{"paper_id":"0a343ef5-bdce-4d48-9ae0-566defe50090","body_text":"The Anti-Müllerian hormone (AMH), a glycoprotein produced by granulosa cells in preantral and small antral follicles, plays a crucial role in follicular growth and development by inhibiting primordial follicle recruitment and regulating granulosa cell differentiation. ( 1 )\nAMH has been widely used in assisted reproductive technologies (ART) for estimating ovarian reserve, predicting ovarian response to stimulation, identifying patients at risk for ovarian hyperstimulation syndrome (OHSS), and guiding gonadotropin dosing in women undergoing in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI). ( 2 )  A simplified diagnostic algorithm by Teede et al. ( 3 )  suggests the inclusion of AMH levels as an alternative to ultrasound in adults. However, access to AMH assays remains limited in many countries, including within Brazil's public healthcare system, potentially limiting the widespread clinical application of AMH-based diagnostic algorithms. ( 4 )\nAMH is considered one of the most reliable predictors of the actual number of primordial follicles. However, the number of available oocytes, especially in those with high AMH concentrations, does not always translate into a high number of fertilized oocytes and well-developed embryos. ( 5 )  Previous studies revealed that high AMH may inhibit FSH-stimulated aromatase expression, causing androgen excess and impaired preantral follicle growth, which may hinder the development of selectable follicles. AMH levels are positively correlated with LH, testosterone, and DHEA and are linked to polycystic ovary syndrome (PCOS) severity and androgen excess. ( 6 - 8 )  In PCOS, elevated AMH in follicular fluid and granulosa cells suggests intrinsic follicular changes that may inhibit fertilization. ( 9 )  However, the impact of excessive AMH on fertilization and embryo development remains unclear due to limited studies.\nPCOS is a common endocrine disorder in women of reproductive age, with a prevalence ranging from 5-18%. ( 10 )  Women with PCOS often experience amenorrhea or oligomenorrhea, excessive androgen production inducing clinical signs of hyperandrogenism such as hirsutism and abnormal insulin sensitivity. ( 11 )\nSerum AMH levels are significantly higher in women with PCOS compared to those with normal ovulatory function. ( 12 )  However, the predictive value of AMH for ART outcomes remains controversial as elevated AMH levels in these women are believed to correlate with disease severity, which may impact AMH's ability to assess ovarian reserve accurately and, consequently, confounding an association between AMH and ART outcomes. ( 13 , 14 )  Therefore, although AMH level is a useful indicator of fertility potential, it should be interpreted alongside other factors influencing reproductive outcomes, particularly maternal age, which remains the strongest predictor of ART success, as AMH levels naturally decline with advancing age. ( 15 )\nTo date, only a limited number of studies have explored the influence of elevated serum AMH levels on IVF/ICSI outcomes, and their findings have been inconsistent. While some studies have identified a negative effect, ( 16 - 18 )  others have observed a positive impact on pregnancy or live birth outcomes. ( 19 , 20 )\nThis review aims to synthesize the literature on the relationship between elevated AMH levels and ART outcomes in women with PCOS. By comparing women with high AMH levels to those with low/normal AMH levels, this study seeks to provide a detailed understanding of how varying AMH levels influence key outcomes of medically assisted reproduction, focusing on live birth rate and other relevant IVF/ICSI success indicators.\n\nThis systematic review and meta-analysis was conducted in accordance with  the Cochrane Handbook for Systematic Reviews of Intervention ( 21 )  and follows the recommendation outlined in the  Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)  guidelines. ( 22 )  The protocol was registered in PROSPERO (CRD42024596056).\nStudies were eligible for inclusion if they met the following criteria: observational studies, including cohort studies or case-control designs. The study population consisted of women of reproductive age with PCOS undergoing IVF or ICSI. Eligible studies had to measure serum AMH levels before ovarian stimulation and report IVF or ICSI outcomes, such as live birth rate, clinical pregnancy rate, miscarriage rate, number of oocytes retrieved, number of mature (MII) oocytes, or fertilization rate. Additionally, studies were required to stratify their populations based on AMH levels, categorizing participants into low, normal, and high or equivalent thresholds. Exclusion criteria included non-human studies, reviews, commentaries, or editorials, and studies that did not provide primary data or involve quantitative analysis were excluded from this review. Studies were also excluded if they did not assess IVF or ICSI outcomes, did not define high AMH levels using specific cut-off values, or focused on oocyte donation programs. No restrictions were applied regarding the language of publication. The search was limited to studies published between 2014-2024, considering the existence of previous reviews covering earlier periods.