Li et al.

Reproductive Biology
Reproductive Biology and Endocrinology (2023) 21:19
https://doi.org/10.1186/s12958-023-01066-w and Endocrinology

RESEARCH Open Access

AMH independently predicts aneuploidy

but not live birth per transfer in IVF PGT‑A cycles
Howard J. Li*, David B. Seifer and Reshef Tal

Abstract
Background While anti-Müllerian hormone (AMH) predicts quantitative IVF outcomes such as oocyte yield, it is not
certain whether AMH predicts markers of oocyte quality such as aneuploidy.
Methods Retrospective case–control analysis of the SART-CORS database, 2014–2016, to determine whether anti-
Müllerian hormone (AMH) predicts aneuploidy and live birth in IVF cycles utilizing preimplantation genetic testing for
aneuploidy (PGT-A).
Results Of 51,273 cycles utilizing PGT-A for all embryos, 10,878 cycles were included in the final analysis; of these,
2,100 cycles resulted in canceled transfer due to lack of normal embryos and 8,778 cycles resulted in primary FET.
AMH levels of cycles with ≥ 1 euploid embryo were greater than those of cycles with no normal embryos, stratifying
by number of embryos biopsied (1–2, 3–4, 5–6, and ≥ 7), P < 0.017 for each stratum. Adjusting for age and number
of embryos biopsied, AMH was a significant independent predictor of ≥ 1 euploid embryo for all age groups: < 35
yrs (aOR 1.074; 95%CI 1.005–1.163), 35–37 years (aOR 1.085; 95%CI 1.018–1.165) and ≥ 38 years (aOR 1.055; 95%CI
1.020–1.093). In comparative model analysis, AMH was superior to age as a predictor of  ≥ 1 euploid embryo for age
groups < 35 years and 35–37 years, but not  ≥ 38 years. Across all cycles, age (aOR 0.945, 95% CI 0.935–0.956) and
number of embryos (aOR 1.144, 95%CI 1.127–1.162) were associated with live birth per transfer, but AMH was not
(aOR 0.995, 95%CI 0.983–1.008). In the subset of cycles resulting in ≥ 1 euploid embryo for transfer, neither age nor
AMH were associated with live birth.
Conclusions Adjusting for age and number of embryos biopsied, AMH independently predicted likelihood of
obtaining ≥ 1 euploid embryo for transfer in IVF PGT-A cycles. However, neither age nor AMH were predictive of live
birth once a euploid embryo was identified by PGT-A for transfer. This analysis suggests a predictive role of AMH for
oocyte quality (aneuploidy risk), but not live birth per transfer once a euploid embryo is identified following PGT-A.
Keywords Anti-Mullerian hormone (AMH), Pre-implantation genetic testing for aneuploidy (PGT-a), Pre-implantation
genetic diagnosis (PGD), Aneuploidy, In-vitro fertilization, Live birth

Background
Anti-Müllerian hormone (AMH) is a well-established
marker for ovarian reserve and found to correlate with
several outcomes in reproductive medicine, most reliably
oocyte yield during ART cycles [1–4]. AMH, by some
*Correspondence: studies, has been shown to have a moderate association
Howard J. Li with implantation, clinical pregnancy, and live birth for
[email protected]
Dept. of Obstetrics, Gynecology & Reproductive Sciences, Yale School
both fresh and frozen transfers [5–11]. It has also been
of Medicine, 333 Cedar Street, New Haven, CT 06510, U.S.A. shown to correlate with amenorrhea and PCOS sever-
ity, risk of premature ovarian insufficiency, and onset of

© The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​
mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 2 of 11

