregular cycle vs
regular cycle
Current OC-use vs
regular cycle
Current pregnancy vs
regular cycle
Current smoking vs
nonsmoking
Parity (per child)
Percentiles
of Change
Standard
Error
P Value
⫺10.8
2.4
⬍.0001
⫺12.7
1.7
⫺17.6
5.1
⫺4.5
1.6
⬍.001
1.4
0.9
.14
Multivariable analysis of the association between reproductive and
lifestyle determinants and age-specific AMH. The association is
quantified as the number of percentiles AMH shifts with the
corresponding standard error and P value.
⫺12.8; P ⫽ .05) as for women who smoked 10 –19
cigarettes daily (␤ ⫽ ⫺12.3; P ⫽ .05) or more than 20
cigarettes daily (␤ ⫽ ⫺11.8; P ⫽ .06, respectively). Although the linear association between the number of packyears smoked and AMH was statistically significant (␤ ⫽
⫺.3; P ⫽ .002), there was evidence for a threshold after
which the number of pack-years becomes significant.
Smoking 5–10 pack-years was not associated with significantly lower age-specific AMH values than smoking less
than 5 pack-years (␤ ⫽ ⫺3.1, P ⫽ .15), but smoking more
than 10 pack-years was associated with significantly
lower age-specific AMH values. The magnitude of the association after 10 –15 pack-years (␤ ⫽ ⫺7.0, P ⫽ .003)
was similar to 15–20 pack-years (␤ ⫽ ⫺8.5, P ⫽ .001) or
20 –25 pack-years (␤⫺6.6, P ⫽ .05). Results are displayed
in Table 3.
Multivariate analysis
Table 4 shows that the variables that were significantly
associated with age-specific AMH levels in the univariable
analysis also remained significantly associated after multivariable adjustment. The magnitude of the effects of current smoking, current OC use, pregnancy, cycle irregularity, and parity remained largely the same in the
multivariable models as in the univariable analysis. Only
parity was no longer significant in the multivariate analysis after correcting for OC use, cycle irregularity, current
pregnancy, and current smoking.
Figure 2 shows the median AMH values according to
the CG-LMS model in the 4 most clinically relevant subgroups: regularly cycling women and OC-taking women
who did and did not smoke at the time of blood sampling
for AMH measurement. In this figure, it is evident that the
difference in AMH values between OC-taking women
who did and did not smoke at the time of AMH measure-
doi: 10.1210/jc.2012-3995
jcem.endojournals.org
2111
Figure 2. Bar graph showing mean (p50) AMH values with increasing age according to the CG-LMS model for women in 4 subgroups.
ment was smaller than for regularly cycling women. A test
for interaction between smoking and OC use was done to
see if this would explain this discrepancy, but no significant interaction effect was found (P ⫽ .07).
Discussion
Main findings
This large cohort study demonstrates that age-specific
AMH levels are influenced by both reproductive factors
and lifestyle determinants. More specifically, we found 2
distinct factors strongly associated with age-specific AMH
levels: current use of oral contraceptives and current
smoking. This knowledge will ameliorate our understanding of AMH dynamics and its usefulness as a screening or
diagnosis tool in clinical practice.
Findings in view of existing literature
OC use at the time of blood sampling was associated
with significantly lower age-specific AMH levels. AMH
was 11 percentiles lower in current OC users compared to
women with a regular cycle and 9 percentiles lower compared to women who had never taken OC. One would
expect the effect of current OC use to be a weighted
average between previous and current OC use (⫺0.3 and
⫺9.0 percentiles). However, due to the occurrence of current pregnancy in the group of women with previous OC
use, the effect of current OC use is slightly lower than
expected. This effect of OC use was seen across all ages and
for all percentiles of AMH. The finding is not entirely
consistent with literature. Several studies have suggested
AMH to remain constant under the influence of exogenous sex steroids used for contraception (15, 21–24).
However, these studies consisted of small numbers, and
none assessed age-specific AMH levels. Other studies have
revealed a significant decrease in AMH during exposure to
OC (13, 14, 16, 35). Last, one prospective longitudinal
study observed an increase in AMH values of approximately 23% after discontinuation of OC, which supports
our notion of the reversibility of the effect of OC (17).
