Anti-Müllerian hormone as a marker of ovarian reserve and premature ovarian insufficiency in children and women with cancer: a systematic review.

Anticancer treatments are known to have substantial impact on ovarian reserve via mechanisms leading to depletion of growing and primordial follicles. This reduction in ovarian reserve may reduce reproductive lifespan and be sufficient in some women to induce premature ovarian insufficiency (POI; Fig. 1 ; Spears et al. , 2019 ). This has profound implications for women’s health and is associated with infertility, cardiovascular disease (CVD), neurological disease, osteoporosis and mood disorders, culminating in an increased risk of premature mortality ( Webber et al. , 2016 ; van Dorp et al. , 2018 ; Panay et al. , 2020 ). Consequently, knowing a patient’s treatment-related risk of POI aids shared decision-making around the choice of anticancer treatment and the need for additional fertility preservation interventions. Putative mechanisms of impact of cancer treatment on ovarian function. ( a ) In premenopausal individuals, circulating AMH is produced by secondary, preantral and early antral growing follicles, and has been shown in animal models to be one of several molecules which contribute to maintenance of ovarian reserve by inhibiting folliculogenesis; ( b ) anticancer treatment can reduce the ovarian pool of primordial follicles either by direct or indirect DNA damage, or by overactivation and subsequent depletion of primordial follicles; ( c ) following treatment, a patient may experience some recovery of the number of AMH-producing growing follicles, depending on the impact of anticancer treatment, or POI. In patients who recover ovarian function, a reduced pool of primordial follicles may still lead to an increased risk of later POI and infertility. AMH, anti-Müllerian hormone; POI, premature ovarian insufficiency; ROS, reactive oxygen species Measurement of circulating anti-Müllerian hormone (AMH), produced by granulosa cells of growing follicles, is widely used as a surrogate for ovarian reserve and has potential as a diagnostic and predictive biomarker of POI ( Anderson and Su 2020 ). AMH levels are often heavily impacted during anticancer treatment, followed by a highly variable recovery ( Anderson and Su 2020 ). The diagnosis of POI is informed by monitoring a combination of biomarkers, including menstrual function and FSH ( Webber et al. , 2016 ; Panay et al. , 2020 ). However, confirming POI is more challenging in patients with cancer, as there is potential for recovery of ovarian/menstrual function, which can occur from several months to years after treatment completion ( Silva et al. , 2016 ). Questions remain about the reliability of AMH for evaluating ovarian reserve in the context of different anticancer treatments, and at what point post-treatment it can provide useful predictive information which may guide later treatment decisions, such as choice of endocrine therapy in women with breast cancer. Our objective was to conduct a systematic review evaluating the value of AMH in assessing ovarian reserve in women with cancer undergoing anticancer treatments, and specifically in the prediction and diagnosis of POI.

This study was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and registered with PROSPERO (CRD42020161717). PubMed, Embase and Cochrane Central Register of Controlled Trials (CENTRAL) were searched for articles published up to 1 April 2021 ( Supplementary Data ) and duplicate articles removed. Search results were screened in stages by title, abstract and full text based on the following inclusion criteria: female, premenopausal patients who have prospectively or retrospectively undergone treatment for any cancer. Exclusion criteria included pathology other than cancer (with the exception of granulosa cell tumours), AMH/ovarian reserve not measured, cohort size <10, duplicate datasets/cohorts, poor quality (e.g. inconsistent reporting of results), review articles and congress abstracts. Publications requiring evaluation of full text to determine inclusion underwent bias review via the Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I) protocol ( Sterne et al. , 2016 ). Bias conclusions were verified by a second author, with any disagreements mediated by a third author. All authors approved the final selection. Data were extracted using the Synthesis Without Meta-analysis (SWiM) in systematic reviews reporting guideline ( Campbell et al. , 2020 ). The main metrics assessed were differences in AMH following anticancer treatment (change versus pre-treatment or control groups), recovery of AMH (in longitudinal cohorts) and correlation of AMH with menstrual function (regular cycles, oligomenorrhoea or amenorrhoea as reported by patient at follow-up). Certainty of evidence was based on vote counting. Extracted data also included description of study types (prospective/retrospective, cross-sectional, longitudinal), year of publication, cancer type, treatment, cohort mean/median age, timing of AMH measurement and assay type. Exploratory subgroups were established according to age (paediatric, adult, </≥35 years), diagnosis (breast cancer or lymphoma) and length of follow-up. The data underlying this article are available in the article and in its online supplementary material , or as referenced publications.

