Low anti-müllerian hormone levels increased early pregnancy loss rate in patients undergoing frozen-thawed euploid single blastocyst transfer: a retrospective cohort study.

This investigation enrolled 1,297 participants who received their first frozen-thawed embryo transfer cycles following PGT. These cycles were then categorized into 3 groups according to quintiles of AMH levels: low AMH group (AMH ≤ 1.872ng/mL, n  = 260), medium AMH group (AMH = 1.872-5.276ng/mL, n  = 779), and high AMH level (AMH ≥ 5.276ng/mL, n  = 258). After PSM, the final cohort comprised 143 participants in each group and the variables demonstrated a well-balanced distribution (Supplemental Fig. S3 ). Despite being matched, equilibrium was not achieved between the three groups of female age and E2 (Table  1 ). For confounders of baseline inequality, we included multivariate logistic regression analyses in the subsequent analyses. In terms of pregnancy outcomes, the matched data revealed a higher rate of EPL in the low AMH level group and a lower rate of clinical pregnancy and live births ( P  < 0.05) (Fig.  2 ). Table 1 Comparison of baseline characteristics and pregnancy outcomes between different AMH levels before and after PSM Before propensity score matching ( n  = 1,297) After propensity score matching( n  = 429) Characteristics Low AMH group Medium AMH group High AMH group p -value q -value 1 Low AMH group Medium AMH group High AMH group p -value q -value 1 N  = 260 N  = 779 N  = 258 N  = 143 N  = 143 N  = 143 Female age (y) , Mean (SD) 31.50 (4.65) 29.96 (3.96) 29.40 (3.62) < 0.001 2 < 0.001 30.99 (4.71) 29.63 (4.29) 29.38 (3.63) 0.011 0.063 Infertility type , n (%) 0.170 3 0.190 0.180 0.280  Primary 173 (67) 500 (64) 152 (59) 53 (37) 53 (37) 40 (28)  Secondary 87 (33) 279 (36) 106 (41) 90 (63) 90 (63) 103 (72) Parity , n (%) 0.011 0.024 0.041 3 0.100  0 165 (63) 579 (74) 204 (79) 97 (68) 112 (78) 101 (71)  1 68 (26) 146 (19) 43 (17) 36 (25) 19 (13) 28 (20)  2 22 (8.5) 45 (5.8) 7 (2.7) 10 (7.0) 10 (7.0) 8 (5.6)  3 4 (1.5) 7 (0.9) 3 (1.2) 0 (0) 2 (1.4) 3 (2.1)  4 1 (0.4) 2 (0.3) 1 (0.4) 0 (0) 0 (0) 3 (2.1) NO. of miscarriage , n (%) 0.049 3 0.074 0.670 0.700  0 125 (48) 374 (48) 142 (55) 70 (49) 59 (41) 66 (46)  1 71 (27) 206 (26) 50 (19) 41 (29) 45 (31) 43 (30)  2 37 (14) 136 (17) 43 (17) 21 (15) 26 (18) 17 (12)  3 16 (6.2) 52 (6.7) 20 (7.8) 7 (4.9) 12 (8.4) 14 (9.8)  4 8 (3.1) 7 (0.9) 3 (1.2) 2 (1.4) 1 (0.7) 2 (1.4)  5 3 (1.2) 4 (0.5) 0 (0) 2 (1.4) 0 (0) 1 (0.7) BMI (kg/m2) , Mean (SD) 22.98 (2.81) 22.78 (2.87) 22.73 (2.87) 0.480 0.510 22.80 (2.80) 23.42 (3.10) 22.98 (3.12) 0.230 2 0.320 E2 (pg/ml) , Mean (SD) 689.31 (1,401.51) 586.25 (1,296.45) 717.85 (1,431.35) 0.053 0.074 673.87 (1,396.93) 547.10 (1,271.54) 1,015.54 (1,639.04) 0.023 2 0.070 FSH (mIU/ml) , Mean (SD) 6.98 (1.72) 6.26 (1.48) 5.70 (1.36) < 0.001 2 < 0.001 6.30 (1.69) 6.33 (1.56) 6.26 (1.51) 0.880 0.880 P (ng/ml) , Mean (SD) 0.44 (0.37) 0.49 (0.37) 0.45 (0.34) 0.019 2 0.037 0.44 (0.36) 0.43 (0.31) 0.47 (0.29) 0.080 0.140 LH (mIU/ml) , Mean (SD) 4.24 (2.07) 4.95 (2.55) 5.60 (3.25) < 0.001 < 0.001 4.64 (2.21) 4.57 (2.33) 4.46 (2.91) 0.250 2 0.320 AFC , Mean (SD) 9.44 (4.28) 16.06 (4.97) 20.27 (5.39) < 0.001 < 0.001 9.53 (4.22) 16.78 (5.15) 18.78 (5.85) < 0.001 2 < 0.001 Endometrial preparation protocols , n (%) < 0.001 < 0.001 0.480 3 0.540  HRT 98 (38) 257 (33) 46 (18) 45 (31) 36 (25) 43 (30)  NC 162 (62) 522 (67) 212 (82) 98 (69) 107 (75) 100 (70) Blastocyst development day , n (%) 0.009 3 0.023 0.350 0.420  D5 164 (63) 558 (72) 192 (74) 93 (65) 103 (72) 93 (65)  D6 96 (37) 221 (28) 66 (26) 50 (35) 40 (28) 50 (35) Biochemical pregnancy rate , n (%) 166 (64) 559 (72) 175 (68) 0.048 0.074 90 (63) 110 (77) 92 (64) 0.021 3 0.070 Clinical pregnancy rate , n (%) 148 (57) 509 (65) 152 (59) 0.023 3 0.041 78 (55) 101 (71) 86 (60) 0.018 0.070 Miscarriage rate , n (%) 22 (14) 80 (16) 23 (15) 0.970 3 0.970 14 (18) 17 (17) 8 (9.3) 0.230 3 0.310 Early pregnancy loss rate , n (%) 133 (51) 344 (44) 126 (49) 0.100 3 0.120 79 (55) 59 (41) 64 (45) 0.048 0.100 Livebirth rate , n (%) 116 (45) 400 (51) 118 (46) 0.090 0.120 57 (40) 76 (53) 74 (52) 0.047 3 0.100 AFC, antral follicle count; AMH, anti-müllerian hormone; BMI, body mass index; FSH, follicle-stimulating hormone; HRT, hormone replacement therapy; LH, luteinizing hormone; NC, natural cycle; PSM, propensity score matching 1 False discovery rate correction for multiple testing; 2 Kruskal-Wallis rank sum test; 3 Pearson’s Chi-squared test Comparison of baseline characteristics and pregnancy outcomes between different AMH levels before