A systematic review of the association between modifiable lifestyle factors and circulating anti-Müllerian hormone

Abstract

BACKGROUND

Levels of anti-Müllerian hormone (AMH) are known to be associated with lifestyle determinants such as smoking and oral contraception (OC) use. When measuring AMH in clinical practice, it is essential to know which factors may influence circulating levels or ovarian reserve in general.

OBJECTIVE AND RATIONALE

To date, there is no systematic review or summarizing consensus of the nature and magnitude of the relation between AMH and modifiable lifestyle factors. The purpose of this review was to systematically assess the evidence on association of lifestyle behaviors with circulating AMH levels.

SEARCH METHODS

We performed a pre-registered systematic review of publications in Embase and PubMed on the lifestyle factors BMI, smoking, OC use, alcohol consumption, caffeine consumption, physical activity, and waist–hip ratio (WHR) in relation to circulating AMH levels up to 1 November 2023. The search strategy included terms such as ‘Anti-Mullerian hormone’, ‘lifestyle’, and ‘women’. Studies were considered eligible if the association between at least one of the lifestyle factors of interest and AMH was assessed in adult women. The quality of included studies was assessed using the Study Quality Assessment Tools of the National Heart, Lung, and Blood Institute. The results were presented as ranges of the most frequently used association measure for studies that found a significant association in the same direction.

OUTCOMES

A total of 15 072 records were identified, of which 65 studies were eligible for inclusion, and 66.2% of the studies used a cross-sectional design. The majority of studies investigating BMI, smoking, OC use, and physical activity reported significant inverse associations with AMH levels. For WHR, alcohol, and caffeine use, the majority of studies did not find an association with AMH. For all determinants, the effect measures of the reported associations were heterogeneous. The mean difference in AMH levels per unit increase in BMI ranged from −0.015 to −0.2 ng/ml in studies that found a significant inverse association. The mean difference in AMH levels for current smokers versus non-smokers ranged from −0.4 to −1.1 ng/ml, and −4% to −44%, respectively. For current OC use, results included a range in relative mean differences in AMH levels of −17% to −31.1%, in addition to a decrease of 11 age-standardized percentiles, and an average decrease of 1.97 ng/ml after 9 weeks of OC use. Exercise interventions led to a decrease in AMH levels of 2.8 pmol/l to 13.2 pmol/l after 12 weeks in women with polycystic ovary syndrome or a sedentary lifestyle.

WIDER IMPLICATIONS

Lifestyle factors are associated with differences in AMH levels and thus should be taken into account when interpreting individual AMH measurements. Furthermore, AMH levels can be influenced by the alteration of lifestyle behaviors. While this can be a helpful tool for clinical and lifestyle counseling, the nature of the relation between the observed differences in AMH and the true ovarian reserve remains to be assessed.

REGISTRATION NUMBER

PROSPERO registration ID: CRD42022322575

Systematic review of the associations between lifestyle factors and anti-Müllerian hormone in women. The figure was partly generated using modified pictures from Servier Medical Art, provided by Servier, licensed under a Creative Commons Attribution 3.0 unported license. OC, oral contraceptive; AMH, anti-Müllerian hormone.

anti-Müllerian hormone, AMH, ovarian reserve, lifestyle, BMI, smoking, oral contraception, physical activity, waist–hip ratio, systematic review

Introduction

In women, anti-Müllerian hormone (AMH) plays an important role in the regulation of folliculogenesis (di Clemente et al., 2021). It is produced by granulosa cells of growing follicles from the primary up to the small antral stage (Moolhuijsen and Visser, 2020). Over the past two decades, AMH has been put forward as a proxy marker of the ‘ovarian reserve’, which is the term used to indicate the number of remaining follicles in the ovaries (Broekmans et al., 2009; Broer et al., 2014). Circulating AMH levels progressively decline with age until a few years prior to the menopause and strongly correlate with the number of recruited growing follicles (Dewailly et al., 2014). The number of recruited growing follicles in turn reflects the number of primordial follicles (Moolhuijsen and Visser, 2020). Hence, although AMH is not produced by primordial follicles, it is the best marker for the ovarian reserve to date (La Marca and Volpe, 2006; Broer et al., 2014).

Testing of ovarian reserve using AMH occurs in several clinical settings, including the assessment of fertility treatment regimen, the assessment of cycle disturbances or, increasingly, as an indicator of the remaining ovarian reserve pool in the context of (postponement of) family planning or fertility preservation (Copp et al., 2023). However, the interpretation of an individual’s AMH level can be complex, as AMH levels are influenced by demographic, socioeconomic, genetic, reproductive, environmental, and lifestyle factors (Jung et al., 2017; Shahrokhi et al., 2018) and do not reliably predict individual age at menopause (Depmann et al., 2018). A good sense of which determinants contribute to AMH levels may help clinicians to interpret AMH levels by taking more into account than just their patient’s age, and provide a better base for any resulting treatment decisions.

A history of smoking, oral contraceptive (OC) use and body weight are routinely assessed in clinical practice and have frequently been studied in relation to AMH levels. However, although there may be a well-known previously established association with these risk factors, to date there does not seem to be a consensus on the effect size of these determinants. Other lifestyle factors that are routinely or easily assessed in a clinical work-up are alcohol consumption, caffeine consumption, physical activity, and waist–hip ratio (WHR), which may additionally be related to AMH. Despite several prior reviews on this topic (Abdolahian et al., 2020; Oldfield et al., 2021; Kloos et al., 2022), to our knowledge, the association of multiple lifestyle factors with AMH in the general or an unselected population has not yet been evaluated systematically. Therefore, the aim of the present review was to systematically assess the associations between the lifestyle factors BMI, smoking, OC use, alcohol consumption, caffeine consumption, physical activity, and WHR with circulating AMH.

Methods

Search strategy

This review was registered in PROSPERO (ID: CRD42022322575). We performed a systematic search of existing publications on lifestyle factors in relation to circulating AMH levels and/or AMH decline available on Embase and PubMed on 4 April 2022, followed by two updates on 25 May 2022 and 1 November 2023. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were followed for this review (Page et al., 2021). The search strategy consisted of Emtree and MeSH terms relating to AMH and the lifestyle factors BMI, smoking, OC use, alcohol consumption, caffeine consumption, physical activity, and WHR, as well as search terms for lifestyle in general to increase sensitivity of the search. Additionally, synonyms and title/abstract keywords of AMH and the lifestyle factors were used. We directed our search string to adult women by including the appropriate Emtree and MeSH terms and title/abstract keywords and excluding Emtree and MeSH terms for men and children. The reference lists of included articles were checked to identify relevant studies not yet included. The search string is presented in Table 1. We intended to include age at natural menopause as secondary outcome and representation of ‘true’ ovarian reserve, but given the high number of papers that the search strategy yielded, we decided to focus the present paper on AMH and made a post hoc exclusion of papers on age at natural menopause.

Table 1.

Systematic review of lifestyle factors and anti-Müllerian hormone; utilized search terms for PubMed and Embase.*

Pubmed Search String
#1	woman [tiab] OR women [tiab] OR female* [tiab] OR Women [Mesh]
#2	Anti-mullerian Hormone [MeSH] OR xercise [tiab] OR ovarian reserve [tiab] OR Ovarian Reserve [MeSH] OR ovarian ag* [tiab] OR reproductive ag* [tiab]
#3	age at menopause [tiab] OR age at natural menopause [tiab] OR age of menopause [tiab] OR menopausal ag* [tiab]
#4	(#2 OR #3) AND #1
#5	smoker* [tiab] OR smoking [tiab] OR cigarette* [tiab] OR tobacco [tiab] OR Tobacco Smoking [MeSH] OR Smoking Cessation [MeSH] OR Tobacco Use Cessation [MeSH] OR Smoking Reduction [MeSH]
#6	Alcohol Drinking [MeSH] OR alcohol [tiab] OR ethanol [tiab]
#7	Caffeine [MeSH] OR caffeine [tiab] OR Coffee [MeSH] OR coffee [tiab] OR Tea [MeSH] OR tea [tiab] OR teas [tiab]
#8	Contraceptives, Oral [MeSH] OR oral contracept* [tiab] OR the pill [tiab]
#9	Exercise [MeSH] OR xercise* [tiab] OR physical activit* [tiab] OR fitness [tiab] OR sport* [tiab] OR training* [tiab]
#10	Body Mass Index [MeSH] OR body mass index [tiab] OR bmi [tiab] OR Weight Loss [MeSH] OR weight loss* [tiab] OR Waist-Hip Ratio [MeSH] OR waist-hip ratio* [tiab] OR waist-to-hip ratio [tiab] OR Waist Circumference [MeSH] OR waist circumference* [tiab] OR hip circumference* [tiab]
#11	Life Style [MeSH] OR life style* [tiab] OR lifestyle* [tiab]
#12	(#5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11) AND #4
#13	#12 NOT infant[MeSH] NOT child[MeSH] NOT men[MeSH]
Embase Search String
#1	‘female’/exp OR ‘female*’:ti, ab, kw OR ‘woman’:ti, ab, kw OR ‘women’:ti, ab, kw
#2	‘muellerian inhibiting factor’/exp OR ‘mullerian’:ti, ab, kw OR ‘muellerian’:ti, ab, kw OR ‘ovarian reserve’:ti, ab, kw OR ‘ovarian ag*’:ti, ab, kw OR ‘reproductive ag*’:ti, ab, kw
#3	‘age at menopause’:ti, ab, kw OR ‘age at natural menopause’:ti, ab, kw OR ‘age of menopause’:ti, ab, kw
#4	#2 OR #3
#5	#1 AND #4
#6	‘tobacco use’/exp OR ‘cigarette’/exp OR ‘smoker*’:ti, ab, kw OR ‘smoking’:ti, ab, kw OR ‘cigarette*’:ti, ab, kw OR ‘tobacco’:ti, ab, kw OR ‘smoking cessation’/exp
#7	‘alcohol consumption’/exp OR ‘alcohol’/exp OR ‘alcohol’:ti, ab, kw OR ‘ethanol’:ti, ab, kw
#8	‘caffeine’/exp OR ‘caffeine’:ti, ab, kw OR ‘coffee’/exp OR ‘coffee’:ti, ab, kw OR ‘tea’/exp OR ‘tea’:ti, ab, kw
#9	‘oral contraception’/exp OR ‘oral contraceptive agent’/exp OR ‘oral contracept*’:ti, ab, kw OR ‘the pill’:ti, ab, kw
#10	‘exercise’/exp OR ‘physical activity’/exp OR ‘exercis*’:ti, ab, kw OR ‘physical activit*’:ti, ab, kw OR ‘fitness’/exp OR ‘fitness’:ti, ab, kw OR ‘sport’/exp OR ‘sport*’:ti, ab, kw OR ‘training’/exp OR ‘training’:ti, ab, kw
#11	‘body mass’/exp OR ‘body mass index’:ti, ab, kw OR ‘bmi’:ti, ab, kw OR ‘body weight loss’/exp OR ‘weight loss’:ti, ab, kw OR ‘waist hip ratio’/exp OR ‘waist hip ratio*’:ti, ab, kw OR ‘waist circumference’/exp OR ‘waist circumference*’:ti, ab, kw OR ‘hip circumference’/exp OR ‘hip circumference*’:ti, ab, kw
#12	‘lifestyle’/exp OR ‘lifestyle modification’/exp OR ‘lifestyle*’:ti, ab, kw OR ‘life style*’:ti, ab, kw
#13	#6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12
#14	#5 AND #13
#15	#5 AND #13 NOT ‘male’/exp NOT ‘juvenile’/exp
#16	#15 AND (‘article’/it OR ‘article in press’/it OR ‘conference abstract’/it OR ‘conference paper’/it OR ‘review’/it) AND [embase]/lim

*

The search string contains search terms for age at menopause. We made the post hoc decision to only include articles with anti-Müllerian hormone as outcome. Last search performed on 1 November 2023.

Eligibility criteria

Studies were considered eligible for inclusion in this review if the relation between AMH levels or AMH decline and at least one of the lifestyle factors of interest was assessed in adult women and this was explicitly mentioned in the aims of the investigation. Studies involving subgroups of women with a medical condition that could influence their AMH levels were included (e.g. women with a polycystic ovary syndrome (PCOS) diagnosis). If publications reported on the same study population, the study with the largest number of participants was selected. However, if publications described overlapping but not identical study populations, both were included and the results were displayed together. Publications in which insufficient information was given about the study design or data analysis necessary for interpretation of the results were excluded. For example, studies using undefined cutoff values for dichotomization of either continuous determinants or AMH levels were excluded. Furthermore, animal studies, review articles, conference abstracts, commentaries, and articles written in another language than English were excluded.