menopause [12–14]. For its ease of use, low menstrual may be far removed from more clinically meaningful
variability, and high predictive value for IVF outcomes, outcomes such as live birth. Indeed, studies that find
AMH has become the most widely used marker of ovar- significant associations between DOR and morpho-
ian reserve particularly in the IVF setting, supplanting logic makers have commonly found no effect on sub-
other markers of ovarian reserve such as antral follicle sequent pregnancy and live birth rates. At the same
count (AFC), clomiphene citrate testing, day 3 FSH, and time, other studies have reported associations between
inhibin B [4, 15]. DOR and clinical pregnancy, miscarriage, and live birth
While the relationship between AMH and quantita- rates (both per cycle and cumulative), but the impact
tive IVF outcomes such as oocyte and embryo yield is of oocyte quality on these multifactorial outcomes are
well-established and reliably reproducible across multi- often intertwined with effects extrinsic to the individ-
ple studies [15, 16], whether AMH is also predictive of ual oocyte (oocyte yield, maternal factors).
oocyte quality remains unclear. It is known that advanc- Embryo aneuploidy is an objective qualitative factor
ing age is associated with a decline in ovarian reserve that is directly affected by intrinsic oocyte quality with
markers (including AMH), oocyte quantity, and oocyte definitive implications for ultimate IVF outcome, and
quality [17, 18]. Whether these trends are biologically- now increasingly assessed via the use of preimplanta-
coupled or independent processes confounded by age is tion genetic testing for aneuploidy (PGT-A). Although
not well understood, and studies attempting to answer its use and clinical utility remain controversial [28–30],
the “quality vs. quantity” question thus far have had the differentiation between PGT and non-PGT cycles
mixed results [15]. in the SART-CORS data base beginning in 2014 pro-
AMH has been shown to predict oocyte morphology, vides an opportunity to correlate AMH and risk of ane-
fertilization, blastocyst formation, implantation, and uploidy as assessed by PGT.
pregnancy rates [5, 19, 20]. Borges et al. (2017) found Katz-Jaffe et al. (2013) found in a prospective cohort
that, in a sample of 4488 oocytes from 408 patients, AMH of 372 patients that patients with AMH < 1 ng/ml had
was predictive of embryo morphology on day 2 and day a higher percentage of aneuploid embryos as assessed
3, fertilization and blastocyst formation, implantation by PGT, though these effects were not necessarily
rates, and clinical pregnancy rates after adjusting for age independent of age [31]. Jaswa et al. (2021) found that
[5]. At the same time, other reports have found no or par- among 1152 women undergoing IVF, women with
tial associations between ovarian reserve and markers of DOR had 24% reduced odds of a single biopsied blas-
embryo quality [21–25]. Morin et al. (2018) found among tocyst being euploid after adjusting for age [32]. Results
2,103 patients undergoing IVF, adjusted odds of blasto- are highly suggestive that, independent of age, DOR is
cyst development, aneuploidy, and live birth after trans- associated with increased risk of aneuploid embryos.
fer were similar between patients with AMH levels under However, in using a comprehensive definition of DOR
the ­10th percentile and patients with AMH between the via the Bologna criteria, the predictive value of AMH
25-75th percentiles [22]. alone was not reported in this study.
In the last two years, recent studies have also noted With the important exception of two recent smaller
negative or mixed results. A 2022 study of 231 patients studies—one comparing women < 38 years with
found only an association between AMH and Day 5 low and normal AMH [22], and another comparing
embryo quality, but not between AMH and Day 3 embryo women ≤ 40 years with and without physician-reported
quality or clinical pregnancy rate [26]. Another study DOR or poor ovarian response [33]—that reported
found that, among 521 patients, AMH was predictive of no association between DOR and aneuploidy rates as
oocytes retrieved, but not predictive of obtaining “good determined by PGT, the evidence thus far suggests that,
quality” embryos [27]. In an analysis of 492 IVF/ICSI at least with the specific concern of embryo ploidy,
cycles, patients with low AMH (< 1.1 ng/ml) had greater DOR appears to be associated with increased risk of
rates of cycle cancelation, fewer oocytes, and slightly aneuploidy. However, larger scale studies are needed to
lower rates of Grade I-II embryo formation, but no differ- further investigate the association between aneuploidy
ences in fertilization, blastocyst formation, implantation risk and DOR, especially as determined by AMH, now
rates, as well rates of miscarriage and livebirth. the most contemporary clinical marker of DOR. Such
The wide heterogeneity of these study results may be findings would have important implications for the
due in part to varying criteria used for diagnosing DOR counseling and management of patients with DOR,
and multiple metrics for assessing oocyte quality. Vari- particularly in identifying which patients would benefit
ables most intrinsic to oocyte quality (fertilization and from PGT-A.
blastocyst formation, oocyte and embryo morphology)
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 3 of 11

Methods patients’ charts. In 2021, records for 1,945 cycles at 33

Data from the Society for Assisted Reproductive Tech- clinics were randomly selected for full validation, along
nology Clinic Outcome Reporting System (SART-CORS) with 262 fertility preservation cycles selected for partial
database, 2014–2016, were analyzed. Data were collected validation. Nine out of ten data fields selected for valida-
through voluntary submission, verified by SART, and tion were found to have discrepancy rates of ≤ 5%. The
reported to the Centers for Disease Control and Pre- exception was the diagnosis field, which, depending on
vention (CDC) in compliance with the Fertility Clinic the diagnosis, had a discrepancy rate between 0.7% and
Success Rate and Certification Act of 1992 (Public Law 9.1%.
102-493). SART maintains HIPAA-compliant business Cycles were included in the analysis if PGT was per-
associates agreements with reporting clinics. In 2004, formed on all embryos, and only cycles utilizing PGT-A
following a contract change with the CDC, SART gained were included. Cycles utilizing other PGT applications
access to the SART CORS data system for the purposes (PGT-M, PGT-HLA) were excluded. Of these, cycles
of conducting research. SART requests the following resulting in the following outcomes were selected for
information be provided in all studies using the analy- a case–control study design: 1. No transfer attempted
sis and publication of SARTCORS data. During 2019, due to no normal embryos after PGT-A, or 2. Trans-
81% of clinics were SART members reporting 90% of all fer attempted (of a presumed euploid embryo) follow-
IVF cycles in the United States. The data in the SART ing PGT-A of all embryos. Cycles were only included
CORS are validated annually with some clinics receiv- in the final analysis if the number of embryos biopsied
ing on-site visits for chart review based on an algorithm was recorded, and if an AMH value was available within
for clinic selection. During each visit, data reported by 1 year of the index oocyte retrieval.
the clinic were compared with information recorded in

Fig. 1 Flowchart of included cycles

Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 4 of 11

Demographic variables (age, race), body mass index Differences in categorical variables are compared using
(BMI), AMH, etiology of infertility, and stimulation char- Chi-squared testing. P-values < 0.05 were considered sta-
acteristics (FSH dosage, number of oocytes retrieved, tistically significant. Statistical analyses were performed
embryos cryopreserved) are reported with summary in GraphPad Prism version 9.4.1 (La Jolla, CA) and R
statistics. v3.4.1 (Vienna, Austria).
Distribution of AMH values were compared using
Mann–Whitney testing between cycles resulting in an Results
attempted transfer and cycles resulting in an aborted trans- Cycles included for analysis
fer. Analysis was stratified by number of embryos biop- Of 533,463 IVF cycles in the SART-CORS database (2014–
sied (1–2, 3–4, 5–6, and ≥ 7 embryos) and by age (< 35, 2016), there were 29,826 primary autologous cycles utiliz-
35–37, ≥ 38 years at time of index cycle start). Likelihood ing PGT-A for all embryos. Of these, 3,809 cycles resulted
of ≥ 1 euploid embryo for transfer and live birth following in a canceled transfer due no normal embryos after PGT-A
PGT-A were modeled with multivariable logistic regression and 16,401 cycles resulted in an attempted transfer of a pre-
using age, AMH, and number of embryos biopsied as inde- sumed euploid embryo after PGT-A. After excluding cycles
pendent variables. Separate models were also fitted for each without the number of embryos biopsied documented or
of 3 age strata: < 35, 35–37, ≥ 38 years at time of cycle start. without an AMH value within 1 year of index cycle start,
Comparative model analysis was performed using Likeli- 10,878 cycles representing 10,020 unique patients were
hood Ratio (LR) testing for nested models, and Akaike included in the final analysis: 2,100 cycles with aborted
information criterion (AIC) for non-nested models. transfers due to no normal embryos after PGT-A, and
Data are expressed as mean with standard deviation 8,778 cycles with an attempted transfer after PGT-A. There
(SD) or median with interquartile range (IQR). Differ- were 4,893 live births (55.8% of transfer cycles). Flowchart
ences between groups in continuous variables are com- of included and excluded cycles is show in Fig. 1. Baseline
pared using Student’s T or Mann–Whitney testing. cycle characteristics are summarized in Table 1.