and after PSM AFC, antral follicle count; AMH, anti-müllerian hormone; BMI, body mass index; FSH, follicle-stimulating hormone; HRT, hormone replacement therapy; LH, luteinizing hormone; NC, natural cycle; PSM, propensity score matching 1 False discovery rate correction for multiple testing; 2 Kruskal-Wallis rank sum test; 3 Pearson’s Chi-squared test Fig. 2 Comparison of clinical pregnancy rates, live birth rates, and early pregnancy loss rates according to anti-Müllerian hormone (AMH) levels. A chi-square test showed low AMH levels are associated with higher pregnancy loss rates and lower clinical pregnancy and live birth rates compared to medium AMH levels. * indicates p  < 0.05, ** indicates p  < 0.01, *** indicates p  < 0.001; NS indicates non-significant differences Comparison of clinical pregnancy rates, live birth rates, and early pregnancy loss rates according to anti-Müllerian hormone (AMH) levels. A chi-square test showed low AMH levels are associated with higher pregnancy loss rates and lower clinical pregnancy and live birth rates compared to medium AMH levels. * indicates p  < 0.05, ** indicates p  < 0.01, *** indicates p  < 0.001; NS indicates non-significant differences Logistic regression analysis identified risk factors for early pregnancy loss in the overall population, ≥ 35 and < 35 age groups. Univariate logistic regression models demonstrated a significantly elevated risk of EPL among participants with low AMH levels compared with those with medium AMH levels, with an unadjusted OR of 1.76 (95% CI, 1.10–2.82) (Supplemental Table S1 ). Following adjustment for potential confounding variables, the multivariate analysis still suggested that low AMH levels were an independent predictor for EPL, with an adjusted OR of 1.85 (95% CI, 1.13–3.04). Supplemental Table S2 showed that all other variables had a variance inflation factor of < 5. The age-stratified analysis demonstrated a statistically significant correlation between diminished AMH levels and EPL. AMH levels and EPL specifically among participants under 35 years of age. However, among the ≥ 35 groups, AMH levels did not significantly affect EPL (Table  2 ). Based on logistic multifactorial regression, the RCS analysis revealed the nonlinear and continuous association between AMH levels and EPL (Fig.  3 ). In the restricted cubic spline, no significant evidence of a nonlinear relationship was identified between AMH levels and EPL (P-non-linear = 0.356). Individuals demonstrating AMH levels below the threshold of 1.813 ng/mL showed an increased risk of EPL. Table 2 Multivariate logistic regression analysis for early pregnancy loss rate in the overall population, in the < 35 and ≥ 35 age groups a Characteristics Overall population ( n  = 429) < 35 ( n  = 372) ≥ 35 ( n  = 57) OR (95% CI) p -value OR (95% CI) p -value OR (95% CI) p -value Female age 1.01 (0.96 to 1.06) 0.744 0.99 (0.93 to 1.06) 0.797 1.34 (0.97 to 1.99) 0.082 Infertility type 0.802 0.644 0.770  Primary Reference Reference Reference  Secondary 1.08 (0.60 to 1.92) 1.16 (0.61 to 2.20) 1.48 (0.10 to 23.2) Parity 0.80 (0.57 to 1.12) 0.192 0.84 (0.57 to 1.20) 0.339 0.68 (0.23 to 1.89) 0.454 NO. of miscarriages 1.14 (0.89 to 1.45) 0.306 1.08 (0.81 to 1.47) 0.592 1.25 (0.75 to 2.12) 0.390 BMI 1.05 (0.98 to 1.12) 0.188 1.06 (0.98 to 1.14) 0.127 0.87 (0.66 to 1.10) 0.254 FSH 1.03 (0.90 to 1.17) 0.713 1.07 (0.92 to 1.24) 0.393 0.97 (0.61 to 1.55) 0.899 E2 1.00 (1.00 to 1.00) 0.159 1.00 (1.00 to 1.00) 0.104 1.00 (1.00 to 1.00) 0.988 P 1.70 (0.85 to 3.46) 0.134 1.40 (0.63 to 3.17) 0.407 2.53 (0.39 to 20.0) 0.337 LH 1.02 (0.93 to 1.11) 0.714 0.97 (0.88 to 1.07) 0.551 1.32 (1.02 to 1.82) 0.038 AMH level 0.041 0.033 0.532  Medium AMH group Reference Reference Reference  Low AMH group 1.85 (1.13 to 3.04) 2.04 (1.18 to 3.54) 1.29 (0.33 to 5.12)  High AMH group 1.20 (0.73 to 1.96) 1.25 (0.74 to 2.14) 3.24 (0.42 to 30.7) Endometrial preparation protocols 0.144 0.141 0.921  NC Reference Reference Reference  HRT 1.40 (0.89 to 2.20) 1.45 (0.89 to 2.39) 0.93 (0.19 to 4.38) Blastocyst development day < 0.001 < 0.001 0.559  D5 Reference Reference Reference  D6 2.21 (1.45 to 3.39) 2.97 (1.86 to 4.80) 0.67 (0.17 to 2.58) AMH, anti-müllerian hormone; BMI, body mass index; CI, confidence interval; FSH, follicle-stimulating hormone; HRT, hormone replacement therapy; LH, luteinizing hormone; NC, natural cycle; OR, odds ratio a Adjusted for female age, BMI, infertility type, parity, No. of miscarriages, BMI, FSH, E2, P, LH, AMH, endometrial preparation protocols and blastocyst development day Multivariate logistic