The observed reduction in AMH among OC users is
unlikely to be due to depletion of the primordial follicle
pool as OC users do not experience natural menopause at
an earlier age (36, 37). Furthermore, in women who were
previous OC users, AMH levels were similar to nonusers,
suggesting that the effect is reversible. The decrease in
AMH among current OC users is therefore more likely to
be due to down-regulation of the hypothalamic-pituitaryovarian axis by OC resulting in diminished FSH and LH
production and consequent altered development of the
antral follicle cohort (14, 38). It has also been argued that
AMH must be partially gonadotropin responsive as FSH
can directly stimulate AMH production from granulosa
cells in vitro (39, 40). However, direct stimulation by endogenous or exogenous FSH rises has not elicited any response in serum AMH levels (41). Therefore, most likely,
2112
Dólleman et al
Determinants of AMH
we may theorize that OC-induced arrest of follicular
growth and decreased support of growing follicles by FSH
causes AMH levels to be diminished. This notion is supported by several studies that observe a decrease in follicle
size and ovarian volume (14, 22, 23, 42) and other studies
that observe a decrease in antral follicle numbers (16, 23).
Furthermore, FSH is an important survival factor for the
AMH-producing preantral and early antral follicles (43).
As these are the stages of follicle development during
which most follicles undergo atresia (44), OC-induced
suppression of FSH may also result in either increased
apoptosis or decreased follicle functionality. This combined effect of smaller numbers of follicles and diminished
follicle functionality from decreased FSH support is a very
plausible explanation for the lower AMH values observed
in this study. All the above arguments may also apply to
the association between pregnancy and reduced AMH observed in this study and a previous longitudinal study as
pregnancy is a transient condition with severely suppressed endogenous FSH (45). However, during pregnancy the GnRH suppression is induced by endogenously
increased levels of estrogens instead of by exogenously
administered drugs (46). Further research is needed to assess whether AMH levels measured under OC use are correlated with natural cycle AMH measures. By doing so it
can be determined whether AMH under OC use is an adequate reflection of individual ovarian reserve and not just
the individual degree of suppression by OC. Interestingly,
van den Berg et al showed AMH levels on day 7 of the
pill-free interval to be highly correlated with AMH levels
after discontinuation of the pill (17). Similarly, Nelson et
al showed pregnant AMH measures to be highly correlated with nonpregnant measures (45), indicating a proportional relationship.
This study showed that cycle irregularity was associated with significantly lower age-specific AMH levels than
AMH levels in women with a regular cycle. This finding is
difficult to interpret, as this group encompasses women
with an oligomenorrhea associated with polycystic ovary
syndrome (PCOS; high age-specific AMH levels), women
with irregular menstrual cycles associated with the menopausal transition (low or undetectable AMH levels),
women with postpartum cycle arrest, or unexplained cycle
irregularity. In our study the mean age of women with an
irregular cycle was 44.6 years (⫾9 years), arguing that in
a proportion of these women cycle irregularity will be
explained by the occurrence of the menopausal transition.
This also explains the discrepancy with a recent article that
showed women with cycle irregularity to have higher
AMH values: in this population the mean age was 26.6
years and therefore high AMH values and cycle irregularity will more likely pertain to the occurrence of PCOS (47).
J Clin Endocrinol Metab, May 2013, 98(5):2106 –2115
Our results also showed that women who have given birth
to more children are likely to have significantly higher
AMH levels. Recent prospective studies have shown that
having a high AMH is associated with a higher age at
menopause (4, 48, 49). As a timed interval of approximately 10 years is suggested to be present between the final
menstrual period and the end of natural fertility, it is conceivable that the time window to conceive is longer in
women with high age-specific AMH levels, thus potentially resulting in more childbirths (50). One small study of
294 women aged 21 to 22 years paradoxically showed
that a lower AMH was associated with a higher parity and
a higher age at menarche. However, in this study contraceptive use was not taken into account, hindering direct
comparison to our study. It is conceivable that women
with a higher parity may be more likely to take contraceptive measures, which may explain lower AMH values.