Table 1 Baseline and stimulation cycle characteristics

No normal embryos (n = 2100) ≥ 1 presumed euploid embryo
(n = 8778)

Age (years) 40.0 (3.2) 37.0 (3.7) P < 0.001

AMH (ng/ml) 1.9 (2.4) 3.5 (3.5) P < 0.001
BMI (kg/m2) 24.9 (5.2) 24.9 (5.3) P = 0.58
Race P < 0.001
White 798 (38.0%) 3382 (38.5%)
American Indian/Alaskan Native 1 (0.05%) 8 (0.1%)
Asian 266 (12.7%) 938 (10.7%)
Black/ African American 84 (4.0%) 246 (2.8%)
Hispanic/Latina 99 (4.7%) 331 (3.8%)
Hawaiian/ Pacific Islander 4 (0.2%) 9 (0.1%)
Multiracial 11 (0.5%) 49 (0.6%)
Unknown 837 (39.9%) 3815 (43.5%)
Infertility etiology
Male factor 617 (29.4%) 2640 (30.1%) P = 0.49
Endometriosis 139 (6.6%) 596 (6.8%) P = 0.82
PCOS 81 (3.9%) 945 (10.8%) P < 0.001
DOR 1289 (61.4%) 2908 (33.1%) P < 0.001
Tubal factor 200 (9.5%) 859 (9.8%) P = 0.75
Uterine factor 283 (13.5%) 833 (9.5%) P < 0.001
Unexplained 116 (5.5%) 978 (11.1%) P < 0.001
Stimulation characteristics
FSH dosage (IU) 4406 (1549) 3669 (1577) P < 0.001
No. eggs retrieved 10.1 (6.7) 16.4 (9.0) P < 0.001
No. embryos cryopreserved 2.5 (1.7) 4.6 (2.4) P < 0.001
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 5 of 11

Table 2 Median AMH of cycles resulting in no normal embryos vs. ≥ 1 euploid embryo after PGT-A
Embryos No normal embryos ≥ 1 euploid embryo Median AMH (No normal Median AMH (≥ 1 euploid P-value
(n) (n) embryos) embryo)

a. All cycles
1—2 1321 1919 1.2 (0.6—2.0) 1.5 (0.8—2.7) 5.86E-13
3—4 522 2463 1.6 (0.9—2.8) 2.1 (1.2—3.5) 1.02E-10
5—6 175 1848 2.3 (1.2—3.6) 2.7 (1.6—4.3) 1.71E-03
7+ 82 2548 2.9 (1.6—5.0) 4.0 (2.6—6.2) 7.71E-05
b. Age < 35
1—2 63 287 1.6 (0.8—2.7) 2.0 (1.0—3.7) 0.03
3—4 32 461 2.0 (1.3—3.3) 2.9 (1.6—5.2) 0.02
5—6 15 441 4.0 (2.5—6.0) 3.2 (1.9—5.4) 0.32
7+ 7 796 5.1 (3.3—7.5) 4.5 (3.1—7.3) 0.86
c. Age 35–37
1—2 162 480 1.2 (0.6—2.3) 1.6 (0.8—2.8) 5.34E-03
3—4 66 682 1.6 (1.1—2.5) 2.2 (1.3—3.7) 6.62E-03
5—6 14 522 2.2 (1.2—3.9) 2.7 (1.7—4.4) 0.384
7+ 13 735 1.8 (1.2—2.8) 4.0 (2.5—6.2) 1.36E-04
d. Age ≥ 38
1—2 1096 1152 1.1 (0.6—2.0) 1.4 (0.7—2.3) 1.28E-05
3—4 424 1320 1.6 (0.9—2.8) 1.8 (1.1—2.9) 1.42E-03
5—6 146 885 2.2 (1.1—3.5) 2.5 (1.5—3.9) 0.014
7+ 62 1017 3.0 (1.7—5.1) 3.7 (2.3—5.6) 0.054

