Antimüllerian hormone as a risk factor for miscarriage in naturally conceived pregnancies

Fecundity is a term that describes the reproductive capacity to both conceive and carry a pregnancy to term ( 1 ). As women age, fecundity declines due to an increase in both time to pregnancy and early pregnancy loss ( 2 ). Concomitant with the decline in fecundity is the decline in oocyte quality and quantity (commonly referred to as ovarian reserve). One measure of oocyte quality is the number of chromosomes it contains, otherwise known as “ploidy”. In oocytes from older women, homologous chromosomes paired during meiosis I have been shown to fail to segregate normally, a process called meiotic non-disjunction ( 3 ). Non-disjunction results in aneuploidy of the mature oocyte and the subsequent embryo, and is thought to be the leading cause of the increased miscarriage rate in women over the age of 35 years ( 4 ). Markers of ovarian reserve, including both serum and ultrasound modalities, have been well-studied in infertile populations and have proven to be efficacious in the quantitative assessment of ovarian reserve ( 5 , 6 ). However, an adequate marker of oocyte quality has yet to be determined. Anti-mϋllerian hormone (AMH) is a hormone produced by the granulosa cells in the pre-antral and early antral follicles. AMH is a marker of oocyte quantity ( 5 ), and declines accordingly with age ( 7 ). AMH has been shown to predict oocyte yield following controlled ovarian hyperstimulation( 8 ) and age at menopause( 9 ). However, whether AMH can be used as a putative marker of oocyte quality, remains uncertain. The ability of AMH to predict conception either with our without assisted reproductive technologies remains controversial and no prior studies have prospectively examined the association between AMH and miscarriage among women without a history of infertility ( 10 , 11 ). With more women electing to defer childbearing, increased inquiry into personal reproductive capacity (fecundity) is becoming common ( 1 ). Having a more robust marker of both oocyte quantity and quality would allow for better counseling in mid to late reproductive aged women. Therefore, we sought to assess the association between AMH and miscarriage in spontaneously conceived pregnancies using a prospective cohort of women. We hypothesized that women with lower AMH would have increased risk of miscarriage, independent of age.

The cohort included observations from 533 women with a positive pregnancy test ( Supplemental Figure 1 ). Eighty-eight percent of women (n=469) had a known pregnancy outcome with 349 (74%) pregnancies resulting in a live singleton or multiple birth, 111 (24%) in miscarriage, and 9 women with ectopic pregnancy, abortion, or stillbirth after 20 weeks. Median AMH for the cohort was 2.9 ng/ml (Interquartile Range (IQR) 1.5, 4.9). The majority of women who conceived were less than 35 years of age (74%), had a normal BMI (55%) and conceived within three cycles of attempt (50%). Over 30% of women were overweight or obese; most were not using hormonal contraceptives in the month prior to study enrollment (66%) and less than 1% had a history of more than two miscarriages ( Table 1 ). In the bivariate analysis, AMH decreased with increasing age (P<0.001) ( Table 1 ). Women who had a history of more than two miscarriages had a lower AMH compared to those who did not (P=0.04). Race, preconception BMI, tobacco, caffeine, alcohol and oral contraceptive use were not significantly associated with AMH ( Table 1 ). However, obese women were more likely to have a miscarriage compared to non-obese women (P=0.01, Table 3 ). After adjusting for age, women with a clinical pregnancy loss had the lowest AMH values and those with a biochemical loss had the highest values (P=0.013, Table 2 ). In unadjusted analyses, women who had a miscarriage were more likely to be older, obese, have a history of recurrent pregnancy loss and have a low AMH ( Table 3 ). Race, tobacco, alcohol, caffeine and oral contraceptive use were not associated with miscarriage in this cohort. After adjusting for age, race, obesity and history of recurrent pregnancy loss, the risk of miscarriage decreased as AMH increased (Risk Ratio (RR) per unit increase in natural log of AMH= 0.83, 95% CI: 0.73, 0.94). Women with significant diminished ovarian reserve (AMH ≤ 0.4ng/mL) had 2.4 times the risk of miscarriage as those with an AMH ≥1ng/mL (Risk Ratio (RR) 2.2; 95% CI 1.5, 3.3). When history of recurrent pregnancy loss was removed from the model, the results did not change (RR 2.2; 95% CI 1.4, 3.3). Women with moderately diminished ovarian reserve (0.4ng/ml<AMH<1.0ng/ml) did not have an increased risk of