regression analysis for early pregnancy loss rate in the overall population, in the < 35 and ≥ 35 age groups a AMH, anti-müllerian hormone; BMI, body mass index; CI, confidence interval; FSH, follicle-stimulating hormone; HRT, hormone replacement therapy; LH, luteinizing hormone; NC, natural cycle; OR, odds ratio a Adjusted for female age, BMI, infertility type, parity, No. of miscarriages, BMI, FSH, E2, P, LH, AMH, endometrial preparation protocols and blastocyst development day Fig. 3 The restricted cubic spline for the relationship between early pregnancy loss and AMH levels. Four knots for the RCS model were set (4 knots were set at the 5th, 35th, 65th, and 95th percentiles of serum AMH levels). The cutoff value was 1.813. The solid red line represented the estimated ORs and the dashed grey lines represented corresponding 95% confidence intervals (CIs). The horizontal dashed grey line and the red dot indicated the reference value. The grey bar represented the frequency. The p-value for overall association < 0.05 manifested a significant association, whatever the shape of the relationship curve was. The p-value for non-linear association < 0.05 indicated a nonmonotonic dose-response curve; otherwise, a monotonic was suggested. The restricted cubic spline for the relationship between early pregnancy loss and AMH levels. Four knots for the RCS model were set (4 knots were set at the 5th, 35th, 65th, and 95th percentiles of serum AMH levels). The cutoff value was 1.813. The solid red line represented the estimated ORs and the dashed grey lines represented corresponding 95% confidence intervals (CIs). The horizontal dashed grey line and the red dot indicated the reference value. The grey bar represented the frequency. The p-value for overall association < 0.05 manifested a significant association, whatever the shape of the relationship curve was. The p-value for non-linear association < 0.05 indicated a nonmonotonic dose-response curve; otherwise, a monotonic was suggested. Subgroup and interaction analyses. Female age, BMI, infertility type, AFC, endometrial preparation protocol, blastocyst development day and AFC were known as confounding factors of the association between AMH levels and EPL. To evaluate the consistency of associations between AMH level and EPL across different subgroups, stratified analyses with interaction assessments were conducted (Table  3 ). Generally, low AMH levels demonstrated significant associations with elevated EPL risk across all stratified subgroups. The effect of AMH levels on EPL was more significant in younger patients, those with primary infertility, AFC ≥ 10, and transferred with D6 blastocysts. Table 3 Impact of different AMH levels on early pregnancy loss rate in each subgroup a Variable Count AMH levels OR (95% CI) P -value P for interaction Overall 429 0 Reference 1 1.84 (1.14, 2.96) 0.012 2 1.21 (0.75, 1.96) 0.426 Age 0.925  < 35 372 0 Reference 1 1.92 (1.14, 3.23) 0.014 2 1.27 (0.76, 2.12) 0.353  ≥ 35 57 0 Reference 1 1.5 (0.42, 5.34) 0.527 2 2.12 (0.32, 14.27) 0.440 Infertility type 0.242  Secondary 283 0 Reference 1 1.36 (0.75, 2.47) 0.306 2 1.11 (0.62, 1.99) 0.715  Primary 146 0 Reference 1 3.23 (1.41, 7.37) 0.005 2 1.44 (0.57, 3.65) 0.437 BMI 0.484  < 25 308 0 Reference 1 1.97 (1.11, 3.47) 0.020 2 1.02 (0.56, 1.83) 0.958  ≥ 25 121 0 Reference 1 1.68 (0.63, 4.43) 0.298 2 1.88 (0.75, 4.68) 0.177 AFC 0.069  < 10 91 0 Reference 1 8.55 (1.81, 40.31) 0.007 2 6.68 (0.9, 49.75) 0.064  ≥ 10 338 0 Reference 1 1.44 (0.81, 2.57) 0.216 2 1.06 (0.64, 1.74) 0.823 Endometrial preparation protocols 0.202  HRT 305 0 Reference 1 1.54 (0.87, 2.71) 0.135 2 1.36 (0.77, 2.4) 0.290  NC 124 0 Reference 1 2.39 (0.9, 6.34) 0.079 2 1.18 (0.43, 3.23) 0.743 Blastocyst development day 0.55  D5 140 0 Reference 1 1.14 (0.47, 2.77) 0.772 2 0.93 (0.39, 2.24) 0.878  D6 289 0 Reference 1 2.22 (1.22, 4.05) 0.009 2 1.27 (0.69, 2.34) 0.449 AFC, antral follicle count; AMH, anti-müllerian hormone; BMI, body mass index; CI, confidence interval; FSH, follicle-stimulating hormone; HRT, hormone replacement therapy; LH, luteinizing hormone; NC, natural cycle. OR, odds ratio a Represents multivariate logistic regression analysis adjusted for parity, NO. of miscarriages, FSH, E2, P, LH 0,1 and 2 represented the medium AMH group, low AMH group and high AMH group, respectively Impact of different AMH levels on early pregnancy loss rate in each subgroup a AFC, antral follicle count; AMH, anti-müllerian hormone; BMI, body mass index; CI, confidence interval; FSH, follicle-stimulating hormone; HRT, hormone replacement