\nA comprehensive literature search was conducted across three electronic databases: MEDLINE (via PubMed), Web of Science, and Scopus on September 17, 2024. Our search strategy is available in  Supplementary Material – Table1S . In addition to database searches, we examined the reference list of included studies to identify further potential studies.\nTwo reviewers carried out study selection independently, initially through title/abstract screening (A.M. and A.M.P.) and subsequently through full-text reading (A.M. and A.M.P.). Rayyan QCRI, a systematic review management software (http://rayyan.qcri.org), was used to expedite the screening of abstracts, further eliminate duplicates and facilitate the selection and organization of full-text articles for inclusion.\nData extraction was conducted independently by two reviewers (A.M. and A.M.P.) using a purpose-built online form. In addition, data were collected on the reported outcomes. The studies were subsequently included in the meta-analysis. Disagreements at any stage of the review process were resolved by consensus.\nThe risk of bias for each included study was independently assessed by two researchers (A.M. and A.M.P.) using the  National Heart, Lung, and Blood Institute Study Quality Assessment Tool  for observational cohort and cross-sectional studies. ( 32 )  These tools evaluated key methodological domains, including research question clarity, population definition, sample size justification, exposure and outcome assessment, blinding, follow-up rates, and adjustment for confounders. Each study received an overall quality rating of \"good,\" \"fair,\" or \"poor,\" determined by adherence to these criteria and the resulting risk of bias. Discrepancies between reviewers were resolved by consensus or, when necessary, by consulting a third assessor to ensure accuracy and consistency. The Robvis tool was employed to visually represent the risk of bias assessments and generate summary plots for transparency and clarity in reporting. ( 33 )\nA narrative synthesis was employed to summarise the data extracted from the included studies, with a focus on highlighting key similarities and differences.\nA meta-analysis was conducted separately for each variable of interest. For continuous outcomes, mean differences (MD) and median differences (MedD) with 95% confidence intervals (95% CI) were used. For categorical outcomes, odds ratios (OR) and risk differences (RD) with 95% CI were calculated based on raw data extracted from the primary studies. A continuity correction of 0.5 was applied to OR estimation whenever a cell had a zero count to ensure statistical validity.\nBoth common-effect and random-effects models were considered based on the level of heterogeneity (I²) and the methodological and study characteristics. Between-study variability was quantified using the I² statistic and the τ² parameter, the latter estimated through the Restricted Maximum Likelihood (REML) method. Heterogeneity was assessed using the Q-Cochran test (p-value) and the I² statistic, with I² > 40% and p-values < 0.10 considered indicative of severe and significant heterogeneity, respectively.\nSources of heterogeneity were assessed using a leave-one-out sensitivity analysis, applied exclusively to study groups with severe/significant heterogeneity (I² > 40%), in which each study was sequentially removed to evaluate its influence on the overall pooled estimate. The overall results of the meta-analysis were visualized using forest plots.\nAll statistical analyses were conducted in R (version 2023.06.1+524, R Foundation for Statistical Computing, Vienna, Austria) using the  meta  and  metafor  packages.\n\nAfter the database search and duplicate removal, a total of 2,998 studies were screened. Following two screening phases a total of 13 published articles were included. ( 13 , 16 , 17 , 19 , 23 - 31 ) . The study selection process is illustrated in  figure 1 . Reasons for exclusion are presented in  Supplementary Material - Table 2S .\nA total of 13 primary studies, comprising 12 retrospective cohorts ( 13 , 16 , 17 , 19 , 23 - 30 )  and one cross-sectional study, ( 31 )  were included in the systematic review. ( Chart 1 )The studies were conducted across diverse geographic regions, such as North America, ( 13 )  Europe, ( 28 )  Asia ( 16 , 19 , 26 , 27 , 29 - 31 )  and Africa, ( 23 , 24 )  with sample sizes ranging from 55 ( 31 )  to 2707 ( 25 )  participants. The mean age of the participants was 30.49 years. All participants were women with PCOS diagnosed based on the Rotterdam criteria, undergoing IVF or ICSI. Serum AMH levels were assessed using various methods, including enzyme-linked immunosorbent assays (ELISA) and chemiluminescent immunoassays, with considerable variation in AMH cut-off values. For instance, some studies categorized AMH levels based on percentiles, ( 16 , 19 , 28 )  others in \"low\", \"normal\" and \"high\" categories, ( 17 , 27 )  and others applied one absolute threshold. ( 13 , 23 - 26 , 29 - 31 )  The high AMH levels thresholds or equivalent ranged from 3.0 ng/mL to 14.30 ng/mL ( Chart 2 ).