Cycles resulting in no euploid embryos are associated and ≥ 38 years (aOR 0.710; 95% CI 0.685–0.736), but not
with lower AMH values for < 35 years (aOR 1.040; 95% CI 0.953–1.130).
Stratifying by number of embryos biopsied (1–2, 3–4,
5–6, ≥ 7), AMH levels of cycles with ≥ 1 euploid embryo Comparative Model Analysis: Age and AMH as predictors
for attempted transfer were greater than those of cycles of aneuploidy risk
with no normal embryos (Table 2a), P < 0.002 for each To compare the incremental predictive value of AMH
stratum of embryos biopsied. Further stratification by age for aneuploidy risk, a multivariable logistic model of
(< 35, 35–37, ≥ 38 years) was also performed with analo- likelihood of ≥ 1 euploid embryo fitted using age, num-
gous results, though differences in AMH between cycles ber of embryos biopsied, and AMH was compared with
resulting in canceled transfer due to no euploid embryos a model incorporating only age and number of embryos
vs. ≥ 1 euploid embryo for attempted transfer were no biopsied (without AMH). In Likelihood Ratio (LR)
longer significant for select strata due to diminished cell testing, the addition of AMH significantly improved
size (Table 2b-d). model performance for all age groups: age < 35 years
(P = 0.034, LR 4.495, AUC​ 1 = 0.806, AUC​2 = 0.800),
Multivariable logistic regression of AMH and likelihood 35–37 years (P = 0.010, LR 6.671, AUC​1 = 0.807, AUC​
of obtaining ≥ 1 euploid embryo for transfer 2 = 0.805), and ≥ 38 years (P = 0.002, LR 10, AUC​
Multivariable logistic models for likelihood of ≥ 1 euploid 1 = 0.791, AUC​2 = 0.790), and all cycles combined
embryo were fitted for each of 3 age groups: < 35, 35–37, (P < 0.0001, LR 15.97, AUC​1 = 0.829, AUC​2 = 0.828).
and ≥ 38 years (Table 3). Adjusting for age and number To compare the relative predictive values of age and
of embryos biopsied, AMH was a significant independent AMH for aneuploidy risk, a multivariable logistical model
predictor of ≥ 1 euploid embryo for all age groups: < 35 of likelihood of ≥ 1 euploid embryo fitted using number of
yrs (aOR 1.074; 95%CI 1.005–1.163), 35–37 years (aOR embryos biopsied and AMH (but not age) was compared
1.085; 95%CI 1.018–1.165) and ≥ 38 years (aOR 1.055; with a model incorporating number of embryos biopsied
95%CI 1.020–1.093). Age was a predictor of ≥ 1 euploid and age (but not AMH) using Akaike information crite-
embryo for 35–37 years (aOR 0.813; 95% CI 0.679–0.969) rion (AIC). Across all cycles, age was superior to AMH as
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 6 of 11

Table 3 Multivariable logistical regression

Model comparison (LR test) Model comparison (AICc test)

Model 1: Age, AMH, Embryos Model 1: AMH, Embryos
Model 2: Age, Embryos Model 2: Age, Embryos
Age < 35
Variable Estimate 95% CI Model 1 ROC AUC​ 0.806 Model 1 ROC AUC​ 0.805
β1 Age 1.040 0.9526 to 1.130 Model 2 ROC AUC​ 0.800 Model 2 ROC AUC​ 0.800
β2 AMH 1.074 1.005 to 1.163 Preferred model Model 1 Preferred model Model 1
β3 Embryos 1.607 1.449 to 1.799 Likelihood ratio 4.495 Model 1 probability 86.42%
P-value P = 0.0340 Model 2 probability 13.58%
ΔAICc -3.702
Age 35–37
Variable Estimate 95% CI Model 1 ROC AUC​ 0.807 Model 1 ROC AUC​ 0.807
β1 Age 0.813 0.679 to 0.969 Model 2 ROC AUC​ 0.805 Model 2 ROC AUC​ 0.805
β2 AMH 1.085 1.018 to 1.165 Preferred model Model 1 Preferred model Model 1
β3 Embryos 1.729 1.584 to 1.897 Likelihood ratio 6.671 Model 1 probability 66.24%
P-value P = 0.0098 Model 2 probability 33.76%
ΔAICc -1.348
Age ≥ 38
Variable Estimate 95% CI Model 1 ROC AUC​ 0.791 Model 1 ROC AUC​ 0.754
β1 Age 0.710 0.686 to 0.736 Model 2 ROC AUC​ 0.790 Model 2 ROC AUC​ 0.790
β2 AMH 1.055 1.020 to 1.093 Preferred model Model 1 Preferred model Model 2
β3 Embryos 1.500 1.447 to 1.558 Likelihood ratio 10 Model 1 probability < 0.01%
P-value P = 0.0016 Model 2 probability > 99.99%
ΔAICc 382.1
All cycles
Variable Estimate 95% CI Model 1 ROC AUC​ 0.829 Model 1 ROC AUC​ 0.789
β1 Age 0.790 0.774 to 0.805 Model 2 ROC AUC​ 0.828 Model 2 ROC AUC​ 0.828
β2 AMH 1.056 1.027 to 1.087 Preferred model Model 1 Preferred model Model 2
β3 Embryos 1.543 1.495 to 1.595 Likelihood ratio 15.97 Model 1 probability < 0.01%
P-value P < 0.0001 Model 2 probability > 99.99%
ΔAICc 691.7

a predictor of ≥ 1 euploid embryo (ΔAICc =  + 691.7, AUC​ Discussion

1 = 0.789, AUC​2 = 0.828). Stratifying by age, AMH was Using a large, national, standardized, multicenter data-
superior to age for age groups < 35 years (ΔAICc = -3.70, base (SART-CORS), these data demonstrate that AMH
AUC​1 = 0.805, AUC​2 = 0.800) and 35–37 years predicts the likelihood of obtaining ≥ 1 euploid embryo
(ΔAICc = -1.35, AUC​1 = 0.807, AUC​2 = 0.805) but age in IVF PGT-A cycles independent of age and number of
was superior to AMH in the subset of women ≥ 38 years embryos biopsied. When directly comparing the predic-
(ΔAICc =  + 382.1, AUC​1 = 0.754, AUC​2 = 0.790). tive values of age or AMH, AMH was a superior predictor
of aneuploidy for patients < 38 years, with age being far
Multiple logistic regression of AMH and live birth more predictive of aneuploidy risk in patients ≥ 38 years.
Across all cycles, age (aOR 0.945, 95%CI 0.935–0.956) Across all cycles, age, but not AMH, was predictive of
and number of embryos (aOR 1.144, 95%CI 1.127–1.162) live birth; however, in the subset of cycles for which ≥ 1
were associated with live birth per transfer, but not AMH euploid embryo was obtained for transfer, neither age nor
(aOR 0.995, 95%CI 0.983–1.008) (Table 4). In the subset AMH predicted live birth. In other words, once a euploid
of cycles resulting in ≥ 1 euploid embryo for transfer, nei- embryo was identified by PGT-A, its chance of successful
ther age (aOR 0.994, 95%CI 0.983–1.006) nor AMH (aOR implantation and progression to live birth was independ-
1.006, 95%CI 0.994–1.019) were associated with live ent of age or AMH.
birth. In this subset, the fitted model incorporating age This study’s greatest strength is its large sample size.
and AMH as predictors was not predictive of live birth With 10,778 included cycles, the sample size of this study
(AUC 0.515, 95%CI 0.503–0.527). is 1–2 orders of magnitude greater than other recent
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 7 of 11