therapy; LH, luteinizing hormone; NC, natural cycle. OR, odds ratio a Represents multivariate logistic regression analysis adjusted for parity, NO. of miscarriages, FSH, E2, P, LH 0,1 and 2 represented the medium AMH group, low AMH group and high AMH group, respectively E-values indicated that when there was an unmeasured confounding with an association of at least 2.06 with the exposure factor and outcome, only then could the currently observed association of OR = 1.85 be fully canceled out by that unmeasured confounding. On the other hand, based on the above results, we categorized the population into low AMH group and medium AMH group by using an AMH level of 1.813 ng/mL as the cutoff point and performed IPTW sensitivity analyses. The results of the IPTW analysis aligned with previous propensity score matching (PSM) analyses, both of which revealed that low AMH levels were significantly correlated with a higher likelihood of EPL (Table  4 ). Table 4 Sensitivity analysis between AMH levels and early pregnancy loss used AMH = 1.813 as a cutoff value Characteristic IPTW IPTW after multiple imputation PSM after multiple imputation OR (95% CI) p -value OR (95% CI) p -value OR (95% CI) p -value Female age 1.00(0.99, 1.01) 0.885 1.00(0.997, 1.00) 0.700 0.99(0.95, 1.04) 0.816 AMH levels  ≥ 1.813 reference reference reference  < 1.813 1.07(0.99, 1.15) 0.086 1.03(1.01, 1.06) 0.002 1.45(1.03, 2.07) 0.034 AMH, anti-müllerian hormone; CI, confidence interval; IPTW, inverse probability of treatment weighting; OR, odds ratio; PSM, propensity score matching a Adjusted for female age, BMI, infertility type, parity, No. of miscarriages, FSH, E2, P, LH, endometrial preparation protocols and blastocyst development day Sensitivity analysis between AMH levels and early pregnancy loss used AMH = 1.813 as a cutoff value AMH, anti-müllerian hormone; CI, confidence interval; IPTW, inverse probability of treatment weighting; OR, odds ratio; PSM, propensity score matching a Adjusted for female age, BMI, infertility type, parity, No. of miscarriages, FSH, E2, P, LH, endometrial preparation protocols and blastocyst development day

The dataset for this retrospective analysis was sourced from the Reproductive Medicine Center at the First Affiliated Hospital of Zhengzhou University. Chart information data were collected on 1982 patients who underwent the PGT cycle between January 2016 and August 2023. Considering the study was retrospective, informed consent was waived. All procedures adhered to the Declaration of Helsinki and institutional guidelines. We recorded the demographic information of the participants in detail. Referring to previous literature, the study cohort was stratified into three groups based on the serum AMH quintile distribution: low (≤ 20th percentile, AMH ≤ 1.872 ng/mL), medium (21st-80th percentile, 1.873–5.276 ng/mL), and high (≥ 81st percentile, ≥ 5.277 ng/mL) [ 16 ]. Study participants were selected according to the following inclusion criteria: 1) patients who underwent PGT and had their first frozen-thawed embryo euploidy blastocyst transfer; 2)BMI ≤ 30 kg/m 2 . The exclusion criteria were: (1) Polycystic ovary syndrome (PCOS); (2) Abnormalities in the uterine environment such as uterine malformations, endometriosis, submucosal uterine fibroids, uterine fluid or heavy adhesions; (3) Hydrosalpinx; (4) Comorbid systemic diseases such as diabetes mellitus, hypertension, immune system diseases, and malignant tumors. A detailed flowchart of patient selection is provided in Fig.  1 . Fig. 1 The flowchart of participants. PCOS = Polycystic Ovary Syndrome; PSM = propensity score matching; PGT = preimplantation genetic testing The flowchart of participants. PCOS = Polycystic Ovary Syndrome; PSM = propensity score matching; PGT = preimplantation genetic testing The study protocol received ethical approval by the Hospital Ethics Committee on October 12, 2024 (reference number: 2024-KY-1226). All serum specimens were measured in the specialized endocrinology laboratory within the Reproductive Medicine Center. Typically, the participants’ hormone levels were measured within 12 months preceding ovarian stimulation initiation, including AMH, FSH, luteinizing hormone (LH), estrogen (E2), and progesterone (P). Blood specimens (2 mL) were obtained under sterile conditions from study participants and maintained at -20 °C for subsequent biochemical analysis. After centrifugation of the samples, serum was analyzed by a fully automated Elecsys ® immunoassay analyzer (Roche, Cobas e601, Canada) to measure serum hormone levels. During