Screening and data extraction

Two reviewers (L.W. and A.d.K.) independently screened the titles and abstracts of all articles retrieved by the search. The full text of potentially eligible articles was screened by one reviewer (L.W.), and all full-text articles which did not clearly meet inclusion or exclusion criteria were screened by two reviewers (L.W. and A.d.K.). All included articles were discussed by both reviewers. Researcher Y.v.d.S. was available for consultation in case of doubt. Screening was performed in Covidence (Melbourne, Australia). Any disagreements about the eligibility of studies were resolved through discussion. Data extraction was performed by one reviewer (L.W.). The following information was extracted from included studies: first author; year of publication; country of data collection; study design; study population including sample size and age (range, mean age ± SD or median (interquartile range)); determinant(s) studied; AMH assay and sample type used; statistical methods (including adjustment for age and other confounders); reported association measure(s); and main findings. In addition, the study years (recruitment or data collection), exclusion criteria for the study population, method and timing of determinant assessment, timing of outcome assessment (if applicable), and missing data methods were extracted (Supplementary Table S1). The study design was extracted according to how the authors characterized their study or presented their methodology. The results were summarized per lifestyle factor and presented as ranges of the most frequently used association measure for studies that found a significant association in the same direction.

Assessment of methodological quality

The Study Quality Assessment Tools developed by the National Heart, Lung and Blood Institute were used to assess the methodological quality of the included studies (National Heart, Lung, and Blood Institute [NHLBI], 2021). Using the appropriate criteria based on the study design, studies received a ‘good’, ‘fair’, or ‘poor’ overall quality rating as judged by one reviewer (L.W.). Other reviewers were consulted in case of doubt. The quality assessment was performed per study. Studies with multiple determinants were assigned the quality assessment rating corresponding to their most extensive analysis. For prospective intervention studies not using randomization and only comparing AMH levels before and after a certain intervention in the same subjects, the quality assessment for ‘observational cohort and cross-sectional studies’ was used instead of that for ‘controlled intervention studies’: the latter focuses on randomization, blinding and similarity of treatment groups at baseline, which are not used or addressed otherwise in these prospective intervention studies. Age adjustment was considered an important criterion to be granted a ‘good’ or ‘fair’ quality rating, as age is strongly associated with AMH levels. Studies not performing age adjustment or not describing the age-matching methods in case of matching therefore received a ‘poor’ quality rating. The manner of age adjustment did not impact the quality rating. Furthermore, studies with a cross-sectional design could not receive a ‘good’ quality rating, since we considered it important for a potential causal relation between lifestyle factors and AMH levels to have the lifestyle factor measured prior to the AMH measurement or to have multiple AMH measurements over time. For the rating of the statistical methods, it was checked whether AMH levels were logarithmically transformed, as AMH levels generally show a right-skewed distribution. Studies not reporting the distribution of AMH in their study population and using parametric statistical methods received a ‘no’ for the question of whether outcome measures were clearly defined, valid, and reliable. For studies describing that logarithmic transformations were performed where data were skewed but not specifying this for AMH levels, it was assumed that AMH levels had been logarithmically transformed and, because of this assumption, these studies also received a ‘no’ for the question of whether outcome measures were clearly defined, valid, and reliable.

Results

The process of study selection and inclusion is depicted in Fig. 1. A total of 15 060 records were identified by the database search. Twelve additional records were identified through reference lists of included articles. After removing duplicates, 11 303 records were screened based on title and abstract. The most common reason for exclusion in this phase was not investigating the lifestyle factors of interest. A total of 313 records for which the full text was available were then screened for eligibility. Following exclusion of 248 articles for various reasons, as shown in Fig. 1, we included 65 publications.

Figure 1.

Flow diagram of the study selection process and inclusion for a systematic review of the association between modifiable lifestyle factors and circulating anti-Müllerian hormone. ANMP: age at natural menopause.

Study characteristics

Table 2 presents the characteristics of the 65 included studies along with the main findings and the quality rating. On three occasions, two publications reported on overlapping study populations. Study results were displayed separately per lifestyle factor, meaning that some studies are displayed in the table multiple times. Figure 2 shows the frequency of study designs used and lifestyle factors studied. The majority of the studies had a cross-sectional design (n = 43); longitudinal designs were less frequently used, i.e. prospective cohort (n = 13); prospective intervention (n = 6); and retrospective cohort (n = 3). One publication reported both a cross-sectional and prospective analysis. The most studied lifestyle factor was BMI (n = 36), followed by OC use (n = 20), smoking (n = 17), alcohol consumption (n = 10), physical activity (n = 7), caffeine consumption (n = 3), and WHR (n = 2). Studies were performed in Europe, the USA, Asia, Africa, South America, and Australia.

Figure 2.

Frequency of study designs used (n = 65) and lifestyle factors studied (n = 94) in studies included in the systematic review. OC, oral contraceptive; WHR, waist–hip ratio.

Table 2.

Systematic review of lifestyle factors and anti-Müllerian hormone; overview of study characteristics and results of included full-text studies (n = 56).