Figure 2 shows the number of citations retrieved, the number after screening and the final number included in the analysis. The searches initially identified over 3700 citations. After removal of duplicates and assessment for quality and relevance to the study question at the title, abstract and full-text level, 110 were included for bias review. Of these, a further 20 were assessed to be at severe or critical risk of bias and excluded ( Supplementary Table SI ), leaving 90 publications including 8916 patients (N) for systematic review, published between 2003 and 2021. During the publication review process, a further two relevant articles that were published after 1 April 2021 were identified (total 112 for bias review), resulting in a total of 92 publications and 9183 patients. An overview of publications included in this review is summarized in Table I , and details of the individual publications are summarized in Supplementary Table SII . Database search results and exclusion flow of publications. AMH, anti-Müllerian hormone. Summary of publications included in review. *Regular menstruation, oligomenorrhoea or amenorrhoea at follow-up. N = 91 as Palinska-Rudzka et al. (2019) included separate analysis for breast cancer and lymphoma. Acute lymphoblastic leukaemia, differentiated thyroid cancer and gestational trophoblastic neoplasia. Mean or median as reported. AMH, anti-Müllerian hormone. Papers reporting AMH before and after treatment (N = 46; N = 4117), and a clear majority of cross-sectional papers comparing post-treatment survivors with control groups (23/26, 88%; n/N = 2283/3088) reported a large reduction in AMH following treatment ( Table I; Supplementary Table SII ). Effect sizes were variable depending on the treatment and diagnosis, but in the 26 papers that reported AMH values at both baseline and ≤3 months from end of treatment ( Lutchman Singh et al. , 2007 ; Decanter et al. , 2010 , 2018 , 2021 ; Rosendahl et al. , 2010 ; Yu et al. , 2010 ; Brougham et al. , 2012 ; Henry et al. , 2014 ; Ben-Aharon et al. , 2015 ; Gharwan et al. , 2016 ; Gupta et al. , 2016 ; Anderson et al. , 2017 , 2018 ; Bi et al. , 2017 ; D'Avila et al. , 2017 ; Leonard et al. , 2017 ; Trapp et al. , 2017 ; Cameron et al. , 2018 ; Evranos et al. , 2018 ; Yaish et al. , 2018 ; Passildas et al. , 2019 ; Silva et al. , 2019 ; Loubersac et al. , 2020 ; Berjeb et al. , 2021 ; Demeestere et al. , 2021 ; Martin et al. , 2021 ), reductions ranged from 42% to below the limit of detection (LOD) and 18 reported mean or median declines of ≥90% ( Rosendahl et al. , 2010 ; Brougham et al. , 2012 ; Dillon et al ., 2013a , b ; Henry et al. , 2014 ; Gharwan et al. , 2016 ; Ben-Aharon et al ., 2015 ; D'Avila et al. , 2017 ; Dezellus et al. , 2017 ; Trapp et al. , 2017 ; Cameron et al. , 2018 ; Decanter et al. , 2018 ; Lambertini et al. , 2019 ; Palinska-Rudzka et al. , 2019 ; Passildas et al. , 2019 ; Silva et al. , 2019 ; Oktay et al. , 2020 ; Ruddy et al. , 2021 ; Yu et al. , 2010 ). The six papers that did not detect a significant difference in AMH versus controls after treatment were all in paediatric populations (i.e. under 18 years of age at cancer diagnosis) and represented long-term follow-up periods of 10–30 years in heterogeneous populations of childhood cancer survivors ( Lie Fong et al. , 2009 ; Nielsen et al. , 2013 ; van den Berg et al. , 2018 ; Nystrom et al. , 2019 ; Nies et al. , 2020 ; Elitzur et al. , 2021 ). A shorter-term longitudinal study (follow-up of up to 43 months) in this population did, however, show clear reductions in AMH following treatment ( Brougham et al. , 2012 ), with large differences in the degree of recovery of AMH by treatment gonadotoxicity: thus in those exposed to high-risk treatment, AMH remained undetectable, while recovering to levels similar to pre-treatment in those exposed to lower risk treatment. In publications examining longitudinal data, post-treatment recovery of AMH was described in 33/42 (79%). With several notable exceptions described below, recovery was typically partial (i.e. to lower than pre-treatment level) or only occurred in a subset of patients. Twelve publications specifically evaluated the association of pre- and post-treatment AMH levels, with all reporting a significant association ( Rosendahl et al. , 2010 ; Dillon et al. , 2013b ; D’Avila et al. , 2015 ; Dezellus et al. , 2017 ; Anderson et al. , 2018 ; Lee et al. , 2018 , 2020 ; Palinska-Rudzka et al. , 2019 ; Silva et al. , 2019 ; Celebi et al. , 2020 ; Loubersac et al. , 2020 ; Decanter et al. , 2021 ). Ten publications specifically evaluated either the value of post-treatment AMH in the diagnosis of POI, or pre-treatment AMH in predicting the likelihood of POI. These studies were either in breast cancer or across multiple diagnoses, representing 