Table 4 AMH and Live Birth

Model comparison (LR test) Model comparison (AICc test)

All cycles Model 1: Age, AMH, Embryos Model 1: AMH, Embryos
Model 2: Age, Embryos Model 2: Age, Embryos
Variable Estimate 95% CI Model 1 ROC AUC​ 0.641 Model 1 ROC AUC​ 0.632
β1 Age 0.945 0.935 to 0.956 Model 2 ROC AUC​ 0.642 Model 2 ROC AUC​ 0.642
β2 AMH 0.995 0.983 to 1.008 Preferred model Model 2 Preferred model Model 2
β3 Embryos 1.144 1.127 to 1.162 Likelihood ratio 0.5256 Model 1 probability < 0.01%
P-value P = 0.469 Model 2 probability > 99.99%
ΔAICc 103.9
Model comparison (LR test) Model comparison (AICc test)
Given ≥ 1 euploid embryo Model 1: Age, AMH Model 1: AMH
Model 2: Age Model 2: Age
Variable Estimate 95% CI Model 1 ROC AUC​ 0.515 Model 1 ROC AUC​ 0.512
β1 Age 0.994 0.983 to 1.006 Model 2 ROC AUC​ 0.509 Model 2 ROC AUC​ 0.509
β2 AMH 1.006 0.994 to 1.019 Preferred model Model 2 Preferred model Model 2
Likelihood ratio 0.8664 Model 1 probability 49.47%
P-value P = 0.352 Model 2 probability 50.53%
ΔAICc 0.0427

studies investigating the relationship between diminished whether for embryo banking, oocyte banking, and/or
ovarian reserve and oocyte quality including embryo frozen oocyte cycles, and thus these cycles were also
aneuploidy [22–27, 31, 32]. Several limitations are excluded. Finally, only 2,100 of 3,809 cycles (55.1%) that
acknowledged. While our primary concern was to inves- resulted in aborted transfers due to no normal embryos
tigate the association of AMH and oocyte quality using following PGT-A, and only 8,778 of 16,401 cycles (53.5%)
embryo aneuploidy as a proxy, the genetic results of indi- that resulted in a documented transfer attempt follow-
vidual embryos were not available, and the dichotomous ing PGT-A had sufficient linked index cycle information
outcomes of obtaining no normal embryos following (such as AMH and number of embryos biopsied) that
PGT-A versus obtaining ≥ 1 presumed euploid embryo would allow for analysis. With nearly half of these cycles
for transfer, controlling for the number of embryos biop- excluded due to incomplete data, data quality remains a
sied, was used as an imperfect proxy of aneuploidy risk. concern that is partially mitigated by our stringent inclu-
Specifically, the outcome of “no normal embryos fol- sion criteria.
lowing PGT” does not sufficiently discriminate between Our results support the recent findings of Jaswa et al.
true euploid, aneuploid, and mosaic results for which (2021) which showed convincing evidence of an associa-
the decision to proceed with embryo transfer may vary tion between DOR, as determined by Bologna criteria,
between patients and institutions [34, 35]. However, we and aneuploidy risk that was independent of age, though
assume that the vast majority of PGT-tested embryo the predictive value of AMH alone was not reported [32].
transfers in the 2014–2016 SART dataset were euploid Interestingly, our results extend this relationship between
rather than mosaic or aneuploid embryos [36, 37]. ovarian reserve and aneuploidy, showing that AMH
Variation and quality of data reporting to the SART- appears to continue to have a predictive range even at
CORS database is another potential limitation. Of 51,273 values above the DOR range, i.e. values above 1.0 ng/ml
cycles utilizing PGT-A for all embryos, only 10,778 cycles (Fig. 2). This suggests that AMH as a quantitative marker
were included in the final analysis due to incomplete or has clinical utility beyond dichotomizing patients into
inconsistent data. It is likely that several cycles initiated groups with and without DOR [3, 9]. By directly assess-
with intent for PGT may have been categorized as PGT ing the predictive role of AMH, currently the most widely
cycles in the SART-CORS database, regardless of whether used marker of ovarian reserve, these data have greater
PGT was ultimately performed. For example, a small frac- clinical applicability to contemporary IVF practice, as
tion of cycles specified as “PGT-A cycles” (1,205 cycles) well as to a broader population of patients (with and
resulted in fresh transfers and were excluded in the final without a diagnosis of DOR).
analysis. The outcome of PGT-A testing (or whether PGT Conversely, our results are not consistent with recent
was ultimately performed) also remained ambiguous studies by Fouks et al. (2021) and Morin et al. (2018)
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 8 of 11