Authors (publication year)	Country	Study design	Participants (incl. age range, mean ± SD or median (IQR) in years)	Sample type, AMH assay	Age-adjustment	Additional factors adjusted for	Reported association measure (mean ± SD; median (IQR); correlation coefficient; (relative/geometric) mean difference [95% CI]	Main finding	Quality rating
BMI
Albu and Albu (2019)	Romania	Retrospective cohort	2204 women (34.6 years ± 4.3) evaluated for infertility	Serum, Gen II AMH ELISA, Beckman Coulter, USA	Added to regression model as continuous covariate	None	Mean difference in AMH per 1 kg/m2 increase in BMI = 0.059 ng/ml [95% CI NR], P = 0.004	↑ AMH	Poor
Bakeer et al. (2018)	Egypt	Cross-sectional	70 women (17–39 years), of whom 53 with PCOS and primary or secondary infertility, and 17 apparently healthy women	Serum, Gen II AMH ELISA, Beckman Coulter, USA	None	None	Spearman rank correlation among women with PCOS: r = −0.031, P = 0.883 Spearman rank correlation among women without PCOS: r = 0.503, P = 0.047	↑ AMH among women without PCOS	Poor
Bernardi et al. (2017)	USA	Cross-sectional	1633 women (23–34 years) participating in a population study	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariates (age and age2)	OC use; history of thyroid condition; abnormal menstrual bleeding; menstrual cycle length	Mean difference in log AMH per 1 kg/m2 increase in BMI = −0.015 ng/ml [−0.021; −0.009], P < 0.0001 When BMI was analyzed categorically, the regression coefficients showed a generally linear pattern, with the strongest association among those with BMI ≥45 kg/m2: mean difference in log AMH vs BMI <25 kg/m2 = −0.372 ng/ml [−0.581; −0.163], P = 0.0005 Mean difference in log AMH per 1 kg/m2 increase in BMI at age 18 = −0.016 ng/ml [−0.024; −0.008], P < 0.0001 When BMI at 18 was analyzed categorically, this association was also generally linear, with the strongest association among those with BMI ≥45 kg/m2: mean difference in log AMH vs BMI <25 kg/m2 = −0.514 ng/ml [−0.887; −0.142], P = 0.0069	↓ AMH	Fair
Buyuk et al. (2011)	USA	Cross-sectional	290 women evaluated for infertility (37.1 years ± 4.8 for NOR group and 38.1 years ± 5.2 for DOR group)	Serum, Beckman Coulter, USA	Added to regression model as covariate	None	Mean difference in log AMH per 1 kg/m2 increase in BMI among women with DOR = −0.2 ng/ml [95% CI NR], P = 0.01	↓ AMH among women with DOR	Fair
Clendenen et al. (2021) and Jung et al. (2017) I	USA, UK, Sweden	Cross-sectional	Clendenen et al.: 3831 women (21–57 years) participating in population studies (10 cohorts), of which 81.5% were controls in breast cancer risk studies Jung et al.: 671 premenopausal women (19–47 years) participating in population studies (9 cohorts) on gynecologic cancer risk	Serum/plasma, picoAMH ELISA kit, Ansh Labs, USA, Gen II AMH ELISA, Beckman Coulter, ultrasensitive AMH ELISA kit, Ansh Labs	Clendenen et al.: Added to regression model as continuous covariates (age and age2) Jung et al.: Added to regression model as continuous covariate	Clendenen et al.: Smoking; cohort; OC use; age at menarche; parity; hysterectomy; oophorectomy; benign breast disease Jung et al.: OC use; age at menarche	Clendenen et al.: Geometric mean AMH [95% CI] for BMI category: <18.5 kg/m2 = 2.20 pmol/l [1.58; 3.08] ≥18.5–<25 kg/m2 = 2.83 pmol/l [2.65; 3.01] ≥25–<30 kg/m2 = 2.85 pmol/l [2.57; 3.15] ≥30–<35 kg/m2 = 2.31 pmol/l [1.95; 2.75] ≥35 kg/m2 = 1.97 pmol/l [1.59; 2.43] P < 0.0001 among women with BMI ≥18.5 kg/m2 Jung et al.: Median AMH of BMI category: ≤20 kg/m2 = 1.03 ng/ml (0.46–1.84) >20–<25 kg/m2 = 0.95 ng/ml (0.34–1.93) ≥25–<30 kg/m2 = 1.05 ng/ml (0.38–2.30) ≥30 kg/m2 = 1.16 ng/ml (0.57–1.96) P = 0.48	Clen-denen et al.: ↓ AMH for BMI ≥18.5  kg/m2 Jung et al.:● AMH	Fair
Cui et al. (2014)	China	Cross-sectional (described by authors as retrospective study)	2200 women (20–47 years) undergoing infertility work-up, of whom 304 with and 1896 without PCOS	Serum, ultrasensitive AMH ELISA, Immunotech, Beckman Coulter	None	None	Pearson correlation women with PCOS: r = −0.148, P = 0.01 Pearson correlation women without PCOS: r = −0.064, P = 0.006 Mean difference in AMH per 1 kg/m2 increase in BMI in: women with PCOS = −0.154 ng/ml, P < 0.05 women without PCOS = −0.051 ng/ml, P < 0.05	↓ AMH	Poor
Dólleman et al. (2013) I	The Nether-lands	Cross-sectional	2230 women (20–59 years) participating in a population study	Serum, Gen II AMH ELISA, Beckman Coulter, USA	Age-standardized percentiles were calculated with the CG-LMS method	None	Relative mean difference in age-standardized AMH per 1 kg/m2 increase in BMI = −0.2 percentiles (SE = 0.2), P = 0.16	● AMH	Fair
Freeman et al. (2007)	USA	Cross-sectional and prospective cohort	122 women (35–47 years) participating in a population study (of the 122 women, 20 women had multiple AMH measurements that were collected prospectively)	Serum, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	Added to regression model as dichotomous covariate (<40 and ≥40 years)	Menopausal status; race	Cross-sectional analysis Geometric mean AMH ratio for BMI ≥30 kg/m2 category (vs BMI <30 kg/m2) = 0.26 [0.11; 0.58] Longitudinal analysis Mean log AMH of BMI category: <30 kg/m2 = 0.566 ng/ml [0.34; 0.94] ≥30 kg/m2 = 0.459 ng/ml [0.28; 0.75] Unadjusted P = 0.016	↓ AMH	Fair
Gouvea et al. (2022) I	Brazil	Cross-sectional	177 climacteric women (40–64 years) from Basic Health Units	Serum, Access II AMH immuno-assay, Beckman Coulter, USA	Added to regression model as covariate	Stages of reproductive aging	Mean difference in log AMH per 1 kg/m2 increase in BMI = −0.010 ng/ml (SE = 0.021), P = 0.627	● AMH	Poor
Grimes et al. (2022)	USA	Prospective cohort	795 nurses participating in a study (mean (IQR): 38 years (36–40))	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariates (age at both blood collections)	Pack-years of smoking; smoking status; duration of OC use; years until cycle became regular; infertility history; age at menarche; alcohol consumption; cumulative total breastfeeding duration; parity	Geometric mean change in AMH (during a period of 11–16 years) for BMI at age 18 category: <18.5 kg/m2 = 4.08 ng/ml (P = 0.10) ≥18.5–<25 kg/m2 = 3.53 ng/ml (ref) ≥25–<30 kg/m2 = 3.78 ng/ml (P = 0.59) ≥30 kg/m2 = 2.29 ng/ml (P = 0.04) Geometric mean change in AMH (during a period of 11–16 years) per 1 kg/m2 increase in BMI at age 18 = −0.02 ng/ml (SE = 0.01), P = 0.02 Geometric mean change in AMH (during a period of 11–16 years) for BMI at baseline category: <18.5 kg/m2 = 3.60 ng/ml (P = 0.61) ≥18.5–<25 kg/m2 = 3.95 ng/ml (ref) ≥25–<30 kg/m2 = 3.22 ng/ml (P = 0.002) ≥30 kg/m2 = 2.88 ng/ml (P = 0.0002) Geometric mean change in AMH (during a period of 11–16 years) per 1 kg/m2 increase in BMI at baseline = −0.03 ng/ml (SE = 0.01), P < 0.0001The authors reported additional results for weight change from age 18 to 1996–1999 and weight change from 1996–1999 to 2010–2012, of which some significant (not reported here).	↓ AMH decline	Good
Halawaty et al. (2010)	Egypt	Cross-sectional	100 women, of whom 50 with BMI ≥30–≤35 kg/m2 (40–48 years) and 50 with BMI <30 kg/m2 (38–48 years)	Serum, AMH ELISA, Diagnostic Systems Laboratories, USA	Age-matching (method not described). Mean age ± SD: 46.2 years ± 6.4 for women with BMI ≥30-≤35 kg/m2 and 46.1 years ± 3.3 for women with BMI <30 kg/m2	None	Mean AMH of BMI category: <30 kg/m2 = 3.39 ng/ml ± 0.24 ≥30–≤35 kg/m2 = 2.55 ng/ml ± 0.81 P = 0.56	● AMH	Poor
Jaswa et al. (2020) and Rios et al. (2020)*	USA	Cross-sectional	Jaswa et al.: 1561 women from two cohorts; women with PCOS and ovulatory women not seeking fertility treatment Rios et al.: 2431 women from three cohorts; women with unexplained fertility, women with PCOS and ovulatory women not seeking fertility treatment	Serum, 2-site sandwich AMH ELISA, Beckman Coulter, Gen I and Gen II AMH ELISA, Beckman Coulter	Added to regression model as continuous covariate	Jaswa et al.: Smoking; race; study center Rios et al.: Study center	Jaswa et al.: Mean difference in log AMH per 1 kg/m2 increase in BMI among women with PCOS = −0.023 ng/ml [−0.030; −0.016], P < 0.001 Mean difference in log AMH per 1 kg/m2 increase in BMI among ovulatory women = −0.019 ng/ml [−0.027; −0.010], P < 0.001 Rios et al.: Mean difference in AMH per 1 kg/m2 increase in BMI among women with unexplained infertility = −0.27 ng/ml [95% CI NR], P = 0.02	↓ AMH	Jaswa et al.: Fair Rios et al.: Poor
Kalem et al. (2019) I	Turkey	Cross-sectional	77 healthy female students (18–28 years) of whom none had ever been pregnant	Plasma, MIS/AMH ELISA kit, NR	None	None	Spearman rank correlation: r = −0.063, P = 0.584	● AMH	Poor
Kriseman et al. (2015)	USA	Cross-sectional	489 women (34.2 years ± 5.4) who presented to a practice at the children’s hospital	Serum, AMH CLIA, NR	Added to Spearman correlation analysis as covariate	None	Spearman rank correlation: r = −0.04, P > 0.05 Spearman rank correlation among women with PCOS: r = −0.29, P = 0.002 Adjusted Spearman rank correlation among women with PCOS: r = −0.31, P = 0.002 Mean AMH of BMI category: ≤25 kg/m2 = 10.9 ng/ml ± 13.4 >25–<30 kg/m2 = 6.4 ng/ml ± 3.8 ≥30–<35 kg/m2 = 5.2 ng/ml ± 5.7 ≥35 kg/m2 = 5.9 ng/ml ± 7.5 P = 0.02 for BMI ≤25 kg/m2 category vs BMI ≥30–<35 kg/m2 category	↓ AMH among women with PCOS	Poor
Lawal and Yusuff (2021)	Nigeria	Cross-sectional	161 women (21–40 years) attending the Gynecology and Immunization clinic of a tertiary hospital	Serum, AMH ELISA, Calbiotech Inc., USA	Added to regression model as continuous covariate	Ethnicity; cycle length; parity	Mean difference in log AMH per 1 kg/m2 increase in BMI = 0.015 ng/ml (SE = 0.005), P = 0.003	↑ AMH	Poor
Lefebvre et al. (2017)	France	Prospective cohort (described by authors as retrospective analysis with prospectively collected data)	691 women (18–36 years) attending the hospital for hyperandrogenism, cycle disorders, infertility, or all three	Serum, AMH-EIA (ref A16507), Beckman Coulter, France	None	None	Median [5th to 95th percentile] AMH of BMI category: <25 kg/m2 = 54.1 pmol/l [15.2–171.4] ≥25 kg/m2 = 53.2 pmol/l [16.0–1407] P > 0.05 Median [5th to 95th percentile] AMH of women with PCOS and BMI category: <25 kg/m2 = 66.0 pmol/l [36.0–186.9] ≥25 kg/m2 = 58.2 pmol/l [25.8–147.1] P < 0.05 Spearman rank correlation PCOS subgroup: r = −0.177, P = 0.0001	↓ AMH among women with PCOS	Poor
Lim et al. (2021)	South Korea	Cross-sectional	448 women (20–45 years) employed in a hospital	Serum, Gen II AMH ELISA, Beckman Coulter & Elecsys AMH assay	None	None	Median AMH of BMI category: <18.5 kg/m2 = 3.60 ng/ml (2.16–5.25) ≥18.5–<23 kg/m2 = 3.86 ng/ml (2.38–5.53) ≥23–<25 kg/m2 = 3.35 ng/ml (1.60–5.47) ≥25 kg/m2 = 3.64 ng/ml (2.01–5.24) P = 0.559 Only a significant association in the 20th percentile of AMH: mean difference in AMH per 1 kg/m2 increase in BMI = −0.087 ng/ml [95% CI NR], P = 0.04	↓ AMH	Poor
Moy et al. (2015)	USA	Cross-sectional (described by authors as a retrospective cohort)	350 women undergoing fertility workup (37.6 years ± 5.2 for African-Americans, 36.3 years ± 6.0 for Caucasians, 38.1 years ± 4.8 for Hispanics, 35.8 years ± 5.2 for Asians)	Serum, Beckman Coulter, USA	Added to regression model as continuous covariate	Smoking status; PCOS	Mean difference in AMH** per 1 kg/m2 increase in one over squared-transformed BMI in Caucasian subgroup = 0.17 (unit NR) [95% CI NR], P = 0.013	↓ AMH among Caucasian women	Poor
Mitchell et al. (2023) I	Ireland	Cross-sectional	96 women (35 years ± 5) attending a reproductive medicine clinic	Serum, Elecsys AMH assay, Roche Diagnostics	Added to regression model as continuous covariate	None	Pearson correlation: r = −0.15, P = 0.15 Mean difference in AMH per 1 kg/m2 increase in BMI = −0.14 pmol/l, P = 0.76	● AMH	Poor
Nardo et al. (2007) I	UK	Cross-sectional	136 women (23–39 years) undergoing fertility work-up prior to ovarian stimulation	Plasma, ultra-sensitive AMH ELISA, Oxford Bio Innovation DSL LTD, UK	None	None	Spearman rank correlation: r = −0.01, P = 0.88	● AMH	Poor
Olszanecka-Glinianowicz et al. (2015)	Poland	Cross-sectional	154 women (25.4 years ± 5.5) of whom 87 with PCOS and 67 without PCOS	Plasma, AMH ELISA, Immunotech, Czech Republic	Added to regression model as continuous covariate	PCOS; HOMA-IR; adiponectin; apelin-36; leptin; omentin-1	Mean AMH of women with PCOS and BMI category: ≥18.5–<25 kg/m2 = 8.9 ng/ml ± 4.4 ≥30 kg/m2 = 7.0 ng/ml ± 4.0 Unadjusted P < 0.05 Mean AMH of women without PCOS and BMI category: ≥18.5–<25 kg/m2 = 5.1 ng/ml ± 2.4 ≥30 kg/m2 = 3.9 ng/ml ± 2.3 Unadjusted P < 0.05 Mean difference in AMH per 1 kg/m2 increase in BMI for all participants = −0.075 ng/ml [95% CI NR], P < 0.05	↓ AMH	Poor
Piouka et al. (2009)	Greece	Cross-sectional	250 women classified into 4 groups based on PCOS phenotype and 1 group with healthy volunteers (mean age only given stratified for 10 subgroups). For all groups, 25 women with BMI ≥20–25 kg/m2 and 25 with BMI > 25 kg/m2 were selected	NR***, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	Age-matching (method and age ranges not described)	Luteinizing hormone; total testosterone; total number of follicles 2–9 mm	Mean difference in AMH** per 1 kg/m2 increase in BMI = −0.708 (unit NR) (SE = 0.137), P < 0.0005	↓ AMH	Poor