762 patients ( Nielsen et al. , 2013 ; Lunsford et al. , 2014 ; Elchuri et al. , 2016 ; Anderson et al. , 2017 ; Nystrom et al. , 2019 ; Passildas et al. , 2019 ; Silva et al. , 2019 ; Zhong et al. , 2019 ; Demeestere et al. , 2021 ; Parissone et al. , 2021 ). Six studies (two of which were in paedia3tric cohorts) reported post-treatment AMH levels <1.0 ng/ml or undetectable in patients with POI ( Nielsen et al. , 2013 ; Lunsford et al. , 2014 ; Anderson et al. , 2017 ; Silva et al. , 2019 ; Zhong et al. , 2019 ; Parissone et al. , 2021 ) but the diagnostic accuracy of AMH for post-treatment POI has only been reported in one study ( Anderson et al. , 2017 ). Two papers found that lower pre-treatment AMH levels were associated with higher risk of POI (as assessed) post-treatment ( Passildas et al. , 2019 ; Zhong et al. , 2019 ), but no studies have assessed post-treatment AMH as a predictor of time to POI. Of 37 papers that specifically investigated menstrual function after treatment, all used prospective evaluation, and 14 (38%) found that reduced post-treatment AMH was associated with oligomenorrhoea or amenorrhoea ( Rosendahl et al. , 2008 ; Su et al. , 2010 ; Lunsford et al. , 2014 ; Ruddy et al. , 2014 ; D'Avila et al. , 2015 ; Gharwan et al. , 2016 ; Palinska- Morarji et al. , 2017 ; Wenners et al. , 2017 ; Decanter et al. , 2018 , 2021 ; Kim et al. , 2018 ; Palinska-Rudzka et al. , 2019 ; Silva et al. , 2019 ; Li et al. , 2020 ). In some of these studies, the observations were nuanced, for example Su et al. (2010) found that post-treatment AMH was associated with amenorrhoea but was not associated with recovery of menses (at 5.2 years post-treatment follow-up). Palinska-Rudzka et al. (2019) found that AMH was lower in patients with amenorrhoea, but that some (7/17) patients with AMH <LOD had menses. Gharwan et al. (2016) found AMH was associated with menstrual status only in woman aged 21−25 years old. Dezellus et al. (2017) found an association between AMH and menstrual status at 6 months of post-treatment follow-up, but not at 12 months of follow-up. Further studies are required to assess with rigour whether measurement of AMH adds to or can replace current diagnostic criteria for POI. A small number of studies have reported pregnancies in some patients despite low or undetectable AMH levels after cancer treatment ( Hamy et al. , 2016 ; Dezellus et al. , 2017 ; Anderson et al. , 2018 ; Loubersac et al. , 2020 ; Demeestere et al. , 2021 ). A recent analysis of AMH in women treated for advanced Hodgkin lymphoma in a randomized controlled trial showed similar pregnancy rates in the two treatment arms after 5 years of follow-up, despite much lower AMH levels and a higher POI rate in the arm receiving higher doses of alkylating agents ( Demeestere et al. , 2021 ). This is consistent with the lack of predictive value of AMH for pregnancy in healthy women ( Steiner et al. , 2017 ), but these limited data on a clinically highly relevant topic highlight the need for further prospective studies to inform on this subject. Of 39 publications that evaluated whether patient age at treatment was associated with post-treatment AMH levels, 30 (77%) found higher post-treatment AMH in younger individuals ( Lutchman Singh et al. , 2007 ; Lie Fong et al. , 2009 ; Charpentier et al. , 2014 ; Henry et al. , 2014 ; Ben-Aharon et al. , 2015 ; D'Avila et al. , 2015 ; Acibucu et al. , 2016 ; Elchuri et al. , 2016 ; Anderson et al. , 2017 , 2018 ; Dezellus et al. , 2017 ; Morarji et al. , 2017 ; Shandley et al. , 2017 ; Wenners et al. , 2017 ; Al-Janabi et al. , 2018 ; Evranos et al. , 2018 ; Lee et al. , 2018 ; Malisic et al. , 2018 ; Yaish et al. , 2018 ; Cameron et al. , 2019 ; Silva et al. , 2019 ; Celebi et al. , 2020 ; Li et al. , 2020 ; Loubersac et al. , 2020 ; Su et al. , 2020 ; van Velsen et al. , 2020 ; Berjeb et al. , 2021 ; Decanter et al. , 2021 ; Elitzur et al. , 2021 ; Martin 2021 ). Thirteen of these studies also stratified patient groups by <35 versus ≥35 years old, with 10 (83%) reporting lower AMH and poorer recovery in the older age group ( Yu et al. , 2010 ; Ben-Aharon et al. , 2015 ; Acibucu et al. , 2016 ; Dezellus et al. , 2017 ; Trapp et al. , 2017 ; Al-Janabi et al. , 2018 ; Anderson et al. , 2018 ; van den Berg et al. , 2018 ; Yaish et al. , 2018 ; Li et al. , 2020 ; Mittica et al. , 2020 ; van Velsen et al. , 2020 ; Decanter et al. , 2021 ). However, only five publications both stratified these two age groups and evaluated menstrual function, with three also showing a reduced chance of recovery of menstruation in older women ( Yu et al. , 2010 ; Dezellus et al. , 2017 ; Decanter et al. , 2021 ). Fifty-two papers specifically investigated gonadotoxicity of treatments either by comparison of groups receiving specific regimens