Furthermore, the sample of women who conceived 2 or
more children was very small (only 9 women had 3 or more
children), which may not be a representative sample from
which to draw such conclusions. With regard to age at
menarche, their finding is paradoxical with most literature, as one would expect women with a high AMH to
experience menarche later due to the occurrence of PCOS.
Without correcting for contraceptive use, however, inferences based on these results may be risky (51).
Of the lifestyle factors investigated, current smoking
was most strongly inversely associated with AMH concentration. This is in line with existing research (13, 18,
19). Although the effect of smoking was dose-dependent
in the sense that one had to smoke at least daily, the effect
was similar in women who smoked 1–10 cigarettes daily
as women who smoked more than 20 cigarettes daily.
There was a similar association with the number of packyears smoked, where a significant association was seen
after 10 –15 years, but after 15 pack-years the magnitude
of the effect did not change very much for women who
smoked more than 20 pack-years. Previous smoking was
not significantly associated with decreased AMH in both
our study and a previous article (18). Translated to a clinical setting, our results thus suggest that if a low age-specific AMH level is measured in a woman of reproductive
age, advising her to stop smoking may be beneficial for her
AMH levels. Our results suggest that the association between AMH and smoking is similar to the association
between age at natural menopause and smoking as both
are dose responsive and both effects diminish after quitting smoking (26, 52, 53). The reversibility of the effect of
smoking on AMH and age at natural menopause is contradictory to the theory that the toxic effect of smoking on
ovarian follicles results in earlier depletion of the primordial follicle pool, as such an effect would be irreversible
doi: 10.1210/jc.2012-3995
(26). Several other explanations could exist. First, although toxicity to human follicles has been confirmed
(54), current smoking may only influence preantral and
antral follicles but not primordial follicles (18). The alternation of the functionality of the antral follicle pool would
explain lower AMH levels and, if the primordial pool remained unaltered by smoking, quitting smoking would
allow both restoration of the growing follicle pool and
normalization of AMH (18). Alternatively, smoking may
alter steroidogenesis or the hypothalamic-pituitary-ovarian axis, as suggested by studies that indicate altered levels
of sex hormones among smokers (19). This could also lead
to lower antral follicle counts and decreased AMH values.
The lower age at menopause in smokers could then be
explained, not by earlier depletion of the primordial follicle pool, but by the number of antral follicles becoming
too small to sustain monthly menstrual cycles. More research is needed to elucidate the exact mechanisms between smoking and measures of ovarian reserve. There
was no association between BMI and AMH. This is in line
with some (55–57) but not all previous studies (9, 15, 58).
It is important to note that the BMI in our study sample
was very homogenous with few obese women (mean
BMI ⫽ 24.3 ⫾ 3.9 SD).
Strengths and weaknesses
The most important strengths of the present cohort are
the large size of the cohort, the fact that it is populationbased, and the extensive measurements and inventories on
lifestyle factors and menstrual cycle characteristics that
were collected starting at a relatively young age of 20
years. Also, the statistical analysis, using the CG-LMS
model, provides advantages over classical approaches.
Usually, linear regression is used with AMH on a log-scale
to model the age related decline of AMH. The effect of
determinants can then be tested with multivariate regression analysis while correcting for age. The LMS model has
an advantage over this linear regression model as it does
not assume that the skewness, the median, and the coefficient of variation are the same at different ages but models them as a function of age. In this way age is more
appropriately corrected for when assessing the influence
of possible determinants on AMH concentrations. Last,
the AMH Gen II assay (Beckman Coulter Ltd) was used,
which is believed to be the newest most generalizable measure of AMH. The samples were determined in a single
laboratory, by the same experienced laboratory technician, and the precision of assay results was validated with
linearity-of-dilution assessment. Samples were frozen
within 12 hours and cryo-preserved within 6 weeks. Efficiency reasons required all samples to go through one
temperature-controlled freeze-thaw cycle before the
jcem.endojournals.org
2113
AMH assay was performed. These are all factors that argue for the presence of homogeneous specimen processing
and supports that we have provided reliable, reproducible
AMH measures (59). However, the Gentech II assay remains a manual ELISA. Future automated ELISAs may
provide even more stable measures of AMH. We did not
time the collection of our blood samples on the same menstrual cycle day for all participants. However, it has been
shown that intracycle fluctuation in AMH does not follow
classical endocrine patterns, but that it occurs randomly
throughout the menstrual cycle, thus supporting our
choice of protocol (60). One weakness of the current study
might be that there is just one measure of AMH per woman; furthermore, the cross-sectional approach used in this
study did not allow us to assess all aspects of AMH suppression by certain determinants. However we sought to
make the situation in our study most comparable to clinical practice in which a physician is usually confronted
with just one AMH measure. In such a situation it is important that the physician is aware of which factors may
have had an influence on the AMH level that he is confronted with.