Sahmay et al. (2012)	Turkey	Cross-sectional	259 women with infertility (32.1 years ± 4.9 for BMI <30 kg/m2 and 32.9 years ± 5.8 for BMI ≥ 30 kg/m2 group)	Serum, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	None	None	Mean AMH of BMI category: <30 kg/m2 = 3.46 ng/ml ± 2.79 ≥30 kg/m2 = 3.79 ng/ml ± 2.93 P > 0.05	● AMH	Poor
Santulli et al. (2016)	France	Prospective cohort	804 women evaluated for infertility or receiving ART, of whom 201 HIV-infected and 603 seronegative (35.7 years ± 4.0 and 35.1 years ± 4.6)	Serum, NR	Added to regression model as continuous covariate	CD4+ count; viral load	Mean difference in log AMH per 1 kg/m2 increase in BMI = −0.057 ng/ml (SE = 0.015), P = 0.025	↓ AMH	Poor
Simões-Pereira et al. (2018)	Portugal	Cross-sectional (described by authors as retrospective analysis)	951 women (35 years (29–41)) whose AMH levels were requested as part of their fertility work-up	Serum, Gen II AMH ELISA, Beckman Coulter	Added to regression model as continuous covariate	Ovarian surgery; ethnicity; smoking	Median AMH of BMI category: <18.5 kg/m2 = 1.92 ng/ml (IQR 2) ≥18.5–<25 kg/m2 = 1.70 ng/ml (IQR 2) ≥25–<30 kg/m2 = 1.77 ng/ml (IQR 2) ≥30–<35 kg/m2 = 2.20 ng/ml (IQR 4) ≥35–<40 kg/m2 = 2.24 ng/ml (IQR 3) Unadjusted P = 0.191 Relative mean difference in AMH (gamma distribution) per 1 kg/m2 increase in BMI = 1.0% [95% CI NR], P = 0.31	● AMH	Poor
Skałba et al. (2011)	Poland	Cross-sectional	137 women (24.8 years ± 4.1) of whom 87 with PCOS and 50 apparently healthy women	Plasma, AMH ELISA, Immunotech a.s., Czech Republic	None	None	Mean AMH of women with PCOS and BMI category: <25 kg/m2 = 9.6 ng/ml ± 3.5 ≥25 kg/m2 = 11.2 ng/ml ± 4.5 P > 0.05 Mean AMH of women without PCOS and BMI category: <25 kg/m2 = 2.5 ng/ml ± 0.8 ≥25 kg/m2 = 2.3 ng/ml ± 0.7 P > 0.05	● AMH	Poor
Soares et al. (2020) I	UK	Prospective cohort	1608 women (48 years (45–51)) participating in a population study	Serum, Elecsys AMH Plus assay, Roche Diagnostics	Added to linear mixed regression model as continuous covariate	Educational achievement; age at menarche; smoking; alcohol consumption	Average predicted means [95% CI] 5 years before the final menstrual period (FMP): <25 kg/m2: 0.19 ng/ml [0.15; 0.25] ≥25–<30 kg/m2: 0.10 ng/ml [0.08; 0.13] ≥30 kg/m2: 0.10 ng/ml [0.07; 0.14] Women with obesity (BMI ≥30 kg/m2) and overweight women (BMI ≥25–<30 kg/m2) had lower rates of decrease in AMH compared to women with normal BMI (BMI <25 kg/m2), based on 4 AMH measurements per participant. Corresponding association measures were not published in the article or supplementary information.	↓ AMH ↓ AMH decline	Good
Steiner et al. (2010)	USA	Prospective cohort (within an exploratory intervention study)	20 women starting OCs, of whom 10 with BMI >30 kg/m2 (29 years ± 4.9) and 10 with BMI <25 kg/m2 (29 years ± 6.3)	Serum, AMH ELISA, Diagnostic Systems Laboratories	None	None	Mean AMH level for BMI category: <25 kg/m2 = 4.4 ng/ml ± 1.8 >30 kg/m2 = 2.9 ng/ml ± 2.1 P < 0.05 Modeling to determine differences in AMH throughout the OC cycles based on obesity status showed that the BMI > 30 kg/m2 group had significantly lower AMH levels across the OC cycles (P < 0.05)	↓ AMH	Poor
Su et al. (2008)	USA	Cross-sectional	36 women (40–52 years) of whom 18 with normal BMI (<25 kg/m2) and 18 obese (>30 kg/m2) participating in a population study	Serum, AMH ELISA, Diagnostic Systems Laboratories, USA	Added to regression model as continuous covariate	Race; smoking status; alcohol consumption	Geometric mean AMH for BMI category: <25 kg/m2 = 0.28 ng/ml [0.12; 0.67] ≥30 kg/m2 = 0.06 ng/ml [0.03; 0.13] P = 0.02	↓ AMH	Poor
Tzeng et al. (2023)	11 regions in Asia	Prospective cohort	4556 women (33 years ± 4.6) visiting the fertility clinic at 12 different IVF centers	Serum, Elecsys AMH assay, Roche Diagnostics	Added to regression model as continuous covariate	Ethnicity; PCOS status; OC pre-treatment; GnRHa pretreatment; day of cycle; freezing-thawing procedure; smoking	Mean difference in AMH in obese (BMI ≥30 kg/m2) vs non-obese (BMI <30 kg/m2) women = −0.41 ng/ml [−0.70; −0.11]	↓ AMH	Poor
Wang et al. (2022)	China	Cross-sectional	8323 women (32 years ± 5.2) attending the Reproductive Medicine Center for ART treatment	Serum, AMH ELISA, Kangrun Biotech, China	Added to regression model as covariate (natural cubic spline function with 3 df)	Race; education; smoking status; alcohol consumption; year and season of hormone examinations; endometriosis; pelvic inflammatory disease	Relative mean difference in AMH per 1 kg/m2 increase in BMI = −1.02% [−1.65%; −0.40%] Relative mean difference in AMH (vs BMI ≥18.5–<25 kg/m2 category) for BMI category: <18.5 kg/m2 = −1.87% [−7.29%; 3.86%] ≥25–<30 kg/m2 = −6.01% [−11.04%; −0.70%] ≥30 kg/m2 = −18.64% [−32.26%; −2.29%]	↓ AMH	Fair
Whitworth et al. (2015)	South Africa	Cross-sectional	425 women (20–30 years) participating in a population study	Plasma, Gen II AMH ELISA, Beckman Coulter, USA	Added to regression model as continuous covariate	Education; parity	Relative mean difference in AMH per 1 kg/m2 increase in BMI = −1% [−2%; 1%]	● AMH	Fair
Yang et al. (2015)	Taiwan	Cross-sectional	186 women, of whom 156 with PCOS, and 30 healthy women (24 years (20–28) for PCOS BMI <27 kg/m2 subgroup, 25 years (30–31) for PCOS BMI ≥27 kg/m2 subgroup, 26 years (24–27) for healthy BMI <27 kg/m2 subgroup)	Serum, Gen II AMH ELISA, Immunotech Beckman Coulter, France	None	PCOS; luteinizing hormone; HOMA-IR; ferritin	Mean difference in log AMH per 1 kg/m2 increase in log BMI = −0.81 pM [95% CI NR], P = 0.003	↓ AMH	Poor
Zhao et al. (2023)	China	Cross-sectional (described by authors as a retrospective cohort)	220 women (20–39 years) visiting the endocrinology clinic	Serum, AMH ELISA, NR	NR¥	NR¥	Pearson correlation: r = −0.208, P = 0.002 Mean difference in AMH per 1 kg/m2 increase in BMI = 0.179 ng/ml (SE = 0.064), P = 0.006	↓ AMH	Poor
WAIST–HIP RATIO
Grimes et al. (2022) II	USA	Prospective cohort	795 nurses participating in a study (mean (IQR): 38 years (36–40))	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariates (age at both blood collections)	Pack-years of smoking; smoking status; duration of OC use; years until cycle became regular; infertility history; age at menarche; alcohol consumption; cumulative total breastfeeding duration; parity	Geometric mean AMH for WHR category: <0.75 = 4.09 ng/ml [3.65; 4.60] ≥0.75–<0.80 = 4.03 ng/ml [3.57; 4.55] ≥0.80–<0.85 = 3.73 ng/ml [3.21; 4.35] ≥0.85–<0.90 = 3.34 ng/ml [2.67; 4.17] ≥0.90 = 3.46 ng/ml [2.65; 4.50] P = 0.08 Geometric mean difference in AMH per 1 unit increase in WHR = −0.93 (SE = 0.6)	● AMH	Good
(Kalem et al., 2019) II	Turkey	Cross-sectional	77 healthy female students (18–28 years) of whom none had ever been pregnant	Plasma, MIS/AMH ELISA kit, NR	None	None	Spearman rank correlation: r = −0.054, P = 0.638	● AMH	Poor
SMOKING
Bhide et al. (2022)	UK	Cross-sectional	101 women undergoing investigation and treatment for subfertility (30 years (25.5–33.0) for current smokers, 32.5 years (31.0–33.5) for former smokers and 31 years (28.0–33.0) for never smokers)	Serum, Access II AMH immuno-assay, Beckman Coulter	Added as covariate to ANCOVA	None	Median AMH** of: never smokers = 27.6 (16.4–39.7) pmol/l former smokers = 26.0 (14.7–32.2) pmol/l current smokers = 38.9 (20.4–66.2) pmol/l Unadjusted P = 0.309 (ANOVA) Age-adjusted P = 0.673 (ANCOVA)	● AMH	Poor
Clendenen et al. (2021) and Jung et al. (2017) II	USA, UK, Sweden	Cross-sectional	Clendenen et al.: 3831 women (21–57 years) participating in population studies (10 cohorts), of which 81.5% were controls in breast cancer risk studies Jung et al.: 671 premenopausal women (19–47 years) participating in population studies (9 cohorts) on gynecologic cancer risk	Serum/plasma, picoAMH ELISA kit, Ansh Labs, USA, Gen II AMH ELISA, Beckman Coulter, ultrasensitive AMH ELISA kit, Ansh Labs	Clendenen et al.: Added to regression model as continuous covariates (age and age2) Jung et al.: Added to regression model as continuous covariate	Clendenen et al.: BMI; OC use; cohort; age at menarche; parity; hysterectomy; oophorectomy; benign breast disease Jung et al.: OC use; age at menarche	Clendenen et al.: Geometric mean AMH of: never smokers = 2.85 pmol/l [2.68; 3.04] former smokers = 2.55 pmol/l [2.31; 2.82] current smokers = 2.24 pmol/l [1.94; 2.59] P = 0.0058 Jung et al.: Median AMH of: never smokers = 1.15 ng/ml (0.49–2.31) former smokers = 1.04 ng/ml (0.36–2.14) current smokers = 1.13 ng/ml (0.47–1.92) P = 0.73	Clen-denen et al.:↓ AMH Jung et al.:● AMH	Fair
Dafopoulos et al. (2010)	Greece	Cross-sectional	137 women (20–49 years) who had delivered at least once after spontaneous conception	Serum, EIA AMH/MIS kit, Immunotech SAS, France	None	None	Mean AMH of: non-smokers = 2.3 ng/ml ± 1.2 current smokers = 1.9 ng/ml ± 1.1 P > 0.05 Pearson correlation of pack-years (among current smokers): r = −0.807, P < 0.001	● AMH ↓ AMH for pack-years among current smokers	Poor
Dólleman et al. (2013) II	The Nether-lands	Cross-sectional	2230 women (20–59 years) participating in a population study	Serum, Gen II AMH ELISA, Beckman Coulter, USA	Age-standardized percentiles were calculated with the CG-LMS method	None	Relative mean difference in age-standardized AMH (vs never smokers) for: former smokers = −0.6 percentiles (SE = 1.5), P = 0.67 current smokers = −3.6 percentiles (SE = 1.5), P = 0.02 Relative mean difference in age-standardized AMH (vs <1 cigarette per month) for: <1 cigarette/day = 1.3 percentiles (SE = 7.2), P = 0.86 1–9 cigarettes/day = −12.8 percentiles (SE = 6.5), P = 0.05 10–19 cigarettes/day = −12.3 percentiles (SE = 6.2), P = 0.05 ≥20 cigarettes/day = −11.8 percentiles (SE = 6.3), P = 0.06 Relative mean difference in age-standardized AMH per 1 year increase in the number of pack-years in current/former smokers = −0.3 percentiles (SE = 0.09), P = 0.001 Relative mean difference in age-standardized AMH (vs 0–5 pack-years) in current/former smokers for: 5–10 pack-years = −3.1 percentiles (SE = 2.1), P = 0.15 10–15 pack-years = −7.0 percentiles (SE = 2.3), P = 0.003 15–20 pack-years = −8.5 percentiles (SE = 2.6), P = 0.001 20–25 pack-years = −6.6 percentiles (SE = 3.3), P = 0.05 ≥25 pack-years = −6.0 percentiles (SE = 2.9), P = 0.04 The authors reported additional results for the number of cigarettes smoked daily in smokers, but this association was not significant.	↓ AMH	Fair
Freour et al. (2008)	France	Retrospective	111 women undergoing IVF-embryo transfer cycles (31.8 years ± 5.1 for current smokers, 31.4 years ± 4.5 for non-smokers)	Serum, ultra-sensitive AMH ELISA, Immunotech, Beckman Coulter, France	Stratified analysis (<30 years and >30 years)	None	Mean AMH of: non-smokers (never or quit ≥1 year ago) <30 years old = 3.93 µg/l  ± 1.9 current smokers <30 years old = 2.99 µg/l  ± 1.32 P < 0.05 Mean AMH of: non-smokers (never or quit ≥1 year ago) >30 years old = 3.82 µg/l  ± 1.9 current smokers >30 years old = 3.1 µg/l  ± 1.89 P < 0.05	↓ AMH	Poor
Hawkins Bressler et al. (2016) I	USA	Cross-sectional	1654 women (23–34 years) participating in a population study	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariates (age and age2)	BMI; OC use	Relative mean difference in AMH (vs never smokers) for: former smokers = 7.5% [−12%; 32%] current smokers = 1.9% [−11%; 17%] The authors reported additional results for age start smoking, current cigarette consumption, maximum consumption of cigarettes/d and years of smoking, but there were no significant associations.	● AMH	Fair
Kalem et al. (2019) III	Turkey	Cross-sectional	77 healthy female students (18–28 years) of whom none had ever been pregnant	Plasma, MIS/AMH ELISA kit, NR	None	None	Median (min-max) AMH of: never smokers = 2.6 ng/ml (0.7–6.0) former smokers = 1.4 ng/ml (1.2–1.5) current smokers = 1.5 ng/ml (1.0–5.6) P = 0.048	↓ AMH	Poor
Kline et al. (2016)	USA	Cross-sectional	477 women (19–45 years) participating in a study examining the relation of highly skewed X-chromosome inactivation to trisomy	Serum, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	Added to regression model as continuous covariate	Karyotype group; day of blood draw; interval between blood draw and assay; alcohol consumption; caffeine consumption	Mean difference in log AMH (vs never smokers) for: former smokers = 0.111 ng/ml [−0.107; 0.329] current smokers = −0.223 ng/ml [−0.539; 0.094]	● AMH	Fair
Mitchell et al. (2023) II	Ireland	Cross-sectional	96 women (35 years ± 5) attending a reproductive medicine clinic	Serum, Elecsys AMH assay, Roche Diagnostics	Added to regression model as continuous covariate	None	Mean difference in AMH (vs non-smokers) for: former smokers = 3.07 pmol/l (SE = 3.67), P = 0.41 current smokers = −1.86 pmol/l (SE = 3.19), P = 0.56	● AMH	Poor
Nardo et al. (2007) II	UK	Cross-sectional	136 women (23–39 years) undergoing fertility work-up prior to ovarian stimulation	Plasma, ultra-sensitive AMH ELISA, Oxford Bio Innovation DSL LTD, UK	Added to regression model as continuous covariate	None	Median log AMH of: non-smokers = 2.1 ng/ml (1.2–3.5) smokers = 1.9 ng/ml (0.8–3.7) Unadjusted P = 0.71 Age-adjusted P = 0.41	● AMH	Poor
Oladipupo et al. (2022)	USA	Cross-sectional	207 women (21–45 years) seeking infertility treatment	Serum, chemiluminescence, Quest Diagnostics, USA	Added to regression model as categorical covariate	Race; PCOS status	Relative mean difference in AMH (vs non-smokers) for: active smokers (based on urinary cotinine levels) = 5.2% [−34.5%; −68.9%] passive smokers (based on urinary cotinine levels) = −14.2% [−45.3%; 34.4%] The authors reported additional results for the number of self-reported cigarettes/d and cumulative lifetime exposure in pack-years, but there were no significant associations.	● AMH	Fair