fresh cycles of ovarian stimulation, participants were mainly treated with a mid-luteal short-acting gonadotropin-releasing hormone (GnRH) agonist long regimen, GnRH antagonist regimen, and mild stimulation regimen. Follicular development was monitored and evaluated by transvaginal ultrasonography. Intramuscular administration of human chorionic gonadotropin (HCG) was initiated when ultrasonographic assessment confirmed the presence of a minimum of three mature follicles, each measuring ≥ 17 mm in diameter (Livzon, Zhuhai, China). Ultrasound-guided ovum retrieval was conducted within 36–38 h of HCG injection. One hour after ovum retrieval, intracytoplasmic sperm injection (ICSI) fertilization was performed, as described in detail in the previous article [ 17 ]. Morphological assessment of embryo quality was performed by experienced embryologists on the mornings of days 3–5 after ovum retrieval. Assessment of blastocyst morphology on day 5 or 6 according to the Gardner and Schoolcraft scoring system [ 18 ]. Typically, blastocysts of grade 3BC and above are selected for biopsy. After blastocysts had completed genome amplification, we conducted biopsies utilizing the laser method described by McArthur et al. [ 19 ]. All patients had a natural or hormone replacement cycle endometrial preparation protocol, as previously described [ 20 ]. Peripheral blood β-hCG was assessed two weeks post-embryo transfer, and the diagnosis of biochemical pregnancy was established if β-hCG ≥ 50 U/L. Biochemical pregnancy loss was characterized by abdominal ultrasonography at day 35 post-transfer. The visualization of an intrauterine gestational sac confirmed the establishment of clinical pregnancy. Abortion was defined as a spontaneous pregnancy loss occurring before the completion of 28 gestational weeks or fetal weight below 1,000 g. Early pregnancy loss (EPL) (including biochemical pregnancy loss and miscarriages), was defined as spontaneous abortion before 10 weeks [ 21 ]. Under the assumption that the data were randomly missing, we first employed the capping method which means imputing the upper outliers with the 95th percentile and imputing the lower outliers with the 5th percentile. The multi-valued imputation was performed using the K-nearest neighbor’s method. Potential confounding variables were identified through previous research literature and the construction of a directed acyclic graph (DAG) illustrating the causal pathways between AMH (exposure) and EPL (outcome) (Supplemental Fig. S1 ). In addition, if the exclusion of a variable resulted in a change in the logarithmic risk ratio of greater than or equal to 10%, the variable was retained in the adjusted model (Supplemental Fig. S2 ) [ 22 ]. Statistical analysis was conducted with ANOVA for continuous variables, whereas categorical variables were performed utilizing Pearson’s or Fisher’s exact tests. To mitigate confounding, TriMatch analysis was implemented. Three different logistic regression models were utilized to determine the propensity scores for each device, and three distances between the propensity scores were computed for each potential matched triplet. We concluded that each pair was distinct and that a participant could only be replicated up to three times. Subsequently, generalized linear models were constructed for univariate and multivariate logistic regression analyses. Moreover, stratified analyses were implemented based on maternal age and we also applied a stepwise variance inflation factor (VIF) to test for covariance between the variables. Based on a generalized multivariate linear model, restricted cubic spline (RCS) regression modeling was implemented in our analytical approach to examine potential nonlinear relationships between AMH levels and early pregnancy loss. Furthermore, subgroup analyses and interaction tests were performed to investigate potential associations between AMH levels and EPL in each subgroup. To reduce the effect of selectivity bias, an inverse probability of treatment weighted (IPTW) model was used to adjust for observed differences in baseline covariates among the 3 groups to eliminate potentially significant differences in characteristics between groups. To further minimize the effect of missing data on the outcome, we applied the Mice software package for multiple imputations and then performed PSM and IPTW analyses after the imputation was completed. Finally, E-value was also reported to assess the potential impact of unmeasured confounders [ 23 ]. All statistical analyses were conducted utilizing R software (version 4.4.1).