Fig. 2 Risk of no normal embryos after PGT-A by AMH and Age

[22, 33]. However, several relevant factors may explain in each sample; across the entire study sample, the ane-
these discrepant results. First, both studies utilized sig- uploidy rate per embryo biopsied was 29–30% for Morin
nificantly smaller sample sizes; while our study captures et al. and 39–42% for Fouks et al. While a per-embryo
10,878 PGT-A cycles (for 10,020 unique patients), Fouks aneuploidy rate could not be directly calculated in our
et al. and Morin et al. report data on 1,150 women (383 sample, we estimate using numerical methods based on a
with DOR and 767 propensity-matched controls without 19.3% rate of no normal embryos and the distribution of
DOR) and 2,103 women (345 with DOR and 1758 with- embryos biopsied in our sample that the aneuploidy rate
out DOR), respectively. Second, both studies exclude per embryo would be approximately 58.6%—consider-
women from advanced age groups; Morin et al. included ably higher than that of both Morin et al. and Fouks et al.
women with < 38 years old and Fouks et al. included Importantly, both studies also model DOR as a dichoto-
women < 40 years old. The combination of smaller sam- mous variable by either parsing the study population by
ple sizes and exclusion of patients at greatest risk of ane- percentile (AMH < ­10th percentile vs. 25-75th percentile
uploidy may reduce the power of each study to detect in Morin et al.), or by presence or absence of a physician-
differences in the effect of AMH of aneuploidy risk. This reported diagnosis of DOR (which the authors validated
is overall reflected in the lower incidence of aneuploidy by a maximum AMH cut-off of 1.1 ng/ml). By modeling
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 9 of 11

AMH as a continuous variable over the entire range rep- suggest that AMH values may play an informative role,
resented in our study population, our data suggest persis- particularly in women 35–37 years old.
tent effects of AMH variation on aneuploidy risks even at
values well above 1.1 ng/ml. Thus, the dichotomization of
Conclusions
patients into groups with and without DOR based on low
While AMH is a predictor of live birth for non-PGT IVF
AMH cut-offs may cause significant effects of AMH vari-
cycles, it is unknown if this is due solely to quantitative
ation above traditionally low cut-offs to evade statistical
factors or if qualitative factors contribute. Consistent with
detection.
other recent studies investigating the relationship between
Comparing the relative predictive abilities of AMH and
diminished ovarian reserve, oocyte quality, and specifically
age for aneuploidy may have important clinical implica-
embryo aneuploidy, this study suggests that AMH indepen-
tions, especially for identifying patients who may benefit
dently predicts likelihood of obtaining ≥ 1 euploid embryo
from PGT-A testing. In patients < 35 years, AMH was
for transfer in IVF PGT-A cycles after adjusting for age and
superior to age in predicting risk of no euploid embryos,
number of embryos biopsied. However, neither age nor
with age being non-predictive. In patients ≥ 35 years,
AMH are predictive of live birth per transfer. This analysis
both age and AMH significantly predict risk of no
further suggests a predictive role for AMH on oocyte qual-
euploid embryos, though AMH was the superior pre-
ity (aneuploidy risk), but not live birth per transfer once a
dictor only in patients 35–37 years, and age was by far
euploid embryo is identified following PGT-A.
the superior predictor in patients ≥ 38 years. Indeed, it is
well-established that the relationship between maternal
age and aneuploidy strengthens at advanced ages [23, 38]. Abbreviations
Notably, our group has previously shown that, in non- AICc Corrected Akaike information criterion
AMH Anti-Müllerian Hormone
PGT cycles from the SART-CORS database, AMH was an aOR Adjusted Odds Ratio
independent predictor of live birth in both fresh and fro- CI Confidence Interval
zen-thawed transfer cycles when controlling for multiple DOR Diminished Ovarian Reserve
FET Frozen Embryo Transfer
confounders, including age, BMI, race, day of transfer, and FSH Follicle Stimulating Hormone
number of embryos transferred [39]. Taken together, this ICSI Intracytoplasmic Sperm Injection
present study’s finding that AMH predicts aneuploidy but IQR Interquartile Range
IVF In-vitro Fertilization
not live birth in PGT-A cycles supports the notion that LR Likelihood Ratio
the association between AMH and live birth in non-PGT PCOS Polycystic Ovarian Syndrome
cycles is not only due to a quantitative effect, but also a PGT-A Preimplantation Genetic Testing for Aneuploid
SART-CORS Society for Assisted Reproductive Technology Clinic Outcome
qualitative effect as reflected by the association between Reporting System
AMH and likelihood of embryo aneuploidy. SD Standard Deviation
The finding that both age and AMH appear to be irrel-
Acknowledgements
evant predictors of live birth once a euploid embryo is As members of SART, the authors thank its membership for providing clinical
identified is consistent with prior studies [22, 32, 33]. It information to the SARTCORS database for use by patients and researchers.
also suggests that embryo aneuploidy (an outcome pre- Without the efforts of our membership this research would not be possi-
ble. The authors also thank Dr. Xiao Xu for helpful discussions and statistical
sumed to be eliminated by normal PGT-A testing) is a guidance.
significant detrimental factor in live birth rates following
transfer of untested embryos for patients of advanced age Authors’ contributions
HJL, DBS, and RT designed and conceived the study. HJL and RT performed
(> 35–38 years), and possibly for patients with diminished the initial data collection and curation. HJL performed the initial data analysis
ovarian reserve as determined by AMH. This is consist- and wrote the manuscript draft. HJL, DBS, and RT reviewed and edited the final
ent with evidence that patients of advanced age, particu- manuscript. All authors read and approved the final version of the manuscript.
larly those considering single embryo transfers to reduce Funding
the risk of multiple pregnancies and its attendant risks, There are no external funding sources to report.
may stand to benefit the most from PGT-A [28, 40, 41].
Availability of data and materials
Translated clinically, especially in the setting of single This study uses data from the SART CORS database, available by request to
embryo transfers, patients ≥ 38 years may benefit from SART CORS members via the SART research portal.
PGT-A testing, while patients 35–37 years may poten-
tially benefit in the setting of diminished ovarian reserve. Declarations
While further studies considering risk–benefit and cost-
Ethics approval and consent to participate
effectiveness analyses are needed to determine which This study was approved by the SART research committee and exempted
patients are likely candidates for PGT-A, our findings from review by the Yale School of Medicine IRB.
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 10 of 11