Clinical implications
The impact of the studied determinants for clinical interpretation depends on whether a woman has a high or a
low AMH for her age to start with. If a woman is above the
75th percentile of AMH for her age, a considerable drop
to a lower percentile may not be problematic, but for a
woman in the 30th percentile, it is conceivable that the
effect may be worrisome. Current smoking was associated
with a decrease of AMH of 4 percentiles. For a 26-year-old
woman this means that if she were in the 50th percentile
of AMH without smoking she would be in the 46th percentile if she did smoke. This is comparable to a woman of
28 years in the 50th percentile. Current smoking thus adds
2 years to her reproductive age in this example. The same
calculation can be done for a 36-year-old woman in the
50th percentile in which case her AMH would be comparable to a 37- to 38-year-old woman if she were to smoke.
For a 40-year-old woman it would thus add 1.5 years to
her reproductive age.
AMH is increasingly applied in clinical practice. Several
studies have presented nomograms for age-specific AMH
values in a variety of populations with AMH measured
with a variety of assays (7–12). This study shows that
AMH concentrations are influenced by several determinants and may thus aid in the interpretation of AMH results. Future studies on the effects of determinants on
AMH would benefit from a prospective longitudinal design to help elucidate the precise mechanisms behind
AMH suppression. The future clinical implications of hav-
2114
Dólleman et al
Determinants of AMH
ing a high or low AMH for one’s age are largely unknown.
More large-scale, long-term prospective studies are
needed to assess the impact of having a low or high agespecific AMH. Until this has been done, there is no concrete place for AMH nomograms in a clinical setting.
Conclusion
In conclusion, our study adds valuable information to the
body of knowledge currently available on AMH in the
general population. The lower AMH levels associated
with OC use and smoking seem reversible, as effects were
confined to current use. It is important to give careful
consideration to the effect of such determinants when interpreting AMH in a clinical setting.
Acknowledgments
We thank J.W.J.M. Cremers for technical assistance with AMH
determinations.
Address all correspondence and requests for reprints to: M.
Dólleman, University Medical Center Utrecht, Room 05.126,
P.O. Box 85500, 3508 GA, Utrecht, The Netherlands. E-mail:
[email protected].
No funding was received for the writing of this manuscript.
Disclosure Summary: M.D., W.M.M.V., M.J.C.E., M.E.T.D.,
E.H.J.M., and Y.T.S. have nothing to disclose. F.J.B. has received fees
and grant support from the following companies: Ferring, Gedeon
Richter, Merck Serono, Medical Specialties Distributors, and Roche.
J Clin Endocrinol Metab, May 2013, 98(5):2106 –2115
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
References
1. Broekmans FJ, Visser JA, Laven JS, Broer SL, Themmen AP, Fauser
BC. Anti-Mullerian hormone and ovarian dysfunction. Trends Endocrinol Metab. 2008;19:340 –347.
2. Broer SL, Mol B, Dolleman M, Fauser BC, Broekmans FJ. The role
of anti-Mullerian hormone assessment in assisted reproductive technology outcome. Curr Opin Obstet Gynecol. 2010;22:193–201.
3. Broer SL, Mol BW, Hendriks D, Broekmans FJ. The role of
antimullerian hormone in prediction of outcome after IVF: comparison with the antral follicle count. Fertil Steril. 2009;91:705–714.
4. Broer SL, Eijkemans MJ, Scheffer GJ, et al. Anti-mullerian hormone
predicts menopause: a long-term follow-up study in normoovulatory women. J Clin Endocrinol Metab. 2011;96:2532–2539.