In this cohort of patients who had their first frozen-thawed embryo euploid blastocyst transfer after PGT, our analysis revealed a statistically significant correlation between low AMH levels and elevated risk of EPL probability, particularly in women below 35 years of age. Noteworthy, the cut-off value obtained by plotting the RCS curve was close to the first quintile of AMH levels (the cut-off between low and medium AMH levels). The threshold value derived from RCS analysis (AMH = 1.813 ng/mL) closely aligned with the 20th percentile (1.872 ng/mL) in our cohort, suggesting its potential utility as a clinical cutoff for risk stratification. Existing literature presents conflicting evidence regarding the AMH-EPL relationship. While Peuranpää et al. reported no association between AMH and miscarriage risk in cohorts with mixed embryo ploidy [ 24 ], and Chinè et al. observed similar null findings in unselected ART cycles [ 15 ]. Arkfeld et al. recently demonstrated that AMH < 1 ng/mL independently predicts miscarriage risk in non-PCOS patients undergoing ART without PGT [ 25 ]. Based on this, we hypothesized that AMH might serve as a potential biomarker for elevated miscarriage risk in non-PCOS patients with limited predictive potential in PCOS patients. Conversely, evidence also exists suggesting the contrary situation. Prior studies have reviewed that serum AMH levels demonstrated a significant independent correlation with miscarriage after IVF-ET [ 5 , 26 ]. A retrospective analysis found that AMH could predict the outcome of IVF and PGT-A cycles as well as the quality of D5 blastocysts [ 27 ]. Existing evidence illustrates that diminished ovarian reserve may correlate with elevated risks of pregnancy-related complications, such as miscarriage when compared to individuals with normal ovarian capacity [ 28 ]. However, several methodological variations may account for these discrepant results. Firstly, heterogeneity in AMH cutoff definitions exists: while some studies classify populations using percentile-based thresholds (e.g., quintiles as in our study), others apply clinical criteria for diminished ovarian reserve (DOR), such as AMH < 1.1 ng/mL [ 5 ]. Secondly, discrepancies in AMH assay platforms (e.g., Gen II vs. Elecsys ® ) contribute to inter-study variability, as different kits may yield divergent results even for the same sample [ 29 ]. Thirdly, a few studies did not exclude women with PCOS and the criteria for exclusion of PCOS were inconsistent. PCOS-associated elevated AMH reflects impaired follicular development rather than oocyte quality, and inconsistent natriuretic criteria introduce bias. Fourthly, the inclusion criteria for participant age varied significantly across studies with substantial heterogeneity in the reported age ranges. Some studies have excluded people aged 35 years and older, and some have been conducted only in younger patients. Finally, the definition of miscarriage varied across individual studies which might limit the generalisability of the research findings. Several investigations have suggested an emerging connection between reduced AMH levels and recurrent pregnancy loss (RPL) [ 25 ]. A systematic review by Bunnewell et al. identified a possible correlation linking DOR with the elevated probabilities of PRL, especially in women under 35 years [ 30 ]. AMH and AFC levels declined progressively with the increasing number of previous miscarriages, suggesting that ovarian reserve depletion may exacerbate the risk of miscarriage [ 31 ]. Atasever et al. indicated that patients with recurrent miscarriages exhibited a threefold higher likelihood of having AMH ≤ 1 ng/mL compared to women with proven fertility [ 32 ]. The inextricable link between AMH and PRL further revealed that AMH levels may be a critical predictor of EPL. However, due to the diverse etiology of recurrent miscarriages [ 33 ], we did not consider this population in our analysis. In light of the well-established correlation between advancing maternal age and pregnancy outcomes [ 34 ], we stratified participants into < 35 and ≥ 35-year subgroups. Consistent with Liu et al. [ 14 ], no significant AMH-EPL association was