Plante et al. (2010)	USA	Cross-sectional	284 women (38–50 years) participating in a population study	Serum, dual monoclonal antibody sandwich enzyme immunoassay, Immunotech, Beckman Coulter	Added to regression model as continuous covariate	BMI	Relative mean difference in AMH (vs never smokers) for: former smokers = −1% [−32%; 44%] current smokers = −44% [−68%; −2%] passive smokers = −17% [−54%; 52%] Relative mean difference in AMH (vs current non-smokers) for: ≤15 cigarettes/day = −66% [−85%; −24%] >15 cigarettes/day = −16% [−58%; 66%] Relative mean difference in AMH (vs never smokers) for: past smokers ≤6 pack-years = 6% [−33%; 66%] past smokers >6 pack-years = −9% [−45%; 51%] Relative mean difference in AMH (vs never smokers) for: past smokers, quit before age 35 years = −9% [−40%; 39%] past smokers, quit after age 35 years = 20% [−33%; 116%]	↓ AMH for current smokers	Fair
Soares et al. (2020) II	UK	Prospective cohort	1608 women (48 years (45–51)) participating in a population study	Serum, Elecsys AMH Plus assay, Roche Diagnostics	Added to linear mixed regression model as continuous covariate	Educational achievement; BMI; alcohol consumption	Average predicted means [95% CI] 5 years before the final menstrual period (FMP): Never smokers: 0.11 ng/ml [0.09; 0.14] Former smokers: 0.15 ng/ml [0.11; 0.19] Current smokers: 0.21 ng/ml [0.12; 0.37] The AMH decline rates were higher in current smokers, followed by former smokers, compared to women who never smoked. Corresponding association measures were not published in the article or supplementary information.	● AMH ↑ AMH decline rate	Good
Tzeng et al. (2023) II	11 regions in Asia	Prospective cohort	4556 women (33 years ± 4.6) visiting the fertility clinic at 12 different IVF centers	Serum, Elecsys AMH assay, Roche Diagnostics	Added to regression model as continuous covariate	Ethnicity; PCOS status; OC pre-treatment; GnRHa pretreatment; day of cycle; freezing-thawing procedure; BMI	Mean difference in AMH (vs non-smokers) for: former smokers = 0.09 ng/ml [−0.45; 0.63] current smokers = −0.10 ng/ml [−0.81; 0.61]	● AMH	Poor
Waylen et al. (2010)	UK	Cross-sectional (described by authors as retrospective analysis)	335 women (24–28 years) who had purchased a test for ovarian assessment	Serum, AMH ELISA, Oxford Bio-Innovation, UK	Added as covariate to ANCOVA	None	Geometric mean AMH of: never smokers = 1.074 ng/ml [0.925; 1.245] former smokers = 0.955 ng/ml [0.760; 1.199] current smokers = 0.869 ng/ml [0.638; 1.183] P > 0.05	● AMH	Poor
White et al. (2016)	USA & Puerto Rico	Cross-sectional	913 women (35–54 years) selected as controls in a case-control study on breast cancer	Serum, ultra-sensitive AMH ELISA, Immunotech, Beckman Coulter, USA	Added to regression model as continuous covariable	Parity; breastfeeding history; race; education; annual household income	Relative mean difference in AMH (vs never smokers) for: former smokers = −13.7% [−47.8%; 42.5%] current smokers = −4.3% [−28.8%; 28.6%] Relative mean difference in AMH (vs never smokers) for: currently smoking <10 cigarettes/day = 89.6% [−10.5%; 301.7%] currently smoking 10–19 cigarettes/day = −21.9% [−66.1%; 80.3%] currently smoking ≥20 cigarettes/day = −55.3% [−79.8%; −0.9%] The authors reported additional results for age start smoking, total pack-years, total years smoked cigarettes, and time since smoking, but there were no significant associations.	↓ AMH for heavy smokers	Fair
ORAL CONTRACEPTIVE USE
Arbo et al. (2007)	Brazil	Prospective intervention study	20 normoovulatory women (23–34 years) with infertility and a BMI of 18–25 kg/m2, starting OCs	Serum, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	N/A (pre- and post-treatment values are measured)	None	Median AMH: baseline (before OC) = 3.02 ng/ml (1.21–6.39) third day of next withdrawal bleeding after OC initiation = 2.22 ng/ml (0.9–3.11) P = 0.04	↓ AMH	Fair
Bentzen et al. (2012)	Den-mark	Prospective cohort	732 female healthcare workers (21–41 years) participating in a population study	Serum, AMH/MIS ELISA kit, Immunotech Beckman Coulter, France	Added to regression model as covariate	None	Relative mean difference in AMH for current users vs non-users¥¥ = −29.8% [−38.5%; −19.9%], P < 0.001 Relative decline per year in AMH (calculated using age (linear) as only one measurement per person was available) for¥¥¥: non-users = 9.2% [7.0%; 11.3%] current users = 7.0% [5.4%; 8.5%] P > 0.05 Relative mean difference in AMH per 1 year increase in hormonal contraceptive use¥¥¥ = 2.3% [−0.6%; 5.0%], P > 0.05	↓ AMH for current users	Poor
Bernardi et al. (2021)	USA	Cross-sectional	1643 women (23–34 years) participating in a population study	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariates (age and age2)	BMI; PCOS history; abnormal menstrual bleeding; thyroid condition; seeking care for difficulty in conceiving	Relative mean difference in AMH (vs non-users) for¶: former users = −4.4% [−16.3%; 9.0%] current users = −25.2% [−35.3%; −13.6%] Relative mean difference in AMH (vs non-users) for¶: current and cumulative use <1 year = −33.5% [−56.1%; 0.8%] current and cumulative use 1–3 years = −21.1% [−36.9%; −1.4%] current and cumulative use ≥4 years = −26.9% [−37.4%; −14.7%] former and cumulative use <1 year = −2.2% [−18.1%; 16.8%] former and cumulative use 1–3 years = −3.4% [−17.0%; 12.5%] former and cumulative use ≥4 years = −6.5% [−19.5%; 8.6%] Relative mean difference in AMH (vs non-users) for¶: current users = −26.1% [−36.2%; −14.5%] <1 year since last use = −5.1% [−22.2%; 15.8%] >1 year since last use = −4.4% [−16.4%; 9.4%] Relative mean difference in AMH for current COC users vs non-users = −24.0% [−36.6%; −8.9%]	↓ AMH for current users	Fair
Birch Petersen et al. (2015)	Denmark	Cross-sectional	887 women (19–46 years) requesting fertility counselling without known fertility problems	Serum, AMH ELISA, Immunotech Beckman Coulter, France	Added to regression model as covariate	BMI; smoking; preterm birth; prenatal exposure to maternal smoking; maternal age of menopause	Relative mean AMH for current users vs non-users = −20% [−30.3%; −8.4%]	↓ AMH	Fair
Clendenen et al. (2021) and Jung et al. (2017) III	USA, UK, Sweden	Cross-sectional	Clendenen et al.: 3831 women (21–57 years) participating in population studies (10 cohorts), of which 81.5% were controls in breast cancer risk studies Jung et al.: 671 premenopausal women (19–47 years) participating in population studies (9 cohorts) on gynecologic cancer risk	Serum/plasma, picoAMH ELISA kit, Ansh Labs, USA, Gen II AMH ELISA, Beckman Coulter, ultrasensitive AMH ELISA kit, Ansh Labs	Clendenen et al.: Added to regression model as continuous covariates (age and age2) Jung et al.: Added to regression model as continuous covariate	Clendenen et al.: Smoking; BMI; cohort; age at menarche; parity; hysterectomy; oophorectomy; benign breast disease Jung et al.: Age at menarche	Clendenen et al.: Geometric mean AMH of: never users = 2.75 pmol/l [2.49; 3.04] former users = 2.78 pmol/l [2.63; 2.95] current users = 1.60 pmol/l [1.24; 2.06] P < 0.0001 Jung et al.: Median AMH of: non-users = 1.15 ng/ml (0.55–2.16) current users = 1.04 ng/ml (<limit of detection-1.55) P = 0.04	↓ AMH	Fair
Deb et al. (2012)	UK	Prospective cohort	70 age-matched healthy women of whom 34 were using COCs for ≥1 year (20–34.6 years) and 36 not using hormonal contraception within the last year (20.3–35 years)	Serum, AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	Age-matching (method not described). Mean age ± SD: 25.35 years ± 4.23 for COC users and 27.16 years ± 4.63 for non-users	None	Mean AMH of: non-users = 3.06 ng/ml ± 1.27 current users = 2.75 ng/ml ± 1.59 P = 0.440	● AMH	Poor
Dólleman et al. (2013) III	The Netherlands	Cross-sectional	2230 women (20–59 years) participating in a population study	Serum, Gen II AMH ELISA, Beckman Coulter, USA	Age-standardized percentiles were calculated with the CG-LMS method	None	Relative mean difference in age-standardized AMH for current users vs non-users = −11.0 percentiles (SE = 1.3), P < 0.0001 Relative mean difference in age-standardized AMH (vs never users) for: former users = −0.3 percentiles (SE = 2.3), P = 0.88 current users = −9.0 percentiles (SE = 2.3), P < 0.0001 Relative mean difference in age-standardized AMH (vs use <1 year) in current/former users for: use 1–5 years = 2.6 percentiles (SE = 3.1), P = 0.39 use 5–10 years = −2.2 percentiles (SE = 3.0), P = 0.46 use 10–15 years = −2.6 percentiles (SE = 3.2), P = 0.42 use 15–20 years = −3.9 percentiles (SE = 3.5), P = 0.26 use >20 years = −8.8 percentiles (SE = 4.2), P = 0.04	↓ AMH	Fair
Hariton et al. (2021) and Nelson et al. (2023)	USA	Cross-sectional	Hariton et al.: 22 248 women (20–46 years) who purchased a fertility hormone test (4877 women using other forms of hormonal contraception not included here) Nelson et al: 42 684 women (21–45 years) who purchased a fertility hormone test (of whom 6850 used OCs, 27 514 used no contraception and 8320 used other forms of contraception)	Serum (either collected by venipuncture or self-collected with dried blood spot collection cards), Access AMH immuno-assay, Beckman Coulter	Hariton et al.: Added to regression model as continuous covariate Nelson et al.: Age-standardized percentiles were calculated with smooth additive quantile regression	Hariton et al.: BMI; number of cigarettes smoked per month; age at menarche; self-reported PCOS; sample collection method; day of sample collection Nelson et al.: only in sensitivity analysis, results only displayed in graphs	Hariton et al.: Relative mean difference in AMH for current users vs non-users = −24% [−28%; −19%], P < 0.001 A subgroup analysis among COC users (n = 1952) revealed no significant association between the duration of COC use and AMH (P = 0.122). Nelson et al: Relative difference in AMH for current users vs non-users = −17% [−18%; −15%] Stronger associations were found among women with lower AMH percentiles: current users 10th AMH percentile = −32% [−35%; −29%] current users 90th AMH percentile = −5% [−8%; −2%]	↓ AMH	Fair
Kallio et al. (2013)	Finland	Prospective intervention study (open label)	42 healthy white women (20–33 years) randomized to either a COC (n = 13), a transdermal contraceptive patch (n = 15) or a vaginal ring (n = 14)	Serum, Gen II AMH ELISA, Beckman Coulter	N/A (pre- and post-treatment values are measured)	None	Mean AMH** at: baseline (before COC) = 3.88 ng/ml ± 3.0 5 weeks after COC initiation = 3.34 ng/ml ± 2.8 9 weeks after COC initiation = 1.91 ng/ml ± 1.5 P < 0.001 for baseline vs 9 weeks after treatment initiation	↓ AMH	Fair
Kucera et al. (2016)	Czech Republic	Cross-sectional	149 women who had either been taking COCs for at least 10 years and terminated their use for ≥ 1 year (27–34 years), or women who had never used hormonal contraception (23–34 years)	Serum, Access AMH immuno-assay, Beckman Coulter, USA	None	None	Median AMH of: never users = 3.37 ng/ml (2.07–4.42) former users = 2.89 ng/ml (1.57–4.46) P = 0.3261	● AMH	Poor
Landersoe et al. (2020a)	Denmark	Prospective cohort	68 women (24–40 years) who are currently using COCs, have used COCs for ≥3 years and are willing to discontinue COCs for 3 months with no other form of hormonal contraception	Serum, Elecsys AMH Plus assay, Roche Diagnostics, Germany	N/A (pre- and post-treatment values are measured) (first analysis) Added to regression model as continuous covariate (second analysis)	BMI; body weight; dose of ethinyl oestradiol in the COC; duration of usage; PCOS status at end of follow-up	Relative mean difference in AMH (before discontinuation of COC use vs 3 months after discontinuation of COC use) = 53% [40%; 68%], P < 0.001 When testing how soon the first changes occurred, linear mixed models showed that median AMH was significantly increased 2 weeks after discontinuation: During COC use = 13 pmol/l (8.4–22) 2 weeks after COC discontinuation = 17 pmol/l (11–27) P < 0.001	↓ AMH	Good
Landersoe et al. (2020b)	Den-mark	Retrospective cohort	1387 women (19–46 years) attending fertility assessment and counselling clinics (161 women using other forms of hormonal contraception not included here)	Serum, Elecsys AMH Plus assay, Roche Diagnostics, Germany	Added to regression model as continuous covariate	Tentative PCOS diagnosis	Relative mean difference in AMH for current users vs non-users = −31.1% [−39.6%; −25.9%], P < 0.001	↓ AMH	Fair
Langton et al. (2021)	USA	Cross-sectional	1398 women (39.1 years ± 2.7) participating in a population study	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariate (age2)	BMI; smoking; pack-years; alcohol consumption; fasting status; season of blood collection; luteal day; paired sample; age at menarche; parity; breast-feeding; tubal ligation; infertility because of an ovulatory disorder	Geometric mean AMH of: never users = 1.88 ng/ml [1.65; 2.14] use 1–23 months = 1.79 ng/ml [1.57; 2.04] use 24–47 months = 1.70 ng/ml [1.51; 1.94] use 48–71 months = 1.59 ng/ml [1.37; 1.86] use 72–95 months = 1.56 ng/ml [1.31; 1.84] use 96–119 months = 1.56 ng/ml [1.25; 1.94] use ≥120 months = 1.60 ng/ml [1.34; 1.90] p for trend = 0.036 Geometric mean AMH of: never users = 1.75 ng/ml [1.51; 2.04] start use at 13–17 years old = 1.74 ng/ml [1.48; 2.06] start use at 18–19 years old = 1.77 ng/ml [1.57; 2.00] start use at 20–21 years old = 1.60 ng/ml [1.39; 1.84] start use at 22–23 years old = 1.77 ng/ml [1.51; 2.10] start use at 24–29 years old = 1.60 ng/ml [1.35; 1.90] start use at 30–34 years old = 1.37 ng/ml [0.90; 2.08] start use ≥35 years old = 1.21 ng/ml [0.51; 2.68] age unspecified = 1.57 ng/ml [0.80; 3.04] p for trend = 0.09	↓ AMH	Fair