or by cumulative toxicity scores (e.g. cyclophosphamide equivalent dose [CED]; summarized in Supplementary Table SIII ). In order to specifically evaluate treatment effect, papers that evaluated patient risk of gonadotoxicity based on other factors, such as diagnosis and disease stage, were not counted. Forty papers (77%) reported a treatment effect on AMH, with higher gonadotoxicity correlating with lower post-treatment AMH. In general, chemotherapy regimens containing alkylating agents, such as cyclophosphamide (e.g. ACVBP, BEACOPP, CHOP, FEC), resulted in lower levels of AMH and poorer recovery of AMH after treatment than non-alkylating treatment regimens ( Rosendahl et al. , 2008 ; Gracia et al. , 2012 ; Thomas-Teinturier et al. , 2015 ; Morarji et al. , 2017 ; Anderson et al. , 2018 ; Leiper et al. , 2020 ; Decanter et al. , 2021 ). Twenty-one publications reported gonadotoxicity scores or ranked treatments as higher/lower toxicity or higher/lower risk, based on treatment type and/or cumulative chemotherapy. In every study where higher versus lower toxicity was assessed, higher toxicity therapies and more treatment cycles resulted in lower post-treatment AMH compared with lower overall toxicity exposure ( Rosendahl et al. , 2010 ; Brougham et al. , 2012 ; Gracia et al. , 2012 ; Hamre et al. , 2012 ; Di Paola et al. , 2013 ; Nielsen et al. , 2013 ; Krawczuk-Rybak et al. , 2013a , 2013b , 2019 ; Charpentier et al. , 2014 ; Lunsford et al. , 2014 ; Thomas-Teinturier et al. , 2015 ; Elchuri et al. , 2016 ; van der Kooi et al. , 2017 , 2019 ; van den Berg et al. , 2018 ; Cameron et al. , 2019 ; George et al. , 2019 ; Su et al. , 2020 ; van Velsen et al. , 2020 ; Parissone et al. , 2021 ). The evidence regarding radiotherapy was less conclusive as its use was only cited in 10 publications ( van Beek et al. , 2007 ; Lie Fong et al. , 2009 ; Brougham et al. , 2012 ; Gracia et al. , 2012 ; Dillon et al. , 2013b ; Miyoshi et al. , 2013 ; Elchuri et al. , 2016 ; van den Berg et al. , 2018 ; George et al. , 2019 ; Nies et al. , 2020 ). However, in studies where radiotherapy did result in significantly lower AMH, this was typically when targeted to pelvic/abdominal regions or total body irradiation ( Lie Fong et al. , 2009 ; Gracia et al. , 2012 ; Miyoshi et al. , 2013 ; Elchuri et al. , 2016 ; van den Berg et al. , 2018 ; George et al. , 2019 ). Thirty-eight papers evaluated AMH levels in women following treatment for breast cancer, representing a total of 3600 patients ( Table II ). All reported a negative treatment effect on AMH, and in 12 papers with data available, reductions in AMH from pre-treatment to <3 months post-treatment ranged from ∼80% to 99% ( Lutchman Singh et al. , 2007 ; Yu et al. , 2010 ; Henry et al. , 2014 ; Ben-Aharon et al. , 2015 ; Anderson et al. , 2017 ; D'Avila et al. , 2017 ; Leonard et al. , 2017 ; Trapp et al. , 2017 ; Decanter et al. , 2018 ; Passildas et al. , 2019 ; Silva et al. , 2019 ; Loubersac et al. , 2020 ). Of those publications with available data, 13/16 found that post-treatment AMH levels were associated with age at baseline ( Lutchman Singh et al. , 2007 ; Henry et al. , 2014 ; Ben-Aharon et al. , 2015 ; D'Avila et al. , 2015; Anderson et al. , 2017 ; Dezellus et al. , 2017 ; Wenners et al. , 2017 ; Al-Janabi et al. , 2018 ; Lee et al. , 2018 ; Silva et al. , 2019 ; Celebi et al. , 2020 ; Li et al. , 2020 ; Loubersac et al. , 2020 ). Some degree of recovery in average AMH levels during follow-up was reported in 14/20 publications; however, the extent of this recovery was marginal (<15% average recovery from nadir levels as a percentage of baseline values) in 12 publications, and only partial in the remainder (range of follow-up 0–5 years), with none demonstrating a return of AMH to pre-treatment levels ( Yu et al. , 2010 ; Henry et al. , 2014 ; Ben-Aharon et al. , 2015 ; Anderson et al. , 2017 ; Dezellus et al. , 2017 ; Leonard et al. , 2017 ; Decanter et al. , 2018 ; Lambertini et al. , 2019 ; Palinska-Rudzka et al. , 2019 ; Silva et al. , 2019 ; Zhong et al. , 2019 ; Loubersac et al. , 2020 ; Oktay et al. , 2020 ; Goldfarb et al. , 2021 ). In all studies reporting no recovery of AMH levels, patients received treatment with alkylating agents, such as cyclophosphamide-based chemotherapy regimens; however, higher pre-treatment AMH and lower age were identified as factors associated with greater recovery after these regimens ( Yu et al. , 2010 ; Anderson and Cameron 2011 ; Anderson et al. , 2013 ; Elgindy et al. , 2013 ; D'Avila et al. , 2017 ; Wenners et al. , 2017 ; Palinska-Rudzka et al. , 2019 ; Zhong et al. , 2019 ). Characteristics of studies in breast cancer cohorts included in the systematic review (N = 38). *Between baseline and follow-up in longitudinal studies, and versus controls in