Consent for publication 19. De Conto E, Genro VK, da Silva DS, Chapon R de CB, Cunha-Filho JSL. AMH
Not applicable. as a Prognostic Factor for Blastocyst Development. JBRA Assist Reprod.
2015;19:131–4.
Competing interests 20. Majumder K, Gelbaya TA, Laing I, Nardo LG. The use of anti-Müllerian
DBS receives royalties from license between Rutgers Medical School and MGH hormone and antral follicle count to predict the potential of oocytes and
with Beckman Coulter for patent of AMH for determining ovarian reserve. embryos. Eur J Obstet Gynecol Reprod Biol. 2010;150:166–70.
21. Smeenk JMJ, Sweep FCGJ, Zielhuis GA, Kremer JAM, Thomas CMG, Braat
DDM. Antimüllerian hormone predicts ovarian responsiveness, but not
Received: 1 December 2022 Accepted: 21 January 2023 embryo quality or pregnancy, after in vitro fertilization or intracyoplasmic
sperm injection. Fertil Steril. 2007;87:223–6.
22. Morin SJ, Patounakis G, Juneau CR, Neal SA, Scott RT, Seli E. Diminished
ovarian reserve and poor response to stimulation in patients <38 years old:
a quantitative but not qualitative reduction in performance. Hum Reprod.
References 2018;33:1489–98.
1. Seifer DB, MacLaughlin DT, Christian BP, Feng B, Shelden RM. Early fol- 23. Thum M-Y, Abdalla HI, Taylor D. Relationship between women’s age and
licular serum müllerian-inhibiting substance levels are associated with basal follicle-stimulating hormone levels with aneuploidy risk in in vitro
ovarian response during assisted reproductive technology cycles. Fertil fertilization treatment. Fertil Steril. 2008;90:315–21.
Steril. 2002;77:468–71. 24. Silberstein T, MacLaughlin DT, Shai I, Trimarchi JR, Lambert-Messerlian G,
2. La Marca A, Sighinolfi G, Radi D, Argento C, Baraldi E, Artenisio AC, et al. Anti- Seifer DB, et al. Mullerian inhibiting substance levels at the time of HCG
Mullerian hormone (AMH) as a predictive marker in assisted reproductive administration in IVF cycles predict both ovarian reserve and embryo
technology (ART). Hum Reprod Update. 2010;16:113–30. morphology. Hum Reprod. 2006;21:159–63.
3. Tal R, Seifer DB. Ovarian reserve testing: a user’s guide. Am J Obstet Gynecol. 25. Scantamburlo VM, von Linsingen R, Centa LJR, Toso KFD, Scaraboto D,
2017;217:129–40. Araujo Júnior E, et al. Association between decreased ovarian reserve and
4. Practice Committee of the American Society for Reproductive Medicine. poor oocyte quality. Obstet Gynecol Sci. 2021;64:532–9.
Electronic address: [email protected], Practice Committee of the American 26. Morales HSG, López GGP, Cortés DV, Torres GCR, Hernández HS, Guiot ML,
Society for Reproductive Medicine. Testing and interpreting measures of et al. Evaluation of the Anti-Müllerian Hormone and its Association with
ovarian reserve: a committee opinion. Fertil Steril. 2020;114:1151–7. Embryo Quality in Advanced Reproductive Treatments in a Latin American
5. Borges E, Braga DPAF, Setti A, Figueira R de C, Iaconelli A. The predictive Population. JBRA Assist Reprod. 2022;26:50–2.
value of serum concentrations of anti-Müllerian hormone for oocyte quality, 27. Sun T-C, Chen X, Shi C, Tian L, Zhou S-J. The Predictive Levels of Serum Anti-
fertilization, and implantation. JBRA Assist Reprod. 2017;21:176–82. Müllerian Hormone and the Combined Index of the Number of Retrieved
6. Li X-L, Huang R, Fang C, Liang X-Y. Basal Serum Anti-Müllerian Hormone Oocytes and Good-Quality Embryos in Advanced-Age Infertile Women. Int J
Level as a Predictor of Clinical Outcomes in Freezing-all Embryo Transfer Endocrinol. 2022;2022:4224417.