\nELISA: Enzyme-linked immunosorbent assay; CLIA: Chemiluminescent Immunoassay; NR: Not Reported\nNine studies were initially included in the meta-analysis for live birth rate. ( 16 , 17 , 19 , 24 , 25 , 27 - 30 )  The pooled OR for live birth was 0.85 (95%CI: 0.71–1.02; p=0.0809), suggesting a 15% reduction in the probability of live birth in women with elevated AMH levels compared to those with normal or low levels, though this difference was not statistically significant ( Figure 2 ). However, high heterogeneity was observed (I²=63.6%; p=0.0049). A leave-one-out sensitivity analysis identified Acharya et al. ( 25 )  as the main contributor. Upon its exclusion, the heterogeneity decreased to a moderate level (I²=43.5%; p=0.089), and the pooled OR shifted to 0.80 (95%CI: 0.64–0.99; p=0.044), indicating a statistically significant 20% reduction in the live birth rate among women with high AMH levels compared to those with normal or low levels.\nForest plot of pooled OR, 95% CI, Random-Effects Model. (a′) Leave-one-out sensitivity analysis\nClinical pregnancy rates were analyzed in 11 studies. ( 16 , 17 , 19 , 23 - 30 )  The pooled OR was 0.98 (95%CI:0.82–1.17; p=0.7896), indicating a reduction of 2% in the clinical pregnancy rate in women with elevated AMH levels compared to those with low or normal levels ( Figure 3 ). Heterogeneity was high (I²=73.6%; p < 0.0001), but leave-one-out sensitivity analysis showed that no single study disproportionately impacted the overall estimate.\nJaralla et al. ( 31 )  was excluded due to their cross-sectional design. However, among women with high AMH, 40% achieved clinical pregnancy versus 25% with normal AMH. Despite limitations, these findings suggest a positive association between high AMH and clinical pregnancy, contrary to the meta-analysis.\nFive studies reported miscarriage rates. ( 19 , 26 , 27 , 29 , 30 )  The pooled OR was 1.25 (95%CI:0.88–1.76; p=0.2131), suggesting a non-significant 25% increase in miscarriage among women with high AMH levels. Heterogeneity was low (I²=23.7%; p=0.2634). One study ( 26 )  reported a markedly higher miscarriage risk (OR: 3.30; 95%CI: 1.27–8.58; p=0.2482).\nEight studies analyzed the number of oocytes retrieved. ( 13 , 16 , 17 , 23 , 27 - 30 )  Five studies reported the outcome as a mean with standard deviation, ( 13 , 16 , 17 , 23 , 28 )  and three studies presented it as a median with interquartile range. ( 27 , 29 , 30 )  To standardize the analysis, the medians and interquartile ranges were converted into means and standard deviations using the method proposed by Hozo et al. ( 34 )  The random-effects meta-analysis, incorporating all studies, yielded a pooled MD of 3.52 (95%CI: 1.70–5.33; p < 0.0001), indicating that women with high AMH levels retrieved, on average, 3.52 more oocytes than those with normal or low AMH levels. However, substantial heterogeneity was observed (I²=96.7%; p < 0.0001), warranting further sensitivity analyses. A leave-one-out sensitivity analysis demonstrated the robustness of the findings, as the overall MD remained within the same range (2.72–4.01) regardless of which study was excluded. The heterogeneity remained high across all iterations, suggesting a cumulative effect of multiple studies.\nGiven the potential impact of median-to-mean conversions on the results, a subgroup analysis was conducted excluding the three studies that initially reported medians. This yielded a pooled MD of 1.51 (95%CI: 0.59–2.42), with a reduction in heterogeneity (I²=52.3%; p=0.0782). A separate meta-analysis, including only the three studies that initially reported medians, was also performed, showing a pooled MedD of 5.34 (95%CI: 1.21–9.46), with substantial heterogeneity (I²= 96.5%; p < 0.0001). This suggests that the inclusion of converted medians may have led to overestimating the effect size.\nOverall, the findings indicate that women with high AMH levels retrieve more oocytes than those with lower levels.\nFertilization rate was reported in six studies, ( 13 , 16 , 17 , 26 - 28 )  but only three ( 13 , 16 , 26 )  were included in the meta-analysis due to the absence of necessary data in the remaining studies. The pooled RD was -0.00 [95%CI: -0.02; p=0.01], with low heterogeneity (I²=8.5%; p=0.3353), indicating no significant difference in fertilization rate between the two groups. Additionally, three studies ( 17 , 27 , 28 )  reported fertilization rates between 63.34% to 80% in the high AMH group and 58% to 89% in the normal/low AMH group. These findings are consistent with the meta-analytic results, but the lack of dispersion measures prevented their inclusion in the quantitative analysis.