Lawal and Yusuff (2021) II	Nigeria	Cross-sectional	161 women (21–40 years) attending the Gynecology and Immunization clinic of a tertiary hospital	Serum, AMH ELISA, Calbiotech Inc., USA	None	None	Mean difference in log AMH for past users vs never users¥¥¥ = 0.029 ng/ml (SE = 0.104), P = 0.781	● AMH	Poor
Li et al. (2011)	China	Prospective cohort	23 women (26–50 years) starting COCs	Serum, Gen II AMH ELISA, Beckman Coulter, France	N/A (pre- and post-treatment values are measured)	None	Median AMH: baseline (before COC) = 27.2 pmol/l (10.3–40.3) 3–4 months after COC initiation = 17.1 pmol/l (8.5–28.6) P > 0.05	● AMH	Fair
Mes-Krowinkel et al. (2014)	The Nether-lands	Cross-sectional	1297 women (28 years ± 5.2) with PCOS attending the fertility clinic	Serum, Gen II AMH ELISA, Beckman Coulter	Added to regression model as continuous covariate	BMI; WHR; previous pregnancy; fertility treatment; genetic ethnicity	Mean AMH** of: never users = 8.3 µg/l ± 2.7 former users who stopped using OCs >10 years ago = 6.8 µg/l ± 1.6 former users who stopped using OCs 5–10 years ago = 7.5 µg/l ± 1.8 former users who stopped using OCs 3–5 years ago = 8.3 µg/l ± 1.8 former users who stopped using OCs 1–3 years ago = 8.1 µg/l ± 1.8 former users who stopped using OCs <1 year ago = 11.8 µg/l ± 2.7 current users = 6.7 µg/l ± 1.5 P ≤ 0.001 for former users who stopped using OCs <1 year ago vs never users P > 0.05 for current users vs never users	↓ AMH	Fair
Somunkiran et al. (2007)	Turkey	Prospective intervention study	45 women starting OCs, of whom 30 with PCOS (25.1 years ± 6.9) and 15 age- and BMI-matched healthy women seeking advice on contraception (24.8 years ± 5.7)	Serum, AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	N/A (pre- and post-treatment values are measured)	None	Mean AMH of women with PCOS: baseline (before COC) = 5.49 ng/ml ± 2.26 6 months after COC initiation = 5.31 ng/ml ± 2.65 P > 0.05 Mean AMH of women without PCOS: baseline (before COC) = 1.93 ng/ml ± 0.51 6 months after COC initiation = 2.11 ng/ml ± 0.56 P > 0.05	● AMH	Fair
van den Berg et al. (2010)	The Nether-lands	Prospective cohort	25 women (18–40 years) who had used hormonal contraceptives for ≥3 months and are willing to discontinue hormonal contraceptive use	Serum, ultrasensitive immuno-enzymo-metric AMH assay kit, Diagnostic Systems Laboratories, USA	N/A (pre- and post-treatment values are measured)	None	Mean AMH**,¶¶: day 7 hormone-free interval (before COC discontinuation) = 2.0 µg/l ± 1.2 natural cycle 1 (after COC discontinuation) = 2.0 µg/l ± 1.3 natural cycle 2 (after COC discontinuation) = 2.6 µg/l ± 1.3 P = 0.001 for natural cycle 2 vs hormone-free interval P = 0.91 for natural cycle 1 vs hormone-free interval P = 0.01 for natural cycle 2 vs natural cycle 1	↓ AMH	Fair
ALCOHOL CONSUMPTION
Dólleman et al. (2013) IV	The Nether-lands	Cross-sectional	2230 women (20–59 years) participating in a population study	Serum, Gen II AMH ELISA, Beckman Coulter, USA	Age-standardized percentiles were calculated with the CG-LMS method	None	Relative mean difference in age-standardized AMH for daily alcohol consumers vs non-consumers = −0.4 percentiles (SE = 1.3), P = 0.74 Relative mean difference in age-standardized AMH per 1 unit increase in the number of daily alcohol beverages = 1.6 percentiles (SE = 1.0), P = 0.12	● AMH	Fair
Gouvea et al. (2022) II	Brazil	Cross-sectional	177 climacteric women (40–64 years) from Basic Health Units	Serum, Access II AMH immuno-assay, Beckman Coulter, USA	Added to regression model as covariate	Stages of reproductive aging	Mean difference in log AMH for alcohol consumers vs non-consumers = −0.496 ng/ml (SE = 0.283), P = 0.081	● AMH	Poor
Hawkins Bressler et al. (2016) II	USA	Cross-sectional	1654 women (23–34 years) participating in a population study	Serum, picoAMH ELISA kit, Ansh Labs, USA	Added to regression model as continuous covariates (age and age2)	BMI; OC use	Relative mean difference in AMH (vs never consumers) for: former consumers = 8% [−22%; 49%] current consumers = 3% [−9%; 16%] Relative mean difference in AMH for frequent (≥2 times p/w) binge drinkers (vs never binge drinkers) among current alcohol consumers = −26% [−44%; −2%] The authors reported additional results for alcohol consumption during years drank, cumulative alcohol exposure, current intake, binging in last year, and current intake in past heavy drinkers, but there were no significant associations.	↓ AMH for frequent binge drinkers	Fair
Kalem et al. (2019) IIII	Turkey	Cross-sectional	77 healthy female students (18–28 years) of whom none had ever been pregnant	Plasma, MIS/AMH ELISA kit, NR	None	None	Median (min-max) AMH of alcohol consumption frequency category: never = 2.6 ng/ml (0.7–6.0) rarely = 2.2 ng/ml (1.0–4.8) 2–3 times/week = 2.3 ng/ml (2.0–2.6) P = 0.880	● AMH	Poor
Kline et al. (2016) II	USA	Cross-sectional	477 women (19–45 years) participating in a study examining the relation of highly skewed X-chromosome inactivation to trisomy	Serum, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	Added to regression model as continuous covariate	Karyotype group; day of blood draw; interval between blood draw and assay; smoking; caffeine consumption	Mean difference in log AMH (vs alcohol consumption <1 day/week) for: alcohol consumption 1 day/week = −0.097 ng/ml [−0.312; 0.117] alcohol consumption 2–7 days/week = −0.067 ng/ml [−0.312; 0.178]	● AMH	Fair
Lawal and Yusuff (2021) III	Nigeria	Cross-sectional	161 women (21–40 years) attending the Gynecology and Immunization clinic of a tertiary hospital	Serum, AMH ELISA, Calbiotech Inc., USA	None	None	Mean difference in log AMH for alcohol consumers vs non-consumers = 0.060 ng/ml (SE = 0.090), P = 0.50	● AMH	Poor
Mitchell et al. (2023) III	Ireland	Cross-sectional	96 women (35 years ± 5) attending a reproductive medicine clinic	Serum, Elecsys AMH assay, Roche Diagnostics	Added to regression model as continuous covariate	None	Mean difference in AMH per 1 unit increase in alcohol score (based on the Simple Lifestyle Indicator Questionnaire (SLIQ)) = −0.94 pmol/l (SE = 3.14), P = 0.77	● AMH	Poor
Nardo et al. (2007) III	UK	Cross-sectional	136 women (23–39 years) undergoing fertility work-up prior to ovarian stimulation	Plasma, ultra-sensitive AMH ELISA, Oxford Bio Innovation DSL LTD, UK	Added to regression model as continuous covariate	None	Median log AMH of: non-consumers = 1.8 ng/ml (1.2–3.3) alcohol consumers = 2.1 ng/ml (1.0–3.5) Unadjusted P = 0.99 The authors report that there was no significant difference in age-adjusted AMH levels between alcohol consumers and non-alcohol consumers (P = 0.94).	● AMH	Poor
Soares et al. (2020) III	UK	Prospective cohort	1608 women (48 years (45–51)) participating in a population study	Serum, Elecsys AMH Plus assay, Roche Diagnostics	Added to regression model as continuous covariate	Educational achievement; BMI; smoking	Average predicted means [95% CI] 5 years before the final menstrual period (FMP): never consumers or <4 times/month: 0.11 ng/ml [0.08; 0.14] alcohol consumption 2–3 times/week: 0.13 ng/ml [0.09; 0.18] The AMH decline rates did not differ by alcohol intake. Corresponding association measures were not published in the article or supplementary information.	● AMH	Good
Whitworth et al. (2015) II	South Africa	Cross-sectional	425 women (20–30 years) participating in a population study	Plasma, Gen II AMH ELISA, Beckman Coulter, USA	Added to regression model as continuous covariate	BMI; education; parity	Relative mean difference in AMH for ever consumers vs never consumers = −21% [−36%; −3%]	↓ AMH	Fair
CAFFEINE CONSUMPTION
Kline et al. (2016) III	USA	Cross-sectional	477 women (19–45 years) participating in a study examining the relation of highly skewed X-chromosome inactivation to trisomy	Serum, DSL-14400 Active AMH/MIS ELISA kit, Diagnostic Systems Laboratories, USA	Added to regression model as continuous covariate	Karyotype group; day of blood draw; interval between blood draw and assay; alcohol consumption; smoking	Mean difference in log AMH (vs caffeine consumption <37 mg/day) for: caffeine consumption ≥37–<122 mg/day = −0.177 ng/ml [−0.312; 0.057] caffeine consumption ≥122–<553 mg/day = −0.097 ng/ml [−0.336; 0.141]	● AMH	Fair
Lawal and Yusuff (2021) IV	Nigeria	Cross-sectional	161 women (21–40 years) attending the Gynecology and Immunization clinic of a tertiary hospital	Serum, AMH ELISA, Calbiotech Inc., USA	None	None	Mean difference in log AMH for caffeine consumers vs non-consumers = 0.039 ng/ml (SE = 0.063), P = 0.541	● AMH	Poor
Whitworth et al. (2015) III	South Africa	Cross-sectional	425 women (20–30 years) participating in a population study	Plasma, Gen II AMH ELISA, Beckman Coulter, USA	Added to regression model as continuous covariate	BMI; education; parity	Relative mean difference in AMH of coffee consumers vs non-consumers = −19% [−31%; −5%]	↓ AMH	Fair
PHYSICAL ACTIVITY
Dólleman et al. (2013) V	The Netherlands	Cross-sectional	2230 women (20–59 years) participating in a population study	Serum, Gen II AMH ELISA, Beckman Coulter, USA	Age-standardized percentiles were calculated with the CG-LMS method	None	Relative mean difference in age-standardized AMH (vs CPAI 1 (inactive) for: CPAI 2 (moderately inactive) = −5.1 percentiles (SE = 4.2), P = 0.23 CPAI 3 (moderately active) = −2.8 percentiles (SE = 4.2), P = 0.50 CPAI 4 (active) = −4.4 percentiles (SE = 4.1), P = 0.29	● AMH	Fair
Kalem et al. (2019) V	Turkey	Cross-sectional	77 healthy female students (18–28 years) of whom none had ever been pregnant	Plasma, MIS/AMH ELISA kit, NR	None	None	Median (min-max) AMH of physical activity frequency category: regular = 2.6 ng/ml (1.1–6.0) irregular = 2.5 ng/ml (0.7–5.6) P = 0.327	● AMH	Poor
Miller et al. (2022)	Israel	Prospective cohort	31 women (29.9 years ± 4.2) engaged in high physical activity for ≥3 years before study recruitment and 31 age- and BMI-matched non-physically active women (31.6 years ± 2.2) working as hospital staff	Serum, Gen II AMH ELISA, Beckman Coulter, Czech Republic	Age-matching (method and age ranges not described)	None	Mean AMH of: Women with high physical activity = 6.26 ng/ml ± 3.3 Women with no physical activity = 4.96 ng/ml ± 3.7 P = 0.16 The authors report that the International Physical Activity Questionnaire (IPAQ) score was not related to AMH in multivariable logistic regression (P = 0.682), but no information about dichotomization of AMH or the strength of the association was provided.	● AMH	Poor
Mitchell et al. (2023) IIII	Ireland	Cross-sectional	96 women (35 years ± 5) attending a reproductive medicine clinic	Serum, Elecsys AMH assay, Roche Diagnostics	Added to regression model as continuous covariate	None	Mean difference in AMH per 1 unit increase in physical activity score (based on the Simple Lifestyle Indicator Questionnaire (SLIQ)) = −2.72 pmol/l (SE = 2.18), P = 0.21	● AMH	Poor
Moran et al. (2011)	Australia	Prospective intervention study (pilot)	15 women (20–40 years) with BMI > 27 kg/m2 participating in a 12-week intensified exercise intervention, of whom 7 with PCOS and 8 with PCOS	Serum, immuno-enzymatic AMH assay, Immuno-tech Beckman Coulter, France	N/A (pre- and post-treatment values are measured)	None	Mean AMH** of women with PCOS: baseline (before exercise intervention) = 59.1 pmol/l ± 20.5 12 weeks after exercise intervention = 45.9 pmol/l ± 15.3 P = 0.025 Mean AMH** of women without PCOS: baseline (before exercise intervention) = 14.9 pmol/l ± 9.9 12 weeks after exercise intervention = 16.3 pmol/l ± 12.0 P = 0.48	↓ AMH in women with PCOS	Fair
Wu et al. (2021)	China	Prospective intervention study	38 women (18–40) with PCOS randomized to either a 12-week aerobic exercise intervention or maintaining their normal lifestyle	NR	N/A (randomization)	N/A (randomization)	Mean AMH of women in the exercise group: baseline (before exercise intervention) = 17.6 ng/ml ± 5.1 12 weeks after exercise intervention = 14.8 ng/ml ± 3.6 P = 0.021 Mean AMH of women in the control group: baseline = 18.4 ng/ml ± 5.8 12 weeks later = 17.9 ng/ml ± 4.7 P = 0.118 There was a significant difference in changes in AMH between the groups (P = 0.014)	↓ AMH in exercise intervention group	Poor
Zubair et al. (2021)	Pakistan	Prospective intervention study	20 healthy women (25–35 years) with a sedentary lifestyle participating in either a 8-week light aerobic exercise or heavy exercise intervention	NR, AMH ECLIA, NR	N/A (pre- and post-treatment values are measured)	None	Mean AMH of women in the light exercise group: baseline (before exercise intervention) = 2.13 ng/dl ± 3.11 8 weeks after exercise intervention = 2.2 ng/dl ± 2.39 P > 0.05 Mean AMH of women in the heavy exercise group: baseline (before exercise intervention) = 4.10 ng/dl ± 2.98 8 weeks after exercise intervention = 3.39 ng/dl ± 2.29 P < 0.05	↓ AMH in heavy exercise group	Poor