miscarriage (RR 1.0; 95% CI 0.6, 1.7) compared to women with normal AMH (AMH ≥ 1ng/ml). Results of the subgroup analysis indicated that age did not modify the association between AMH and miscarriage (Pinteraction=0.7). Among women less than 35 years of age, women with an AMH ≤0.4 ng/ml had 2.3 (95% CI: 1.3–4.1) times the risk of miscarriage compared to women with AMH ≥1ng/mL. The estimate was similar among women 35 years of age and older (RR 2.2, 95%CI: 1.1–4.4). Using unadjusted Kaplan Meier curves we examined the risk of miscarriage over time (days since the positive pregnancy test) by AMH category. Women with an AMH of ≤ 0.4 ng/mL had a higher risk of miscarriage over time. ( Figure 1 ; Log Rank P=0.008). Interestingly, the shape of the curves suggests that the increase in risk occurred between 6 and 7 weeks gestational age. The multivariable Cox model adjusting for maternal age, race, obesity and recurrent pregnancy loss suggested that risk of miscarriage showed a non-significant trend to decrease with increasing AMH (Hazard Ratio (HR) per unit increase in natural log of AMH = 0.85, 95% CI: 0.72, 1.01). Women with an AMH ≤ 0.4 ng/ml miscarried at over twice the rate of women with an AMH ≥ 1ng/mL (HR 2.3; 95% CI 1.3, 4.3). Including the 64 women without a known pregnancy outcome did not significantly change the results (HR 2.4, 95% CI 1.3, 4.4).

In this cohort of naturally conceived pregnancies, low AMH was associated with increased risk of miscarriage. Women with an AMH of ≤ 0.4 had over twice the risk of miscarriage with the increased risk primarily associated with clinical pregnancy loss. In this study of women, with no known history of infertility and who conceived naturally, very low AMH (≤ 0.4 ng/ml) was associated with a higher risk of miscarriage. Some retrospective studies have shown that among women with infertility, diminished ovarian reserve may increase the risk of pregnancy loss ( 13 – 15 ). One recent study showed that women over the age of 34 years with an AMH in the 20 th percentile (≤ 1.6ng/mL) had a two-fold increase in miscarriage rate compared to women with an AMH >1.6ng/mL ( 16 ). Additional research has been conducted among women with recurrent pregnancy loss. Atasaver et al showed that women with a history of miscarriage were three times as likely to have a low AMH (≤ 1ngmL) compared to age-matched, fertile controls ( 17 ). Finally, when comparing women with an identified cause of recurrent miscarriage to those with idiopathic miscarriage, median AMH was noted to be significantly lower (median AMH 1.2ng/mL versus 2ng/mL) in those whom an identifiable cause of miscarriage could not be found ( 18 ). In contrast, two other studies have suggested AMH is not associated with a higher rate of miscarriage ( 19 – 21 ). A secondary analysis of the EAGER trial, which included women with at least one prior pregnancy loss attempting to conceive naturally, suggested women with a low AMH value (<1ng/mL) did not have an increased risk of pregnancy loss ( 19 ). Thus, it possible, that AMH is not a predictor of recurrent pregnancy loss, as suggested by Zarek at al ., but may play a role in spontaneous loss, as suggested by our cohort. In addition, Zarek et al , used pregnancy loss as their outcome, including pre-embryonic or embryonic loss, fetal loss, stillbirth, ectopic or pregnancy of unknown location, while our study focuses specifically on miscarriage. We defined miscarriage as a loss prior to 20 weeks gestational age and excluded any loss that could be reasonably attributed to another mechanism (e.g. ectopic pregnancy and history of pelvic inflammatory disease). A second study evaluating AMH and pregnancy loss in women less than 35 years found similar rates of miscarriage in women with AMH ≤0.5ng/mL compared to women with AMH >0.5ng/mL ( 20 ). However, this cohort included women with infertility undergoing IVF. AMH may play less of a role following ART, compared to spontaneously conceived pregnancies. AMH may be less correlated with miscarriages following ART, compared to spontaneously conceived pregnancies. In our study, women experiencing a biochemical loss had higher AMH values. Prior studies have suggested an increased risk of poor pregnancy outcome including decreased fecundability and increased risk of miscarriage among women with AMH values greater than 2.5 ng/ml ( 10 , 22 , 23 ). Similar trends in AMH and pregnancy outcomes have been observed in women with PCOS, a syndrome traditionally characterized by irregular cycles and hormonal imbalances ( 24 , 25 ). Interestingly, women in our cohort experiencing biochemical losses had higher AMH values and were more likely to have irregular (long) menstrual cycles suggesting