cross-sectional studies. Presence of regular cycles, oligomenorrhoea or amenorrhoea. AMH: anti-Müllerian hormone; BL, baseline; LOD, limit of detection; N/A, not applicable; PTFU, post-treatment follow-up. In patients with breast cancer, post-treatment AMH was also linked to menstrual function at follow-up in 11/18 publications ( Su et al. , 2010 ; Ruddy et al. , 2014 ; D'Avila et al. , 2015 , 2017 ; Morarji et al. , 2017 ; Wenners et al. , 2017 ; Decanter et al. , 2018 ; Kim et al. , 2018 ; Palinska-Rudzka et al. , 2019 ; Silva et al. , 2019 ; Li et al. , 2020 ). Patients with undetectable AMH generally had a higher risk of amenorrhoea, although several studies reported patients with low or undetectable AMH who did recover menstrual function. The impact of BRCA status was investigated in two studies ( Lambertini et al. , 2019 ; Oktay et al. , 2020 ), with one identifying increased loss of ovarian reserve and greater reduction in AMH post-treatment recovery in BRCA + individuals ( Oktay et al. , 2020 ). Of 11 studies in lymphoma ( Table III ), all nine publications evaluating treatment effect on AMH in patients with lymphoma (including Hodgkin lymphoma), representing a total of 521 patients, found a significant impact ( van Beek et al. , 2007 ; Decanter et al. , 2010 , 2021 ; Nitzschke et al. , 2010 ; Demeestere et al. , 2016 , 2021 ; Gharwan et al. , 2016 ; Anderson et al. , 2018 ; Palinska-Rudzka et al. , 2019 ). Eight papers reported significant decline in AMH values from pre-treatment to ≤3 months’ post-treatment ( Decanter et al. , 2010 , 2021 ; Demeestere et al. , 2016 , 2021 ; Gharwan et al. , 2016 ; Anderson et al. , 2018 ; Palinska-Rudzka et al. , 2019 ). An additional study also reported significant decline in post-treatment AMH compared with controls ( Nitzschke et al. , 2010 ). All eight papers evaluating gonadotoxicity found a regimen- or dose-dependent effect on AMH following treatment and during recovery ( van Beek et al. , 2007 ; Decanter et al. , 2010 , 2021 ; Hamre et al. , 2012 ; Behringer et al. , 2013 ; Anderson et al. , 2018 ; Palinska-Rudzka et al. , 2019 ; Demeestere et al. , 2021 ). Five of these studies identified that patients receiving doxorubicin, bleomycin, vinblastine, dacarbazine (ABVD)-based regimens had at least partial post-treatment recovery of AMH at follow-up compared with those receiving cyclophosphamide-containing regimens ( Decanter et al. , 2010 , 2021 ; Anderson et al. , 2018 ; Palinska-Rudzka et al. , 2019 ; Demeestere et al. , 2021 ). One paper described limited recovery of AMH in older ABVD-treated patients ( Anderson et al. , 2018 ) whereas another found that the impact of age on recovery was only found in women treated with alkylating agents ( Decanter et al. , 2021 ). Characteristics of studies in lymphoma cohorts included in the systematic review (N = 11). *Between baseline and follow-up in longitudinal studies, and vs controls in cross-sectional studies. Presence of regular cycles, oligomenorrhoea or amenorrhoea. AMH: anti-Müllerian hormone; BL, baseline; LOD, limit of detection; N/A, not applicable; PTFU, post-treatment follow-up. The majority (20/29; 69%) of papers including paediatric cancer patients were cross-sectional studies reporting on adult childhood cancer survivors ( Bath et al. , 2003 ; van Beek et al. , 2007 ; Lie Fong et al. , 2009 ; Gracia et al. , 2012 ; Hamre et al. , 2012 ; El-Shalakany et al. , 2013 ; Miyoshi et al. , 2013 ; Nielsen et al. , 2013 ; Krawczuk-Rybak et al. , 2013b , 2019 ; Charpentier et al. , 2014 ; Lunsford et al. , 2014 ; Thomas-Teinturier et al. , 2015 ; Elchuri et al. , 2016 ; van der Kooi et al. , 2017 ; van den Berg et al. , 2018 ; Cameron et al. , 2019 ; George et al. , 2019 ; Nystrom et al. , 2019 ; Nies et al. , 2020 ), of which 19 had follow-up periods of >5 years. Reductions in AMH compared with pre-treatment levels or versus controls were reported by 14/20 (70%) publications ( Bath et al. , 2003 ; van Beek et al. , 2007 ; Brougham et al. , 2012 ; El-Shalakany et al. , 2013 ; Krawczuk-Rybak et al. , 2013a , 2013b , 2019 ; Elchuri et al. , 2016 ; Gupta et al. , 2016 ; Morse et al. , 2016 ; van der Kooi et al. , 2017 , 2019 ; Cameron et al. , 2019 ; George et al. , 2019 ). As with adult cancer patients, the degree of reduction varied with treatment received, with complete loss of AMH in some patients. Low AMH levels were found in some cancer survivors despite normal FSH levels (e.g. in 20% of survivors ( George et al. , 2019 )), indicating the added value of AMH in detecting partial ovarian damage. Additional studies reported trends to lower AMH, or low AMH levels in patients following childhood cancer that were not compared with a control group or baseline data ( Hamre et al. , 2012 ; Miyoshi et al. , 2013 ; Nielsen et al. , 2013 ; Charpentier