Explanation symbols ‘Main findings’: an upwards arrow means a significantly increased anti-Müllerian hormone (AMH) level (or higher AMH decline) in presence of the lifestyle factor, a downwards arrow means a significantly decreased AMH level (or lower AMH decline) in presence of the lifestyle factor, and a dot means no significant association between AMH level/decline and the lifestyle factor.

The Roman numbers are used for studies including multiple determinants and indicate the time the study was mentioned in this table. For example, II indicates that the study is mentioned for the second time.

IQR, interquartile range; NR, not reported; OC, oral contraceptive; NOR, normal ovarian reserve; DOR, diminished ovarian reserve; ref, reference; CLIA, chemiluminescent immunoassay; df, degrees of freedom; COC, combined oral contraceptives; CPAI, Cambridge Physical Activity Index; GnRHa, GnRH agonist; HOMA-IR, homeostatic model assessment for insulin resistance; WHR, waist–hip ratio.

*

Jaswa et al. and Rios et al. use data from the same two cohorts. Rios et al. included one additional cohort. As Jaswa et al. performed a logarithmic transformation of AMH, we chose to include their data for the overlapping cohorts. The data from the additional cohort from Rios et al. have been included separately.

**

It is not completely certain that AMH has been log-transformed. Although the authors reported that variables with a skewed distribution were log-transformed, it was not mentioned whether this applied to AMH. Since the distribution of AMH is generally right-skewed, we assumed the authors performed a log-transformation. However, it remains unclear whether the given values had been back-transformed, especially when the unit of AMH concentration used in the analyses was not reported. These results should be interpreted carefully.

***

The terms ‘serum’ and ‘plasma’ were used simultaneously.

¥

Multiple linear regression was performed, but it was not clearly specified which variables were added to the model(s).

¥¥

In these analyses, no distinction was made between different types of hormonal contraceptives. Therefore, the analyses also include women using a vaginal ring (n = 11).

¥¥¥

In these analyses, no distinction was made between different types of hormonal contraceptives.

¶

In these analyses, no distinction was made between different types of hormonal contraceptives. Therefore, the analyses also include women using hormonal IUDs, implants, vaginal rings, transdermal patches, and injections. COCs were the most commonly used type of hormonal contraceptives.

¶¶

In these analyses, no distinction was made between different types of hormonal contraceptives. Therefore, the analyses also include women using a transdermal patch and a vaginal ring (n = 2).

Quality assessment

The results of the quality assessment are presented in Supplementary Table S2 and Supplementary Table S3 for observational and intervention studies, respectively. Three studies were judged as having ‘good’ quality, 27 as ‘fair’ quality, and 35 as ‘poor’ quality. The studies with a perceived ‘good’ quality were all prospective cohorts and clearly stated the timing of the lifestyle factor and AMH measurements. In addition, AMH levels were logarithmically transformed and age adjustment was applied using chronological age as continuous variable for all the results. In studies with a perceived ‘fair’ quality, logarithmic transformation of AMH and age adjustment was mostly performed, but the study design was cross-sectional. Studies received a ‘poor’ quality rating for various reasons, the most common being lack of age adjustment. Additional reasons that contributed to a ‘poor’ rating were: no description of age-matching methods; unclear description of how and when lifestyle factors and/or AMH were measured; using parametric methods but not performing a logarithmic transformation of AMH levels; and an unclear description of the (recruitment of the) study population.

BMI

Thirty-six studies investigated the relation between AMH levels or AMH decline and BMI, of which 28 were cross-sectional, seven were prospective cohort studies and one was a retrospective cohort study. Twenty-two studies found an inverse association between AMH levels and BMI (Freeman et al., 2007; Su et al., 2008; Piouka et al., 2009; Steiner et al., 2010; Buyuk et al., 2011; Cui et al., 2014; Kriseman et al., 2015; Moy et al., 2015; Olszanecka-Glinianowicz et al., 2015; Yang et al., 2015; Santulli et al., 2016; Bernardi et al., 2017; Lefebvre et al., 2017; Jaswa et al., 2020; Rios et al., 2020; Soares et al., 2020; Clendenen et al., 2021; Lim et al., 2021; Grimes et al., 2022; Wang et al., 2022; Tzeng et al., 2023; Zhao et al., 2023), whereas three studies found a positive association (Bakeer et al., 2018; Albu and Albu, 2019; Lawal and Yusuff, 2021). Eleven studies found no significant association (Nardo et al., 2007; Halawaty et al., 2010; Skałba et al., 2011; Sahmay et al., 2012; Dólleman et al., 2013; Whitworth et al., 2015; Jung et al., 2017; Simões-Pereira et al., 2018; Kalem et al., 2019; Gouvea et al., 2022; Mitchell et al., 2023).

Among the studies that reported a significant inverse association, i.e. higher BMI associating with lower AMH, the absolute mean differences in serum log AMH levels in ng/ml per 1 kg/m² increase in BMI ranged from −0.015 ng/ml [95% CI: −0.021; −0.009] (Bernardi et al., 2017) to −0.2 ng/ml [95% CI NR] (P = 0.01) (Buyuk et al., 2011). The highest mean difference was found in a cross-sectional study including 152 women classified as having a diminished ovarian reserve based on their FSH level (>10 IU/l) (Buyuk et al., 2011). This study did not detect a significant relation between AMH level and BMI in the 138 women with normal ovarian reserve (FSH ≤10 IU/l), however.

Two prospective cohort studies investigated the association between BMI and AMH decline. In a study including 795 female registered nurses, two AMH measurements per person were performed with a time interval of 11–16 years (Grimes et al., 2022). Geometric AMH levels were calculated in order to compare the change in geometric mean AMH levels over this period among underweight (BMI <18.5 kg/m²), healthy weight (BMI ≥18.5–<25 kg/m²), overweight (BMI ≥25–<30 kg/m²) and obese (BMI ≥30 kg/m²) women at baseline. The analyses were adjusted for age at both blood collections to account for individual differences in the time interval between measures. In this study, obese and overweight women had slower rates of AMH decline compared to women with a healthy weight (geometric mean change_obesity = 2.88 ng/ml (P = 0.0002), geometric mean change_overweight = 3.22 ng/ml (P = 0.002), geometric mean change_{normal weight} = 3.95 ng/ml (Grimes et al., 2022)).

The other prospective cohort study was a population-based study among 1608 individuals in which AMH levels were measured on four occasions (Soares et al., 2020). Obese and overweight women also appeared to have lower rates of decrease in AMH compared to women with a healthy BMI in this study. At 5 years before the final menstrual period (FMP), women with a healthy weight had the highest mean predicted AMH levels in comparison with overweight and obese women (0.19 [95% CI: 0.15; 0.25] compared to 0.10 [95% CI: 0.08; 0.13]), which in all cases declined to 0.02 ng/ml [95% 0.01; 0.02] at the time of the FMP.

Several studies performed a subgroup analysis for women with and without PCOS, a syndrome relating to both BMI and AMH levels (Piouka et al., 2009; Cui et al., 2014; Olszanecka-Glinianowicz et al., 2015; Lefebvre et al., 2017; Bakeer et al., 2018). Overall, the relation of BMI with AMH appeared stronger or more apparent in women with PCOS in comparison with women who did not meet this classification; in some studies, the relation between BMI and AMH was only significant in women with PCOS (Kriseman et al., 2015; Lefebvre et al., 2017), while another study found that PCOS increased the effect size of the association (Cui et al., 2014).

The studies that did not find an association between BMI and AMH were, similar to the studies that did report an association, mostly performed in a population-based sample of women or unselected women attending a reproductive care clinic.

Waist–hip ratio

Two studies, one of which was cross-sectional and the other a prospective cohort, included WHR as a possible determinant of AMH level. The cross-sectional study including 77 healthy female students performed a Spearman rank order correlation analysis (Kalem et al., 2019), while the prospective cohort study among 795 nurses calculated geometric means for WHR categories (Grimes et al., 2022). Both studies did not report a significant association.

Smoking

Among the 17 studies assessing the association between AMH levels or AMH decline and smoking, 14 used a cross-sectional design, two were prospective cohorts, and one was a retrospective cohort. Seven studies reported that smoking was related to lower AMH levels (Freour et al., 2008; Dafopoulos et al., 2010; Plante et al., 2010; Dólleman et al., 2013; White et al., 2016; Kalem et al., 2019; Clendenen et al., 2021), while 10 studies found no relation (Nardo et al., 2007; Waylen et al., 2010; Hawkins Bressler et al., 2016; Kline et al., 2016; Jung et al., 2017; Soares et al., 2020; Bhide et al., 2022; Oladipupo et al., 2022; Mitchell et al., 2023; Tzeng et al., 2023).

The largest study analyzed 3831 subjects participating in 10 different cohorts from the USA, the UK, and Sweden, and calculated geometric means of AMH concentrations for current, former and never smokers (2.24 pmol/l [95% CI: 1.94; 2.59], 2.55 pmol/l [95% CI: 2.31; 2.82] and 2.85 pmol/l [95% CI: 2.68; 3.04], respectively, P = 0.0058) (Clendenen et al., 2021). Other studies found current smokers had 44% lower AMH levels (Plante et al., 2010) and 3.6 lower age-standardized AMH percentiles (Dólleman et al., 2013). In two studies that did not report a significant difference between AMH levels of smokers and non-smokers, the amount of smoking was found to be relevant; currently smoking ≥20 cigarettes/day was associated with lower AMH levels (−55.3% [95% CI: −79.8%; −0.9%]) compared to never smokers (White et al., 2016), and AMH levels were correlated with the number of pack-years of smoking among smokers (r = −0.807, P < 0.001) (Dafopoulos et al., 2010).