an underlying hormonal imbalance like PCOS might be a contributing factor. Recently, Broughton et al. proposed that many of the similarities in poor pregnancy outcomes seen between obese women and those defined by PCOS may be caused by direct effects on oocyte competence leading to downstream effects on endometrial receptivity and embryo implantation( 26 ). An alternative explanation for our findings is that the observation of more biochemical pregnancies was due to our standardized pregnancy testing. We may have simply detected pregnancies that otherwise would not have been detected in women with irregular or long cycles given that women were instructed to test every three days starting at cycle day 28. Though studies have previously observed a relationship between AMH and miscarriage in select populations, our study represents one of the largest prospective cohorts to examine the association between AMH and spontaneous pregnancies in women without a history of infertility. Furthermore, the prospective study design allowed for characterization of the timing of miscarriage. Limitations of our study include timing of assessment of AMH and the reliance of patient interpretation of the definition of ‘miscarriage’. AMH was obtained at study enrollment. Women may have varied in the time from AMH assessment to pregnancy detection. However, all pregnancies were conceived within one year of enrollment. AMH values have been found to be relatively stable over a year (decline of 5.6% per year)( 27 ). In addition, the etiology of the miscarriage is unknown. Karyotypes were not performed on all of the miscarriages. Finally, our data reflects a population of older reproductive age women with increased risk of miscarriage only associated with severely diminished ovarian reserve (DOR) and may not necessarily be directly applicable to “younger” women or those with only mild to moderate DOR. In conclusion, our study suggests that significantly diminished AMH is associated with increased risk of miscarriage in spontaneously conceived pregnancies. While AMH may not be an independent marker of fecundability, low AMH or diminished ovarian reserve may be a marker of reduced reproductive potential due to its association with pregnancy loss. Further research is needed to identify the biological mechanism underlying this association.

Time to Conceive (TTC) was a prospective, time-to-pregnancy cohort study conducted between 2008 and 2016 that enrolled women between the ages of 30 and 44 years who were trying to conceive naturally. The cohort was constructed and previously used to examine the association between biomarkers of ovarian reserve and fecundability( 12 ). Women were recruited from the Chapel Hill-Raleigh- Durham area of North Carolina. Eligible women had been attempting to conceive for 3 months or less (self-reported). Women were excluded if they reported a history of infertility, polycystic ovarian syndrome (PCOS), endometriosis, had a partner with infertility, were currently breastfeeding or did not speak English. Institutional Review Board approval was obtained for this research. In the first menstrual cycle after enrollment participants provided a blood sample on second, third or fourth menstrual day. Serum samples were stored at −30°C until analysis. They were assayed using sensitive and specific assays for AMH (Ultrasensitive AMH ELISA, Ansh). Interassay coefficients of variation ranged from 9–11% (lower limit of detection 0.078 ng/ml). Women completed a baseline questionnaire including information such as demographics, medical history, reproductive history, and lifestyle behaviors such as tobacco, alcohol, and caffeine use. They were given home pregnancy tests (sensitivity of 20 mIU human chorionic gonadotropin (hCG) per mL). Women enrolled before April 2011 were instructed to perform the pregnancy test with missed menses. From April 2011 forward, women were instructed to test starting on menstrual cycle day 28 and every 3 days thereafter until a positive pregnancy test or menses occurred. Participants were asked to notify study staff when they observed a positive pregnancy test. They were scheduled for an endovaginal ultrasound between 6 0/7 and 8 0/7 weeks of gestation to confirm estimated date of delivery and fetal viability. Participants who became pregnant completed a pregnancy outcome report at the end of their pregnancy. For women reporting a miscarriage, the survey queried the date the miscarriage was identified and whether or not dilation and curettage (D&C) was performed. Women, who did not report a pregnancy loss, were contacted between 20 and 24 weeks gestation to confirm continued viability of the pregnancy and to update contact