et al. , 2014 ; Lunsford et al. , 2014 ; van den Berg et al. , 2018 ; George et al. , 2019 ). While these studies support that long-term AMH levels can be reduced in survivors of childhood cancer, important outcomes such as relationships with fertility and age at menopause have not been evaluated. Most studies (10/12 (90.9%)) ( van Beek et al. , 2007 ; Lie Fong et al. , 2009 ; Brougham et al. , 2012 ; Hamre et al. , 2012 ; Miyoshi et al. , 2013 ; Krawczuk-Rybak et al. , 2013a , 2019 ; Charpentier et al. , 2014 ; Thomas-Teinturier et al. , 2015 ; van der Kooi et al. , 2019 ) which had data for both pre- and post-menarchal patients did not detect significant differences in post-treatment AMH by this parameter, and one reported better AMH recovery in pre-menarchal patients ( van der Kooi et al. , 2019 ). Two publications also reported delayed puberty in patients with low post-treatment AMH ( El-Shalakany et al. , 2013 ; Lunsford et al. , 2014 ). To assess the impact of cancer treatment on AMH levels during long-term follow-up, studies with <1 year of follow-up were excluded owing to the potential for a residual effect of cancer treatment and incomplete recovery. Based on median reported follow-up post-treatment periods, there were 35 studies with 1- to 2-year follow-up ( Bath et al. , 2003 ; Rosendahl et al. , 2010 ; Yu et al. , 2010 ; Brougham et al. , 2012 ; Anderson et al. , 2013 , 2017 ; Elgindy et al. , 2013 ; Iwase et al. , 2013 ; Dillon et al. , 2013b ; Henry et al. , 2014 ; Ruddy et al. , 2014 , 2021 ; Ben-Aharon et al. , 2015 ; Gharwan et al. , 2016 ; Gupta et al. , 2016 ; Dezellus et al. , 2017 ; Leonard et al. , 2017 ; Morarji et al. , 2017 ; Trapp et al. , 2017 ; Wenners et al. , 2017 ; Decanter et al. , 2010 , 2018 ; Evranos et al. , 2018 ; Lee et al. , 2018 , 2020 ; Yaish et al. , 2018 ; Silva et al. , 2019 ; van der Kooi et al. , 2019 ; Zhong et al. , 2019 ; Celebi et al. , 2020 ; Loubersac et al. , 2020 ; Oktay et al. , 2020 ; Berjeb et al. , 2021 ; Goldfarb et al. , 2021 ; Martin, 2021 ). Most (33/35; 96%) found that AMH was reduced after treatment, 19/25 (76%) later then identified some degree of increase in AMH at follow-up. Although only a few studies found a (very modest) increase in AMH after 12 months, several studies with a maximum post-treatment follow-up period of 12 months showed the highest post-treatment AMH value at 12 months, raising the possibility that further recovery could have been observed ( Brougham et al. , 2012 ; Ben-Aharon et al. , 2015 ; Gupta et al. , 2016 ; Decanter et al. , 2018 ; Yaish et al. , 2018 ). It therefore appears that 12 months after completion of treatment is a minimum period to allow complete or near-complete recovery of ovarian function: little recovery is seen thereafter and becomes confounded with the normal decline in AMH with increasing age. Of 18 studies with a median >2–5.5 years of follow-up, 15/15 with available data (100%) identified AMH reductions in cancer survivors ( Nitzschke et al. , 2010 ; Partridge et al. , 2010 ; Su et al. , 2010 ; Anderson and Cameron 2011 ; El-Shalakany et al. , 2013 ; Acibucu et al. , 2016 ; Demeestere et al. , 2016 , 2021 ; Morse et al. , 2016 ; Anderson et al. , 2018 ; Cameron et al. , 2019 ; Lambertini et al. , 2019 ; Palinska-Rudzka et al. , 2019 ; Chemerinski et al. , 2020 ; van Velsen et al. , 2020 ), 8/10 (80%) identified at least partial recovery at follow-up in ≥1 patient group ( Demeestere et al. , 2016 , 2021 ; Morse et al. , 2016 ; Anderson et al. , 2018 ; Malisic et al. , 2018 ; Cameron et al. , 2019 ; Lambertini et al. , 2019 ; Palinska-Rudzka et al. , 2019 ) and 3/5 (60%) reported correlation of post-treatment AMH with menstrual function in some patients ( Rosendahl et al. , 2008 ; Su et al. , 2010 ; Palinska-Rudzka et al. , 2019 ). Of 24 cross-sectional studies with follow-up >5.5 years, 11/17 (65%) found AMH reductions with treatment when compared with controls ( van Beek et al. , 2007 ; Gracia et al. , 2012 ; Di Paola et al. , 2013 ; Krawczuk-Rybak et al. , 2013a , 2013b , 2019 ; Thomas-Teinturier et al. , 2015 ; Elchuri et al. , 2016 ; Leiper et al. , 2020 ; Mittica et al. , 2020 ; Roshandel et al. , 2021 ) and 1/4 (25%) linked AMH to menstrual function in some patients ( Lunsford et al. , 2014 ). There are very limited data regarding continuing changes in AMH after cancer treatment and whether cancer survivors show an accelerated decline in AMH in the later reproductive years. In a mixed longitudinal/cross-sectional analysis, AMH levels were shown to be reduced according to gonadotoxicity but maintained a plateau for a prolonged period after treatment ( Su et al. , 2020 ). Supporting that, cancer survivors showed no evidence of an increased rate of decline in AMH compared to age-matched controls ( Cameron et al. , 2019 ).