observed in women ≥ 35 years. This may reflect the dominant role of age-related aneuploidy, which accounts for > 50% of pregnancy losses in this population [ 35 ]. However, Jiang et al. proposed higher aneuploidy rates in older women with low AMH [ 9 ], whereas Fouks et al. found comparable euploidy rates between young DOR patients and controls [ 36 ]. Notably, our ≥ 35-year subgroup comprised only 57 patients, necessitating validation in larger cohorts to clarify AMH’s predictive value in advanced maternal age. The observed link between low AMH and EPL may involve both chromosomal and non-chromosomal mechanisms. AMH exerts inhibitory effects on primordial follicle activation, thereby preventing premature exhaustion of the ovarian reserve [ 37 ]. A decline in AMH levels attenuates this inhibitory effect, prompting premature recruitment of primordial follicles into the growth phase and potentially bypassing normal selection and maturation processes. Prematurely activated follicles may initiate development without achieving full maturation competence, resulting in compromised oocyte quality. This manifests as mitochondrial dysfunction, aberrant spindle assembly, and consequently elevated risk of meiotic chromosome nondisjunction, ultimately generating aneuploid embryos [ 14 , 38 ]. While embryonic aneuploidy remains the primary cause of early miscarriage [ 35 ], our use of PGT-confirmed euploid embryos excluded this confounding factor. Intriguingly, Jaswa et al. reported reduced aneuploidy rates in women with DOR [ 39 ], whereas Pipari et al. found no AMH-aneuploidy association [ 8 ]. Fouks et al. further demonstrated that young DOR patients achieve comparable live birth rates per euploid transfer as controls [ 36 ]. Therefore, further follow-up studies are needed to observe pregnancy outcomes in patients with different AMH levels. Our research suggested that the increased miscarriage rates in the low AMH population may also be due to non-chromosomal causes, such as altered hormonal dynamics (e.g., luteal phase defects) or endometrial receptivity, which may mediate the AMH-EPL relationship in euploid cycles [ 40 ]. This study’s strengths include: (1) the first study to date to compare the relationship between AMH levels and EPL in a population with the first frozen-thawed embryo transfer of euploid blastocysts who underwent PGT and had their first frozen-thawed embryo euploidy blastocyst transfer; (2) exclusion of PCOS patients, eliminating bias from aberrant AMH secretion patterns [ 41 ]; (3) rigorous adjustment for confounders using directed acyclic graphs (DAGs) and propensity score matching (PSM); (4) avoidance of inter-cycle variability by restricting analysis to the first transfer. Limitations must also be acknowledged: (1) Retrospective design precluded assessment of unmeasured confounders (e.g., lifestyle factors, sperm DNA fragmentation); (2) Single AMH measurement prior to ovarian stimulation, though AMH exhibits low intra-individual variability (mean annual decline: 5.6%) [ 42 ]; (3) Single-center data, which may limit generalizability despite standardized laboratory protocols; (4) Quintile-based AMH categorization, which differs from clinical DOR thresholds; (5) Only individuals with a BMI ≤ 30 that the conclusions may not extend to obese populations (BMI > 30). In conclusion, our study suggested that in non-PCOS women < 35 years undergoing euploid blastocyst transfer, low AMH (≤ 1.8 ng/mL) independently predicts EPL risk, underscoring its role as a biomarker of oocyte competence beyond chromosomal integrity. These findings advocate for AMH-guided counseling in young patients with diminished ovarian reserve and highlight the need for mechanistic studies (e.g., follicular microenvironment analyses) to elucidate non-chromosomal pathways linking AMH to pregnancy loss. Young non-PCOS patients with low AMH levels may benefit from intensified early pregnancy monitoring and enhanced pregnancy risk counseling. Although advanced maternal age remains the most robust predictor of early pregnancy loss, both clinicians and patients should recognize that AMH possesses independent prognostic value for pregnancy outcomes.