5. Knauff EA, Eijkemans MJ, Lambalk CB, et al. Anti-Mullerian hormone, inhibin B, and antral follicle count in young women with
ovarian failure. J Clin Endocrinol Metab. 2009;94:786 –792.
6. Pigny P, Jonard S, Robert Y, Dewailly D. Serum anti-Mullerian
hormone as a surrogate for antral follicle count for definition of the
polycystic ovary syndrome. J Clin Endocrinol Metab. 2006;91:941–
945.
7. Lie FS, Visser JA, Welt CK, et al. Serum anti-Mullerian hormone
levels in healthy females: a nomogram ranging from infancy to adulthood. J Clin Endocrinol Metab. 2012;97:4650 – 4655.
8. La Marca A, Spada E, Grisendi V, et al. Normal serum anti-Mullerian hormone levels in the general female population and the relationship with reproductive history. Eur J Obstet Gynecol Reprod
Biol. 2012;163:180 –184.
9. La Marca A, Sighinolfi G, Giulini S, et al. Normal serum concen-
21.
22.
23.
24.
25.
26.
27.
28.
trations of anti-Mullerian hormone in women with regular menstrual cycles. Reprod Biomed Online. 2010;21:463– 469.
van Disseldorp J, Faddy MJ, Themmen AP, et al. Relationship of
serum antimullerian hormone concentration to age at menopause.
J Clin Endocrinol Metab. 2008;93:2129 –2134.
Nelson SM, Messow MC, McConnachie A, et al. External validation of nomogram for the decline in serum anti-Mullerian hormone
in women: a population study of 15,834 infertility patients. Reprod
Biomed Online. 2011;23:204 –206.
Nelson SM, Messow MC, Wallace AM, Fleming R, McConnachie
A. Nomogram for the decline in serum antimullerian hormone: a
population study of 9,601 infertility patients. Fertil Steril. 2011;95:
736 –741.
Shaw CM, Stanczyk FZ, Egleston BL, et al. Serum antimullerian
hormone in healthy premenopausal women. Fertil Steril. 2011;95:
2718 –2721.
Arbo E, Vetori DV, Jimenez MF, Freitas FM, Lemos N, Cunha-Filho
JS. Serum anti-mullerian hormone levels and follicular cohort characteristics after pituitary suppression in the late luteal phase with
oral contraceptive pills. Hum Reprod. 2007;22:3192–3196.
Steiner AZ, Stanczyk FZ, Patel S, Edelman A. Antimullerian hormone and obesity: insights in oral contraceptive users. Contraception. 2010;81:245–248.
Kristensen SL, Ramlau-Hansen CH, Andersen CY, et al. The association between circulating levels of antimullerian hormone and follicle number, androgens, and menstrual cycle characteristics in
young women. Fertil Steril. 2012;97:779 –785.
van den Berg MH, van Dulmen-den Broeder E, Overbeek A, et al.
Comparison of ovarian function markers in users of hormonal contraceptives during the hormone-free interval and subsequent natural
early follicular phases. Hum Reprod. 2010;25:1520 –1527.
Plante BJ, Cooper GS, Baird DD, Steiner AZ. The impact of smoking
on antimullerian hormone levels in women aged 38 to 50 years.
Menopause. 2010;17:571–576.
Schuh-Huerta SM, Johnson NA, Rosen MP, Sternfeld B, Cedars MI,
Reijo Pera RA. Genetic variants and environmental factors associated with hormonal markers of ovarian reserve in Caucasian and
African American women. Hum Reprod. 2012;27:594 – 608.
Dafopoulos A, Dafopoulos K, Georgoulias P, et al. Smoking and
AMH levels in women with normal reproductive history. Arch Gynecol Obstet. 2010;282:215–219.
Streuli I, Fraisse T, Pillet C, Ibecheole V, Bischof P, de Ziegler D.
Serum antimullerian hormone levels remain stable throughout the
menstrual cycle and after oral or vaginal administration of synthetic
sex steroids. Fertil Steril. 2008;90:395– 400.
Deb S, Campbell BK, Pincott-Allen C, Clewes JS, Cumberpatch G,
Raine-Fenning NJ. Quantifying effect of combined oral contraceptive pill on functional ovarian reserve as measured by serum antiMullerian hormone and small antral follicle count using three-dimensional ultrasound. Ultrasound Obstet Gynecol. 2012;39:574 –
580.