In this systematic review, we confirm that anticancer treatment substantially impacts AMH, with large reductions during treatment followed by a period of partial recovery in some women, peaking within 1–2 years ( Fig. 3 ). Low post-treatment AMH was predicted by lower pre-treatment AMH, as well as by older age and gonadotoxicity of treatment. Consequently, discussions around treatment and ovarian outcomes should consider the risk of POI and need for family planning and fertility preservation, informed by AMH. While low post-treatment AMH correlated with increased likelihood of POI, the evidence with respect to oligo- or amenorrhoea was less clear and there were very little data relating AMH to fertility or time to/age at POI or menopause. AMH is therefore a clinically useful biomarker of ovarian reserve before and after cancer treatment and it is likely to be of value in the diagnosis of POI after cancer treatment but relationships with post-treatment fertility and reproductive lifespan require further investigation.

Anticancer treatment has a clear negative impact on women’s ovarian reserve, with reduced AMH levels observed across studies and diagnoses, and reductions of ≥90% commonly identified. Some degree of post-treatment recovery in AMH levels was described in many patient groups, although full- or near-complete restoration to pre-treatment levels was rare and typically occurred in patients receiving milder treatment. Figure 3 gives a representation of the key findings of this review. It is notable that there are relatively few prospective studies, and among those with longitudinal data, several had a maximum 1 year of follow-up. While AMH levels measured 1–2 years after treatment are likely to be reliable at indicating longer-term menstrual outcomes, further research is required to confirm this, especially in older women where recovery is slower and more limited ( van Beek et al. , 2007 ; Decanter et al. , 2010 ; Behringer et al. , 2013 ; Anderson et al. , 2018 ; Palinska-Rudzka et al. , 2019 ). Summary of findings of systematic review represented graphically. The three lines represent women with high (green), average (blue) and low (red) AMH levels before treatment, with treatment represented by the shaded area and a threshold indicating POI is also represented. (1) AMH concentrations are reduced by cancer treatments. (2) Recovery is variable, depending on patient age, treatment regimen and pre-treatment AMH levels. Recovery can be near complete or absent, with the latter resulting in permanent POI. The relationship of post-treatment AMH to POI needs to be explored further. (3) Prediction of permanent POI at the time of end of treatment may be possible in some situations, but further studies are required to determine the patient groups for which this may and may not be possible. (4) There are insufficient data to be able to predict the duration of post-recovery ovarian function using AMH levels before or after treatment, which will interact with the physiological decline in AMH with increasing age. Reproduced from Jayasinghe et al. (2018) with permission. Treatments known to be more gonadotoxic, whether this be drug or cumulative exposure related, generally caused a greater reduction in AMH. The clearest evidence comes from lymphoma, where patients receiving ABVD-based regimens showed more complete AMH recovery and resumption of menstruation compared to more severe treatment strategies ( van Beek et al. , 2007 ; Decanter et al. , 2010 , 2021 ; Behringer et al. , 2013 ; Anderson et al. , 2018 ; Palinska-Rudzka et al. , 2019 ). Studies in breast cancer also identified taxanes as contributing to lower AMH ( Hamy et al. , 2016 ; Lambertini et al. , 2019 ; Silva et al. , 2019 ). However, studies evaluating cumulative dose of chemotherapy, regardless of regimen, show this to be a key factor in ovarian outcomes ( Charpentier et al. , 2014 ; Elchuri et al. , 2016 ; van den Berg et al. , 2018 ; Cameron et al. , 2019 ; George et al. , 2019 ). Pre-treatment AMH values were associated with post-treatment AMH in all 11 studies that evaluated this, supporting the predictive value of AMH at baseline. All of these studies were in post-menarchal patients. Several studies also identified low pre-treatment AMH in some patients ( Lutchman Singh et al. , 2007 ; Cameron et al. , 2018 ; Palinska-Rudzka et al. , 2019 ; van der Kooi et al. , 2019 ), which was not limited to a particular diagnosis. An important question is whether the lower AMH levels after cancer treatment then decline at a different rate compared to other women. Three studies found no significant difference in rates between patients and age-matched controls ( van der Kooi et al. , 2017 ; Cameron et al. , 2019 ; Elitzur et al. , 2021 ). However, there is a higher rate of decline in older women, both cancer survivors and controls ( Dezellus et al. , 2017 ; Cameron et al. , 2019 ). These studies