The AMH decline rate was highest for current smokers, followed by former smokers and non-smokers (Soares et al., 2020). Passive smoking did not associate with AMH levels in two studies (Plante et al., 2010; Oladipupo et al., 2022). The studies which did not find a significant association of smoking with AMH levels reported an overall trending inverse relation, which did not reach statistical significance. There were no studies relating smoking to higher AMH levels.

Oral contraceptive use

Of the 20 studies exploring the relation between AMH and OC use, 11 were cross-sectional, five were prospective cohort studies, three were prospective intervention studies and one was a retrospective cohort study. Fifteen studies reported an inverse association of OC use with AMH levels (Arbo et al., 2007; van den Berg et al., 2010; Bentzen et al., 2012; Dólleman et al., 2013; Kallio et al., 2013; Mes-Krowinkel et al., 2014; Birch Petersen et al., 2015; Jung et al., 2017; Landersoe et al., 2020a,b; Bernardi et al., 2021; Clendenen et al., 2021; Hariton et al., 2021; Langton et al., 2021; Nelson et al., 2023). Five studies did not find a significant association (Somunkiran et al., 2007; Deb et al., 2012; Kucera et al., 2016; Lawal and Yusuff, 2021). The studies which did not report a significant association of OC use with AMH were all performed in small study populations, with the sample size not exceeding 149. The remaining studies generally included a larger population-based or unselected group of women, of which the largest sample size was 22 248. Four small studies (with a largest sample size of 68) compared intra-individual AMH levels before and after OC initiation, and two studies compared AMH levels before and after OC discontinuation.

Among the studies that observed a significant difference between current OC users and non-users, the relative mean difference in AMH levels ranged between −17% [95% CI: −18%; −15%] (Nelson et al., 2023) and −31.1% [95% CI: −39.6%; −25.9%] (Landersoe et al., 2020b) or a 42% lower geometric mean level for OC users (Clendenen et al., 2021). One study assessed whether there was effect modification by AMH level and found that women with the lowest age-specific AMH percentiles experienced a greater difference of AMH levels for OC users versus non-users, in comparison with women with high age-specific AMH percentiles (Nelson et al., 2023). The AMH decline rate was not reported to be affected by OC use (Bentzen et al., 2012). The duration of OC use was not associated with degree of AMH difference in any of the studies and former OC users did not have differing levels from women without hormonal contraception (Kucera et al., 2016; Bernardi et al., 2021; Clendenen et al., 2021). Initiation of OC use led to a swift decline in AMH levels after 1 month and no difference in the levels thereafter (Langton et al., 2021). Discontinuation of OC led to a 53% increase of AMH levels after 3 months (Landersoe et al., 2020a). Another study found OC discontinuation led to increased AMH levels after the second natural cycle (van den Berg et al., 2010).

Alcohol consumption

Of the 10 studies investigating the relation between alcohol consumption and AMH, one was a prospective study, and the remaining studies used a cross-sectional design.

One population-based study of 425 women found an inverse association, where the AMH levels of women who had ever consumed alcohol were 21% lower [95% CI: −36%; −3%] than never alcohol consumers (Whitworth et al., 2015). Another study did not find a difference between ever and never consumers, but did find that among alcohol consumers, frequent binge drinkers on average had 26% lower AMH levels [95% CI: −44%; −2%] (Hawkins Bressler et al., 2016).

The remaining studies did not find an association between alcohol intake and AMH levels. Four of these studies compared current alcohol consumption to no current consumption (Nardo et al., 2007; Dólleman et al., 2013; Lawal and Yusuff, 2021; Gouvea et al., 2022), three studies used categories based on intake frequency (Kline et al., 2016; Kalem et al., 2019; Soares et al., 2020), and one study measured alcohol intake on a standardized scale (Mitchell et al., 2023). The only prospective cohort study did not find any significant differences between the average predicted mean AMH levels by time to the FMP in women who consumed alcohol 2–3 times per week and women who had never consumed alcohol or consumed alcohol <4 times per month (Soares et al., 2020).

Caffeine consumption

All three studies assessing the relation between AMH levels and caffeine consumption made use of a cross-sectional design. One study, which included 425 women, found a 19% [95% CI: −31%; −5%] lower mean AMH level in subjects who drank coffee than in subjects who did not drink coffee (Whitworth et al., 2015). The other two studies, which included other caffeinated drinks besides coffee, compared caffeine consumers to non-consumers (Lawal and Yusuff, 2021) or tertiles based on caffeine consumption in mg/day (Kline et al., 2016), and they found no association in their samples of 477 and 161 women, respectively.

Physical activity

Among the seven studies exploring the association between AMH and physical activity, three were prospective intervention studies, three were cross-sectional studies and one was a prospective cohort study.

The intervention studies included 15 women with and without PCOS and a BMI >27 kg/m² (Moran et al., 2011), 38 women with PCOS (Wu et al., 2021) and 20 healthy women with a sedentary lifestyle (Zubair et al., 2021) and all found that increased physical activity was associated with decreased AMH levels. In the largest intervention study, a decrease in AMH level of 2.8 ng/ml (P = 0.025) was observed in women with PCOS who underwent a 12-week aerobic exercise intervention (Wu et al., 2021). The observational studies, which included a large population study with 2230 women (Dólleman et al., 2013) and three smaller studies (<100 participants) (Kalem et al., 2019; Miller et al., 2022; Mitchell et al., 2023), did not detect an association.

Discussion

With this systematic review, we are able to give a comprehensive summary of the relations between seven routinely assessed lifestyle factors and circulating AMH levels and AMH decline. Based on a substantial number of studies with poor methodological quality, in combination with the cross-sectional designs of most studies, we perceived the overall quality of evidence for the relations between lifestyle and AMH to be fairly low. Additionally, heterogeneity in the design of the studies and statistical methods did not allow for data synthesis or meta-analysis of the results. Therefore, although the majority of studies found that higher BMI, smoking, OC use, and increased physical activity were associated with lower AMH levels, and alcohol consumption, caffeine consumption, and WHR were not associated with AMH, we believe that the current evidence is not sufficient to draw definitive causal conclusions.

Our findings are analogous to the results of a recent meta-analysis, which created a pooled effect estimate for the relation between BMI and AMH with data from 16 studies, confirming a significant inverse relation of BMI with AMH (Moslehi et al., 2018). Although the effect size was larger in women with PCOS compared to women without PCOS, this difference did not reach statistical significance. As obesity has detrimental effects on multiple aspects of reproductive health (Lainez and Coss, 2019), it is not implausible that there is a physiological relation between BMI and AMH. Several possible explanations exist for an effect of obesity on circulating AMH levels. It is speculated that increased adiposity might reduce AMH production per antral follicle (Jaswa et al., 2020; Oldfield et al., 2021). This hypothesis is supported by in vitro studies suggesting a role for several adipokines associated with obesity. For instance, increased leptin production could directly suppress follicular AMH production and signaling (Procaccini et al., 2012; Merhi et al., 2013). Indeed, higher BMI levels were found to be associated with lower expression of the AMH and its type 2 receptor genes in granulosa cells (Nouri et al., 2016). Another potential mechanism involves altered secretion of reproductive hormones in obese women at the level of the hypothalamus and pituitary, thereby potentially influencing the rate of follicle recruitment (Lainez and Coss, 2019). Moreover, as higher BMI appears to be associated with a potential modest increase in age at natural menopause (Tao et al., 2015; Zhu et al., 2018), it does not seem likely that lower AMH levels associated with BMI are primarily due to an accelerated decrease of ovarian reserve.

The effects of cigarette smoking on AMH concentrations are not fully understood. Most evidence points to the direction of smoke-induced oxidative stress affecting cellular mechanisms, including inflammation, apoptosis and mitochondrial damage, and thereby accelerating ovarian aging and loss of ovarian reserve (Sobinoff et al., 2013; Yang et al., 2020; Yan et al., 2022). Smoking is related to an earlier age at menopause, supporting the potential link between exposure to smoke and accelerated loss of ovarian reserve (Kinney et al., 2006; Parente et al., 2008). However, the inconsistencies in the current evidence are too large to definitively conclude that there is a causal effect of smoking on circulating AMH.

The findings with regard to OC use and AMH are mostly consistent. The majority of studies found circulating AMH levels to be lower in current but not in former OC users compared to never users, and also reported that discontinuation of OC use was associated with a subsequent increase of AMH levels within a period of 6 months at most. These findings suggest that the exogenous hormones in OCs exert their effects on AMH through temporary hormonal changes owing to pituitary suppression (De Leo et al., 1991; van Heusden et al., 2002). It is hypothesized that the FSH downregulation caused by OC use leads to decreased AMH production by granulosa cells and that this effect is likely reversible (Kristensen et al., 2012; Landersoe et al., 2020a; Yin et al., 2022). Moreover, no relation was found between OC use and total adjusted ovarian non-growing follicle count among 133 premenopausal women undergoing oophorectomy, further supporting a relation between OC use and follicle growth, rather than true ovarian reserve (Peck et al., 2016).

The number of studies investigating alcohol consumption, physical activity, caffeine consumption, and WHR was low. Therefore, we consider the current evidence insufficient to speak of either presence or absence of a relation with AMH. Interestingly, all three studies investigating the effect of physical activity interventions reported lower AMH levels post-intervention. It should be noted that the majority of study subjects in these exercise intervention studies were women with PCOS, whereas in the larger population-based study (Dólleman et al., 2013) there was no correlation between AMH and physical exercise. Women with PCOS are known to be at higher risk of obesity, likely in part through genetic susceptibility, as well as insulin resistance, hyperinsulinemia and hyperandrogenism (Joham et al., 2022). A lifestyle intervention study (including 33 women) found that physical activity led to a better improvement of insulin resistance than diet in women with PCOS (Jurczewska et al., 2023). Considering that levels of AMH in both women with and without PCOS have been found to be positively correlated to insulin, insulin resistance, and androgen concentration (Nardo et al., 2009), it is possible that the improvement of these parameters through physical exercise can lead to a decrease in AMH to more ‘normal’ levels (Gu et al., 2022).

Naturally, in addition to the factors we have included in our review, there are other lifestyle factors that may influence AMH levels or ovarian reserve. In recent years, dietary supplements, such as vitamin D, have frequently been associated with AMH levels, as summarized in a recent review (Moridi et al., 2020). It stands to reason that other dietary supplements or nutritional components may also be associated with AMH. Furthermore, factors such as stress and sleep, be it through hypothalamic regulation or otherwise, may additionally be related to AMH (Lim et al., 2016). Although these are relevant lifestyle facets that may play a part in the expression or regulation of AMH or ovarian aging physiology, their lack of standardization and need for large-scale intervention studies place them outside the scope of our current review.

Collectively, the evidence presented in this review can be a helpful tool for clinicians to incorporate lifestyle assessment into clinical AMH interpretation. While the current evidence does not support a relation with caffeine and alcohol intake, it can be helpful to take into account OC use, smoking, body weight, and physical activity when measuring AMH. For example, a patient with endometriosis who may not want to discontinue her hormonal treatment for an AMH measurement, can get information of the expected relative effect range of OC use on her AMH level. Likewise, if a patient who smokes has her AMH measured, an indication of the effect of smoking on her AMH levels can be provided, potentially as an additional motivational tool to stop smoking. However, we consider the summarized evidence regarding lifestyle factors and AMH levels insufficient to make causal inferences, and there is a lack of knowledge on the relations between lifestyle factors and the true ovarian reserve. Future studies on the associations between lifestyle factors and AMH levels may include methods such as Mendelian randomization to further address the causality of the reported associations. Furthermore, more research on alteration of lifestyle behaviors in relation to circulating AMH could help establish more clinically meaningful recommendations.

This is the first systematic review studying the effect of multiple lifestyle factors on AMH. As subgroups based on medical conditions were not excluded, the findings in this review are broadly clinically applicable. Owing to the lack of feasibility, we have not contacted corresponding authors of conference abstracts of which we could not find the full-text publications or otherwise inaccessible papers (n = 44), which mainly concerned BMI. This could introduce some publication bias to the results.

In conclusion, this review helped gain a comprehensive insight into the relation between lifestyle and circulating AMH. Several lifestyle factors are associated with differences in AMH levels and could therefore be incorporated into the interpretation of individual AMH measurements. However, we also demonstrated that there are insufficient studies using adequate methodological approaches to determine causality. Until more prospective studies are carried out, it is not yet possible to provide recommendations with regard to lifestyle changes and their effects on AMH levels and ovarian reserve.

Data availability

Data sharing not applicable. No new datasets were used during this study.

Authors’ roles

Review conception and design: L.W., A.d.K., Y.v.d.S.; study search and selection: L.W., A.d.K.; data extraction: L.W., A.d.K.; writing of draft manuscript: L.W.; supervision and reviewing manuscript: A.d.K., Y.v.d.S.

Funding

No external funding was obtained in support of this study.

Conflict of interest

The authors have no conflicts of interest to declare.