With the continuous development of assisted reproductive technologies and the growing proportion of women of advanced maternal age, adverse pregnancy outcomes following in vitro fertilization (IVF) and embryo transfer have become a growing concern. Studies have shown that an increase in maternal age is associated with a decline in fertility and a higher risk of miscarriage[1–3]. Miscarriage can harm women's physical health and impose psychological and financial burdens, especially for those undergoing assisted reproduction. Consequently, timely assessment of miscarriage risk and appropriate interventions are vital for these patients.

Anti-Müllerian hormone (AMH), produced by ovarian granulosa cells, plays a pivotal role in folliculogenesis[4–6]. Ovarian reserve, which encompasses the quantity and quality of follicles in the ovarian pool, diminishes with advancing age[7–9]. Serum AMH levels serve as an accurate marker of growing follicle count and are considered a reliable indicator of ovarian reserve and response[10–12]. Consequently, AMH measurement has become a routine assessment in the field of assisted reproduction. Nevertheless, the extent to which AMH correlates with oocyte and embryo quality, as well as its potential influence on pregnancy outcomes, particularly miscarriage, remains to be fully elucidated.

Current studies on the correlation between serum AMH levels and miscarriage risk have conflicting conclusions. Some studies indicate that pre-pregnancy low AMH levels are linked to higher miscarriage risk. A systematic review and meta - analysis of 13 retrospective studies (N = 12,042) shows that, compared to women with normal or high AMH levels, those with low AMH have a significantly increased miscarriage risk after assisted reproduction (P = 0.004, OR = 1.35, 95% CI = 1.10–1.66)^[13]. A 2017 cohort study with 1,060 patients found that after adjusting for age and oocyte number, low baseline serum AMH raises the risk of early miscarriage after fresh embryo transfer (P = 0.021)[14], a prospective study on natural conception found that compared to women with normal AMH (≥ 1.0 ng/mL), those with severe ovarian reserve decline (AMH ≤ 0.4 ng/mL) have a higher miscarriage risk (HR = 2.3, CI = 1.3–4.3)[15]. However, other studies suggest high AMH levels may also be an independent miscarriage risk factor, possibly due to endocrine changes, similar to those in polycystic ovary syndrome (PCOS). A domestic retrospective study found that in young women, high pre-pregnancy AMH levels (> 3.99 ng/mL) increase early miscarriage risk after fresh embryo transfer, and this was consistent in the non - PCOS subgroup[16]. Additionally, a prospective study on patients with natural conception showed that both low (AMH < 1.1 ng/mL) and high (AMH > 4.5 ng/mL) serum AMH levels during pregnancy are independent risk factors for early natural miscarriage (both P < 0.05)[17]. In conclusion, despite extensive research on AMH and miscarriage risk, the clinical predictive value of AMH remains controversial, due to inconsistent findings.

Moreover, existing studies have some limitations. On the one hand, the predictive ability of AMH may be affected by age, the type of conception (natural conception and assisted reproductive technology), the cause of miscarriage (embryo-origin and maternal-origin) and so on, while some studies fail to conduct stratified analyses of the study population. On the other hand, some studies inadequately control for confounding factors. For example, certain studies may include PCOS patients so that the association between their high AMH levels and miscarriage might be mediated by obesity and metabolic disorders rather than AMH itself.

Given that, the retrospective cohort study included the patients from our hospital who underwent IVF and achieved clinical pregnancy. By analyzing pre - pregnancy serum AMH levels, it explores the link between the levels of AMH and miscarriage risk following embryo transfer across different age groups and different types of embryo transfer. The study aims to illuminate the impact of diminished ovarian reserve on assisted reproductive outcomes. It seeks to supply robust evidence for pre - treatment evaluation and pregnancy management in assisted reproductive patients and to advance the clinical application of serum AMH levels.

The retrospective cohort study was conducted at Peking University Third Hospital Reproductive Center from January 1, 2017, to December 31, 2020. The study consisted of patients who underwent IVF/ICSI and achieved singleton clinical pregnancies through fresh or frozen-thawed embryo transfer. For each patient, only their first clinical pregnancy within the study period was included in the analysis to evaluate miscarriage outcomes. The process for selecting participants is shown in Fig. 1.

The study included participants who met the following criteria:

Patients presenting any of the following were excluded:

Based on clinical practice and the data characteristics of this study, this study used the 2019 ACOG criteria[19], defining AMH < 1.00 ng/mL as low. Participants were categorized into two groups: low AMH (AMH < 1.00 ng/mL) and normal AMH (AMH ≥ 1.00 ng/mL), according to their serum AMH levels measured before ovarian stimulation.