\nFour studies analyzed the number of MII oocytes. ( 13 , 16 , 24 , 27 )  Jabarpour et al. ( 27 )  reported this outcome as a median in the high AMH group - 13(10, 17) - and in the low/normal group - 7(5,9). Guo et al. ( 16 )  reported a mean number of 14.83±7.58 in the high AMH group and 13.41±5.81 in the normal/ low AMH group. Mohamed et al. ( 24 )  reported a significant positive correlation between AMH concentration and the number of MII oocytes (Spearman r=0.276; p=0.006). Finally, Tal et al. ( 13 )  presented the absolute number of 336 MII oocytes across the 21 cycles performed in the high AMH group and 499 MII oocytes across the 37 cycles of the normal AMH group.\nThe selected articles’ bias risk is displayed in  figures 4  and  5 . Among the included studies, ten were classified as having a low risk of bias, ( 13 , 16 , 17 , 19 , 23 , 24 , 26 , 27 , 29 , 30 )  indicating strong methodological quality. Two studies ( 25 , 28 )  had an unclear risk of bias due to methodological limitations, while one ( 31 )  was rated as having a high risk of bias, suggesting significant concerns regarding its validity. The items most frequently associated with a high risk of bias included consistent subject selection, valid exposure measures, and repeated exposure assessment. These domains exhibited a notable proportion of studies classified as unclear or high risk, indicating methodological weaknesses.\n\nThe findings of this systematic review and meta-analysis contribute to the ongoing debate on the impact of AMH levels on IVF and ICSI outcomes in women with PCOS. Our results confirm that elevated serum AMH level is inversely associated with live birth and clinical pregnancy rates while also being linked to higher miscarriage rates in this population, supporting previous research on the detrimental effects of excessive AMH on reproductive success. ( 5 , 35 )  Additionally, we found that high AMH levels correlate with an increased number of oocytes retrieved and similar fertilization rates; similar results were reported by Yuwen et al. ( 35 )  These findings emphasize the potential role of AMH as a prognostic biomarker in ART for PCOS patients and highlight the need for personalized treatment strategies.\nOur meta-analysis demonstrated significantly higher LBR in women with low/normal AMH than those with high AMH. A similar trend was observed for CPR, though not statistically significant. These results diverge from studies in the general population, where higher AMH levels are often correlated with better ART outcomes. ( 20 , 36 , 37 )  Previous meta-analyses indicated AMH´s moderate predictive value for clinical pregnancy in a general IVF population, whereas in PCOS patients, the predictive utility appears to be different. ( 38 )  The weaker predictive ability of AMH for pregnancy outcomes in women with PCOS may result from its close association with disease severity, as elevated levels correlate with the three PCOS hallmarks—polycystic ovarian morphology, oligo/anovulation, and hyperandrogenism— thereby confounding its role as a predictor of ART success. ( 39 , 40 )\nVale-Fernandes et al. ( 28 )  and other studies observed an inverse relationship between FSH and AMH levels and a positive trend between LH and AMH in PCOS women. ( 16 , 18 , 28 )  These findings support the role of elevated AMH in disrupting folliculogenesis through two key mechanisms. First, it inhibits FSH-dependent dominant follicle selection, reducing granulosa cell sensitivity to FSH and potentially causing follicular arrest. Second, AMH is positively correlated with hyperandrogenemia, suppressing FSH-stimulated aromatase mRNA expression in granulosa-luteal cells, thereby reducing aromatase activity and inducing intraovarian androgen accumulation. ( 17 , 28 , 41 )  Androgens, in turn, promote early follicular growth independently of FSH, further increasing AMH production. In fact, AMH levels in PCOS women are 2–4 times higher than in those without the condition, primarily due to granulosa cells within individual follicles overproducing AMH by up to 75 times the normal amount rather than a higher follicle count. ( 42 - 44 )  Tal et al. ( 17 )  also found that total testosterone levels were significantly higher in the high serum AMH group compared to the average/low AMH groups, highlighting the complex relationship between AMH levels and ART outcomes across different PCOS phenotypes, with hyperandrogenic phenotypes in PCOS women shown to be associated with poorer CPR and LBR. ( 17 )\nOur results also found a non-significant 25% increase in miscarriage rates among women with high AMH levels. Beyond oocyte-related factors, the detrimental effects of high AMH and androgens on endometrial homeostasis and receptivity may further explain implantation failure and recurrent miscarriage in women with PCOS. ( 16 , 28 )\nIn our