With the global trend toward delayed childbearing in recent decades, there is growing emphasis on identifying biomarkers that predict not only conception success but also pregnancy sustainability, a critical component of female reproductive potential [ 1 ]. Fertility encompasses both the ability to conceive and to achieve live birth. In this context, early pregnancy loss (EPL) rates have been considered as a potential key indicator of reproductive competence [ 2 ]. Approximately 20% of pregnancies achieved through in vitro fertilization (IVF) end in miscarriage, of which 50% are biochemical losses [ 3 ]. EPL represents a pivotal endpoint for assessing female reproductive potential beyond conception alone. Anti-Müllerian hormone (AMH), a glycoprotein within the transforming growth factor-β (TGF-β) superfamily, is predominantly secreted by granulosa cells located in the ovary’s anterior and lesser antral lobe follicles [ 2 ]. Compared to antral follicle count (AFC) or follicle-stimulating hormone (FSH), AMH exhibits minimal intra- and inter-cycle variability, establishing it as a stable biomarker of ovarian reserve and oocyte competence [ 4 ]. The association between the AMH level and EPL is also supported by several biological mechanisms. Compromised oocyte quality, particularly meiotic errors during chromosomal segregation, underlies 50–70% of embryonic aneuploidies and serves as a major contributor to pregnancy loss [ 5 ]. Furthermore, AMH plays a regulatory role in folliculogenesis by modulating FSH sensitivity in granulosa cells; its deficiency may lead to inadequate luteinization and progesterone synthesis (< 10 ng/mL) - a critical factor for early pregnancy maintenance [ 6 ]. One indicator of oocyte quality is the number of chromosomes it contains, also known as ploidy, which is believed to be the main factor contributing to the rise in miscarriages of women over 35 [ 7 ]. The relationship between serum AMH levels and embryonic aneuploidy remains controversial. According to findings from a retrospective single-institution cohort analysis, no significant predictive relationship was detected between AMH levels and aneuploidy rates in embryos [ 8 ]. Nevertheless, most studies have demonstrated that AMH was an important predictor of embryo aneuploidy [ 9 – 11 ]. Jiang et al. have suggested that diminished ovarian reserve (AMH < 1.5 ng/mL) correlated with a 1.5-fold increased risk of aneuploidy [ 9 ]. To isolate non-chromosomal mechanisms, this study exclusively enrolled women undergoing preimplantation genetic testing (PGT) with euploid blastocyst transfer, thereby controlling for embryonic ploidy status. Advanced maternal age is inextricably linked to both diminished ovarian reserve and elevated EPL risk, creating a complex interplay between chronological aging and reproductive physiology. Although EPL is generally induced by embryonic aneuploidy [ 12 ], most meiotic errors leading to chromosomal anomalies primarily occur during oocyte maturation [ 13 ]. Based on this, it has been hypothesized that the incidence of early miscarriage indirectly reflected the quality of oocytes [ 14 ]. The influence of serum AMH levels on EPL remains an understudied and unresolved issue. Emerging data suggest that AMH levels might serve as age-independent predictors of pregnancy outcomes [ 15 ]. Several investigations have demonstrated a possible association between diminished AMH levels and elevated risk of early miscarriage [ 4 ]. These knowledge gaps underscore the need for rigorous research on how diminished ovarian reserve quantified by AMH affects pregnancy sustainability independently of embryonic ploidy. In summary, the primary objective of this research was to shed light on the relationship between low AMH levels and EPL in the infertile cohort who had their first frozen-thawed embryo euploid blastocyst transfer after PGT. By stratifying analyses by maternal age (< 35 vs. ≥35 years), we further explored whether AMH’s predictive value persists across reproductive aging stages, thereby advancing its potential as a biomarker of oocyte competence.

Below is the link to the electronic supplementary material. Supplementary Material 1: Supplemental table S1. Univariate logistic regression analysis for early pregnancy loss rate in the overall population, in the < 35 and ≥ 35 age groups. Supplementary Material 1: Supplemental table S1. Univariate logistic regression analysis for early pregnancy loss rate in the overall population, in the < 35 and ≥ 35 age groups. Supplementary Material 2: Supplemental table S2. Variance inflation factors (VIFs) of the independent variables. Supplementary Material 2: Supplemental table S2. Variance inflation factors (VIFs) of the independent variables. Supplementary Material 3: Supplemental figure S1. A simplified directed acyclic graph describing the relationship between AMH (the exposure) and early pregnancy loss (the outcome). AFC = antral follicle count; AMH = anti-müllerian hormone; BMI = body mass index; FSH = follicle-stimulating hormone; LH = luteinizing hormone. Supplementary Material 3: Supplemental figure S1. A simplified directed acyclic graph describing the relationship between AMH (the exposure) and early pregnancy loss (the outcome). AFC = antral follicle count; AMH = anti-müllerian hormone; BMI = body mass index; FSH = follicle-stimulating hormone; LH = luteinizing hormone. Supplementary Material 4: Supplemental figure S2. The changes in the effect estimates when each variable is added to the model sequentially in a step-wise fashion. The inclusion or exclusion of a variable that currently changes the effect value of the equation by more than 10% is an adjustment for the need for confounders. AFC = antral follicle count; BMI = body mass index; FSH = follicle-stimulating hormone; LH = luteinizing hormone. Supplementary Material 4: Supplemental figure S2. The changes in the effect estimates when each variable is added to the model sequentially in a step-wise fashion. The inclusion or exclusion of a variable that currently changes the effect value of the equation by more than 10% is an adjustment for the need for confounders. AFC = antral follicle count; BMI = body mass index; FSH = follicle-stimulating hormone; LH = luteinizing hormone. Supplementary Material 5: Supplemental figure S3. Balance plot for the matching variables covariate. Treatment and comparison units are divided into five stratum so that treated and comparison are similar within each stratum. Three logistic regression models were created to see the match between the three groups. The horizontal coordinate represents a standardized mean difference and < 0.1 represents a good match. Supplementary Material 5: Supplemental figure S3. Balance plot for the matching variables covariate. Treatment and comparison units are divided into five stratum so that treated and comparison are similar within each stratum. Three logistic regression models were created to see the match between the three groups. The horizontal coordinate represents a standardized mean difference and < 0.1 represents a good match.