Somunkiran A, Yavuz T, Yucel O, Ozdemir I. Anti-Mullerian hormone levels during hormonal contraception in women with polycystic ovary syndrome. Eur J Obstet Gynecol Reprod Biol. 2007;
134:196 –201.
Li HW, Wong CY, Yeung WS, Ho PC, Ng EH. Serum anti-mullerian
hormone level is not altered in women using hormonal contraceptives. Contraception. 2011;83:582–585.
Dossus L, Kvaskoff M, Bijon A, et al. Determinants of age at menarche and time to menstrual cycle regularity in the French E3N
cohort. Ann Epidemiol. 2012;22:723–730.
Gold EB. The timing of the age at which natural menopause occurs.
Obstet Gynecol Clin North Am. 2011;38:425– 440.
Morris DH, Jones ME, Schoemaker MJ, McFadden E, Ashworth A,
Swerdlow AJ. Body mass index, exercise, and other lifestyle factors
in relation to age at natural menopause: analyses from the breakthrough generations study. Am J Epidemiol. 2012;175:998 –1005.
Ganesh SK, Tragante V, Guo W, et al. Loci influencing blood pres-
doi: 10.1210/jc.2012-3995
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
jcem.endojournals.org
sure identified using a cardiovascular gene-centric array [published
online February 7, 2013]. Hum Mol Genet. doi:10.1093/hmg/
dds555.
Verschuren WM, Blokstra A, Picavet HS, Smit HA. Cohort profile:
the Doetinchem Cohort Study. Int J Epidemiol. 2008;37:1236 –
1241.
WHO. Waist circumference and waist– hip ratio: report of a WHO
expert consultation. Geneva: World Health Organization; 2008.
Pols MA, Peeters PH, Ocke MC, et al. Estimation of reproducibility
and relative validity of the questions included in the EPIC Physical
Activity Questionnaire. Int J Epidemiol 26 Suppl. 1997;1:S181–
S189.
Wareham NJ, Jakes RW, Rennie KL, et al. Validity and repeatability
of a simple index derived from the short physical activity questionnaire used in the European Prospective Investigation into Cancer
and Nutrition (EPIC) study. Public Health Nutr. 2003;6:407– 413.
Cole TJ, Green PJ. Smoothing reference centile curves: the LMS
method and penalized likelihood. Stat Med. 1992;11:1305–1319.
Stasinopoulos DM, Rigby RA. Generalized Additive Models for
Location Scale and Shape (GAMLSS) in R. J Statist Software 2007;
23:2891–2911.
Bentzen JG, Forman JL, Larsen EC, et al. Maternal menopause as a
predictor of anti-Mullerian hormone level and antral follicle count
in daughters during reproductive age. Hum Reprod. 2012;28:247–
255.
van Noord PA, Dubas JS, Dorland M, Boersma H, te Velde E. Age
at natural menopause in a population-based screening cohort: the
role of menarche, fecundity, and lifestyle factors. Fertil Steril. 1997;
68:95–102.
Gold EB, Bromberger J, Crawford S, et al. Factors associated with
age at natural menopause in a multiethnic sample of midlife women.
Am J Epidemiol. 2001;153:865– 874.
Pache TD, Wladimiroff JW, de Jong FH, Hop WC, Fauser BC.
Growth patterns of nondominant ovarian follicles during the normal menstrual cycle. Fertil Steril. 1990;54:638 – 642.
Taieb J, Grynberg M, Pierre A, et al. FSH and its second messenger
cAMP stimulate the transcription of human anti-Mullerian hormone in cultured granulosa cells. Mol Endocrinol. 2011;25:645–
655.
Hagen CP, Sorensen K, Anderson RA, Juul A. Serum levels of antimullerian hormone in early maturing girls before, during, and after
suppression with GnRH agonist. Fertil Steril. 2012;98:1326 –1330.
van Rooij I, Broekmans FJ, te Velde ER, et al. Serum anti-Mullerian
hormone levels: a novel measure of ovarian reserve. Hum Reprod.
2002;17:3065–3071.