study, women with high AMH levels showed a trend toward an increased number of oocytes retrieved, consistent with previous research findings. However, this increase did not translate into improved clinical pregnancy rate although there were no differences in fertilization rates. One possible explanation is that while elevated AMH levels indicate a greater ovarian reserve, they are often associated with a decline in oocyte quality. ( 38 )  As a result, the higher number of oocytes retrieved leads to a larger pool of embryos for selection but does not necessarily enhance clinical pregnancy rate. Several studies have suggested that higher AMH levels are associated with lower oocyte maturity rates. ( 11 , 45 )  Dai et al. ( 46 )  reported that older women with lower AMH levels had a higher proportion of MII oocytes, possibly due to differences in follicular development. Compared with a large cohort of developing follicles, a smaller number of oocytes may receive adequate nutritional support from the ovary, promoting their maturation. ( 46 )  Despite this, our findings do not show a decrease in MII oocytes among women with high AMH levels. On the contrary, our results reported a positive association between higher AMH levels and the number of mature oocytes. ( 16 , 27 )  This discrepancy may point to differing sample characteristics or methodological factors that could influence outcomes related to oocyte maturity.\nThis study is the most recent and comprehensive systematic review and meta-analysis assessing the impact of high AMH levels on ART outcomes in women with PCOS. The inclusion of studies from diverse geographic regions enhances the generalizability of findings. Additionally, rigorous methodological approaches in compliance with recent guidelines, ( 22 )  including risk-of-bias assessment, sensitivity analyses, and stratified meta-analyses, strengthen the reliability of the results. Although the process of systematic literature review and meta-analysis provide stronger effect estimates with reduced random error, it does come with limitations. Overall, while most included studies demonstrated low risk of bias and strong methodological quality, the presence of studies with unclear or high risk highlights potential limitations in the consistency of subject selection and exposure assessment, which should be considered when interpreting the results. First, the included studies reported different effect measures, and our pooled analysis for each outcome did not incorporate data from all eligible studies, which may have introduced bias. Second, some studies did not report adjusted effect estimates, which are less prone to confounding than crude estimates - the measures used in our meta-analysis. Third, the variation in study populations, including differences in baseline characteristics such as age, BMI, and clinical conditions, and differences in stimulation protocols and AMH assays, including enzyme-linked immunosorbent assays and chemiluminescent immunoassays, with considerable variation in AMH cut-off values, may have contributed to the observed heterogeneity; however, a universal threshold for AMH is not clinically appropriate, which further underscores the variability across studies. As in the study by Acharya et al., ( 25 )  the observed heterogeneity may also be explained by several methodological aspects, including its retrospective cohort design based on the large multicenter SART-CORS database, in contrast to the predominantly unicentric cohorts of the other studies. Additional contributing factors include its unclear risk of bias, the use of exposure measures not consistently or clearly defined across participants, statistical differences in analysis, and its comparatively larger sample size, all of which may have amplified variability relative to the remaining studies. Lastly, the limited number of studies and small sample sizes may have reduced the statistical power, further constraining the interpretation of the findings.\n\nOur study confirms that high AMH levels are associated with lower live birth and clinical pregnancy rates in women with PCOS, alongside a higher risk of miscarriage. While elevated AMH correlates with a greater number of retrieved oocytes, it does not improve pregnancy rate, raising concerns about oocyte quality. These findings underscore the need for individualized ART strategies, including optimized stimulation protocols (milder), ‘freeze-all’ and tailored patient counseling, to improve outcomes in PCOS patients with high AMH. Further research is required to refine AMH-based prognostic models and understand its impact on both oocyte and endometrial quality, with particular attention to tailoring stimulation protocols according to PCOS phenotypes. Dedicated studies specifically designed to address these aspects will be crucial to advancing personalized approaches and improving reproductive outcomes in women with PCOS.","source_license":"CC-BY-4.0","license_restricted":false}