highlight the need for research investigating the relationships between post-treatment AMH, age and age of POI/menopause, which would be of substantial clinical value. While the picture regarding treatment impact on AMH is clear, there was no clear relationship between AMH and menstrual status at follow-up, with menstruation reported in patients with undetectable AMH across different diagnoses and treatments ( Decanter et al. , 2010 ; Rosendahl et al. , 2010 ; Ben-Aharon et al. , 2015 ). The reliability of AMH in predicting menstrual function may be better in older patients, consistent with AMH becoming undetectable in advance of natural menopause in healthy women. In contrast, in younger women, menstrual cycles may continue despite a very limited ovarian reserve and undetectable AMH. Although several studies have reported a clear relationship between low/undetectable AMH and diagnosis of POI, the diagnostic validity of AMH for post-cancer treatment POI has not been sufficiently evaluated, either alone or in combination with established criteria. Interpretation of AMH in paediatric cohorts is potentially complicated by the natural arc of AMH, which increases during childhood and through the early adult years before a decline to the menopause ( Kelsey et al. , 2011 ). With the exception of one publication ( van der Kooi et al. , 2019 ), there were no differences in the effect of treatment on post-treatment AMH levels between pre- and post-menarchal patients, suggesting the reliability of AMH as a biomarker in younger populations. It should also be noted that all six studies that did not find a treatment influence on AMH were in adult survivors of childhood cancer ( Lie Fong et al. , 2009 ; Nielsen et al. , 2013 ; van den Berg et al. , 2018 ; Nystrom et al. , 2019 ; Nies et al. , 2020 ; Elitzur et al. , 2021 ), whereas other studies did find an effect. Further longitudinal studies, from diagnosis through to adulthood, are required to further explore the value of AMH for predicting late ovarian function shortly after completion of therapy in paediatric and adolescent patients with cancer. Across the different publications, spanning nearly two decades, multiple AMH assay types, most commonly ELISAs or electrochemiluminescence immunoassays (ECLIAs), from a variety of manufacturers have been used. There is currently no internationally agreed calibration standardization for AMH, and the improvements in sensitivity afforded by automated assays in comparison with historical methods, as well as inherent variation and non-linear conversion between assays, can complicate interpretation of the data ( Li et al. , 2021 ). Whether variations in assay calibration and performance influence the clinical value of AMH assessment was not formally assessed, but the overall consistency of the main findings indicates that this is not a major issue. It is possible though that the use of assays with greater sensitivity ( Chai et al. , 2014 ; Decanter et al. , 2014 ) will impact on analysis of AMH as a diagnostic test for POI. A variety of additional independent factors can influence levels of AMH ( Dafopoulos et al. , 2010 ; Dólleman et al. , 2013 ). Although publications featuring potentially AMH-modifying conditions, such as systemic illness, endometriosis, polycystic ovary syndrome and granulosa cell tumours, were excluded, studies with patients who use cigarettes or oral contraceptives ( Elitzur et al. , 2021 ), as well as BRCA mutation carriers ( Lambertini et al. , 2019 ; Oktay et al. , 2020 ), were included. The studies described here represent highly heterogeneous patient populations, incorporating a wide range of ages, diagnoses and treatment types. While this may limit the ability of this review to make specific recommendations, it reinforces the wider applicability of our observations. This review confirms AMH as a marker of ovarian reserve but does not specifically address assessment of time to menopause, which may be clinically useful. This has previously been addressed in relation to natural menopause via probability ( Finkelstein et al. , 2020 ) and estimated rate of decline-based models ( de Kat et al. , 2019 ; Ramezani Tehrani et al. , 2020 ). As AMH has been shown to correlate with patient age and treatment gonadotoxicity and declines at a similar rate to healthy individuals following any post-treatment recovery, it should be possible to develop a model for menopause risk over time based on pre- and post-treatment AMH measurements. Additional longitudinal studies would support such a model and help to further establish the time course of any AMH recovery, and in which patient groups it occurs, especially whether a time point <2 years’ post-treatment can reliably provide a long-term predictive marker.

Supplementary data are available at Human Reproduction Update online.

There are no new data associated with this article.

Click here for additional data file.