Jensen AM, Brocks V, Holm K, Laursen EM, Muller J. Central
precocious puberty in girls: internal genitalia before, during, and
after treatment with long-acting gonadotropin-releasing hormone
analogues. J Pediatr. 1998;132:105–108.
Markstrom E, Svensson EC, Shao R, Svanberg B, Billig H. Survival
factors regulating ovarian apoptosis— dependence on follicle differentiation. Reproduction. 2002;123:23–30.
2115
44. Gougeon A. Regulation of ovarian follicular development in primates: facts and hypotheses. Endocr Rev. 1996;17:121–155.
45. Nelson SM, Stewart F, Fleming R, Freeman DJ. Longitudinal assessment of antimullerian hormone during pregnancy-relationship
with maternal adiposity, insulin, and adiponectin. Fertil Steril.
2010;93:1356 –1358.
46. Weghofer A, Dietrich W, Barad DH, Gleicher N. Live birth chances
in women with extremely low-serum anti-Mullerian hormone levels.
Hum Reprod. 2011;26:1905–1909.
47. Hagen CP, Vestergaard S, Juul A, et al. Low concentration of circulating antimullerian hormone is not predictive of reduced fecundability in young healthy women: a prospective cohort study. Fertil
Steril. 2012;98:1602–1608.
48. Tehrani FR, Shakeri N, Solaymani-Dodaran M, Azizi F. Predicting
age at menopause from serum antimullerian hormone concentration. Menopause. 2011;18:766 –770.
49. Freeman EW, Sammel MD, Lin H, Gracia CR. Anti-Mullerian Hormone as a Predictor of Time to Menopause in Late Reproductive Age
Women. J Clin Endocrinol Metab. 2012;97:1673–1680.
50. Broekmans FJ, Knauff EA, te Velde ER, Macklon NS, Fauser BC.
Female reproductive ageing: current knowledge and future trends.
Trends Endocrinol Metab. 2007;18:58 – 65.
51. Bragg JM, Kuzawa CW, Agustin SS, Banerjee MN, McDade TW.
Age at menarche and parity are independently associated with AntiMullerian hormone, a marker of ovarian reserve, in Filipino young
adult women. Am J Hum Biol. 2012;24:739 –745.
52. Adena MA, Gallagher HG. Cigarette smoking and the age at menopause. Ann Hum Biol. 1982;9:121–130.
53. Midgette AS, Baron JA. Cigarette smoking and the risk of natural
menopause. Epidemiology. 1990;1:474 – 480.
54. Mattison DR, Thorgeirsson SS. Smoking and industrial pollution,
and their effects on menopause and ovarian cancer. Lancet. 1978;
1:187–188.
55. Park HT, Cho GJ, Ahn KH, et al. Association of insulin resistance
with anti-Mullerian hormone levels in women without polycystic
ovary syndrome (PCOS). Clin Endocrinol (Oxf). 2010;72:26 –31.
56. Skalba P, Cygal A, Madej P, et al. Is the plasma anti-Mullerian
hormone (AMH) level associated with body weight and metabolic,
and hormonal disturbances in women with and without polycystic
ovary syndrome? Eur J Obstet Gynecol Reprod Biol. 2011;158:
254 –259.
57. Halawaty S, ElKattan E, Azab H, ElGhamry N, Al-Inany H. Effect
of obesity on parameters of ovarian reserve in premenopausal
women. J Obstet Gynaecol Can. 2010;32:687– 690.
58. Freeman EW, Gracia CR, Sammel MD, et al. Association of antimullerian hormone levels with obesity in late reproductive-age
women. Fertil Steril. 2007;87:101–106.
59. Rustamov O, Smith A, Roberts SA, et al. Anti-Mullerian hormone:
poor assay reproducibility in a large cohort of subjects suggests
sample instability. Hum Reprod. 2012;27:3085–3091.
60. Overbeek A, Broekmans FJ, Hehenkamp WJ, et al. Intra-cycle fluctuations of anti-Mullerian hormone in normal women with a regular
cycle: a re-analysis. Reprod Biomed Online. 2012;24:664 – 669.
Save the Date for ENDO 2013
June 15 - 18, 2013, San Francisco, California
www.endo-society.org/endo2013