information. Participants who were enrolled prior to April 2011 were not initially required to fill out a pregnancy outcome report, but were contacted later, in 2016, to determine the outcome of their pregnancy. All women with a reported positive pregnancy test were included in the analysis (n=533). For the purposes of our study, miscarriage was defined as a pregnancy loss prior to 20 weeks gestation. Estimated date of delivery was defined by last menstrual period unless there was a greater than two week discrepancy when compared to first trimester ultrasound. If a greater than two-week discrepancy existed, estimated date of delivery was defined by first trimester ultrasound. Miscarriage was categorized into the following groups: biochemical pregnancy (positive pregnancy test followed by a negative pregnancy test or menses ≤ 4 days after first positive pregnancy); early pregnancy loss (negative pregnancy test or menses > 4 days after first test and before the pregnancy ultrasound); clinical pregnancy loss (pregnancies lost after documented viable pregnancy at the pregnancy ultrasound and before 20 weeks gestational age). AMH values below the limit of detection (0.078ng/mL) were assigned a value of the limit of detection divided by the square root of two. AMH was log transformed and age-adjusted, bivariate analyses were conducted to examine relationships between AMH and covariates and using the Student T test and ANOVA for continuous variables and χ 2 for discrete variables. Age-adjusted AMH values were compared among the types of miscarriage (biochemical pregnancy, early pregnancy loss, clinical pregnancy loss). Biochemical losses did not follow trends in the rest of our data. Participants who reported a biochemical loss (n=9) were more likely to be overweight or obese and have an elevated AMH (median 4, range 1.8, 26.6). In addition, these women tended to have a long menstrual cycle length (median 40 days, range 27, 101). Bivariate analyses were conducted to assess the relationship between 1) AMH and pregnancy outcome (miscarriage or live birth) and 2) covariates and pregnancy outcome using the Student’s T-test and ANOVA for continuous variables and chi-square tests for discrete variables. Multivariable binomial regression was used to calculate risk ratios (RR) and 95% confidence intervals (CI) for the association between AMH and clinical miscarriage (compared with live birth). This regression model excludes biochemical pregnancies (n=9), pregnancies that end in stillbirth (n=2), or pregnancies that end in induced abortion (n=6). AMH was analyzed both as a linear variable (lnAMH) and categorical: AMH ≤ 0.4 (severe diminished ovarian reserve), AMH >0.4 and <1 (compromised ovarian reserve); AMH ≥ 1 (normal ovarian reserve). To adjust for potential confounders, the model incorporated covariates predictive of miscarriage including age, race, obesity and history of recurrent pregnancy loss (history of 3 or more miscarriages) based on bivariate analyses (P ≤ 0.1). Covariates were categorized as follows: maternal age at study enrollment (40 years); white race (yes / no); obese (BMI ≥ 30) (yes / no); and history of recurrent pregnancy loss (yes / no). A subgroup analysis was performed to determine if the relationship between AMH and miscarriage differed for women who were < 35 years of age compared to women who were 35 years of age and older. A Cox model including all women who had a pregnancy outcome (N=469) was created to explore the association between AMH and time to miscarriage. Time (in days) was defined as follows: miscarriage (date of positive pregnancy test to date of miscarriage diagnosis); ectopic pregnancy/termination (date of positive pregnancy test to date of ectopic pregnancy diagnosis/date of termination); and live birth and stillbirth (date of positive pregnancy test to 20 weeks gestation). Kaplan Meier curves were first created to visually demonstrate the relationship between AMH and miscarriage. A multivariable Cox model was then created to adjust for age, obesity, race and prior history of miscarriage. The model was tested using Schoenfeld’s residuals test to ensure the proportional hazards assumption was not violated (P=0.30). A sensitivity analysis was performed including the 64 women without a known pregnancy outcome in the Cox proportional hazards model. For this analysis (N=533) women were censored at the latest time point that an ongoing pregnancy was observed. For most women, this was the date of their first trimester ultrasound. All statistical analyses were performed using Stata 14. A P-value of <0.05 was considered statistically significant.

Supplemental Figure 1 . Flow diagram of participants in the cohort