All participants underwent controlled ovarian hyperstimulation (COH) following the center's standard protocol. Clinicians selected COH protocols and drug doses by considering patients' age, AFC, AMH levels, basal sex hormone levels, prior IVF retrieved oocytes yield. The primary COH protocols included the GnRH - a long protocol, the ultralong protocol, the short protocol, or the GnRH antagonist protocol. Less than 5% of participants used non- traditional COH protocols, such as mild stimulation protocol, natural cycle protocol, or luteal phase stimulation protocol.

The starting dose of Gn was determined by the patient's age, AFC, basal FSH, and body surface area, usually 112.5–300 IU/day. After 4–5 days, ultrasound and hormone tests monitored follicle growth, and Gn dosage was adjusted When at least two follicles reached ≥ 18 mm, hCG was administered, and oocyte retrieval followed 34–38 hours later. Fertilization was via IVF or ICSI based on sperm quality or the need for pre - implantation genetic testing.

Baseline clinical data, basal hormone levels (including serum AMH), COH protocols, drug doses, details of embryo transfer, and pregnancy outcomes were collected for all participants. All clinical pregnancy cohort data were obtained from the Reproductive Center's clinical database.

The study's outcome measures were as follows:

The primary outcome was the miscarriage rate, calculated as:

(1) Miscarriage rate = Number of miscarriages/Total clinical pregnancies×100%

Secondary outcomes were the early and late miscarriage rates, calculated as:

Our analysis included 10,260 embryo transfer cycles from Peking University Third Hospital, between January 2017 and December 2020, after applying the inclusion and exclusion criteria (Fig. 1). Of the included cycles, 4,982 were fresh embryo transfers (48.56%) and 5,278 were frozen - thawed embryo transfers (51.44%).

Baseline characteristics, relevant IVF data and pregnancy outcomes of patients following fresh or frozen-thawed embryo transfer

The baseline characteristics and pregnancy outcomes of patients undergoing fresh or frozen - thawed embryo transfer are summarized in Table 1 and Table 2. Among 10,260 eligible cycles, 4,982 (48.6%) were fresh embryo transfers (FET) and 5,278 (51.4%) frozen-thawed embryo transfers (FET).

Significant baseline differences existed between AMH groups (Table 1) in fresh embryo transfer cycles: The low AMH group (< 1.00 ng/mL) was older (median 35 vs. 32 years, P < 0.001), had higher FSH (7.81 vs. 6.51 IU/L, P < 0.001), and lower AFC (6 vs. 10, P < 0.001). Notably, the low AMH group had higher rates of endometriosis (12.0% vs. 8.5%, P = 0.001) and prior IVF failure (30.5% vs. 19.9%, P < 0.001), suggesting residual confounding despite stratification. In IVF - related parameters, the low AMH group used more Gn (Z = 17.84, P < 0.001) but for a shorter time (Z=-2.06, P = 0.039), had a lower ICSI rate (χ²=33.44, P < 0.001), and a higher cleavage stage embryo or blastocyst transfer rate (χ²=68.46 and 13.89, both P < 0.001). Regarding pregnancy outcomes, the low AMH group had higher miscarriage and early miscarriage rates (χ²=22.45 and 22.74, both P < 0.001), with no significant difference in late miscarriage rates (χ²=0.393, P = 0.531).

As shown in Table 2, in frozen - thawed embryo transfer cycles, the low AMH group (N = 486) had higher age (Z = 11.34, P < 0.001), longer infertility duration (Z = 2.54, P = 0.011), higher BMI (Z = 2.23, P = 0.026) and a slightly lower rate of primary infertility (χ² = 1.88, P = 0.170) than the normal AMH group (N = 4,792). Basal FSH was higher (Z = 10.23, P < 0.001) and AFC was lower (Z = -21.93, P < 0.001) in the low AMH group. In terms of infertility causes, the low AMH group had fewer tubal - factor cases (χ² = 10.32, P = 0.001) but more endometriosis cases (χ² = 18.52, P < 0.001), with no significant difference in uterine - fibroid cases (χ² = 2.95, P = 0.086). Prior IVF failure was more common in the low AMH group (χ² = 51.08, P < 0.001). There were also differences in protocols of COS (χ²=105.35, P < 0.001), with the low AMH group using more Gn (Z = 14.92, P < 0.001) but for a shorter time (Z=-2.25, P = 0.025). The low AMH group had a lower ICSI rate (χ²=39.76, P < 0.001) and higher rates of cleavage stage embryo (χ²=5.89, P = 0.015) and blastocyst transfer (χ²=7.62, P = 0.006). In pregnancy outcomes, the low AMH group had higher miscarriage (χ²=5.62, P = 0.018) and early miscarriage rates (χ²=5.45, P = 0.020), with no significant difference in late miscarriage rates (χ²=0.09, P = 0.769).

Comparison of fresh embryo transfer pregnancy outcomes in women aged <35 and ≥ 35 with Different AMH Levels

The study showed that in both fresh and frozen - thawed embryo transfer cycles, the low AMH group exhibited higher miscarriage and early miscarriage rates than the normal AMH group. However, significant differences in baseline characteristics and IVF related data were found between the different AMH groups across these cycles. To better analyze how AMH levels relate to pregnancy outcomes, it is necessary to better match the differences between groups. Patients were stratified into two age subgroups (≥ 35 and < 35 years) to account for differences in baseline data and IVF treatment processes. Tables 3 and 4 provide detailed results of the comparisons within these subgroups.

As indicated in Supplementary table 1, among patients who were 35 years or older and achieved clinical pregnancies following fresh embryo transfer, there were 518 in the low AMH group and 1,149 in the normal AMH group. In terms of baseline data, the low AMH group had a significantly higher age (Z = 4.20, P < 0.001) but comparable BMI (P = 0.995). The low AMH group also exhibited higher basal FSH (Z = 8.59, P < 0.001) and lower AFC (Z = -17.13, P < 0.001). They had a lower proportion of tubal factors (χ² = 12.87, P < 0.001). IVF failure was more common in the low AMH group (χ² = 11.91, P = 0.001). Additionally, the low AMH group used more Gn (Z = 6.11, P < 0.001) but for a shorter duration (Z = -2.46, P = 0.014), had a lower ICSI rate (χ² = 4.80, P = 0.028), and a higher rate of cleavage stage embryo (χ² = 34.15, P < 0.001) and blastocyst transfer (χ² = 5.68, P = 0.017). Regarding pregnancy outcomes, the miscarriage rate was 30.9% in the low AMH group and 28.5% in the normal AMH group, with no significant difference (χ² = 0.90, P = 0.342). Similarly, no significant differences were found in early (26.3% vs. 23.2%, χ² = 1.71, P = 0.190) and late miscarriage rates (4.6% vs. 5.3%, χ² = 0.21, P = 0.645).

Supplementary table 2 shows that among patients under 35 with clinical pregnancies from fresh embryo transfer, 499 were in the low AMH group and 2,816 in the normal AMH group. The low AMH group had higher age (Z = 5.88, P < 0.001), slightly higher BMI (Z = 2.21, P = 0.027), and longer infertility duration (Z = 2.75, P = 0.006). They also had higher basal FSH (Z = 8.39, P < 0.001), lower AFC (Z = -20.51, P < 0.001), and a higher proportion of endometriosis (χ² = 20.83, P < 0.001). IVF failure was more common in the low AMH group (χ² = 16.05, P < 0.001). The low AMH group used more Gn (Z = 15.47, P < 0.001), had a lower ICSI rate (χ² = 17.92, P < 0.001), and a higher rate of cleavage stage embryo (χ² = 20.11, P < 0.001) and blastocyst transfer (χ² = 7.10, P = 0.008). In pregnancy outcomes, the miscarriage rate was significantly higher in the low AMH group (20.8% vs. 15.4%, χ² = 8.93, P = 0.003), as was the early miscarriage rate (17.4% vs. 12.5%, χ² = 8.56, P = 0.003). No significant difference was found in late miscarriage rates (3.4% vs. 2.8%, χ² = 0.30, P = 0.584), suggesting AMH’s effect may be limited to early pregnancy loss.

These results indicate that in patients under 35, low AMH levels are significantly associated with a higher risk of miscarriage following fresh embryo transfer, particularly in the case of early miscarriage.

Comparison of frozen-thawed embryo transfer pregnancy outcomes in women aged ≥ 35 and <35 with Different AMH Levels

Among the clinical pregnancies from frozen - thawed embryo transfer, there were 486 in the low AMH group and 4,792 in the normal AMH group. Subgroup analysis was performed based on age (≥ 35 years and < 35 years), and the results of the comparison between the two groups are presented in Supplementary Tables 3 and 4.

As shown in Supplementary table 3, among women aged 35 and older with clinical pregnancies from frozen - thawed embryo transfer, there were 229 in the low AMH group and 1,263 in the normal AMH group. The low AMH group had significantly higher age (Z = 4.91, P < 0.001), higher basal FSH (Z = 6.93, P < 0.001), and lower AFC (Z = -12.57, P < 0.001). They also had a lower rate of tubal factors (χ² = 4.73, P = 0.030) but a higher rate of endometriosis (χ² = 6.71, P = 0.010). IVF failure was more common in the low AMH group (χ² = 9.28, P = 0.002). The low AMH group used more Gn (Z = 5.89, P < 0.001) but for a shorter duration (Z = -2.56, P = 0.010), had a lower ICSI rate (χ² = 18.28, P < 0.001), and a higher rate of cleavage stage embryo (χ² = 5.77, P = 0.016) and blastocyst transfer (χ² = 7.32, P = 0.007). The differences in BMI (P = 0.707), primary infertility rate (P = 0.442), and infertility duration (P = 0.486) between the two groups were not statistically significant. Among women aged ≥ 35 years, there was a higher miscarriage rate in the low AMH group (30.6%) compared to the normal AMH group (26.4%), but the difference was not statistically significant (χ² = 1.47, P = 0.226). Similarly, no significant differences were found in early miscarriage rates (27.1% vs. 23.1%, P = 0.226) or late miscarriage rates (3.5% vs. 3.3%, P = 1.000).

As shown in Supplementary table 4, among women under the age of 35, there were 257 patients in the low AMH group and 3,529 in the normal AMH group. The low AMH group demonstrated significantly higher age (Z = 5.43, P < 0.001), significantly higher basal FSH levels (Z = 6.86, P < 0.001), and significantly lower AFC (Z = -16.76, P < 0.001). Additionally, the low AMH group had a higher proportion of endometriosis cases (χ² = 11.09, P = 0.001), used a higher dosage of Gn (Z = 12.55, P < 0.001), and had a lower ICSI rate (χ² = 16.35, P < 0.001). However, differences in infertility duration (P = 0.669), cleavage stage embryo transfer rate (P = 0.170), and blastocyst transfer rate (P = 0.426) between the two groups were not statistically significant. In terms of pregnancy outcomes, the low AMH group had a miscarriage rate of 16.3% compared to 15.7% in the normal AMH group (χ² = 0.04, P = 0.844). Similarly, no significant differences were found in early miscarriage rates (14.0% vs. 13.4%, P = 0.847) or late miscarriage rates (2.3% vs. 2.3%, P = 1.000).

The study analyzed 10,260 clinical pregnancies from embryo transfer, including 4,982 fresh and 5,278 frozen - thawed embryo transfer cycles, using serum AMH levels of 1.00 ng/mL as the classification threshold. Results showed that low AMH was linked to higher miscarriage risk in both fresh and frozen - thawed cycles. Subgroup analysis indicated this was significant only in women under 35 following fresh embryo transfer. Binary logistic regression revealed that low AMH was an independent risk factor for miscarriage and early miscarriage in fresh cycles for women under 35, even after adjusting for potential confounders. However, no significant association was found between AMH levels and miscarriage risk in women aged ≥ 35 following fresh embryo transfer or those following frozen-thawed embryo transfer.

Our findings suggest a correlation between serum AMH levels and the risk of early miscarriage following fresh embryo transfer, which varies with patient age. Advanced maternal age is the primary factor for early miscarriage in older patients. Conversely, in younger patients, low AMH levels (< 1.00 ng/mL) emerge as an independent risk factor for miscarriage. These results highlight the potential of AMH as a predictive biomarker for miscarriage risk in young women undergoing fresh embryo transfer, suggesting that pre-treatment AMH screening could guide personalized clinical interventions, such as intensified luteal support or preimplantation genetic testing, to mitigate early pregnancy loss.

Some previous studies align with parts of our findings. Tarasconi et al.[14] analyzed 1,060 fresh - embryo transfer pregnancies and found low AMH (AMH < 1.6 ng/mL) linked to higher miscarriage risk (OR = 1.07, 95% CI = 1.03–1.12, P = 0.021), but only in women aged > 33. Our age-stratified analysis contradicts this. This discrepancy may stem from different AMH thresholds (1.6 ng/mL in their study vs. 1.0 ng/mL in ours per ACOG guidelines), which can affect risk stratification. Moreover, in the study, even within age groups, significant age differences between AMH groups may confound results. Notably, Hong et al.[20]studied 848 infertile women, 206 of whom had early miscarriages. They found both age and serum AMH independently associated with early miscarriage (P = 0.022). Other studies suggest no link between AMH levels and miscarriage. Zhang et al.'s[21] large retrospective study of 9,431 first - time IVF patients found no independent correlation between AMH levels (using the 25th and 75th percentiles as cutoffs) and miscarriage rates after fresh embryo transfer. Discrepancies between this study and ours may arise from different AMH cutoffs and the inclusion of PCOS patients or other disease - related factors in their study that could impact results.

Although our study shows low AMH is associated with miscarriage in women under 35 after fresh embryo transfer, the underlying mechanism is unclear. This may relate to several hypotheses combining AMH's physiological role and clinical application. Firstly, it might be associated with decreased oocyte quality. AMH, a key regulator secreted by granulosa cells in small follicles, reflects ovarian reserve depletion. Its reduction may indirectly impair oocyte maturation. Specific factors secreted by oocytes, such as GDF − 9 and BMP15, can act on granulosa cells and downregulate AMH expression[22]. Aged ≥ 37 show downregulated TGF - β-related protein expression (including AMH) in cumulus cells, which affects follicle development and oocyte maturation[23]. AMH levels in young women may reduce oocyte quality and increase miscarriage risk. Notably, oocyte aging in women ≥ 35 may mask the impact of AMH, while young women with rapid ovarian reserve depletion are more prone to oocyte quality defects. Second, it might relate to AMH's impact on endometrial receptivity. Evidence shows that endometrial cells express AMH, AMHRⅡ, and molecules in the AMH - pathway like ALK3, Smads1, and Smads9, indicating a functional AMH signaling system in the endometrium[24–26]. However, the exact physiological role and mechanism of AMH are still unclear. Paulson et al.[27] reported that AMHRⅡ expression in the endometrium varies with the menstrual cycle, and has a substantial increase in the luteal phase. Fu et al.[28] found that in women with recurrent implantation failure, AMHRⅡ expression in endometrial cells is higher, with downregulated estrogen and prolactin receptors and increased apoptosis in luteal - phase cells. Thus, AMH may affect pregnancy outcomes by binding to AMHRⅡ, inducing apoptosis, and disrupting decidualization. Lastly, ovarian stimulation protocols may play a role. Serum AMH is a reliable predictor of ovarian response and influences the choice of stimulation protocol and drug dose. Our data show that patients with low AMH levels often use the antagonist protocol and higher Gn doses. Our study controlled for the impact of ovarian stimulation protocols using binary logistic regression but did not account for Gn dosage. Wang et al.[29] reported that in antagonist - protocol patients, the aneuploidy rate in miscarriage tissue and Preimplantation Genetic Testing for Aneuploidy (PGT - A) cycles was higher than in the GnRH - a long protocol group. Further research is needed to confirm the link between Gn dosage and aneuploidy or miscarriage.

The study has several strengths over previous research. First, it has the largest sample size to date investigating the link between serum anti-Müllerian hormone (AMH) levels and miscarriage risk, which enhances the credibility of our findings. Second, while prior studies have primarily focused on AMH as an ovarian reserve marker, our study extends this by examining its association with miscarriage risk, further filling a knowledge gap in this field. Moreover, our results have significant implications for clinical practice and policymaking. Our data suggest that young women with low AMH levels should be informed of their increased miscarriage risk during IVF, and clinicians should consider AMH levels when devising individualized treatment plans.

However, our study has several limitations. First, even though it has a large sample size, as a retrospective study itself, it might lead to selection bias. This could impact how widely the results can be applied and how reliable they are. Second, the absence of wet lab validation restricts the discussion of biological mechanisms underlying the results. Third, despite adjusting for confounders, residual confounding by unmeasured factors (e.g., embryonic aneuploidy rates) cannot be excluded. Lastly, the study cohort doesn't fully represent different races and ages, which might affect the reproducibility of the findings.

Future studies should aim to strengthen the current evidence through prospective, multicenter designs with larger sample sizes. In addition, further research is needed to elucidate the underlying mechanisms through which low AMH levels contribute to an increased risk of miscarriage. It would also be valuable to investigate dynamic changes in AMH levels—such as those occurring before and after ovarian stimulation or during early pregnancy—to determine their potential predictive value for miscarriage.

This retrospective study of 10,260 embryo transfer cycles shows that in women under 35, following fresh embryo transfer, low serum AMH (< 1.00 ng/mL) predicts increased miscarriage risk, especially in early pregnancy. It offers new insights into AMH as a biomarker and provides guidance for future clinical practice.

In conclusion, AMH levels showed no link to miscarriage risk in women over 35 following fresh embryo transfer or those aged 20–45 but following frozen embryo transfer. However, in women under 35 following fresh embryo transfer, low serum AMH (< 1.00 ng/mL) was found to predict higher miscarriage risk, particularly in early pregnancy. Young women with low AMH undergoing fresh embryo transfer may benefit from enhanced luteal support or preimplantation genetic testing. Future research should focus on exploring AMH's clinical potential and its applicability across different populations to improve women's reproductive health management.