A, et al. PGT-A: who and when? Α systematic review and network meta-
Program. Curr Med Sci. 2018;38:861–7. 28. Simopoulou M, Sfakianoudis K, Maziotis E, Tsioulou P, Grigoriadis S, Rapani
7. Tal R, Seifer DB, Tal R, Granger E, Wantman E, Tal O. AMH Highly Correlates
With Cumulative Live Birth Rate in Women with Diminished Ovarian analysis of RCTs. J Assist Reprod Genet. 2021;38:1939–57.
Reserve Independent of Age. J Clin Endocrinol Metab. 2021;106:2754–66. 29. Bedrick BS, Tipping AD, Nickel KB, Riley JK, Jain T, Jungheim ES. State-
8. Iliodromiti S, Kelsey TW, Wu O, Anderson RA, Nelson SM. The predictive Mandated Insurance Coverage and Preimplantation Genetic Testing in the
accuracy of anti-Müllerian hormone for live birth after assisted conception: a United States. Obstet Gynecol. 2022;139:500–8.
systematic review and meta-analysis of the literature. Hum Reprod Update. 30. Lee M, Lofgren KT, Thomas A, Lanes A, Goldman R, Ginsburg ES, et al. The
2014;20:560–70. cost-effectiveness of preimplantation genetic testing for aneuploidy in the
9. Tal R, Tal O, Seifer BJ, Seifer DB. Antimüllerian hormone as predictor of United States: an analysis of cost and birth outcomes from 158,665 in vitro
implantation and clinical pregnancy after assisted conception: a systematic fertilization cycles. Am J Obstet Gynecol. 2021;225:55.e1-55.e17.
review and meta-analysis. Fertil Steril. 2015;103:119-130.e3. 31. Katz-Jaffe MG, Surrey ES, Minjarez DA, Gustofson RL, Stevens JM, Schoolcraft
10. Seifer DB, Tal O, Wantman E, Edul P, Baker VL. Prognostic indicators of assisted WB. Association of abnormal ovarian reserve parameters with a higher
reproduction technology outcomes of cycles with ultralow serum antimül- incidence of aneuploid blastocysts. Obstet Gynecol. 2013;121:71–7.
lerian hormone: a multivariate analysis of over 5,000 autologous cycles from 32. Jaswa EG, McCulloch CE, Simbulan R, Cedars MI, Rosen MP. Diminished
the Society for Assisted Reproductive Technology Clinic Outcome Report- ovarian reserve is associated with reduced euploid rates via preimplanta-
ing System database for 2012–2013. Fertil Steril. 2016;105:385-393.e3. tion genetic testing for aneuploidy independently from age: evidence
11. Seifer DB. Connecting the dots between oocyte quantity and quality in for concomitant reduction in oocyte quality with quantity. Fertil Steril.
diminished ovarian reserve. Fertil Steril. 2021;115:890. 2021;115:966–73.
12. Tal R, Seifer DB, Khanimov M, Malter HE, Grazi RV, Leader B. Characterization 33. Fouks Y, Penzias A, Neuhausser W, Vaughan D, Sakkas D. A diagnosis of
of women with elevated antimüllerian hormone levels (AMH): correlation of diminished ovarian reserve does not impact embryo aneuploidy or live
AMH with polycystic ovarian syndrome phenotypes and assisted reproduc- birth rates compared to patients with normal ovarian reserve. Fertil Steril.
tive technology outcomes. Am J Obstet Gynecol. 2014;211(59):e1-8. 2022;118:504–12.
13. Bentzen JG, Forman JL, Larsen EC, Pinborg A, Johannsen TH, Schmidt L, et al. 34. Gleicher N, Albertini DF, Barad DH, Homer H, Modi D, Murtinger M, et al.
Maternal menopause as a predictor of anti-Mullerian hormone level and The 2019 PGDIS position statement on transfer of mosaic embryos within a
antral follicle count in daughters during reproductive age. Hum Reprod. context of new information on PGT-A. Reprod Biol Endocrinol. 2020;18:57.
2013;28:247–55. 35. Practice Committee and Genetic Counseling Professional Group (GCPG) of
14. Dólleman M, Depmann M, Eijkemans MJC, Heimensem J, Broer SL, van the American Society for Reproductive Medicine. Electronic address: asrm@
der Stroom EM, et al. Anti-Mullerian hormone is a more accurate predictor asrm.org. Clinical management of mosaic results from preimplantation
of individual time to menopause than mother’s age at menopause. Hum genetic testing for aneuploidy (PGT-A) of blastocysts: a committee opinion.
Reprod. 2014;29:584–91. Fertil Steril. 2020;114:246–54.
15. Kotlyar A, Seifer DB. Anti-Müllerian hormone as a qualitative marker - or just 36. Kim TG, Neblett MF, Shandley LM, Omurtag K, Hipp HS, Kawwass JF.
quantity? Curr Opin Obstet Gynecol. 2020;32:219–26. National mosaic embryo transfer practices: a survey. Am J Obstet Gynecol.
16. Granger E, Tal R. Anti-Müllerian Hormone and Its Predictive Utility in 2018;219:602.e1-602.e7.
Assisted Reproductive Technologies Outcomes. Clin Obstet Gynecol. 37. Abhari S, Kawwass JF. Pregnancy and Neonatal Outcomes after Transfer of
2019;62:238–56. Mosaic Embryos: A Review. J Clin Med. 2021;10:1369.
17. Baird DT, Collins J, Egozcue J, Evers LH, Gianaroli L, Leridon H, et al. Fertility 38. Cuckle H, Morris J. Maternal age in the epidemiology of common autoso-
and ageing. Hum Reprod Update. 2005;11:261–76. mal trisomies. Prenat Diagn. 2021;41:573–83.
18. Broekmans FJ, Soules MR, Fauser BC. Ovarian aging: mechanisms and clinical 39. Tal R, Seifer DB, Wantman E, Baker V, Tal O. Antimüllerian hormone as a
consequences. Endocr Rev. 2009;30:465–93. predictor of live birth following assisted reproduction: an analysis of 85,062
Li et al. Reproductive Biology and Endocrinology (2023) 21:19 Page 11 of 11

fresh and thawed cycles from the Society for Assisted Reproductive Tech-
nology Clinic Outcome Reporting System database for 2012–2013. Fertil
Steril. 2018;109:258–65.
40. Rubio C, Bellver J, Rodrigo L, Castillón G, Guillén A, Vidal C, et al. In vitro fertili-
zation with preimplantation genetic diagnosis for aneuploidies in advanced
maternal age: a randomized, controlled study. Fertil Steril. 2017;107:1122–9.
41. Ubaldi FM, Capalbo A, Colamaria S, Ferrero S, Maggiulli R, Vajta G, et al.
Reduction of multiple pregnancies in the advanced maternal age popula-
tion after implementation of an elective single embryo transfer policy
coupled with enhanced embryo selection: pre- and post-intervention
study. Hum Reprod. 2015;30:2097–106.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub-
lished maps and institutional affiliations.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission

• thorough peer review by experienced researchers in your field
• rapid publication on acceptance
• support for research data, including large and complex data types
• gold Open Access which fosters wider collaboration and increased citations
• maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions