Clinical Relevance of Combined FSH and AMH Observations in ...

ORIGINAL E n d o c r i n e

ARTICLE R e s e a r c h

Clinical Relevance of Combined FSH and AMH Observations in Infertile Women Norbert Gleicher, Ann Kim, Vitaly Kushnir, Andrea Weghofer, Aya Shohat-Tal, Emanuela Lazzaroni, Ho-Joon Lee, and David H. Barad The Center for Human Reproduction (N.G., A.K., V.K., A.W., A.S.-T., E.L., H.-J.L., D.H.B.) New York, New York 10021; Foundation for Reproductive Medicine (N.G., D.H.B.), New York, New York 10021; and Department of Gynecologic Endocrinology and Reproductive Medicine (A.W.), Medical University Vienna, 1090 Vienna, Austria

Context: FSH and anti-Müllerian hormone (AMH) are, individually, widely used to assess functional ovarian reserve (FOR) but demonstrate discrepancies in efficacy. How predictive they are combined is unknown. Objective: The purpose of this study was to assess predictive values of different FSH and AMH combinations on in vitro fertilization (IVF). Design and Setting: FSH and AMH levels in patients were categorized as low, normal, and high, based on age-specific 95% confidence intervals. This allowed for establishment of nine combinations of low, normal, or high FSH/AMH patient categories. With use of various statistical methods, patients in individual categories were then compared in outcomes. Patients: We investigated 544 consecutive infertility patients in their first IVF cycles. Interventions: IVF cycles were managed. Main Outcome Measures: Oocyte yields and implantation and pregnancy rates, adjusted for age and fragile X mental retardation 1 (FMR1) genotypes/subgenotypes, were measured. Results: The most notable repeated finding was a strong statistical association of the FSH/AMH high/high category (characterized by abnormally high FSH and AMH levels) with favorable IVF outcomes compared with outcomes for other FSH/AMH variations (4.34 times odds of high oocyte yields and 1.93 times odds of clinical pregnancy). Addition of age to the model only minimally further improved the odds of pregnancy to 2.03 times. The positive association with high oocyte yields, however, turned negative (0.75 times lower yields) with addition of FMR1 to the model for women with FSH/AMH high/high and the het-norm/low FMR1 subgenotype compared with women with the norm FMR1 genotype and other FSH/AMH categories. Conclusions: In the absence of het-norm/low FMR1, abnormally high FSH and AMH, a seemingly contradictory combination, reflects highly beneficial outcomes in IVF compared with the other FSH/AMH categories, suggesting greater importance of FSH in early follicle maturation than currently recognized. The study also confirms adverse outcome effects of het-norm/low FMR1 and, therefore, the gene’s importance for reproductive success. (J Clin Endocrinol Metab 98: 2136 –2145, 2013)

ISSN Print 0021-972X ISSN Online 1945-7197 Printed in U.S.A. Copyright © 2013 by The Endocrine Society Received January 4, 2013. Accepted February 27, 2013. First Published Online March 26, 2013

2136

jcem.endojournals.org

Abbreviations: AMH, anti-Müllerian hormone; AFC, antral follicle count; BMI, body mass index; CI, confidence interval; DFOR, diminished functional ovarian reserve; DHEA, Dehydroepiandrosterone; FMR1 gene, fragile X mental retardation1 gene; FOR, functional ovarian reserve; het, heterozygous; hMG, human menopausal gonadotropin; hom, homozygous; IVF, in vitro fertilization; norm, normal; PCO, polycystic ovary; PCOS, polycystic ovary syndrome.

J Clin Endocrinol Metab, May 2013, 98(5):2136 –2145

doi: 10.1210/jc.2013-1051

doi: 10.1210/jc.2013-1051

woman’s functional ovarian reserve (FOR) is reflective of oocyte yields and pregnancy outcomes in association with in vitro fertilization (IVF) (1). There are various methods available to evaluate FOR but, aside from antral follicle count (AFC), FSH and anti-Müllerian hormone (AMH) have evolved as the principal tools (2). Although FSH and AMH correlate quite well in this function (3), they do demonstrate distinct differences, especially at extreme ages. At younger ages, abnormally elevated FSH levels have lower significance in the presence of good AMH levels (2, 4), whereas in older women, especially those older than age 42, AMH loses specificity in the presence of still decent FSH levels (4, 5). FSH and AMH have been individually evaluated for their ability to predict IVF outcomes in many studies (1, 2, 6). What, however, has so far not been explored is how FSH and AMH in combination are predictive of IVF outcomes. This may be particularly important in association with polycystic ovary syndrome (PCOS), in which AMH is usually high (2, 7) and FSH is usually low (8), but both hormones may also be found to be elevated in some patients (Gleicher, N., V. Kushnir, D. H. Barad, unpublished observations). To clinicians, the combination of abnormally high FSH and AMH proves to be potentially misleading because abnormally elevated FSH levels usually indicate abnormally low FOR (1, 9), whereas abnormally high AMH levels denote the opposite (2, 7, 10). Therefore, this study, in a number of statistical models, explores the combined utility of FSH and AMH in predicting IVF outcomes, with particular attention to the clinical circumstance of combined elevated FSH and high AMH levels and to appropriate adjustments for patient age and recent newly described genotypes and subgenotypes of the fragile X mental retardation 1 (FMR1) gene. Female age is well established as a crucial contributor to FOR (1, 2), and impacts of the FMR1 gene on ovarian aging, beyond the long-known association of the premutation-range genotype (55–200 CGG repeats) with primary ovarian insufficiency (11), have been recently documented in humans in a series of studies from our center (12, 13) and by others in a mouse model (14).

A


Patient population This study involved the retroactive analysis of 544 consecutive infertile women, who presented to our center between January 2005 and May 2012 for a first fresh nondonor IVF cycle. Cycle data were extracted from the center’s anonymized electronic research data bank, which contains selected patient data from at least 1 IVF cycle. Table 1 summarizes patient characteristics. As Table 1 demonstrates, the mean ⫾ SD age of the patient population was 38.4 ⫾ 5.0 years. The body mass index (BMI) was 24.5 ⫾ 5.1 kg/m2, and their most common primary infertility diagnosis was diminished FOR (DFOR) in 315 (57.9%) of the women investigated. PCOS was a primary diagnosis in only 48 women (8.8%). By race, assigned according to National In-

Table 1. IVF Cycle Characteristics and Patient Demographics

Materials and Methods In this study, we explored the hypothesis that different combinations of FSH and AMH, based on age-specific levels, defined as normal, high, and low, using previously reported age-specific 95% confidence intervals (CIs) (9, 10), were predictive of oocyte yields and pregnancy chances in association with IVF.

2137

Characteristic

Value

Cycles, n Age, y BMI, kg/m2 Estradiol, pg/mL FSH, mIU/mL AMH, ng/mL Gonadotropin dosage, IU Oocytes retrieved, n Embryos transferred, n Embryos cryopreserved, n Implantation rate Clinical pregnancy rate, n (%) Miscarriage rate, n (%)a Live birth rate, n (%)a Infertility diagnoses, n (%)b Diminished ovarian reserve Male factor Tubal factor Polycystic ovary syndrome Uterine factor Endometriosis Race, n (%) Caucasian Asian African FSH/AMH in categories, n (%) High/low High/normal High/high Low/high Low/normal Low/low Normal/low Normal/high Normal/normal FMR1, n (%) Norm Het-norm/low Het-norm/high Hom-low/low Hom-low/high Hom-high/high

544 38.4 ⫾ 5.0 24.5 ⫾ 5.1 52.0 ⫾ 27.8 10.7 ⫾ 6.9 1.6 ⫾ 1.9 5644 ⫾ 2507 8.5 ⫾ 6.4 2.4 ⫾ 1.0 2.0 ⫾ 3.5 0.5 ⫾ 0.2 142/544 (26.1) 28/130 (21.5) 102/130 (78.5) 315 (57.9) 134 (24.6) 117 (21.5) 48 (8.8) 36 (6.6) 35 (6.4) 385 (70.8) 98 (18.0) 61 (11.2) 194 (35.7) 80 (14.7) 65 (11.9) 61 (11.2) 43 (7.9) 41 (7.5) 27 (5.0) 20 (3.7) 13 (2.4) 303 (55.7) 117 (21.5) 83 (15.3) 20 (3.7) 14 (2.6) 7 (1.3)

Unless otherwise noted, values shown are means ⫾ SD. a

Outcomes for 12 pregnancies were unknown.

b

Patients had more than one infertility diagnosis.

2138

Gleicher et al

Combined FSH and AMH in Infertile Women

stitutes of Health guidelines, 385 women (70.8%) were Caucasian, 98 (18.0%) were Asian, and 61 (11.2%) were African. For further patient details, see Table 1.

Laboratory testing All laboratory tests reported here were performed at the first routine evaluation of patients after initial presentation to our center. FSH and estradiol were obtained on day 2 or 3 of menstrual cycles; AMH and other tests were performed at random, unrelated to cycle day. Laboratory tests were performed either in house or, based on patient demands/insurance circumstances, via commercial testing as reported previously (12, 13). All fragile X mental retardation 1 (FMR1) testing was performed using commercial tests, as reported (12, 13). Results were reported as number of CGG repeats per allele. We then further classified patients according to recently defined FMR1 genotypes into normal (norm), heterozygous (het), and homozygous (hom) and into subgenotypes high and low, depending on whether abnormal CGG counts were above or below the normal range of 26 to 34 CGG repeats (12, 13).

Clinical management Our center serves a very adversely selected patient population, which in 2011 for the first time exceeded a mean age of 40 years at the time of first presentation. This fact is reflected in the observation that 57.9% of patients in this study had a primary infertility diagnosis of DFOR, mean ⫾ SD FSH was 10.7 ⫾ 6.9 mIU/mL, and AMH was 1.6 ⫾ 1.9 ng/mL (Table 1). A diagnosis of DFOR is reached if a patient demonstrates abnormal agespecific ovarian reserve, based on either abnormally high FSH and/or abnormally low AMH levels (9, 10). Because additional patients had this diagnosis as a secondary, tertiary, or other diagnosis, the overall number of women affected by DFOR exceeded two thirds of patients reported here. A diagnosis of DFOR was not reached in the presence of abnormally elevated FSH and AMH levels. Only 48 (8.8%) of the study population carried a primary diagnosis of PCOS at IVF cycle initiation. Considering the high mean age for the study population noted above, it is reasonable to assume that an additional proportion of patients carried such a diagnosis at younger ages because the clinical phenotypes of PCOS recede with advancing female age (2). Patients who enter IVF treatment receive standardized ovarian stimulation protocols. Women younger than age 40 with normal age-specific FOR usually are, based on age, stimulated in long agonist cycles with daily administration of 150 to 350 IU of gonadotropins, almost always in the form of a human menopausal gonadotropin (hMG) product (our center uses products from different manufacturers, based on patient convenience and/or insurance requirements). In contrast, women with DFOR at young ages, ie, women with premature ovarian aging (also often called occult primary ovarian insufficiency) and women older than age 40, considered to have physiological DFOR, are stimulated in a microdose agonist cycle, with daily gonadotropin doses of 450 to 600 IU, depending on the age and severity of DFOR, all given as pure FSH except 150 IU, administered as hMG. The only exceptions to this premature ovarian aging/occult primary ovarian insufficiency protocol were young women, aged less than 38 years, who demonstrate abnormally high AMH levels (⬎95% CI for age) in combination with abnormally elevated


FSH levels (⬎95% CI for age). In such patients, our empiric clinical impression has been that abnormally high FSH levels do not reflect DFOR because we have not observed low oocyte yields, usually seen in association with high age-specific FSH levels (in the presence of normal or high AMH levels). Patients with abnormally high FSH as well as AMH, therefore, still received microdose agonist stimulations. Depending on age, daily gonadotropin dosages were, however, only in the range of 225 to 450 IU, with 150 IU given as hMG. IVF cycles were otherwise managed routinely. The number of embryos transferred generally followed professional guidelines (mean ⫾ SD, 2.4 ⫾ 1.0) (Table 1), the last such guideline being issued in 2012 (15). Our center routinely uses dehydroepiandrosterone (DHEA) supplementation in women with DFOR, as described in detail previously (16). Because DHEA supplementation beneficially affects a variety of IVF outcomes (16), the data reported here may not necessarily be applicable to unsupplemented IVF cycles in similar patient populations. Because more than two thirds of patients reported here carried a diagnosis of DFOR, more than two thirds were supplemented with DHEA. However, once again women younger than age 40 with abnormally elevated FSH as well as AMH were excluded because they were not considered to have DFOR. Such patients were supplemented with DHEA only in exceptional cases if they demonstrated abnormally low testosterone levels (less than or equal to the lower third of normal range), which appear to be associated with DFOR (17).

Data analysis FSH and AMH levels were defined as low, normal, or high for each patient, based on age-specific 95% CIs. This means that 9 possibilities for the combinations of the 2 variables, FSH and AMH, exist: normal/normal, high/high, low/low, normal/high, normal/low, high/low, high/normal, low/high, and low/normal. As Table 1 demonstrates, of the 544 patients reported here, 65 (11.9%) presented with the FSH/AMH high/high category, and 146 (26.8%) presented with abnormally high age-specific AMH, potentially suggestive of a polycystic ovary (PCO)-like ovarian phenotype (Table 1). Table 1 also summarizes the prevalence of the other categories. A new variable, FSH/AMH, was created to identify patients who fell into any of these 9 possible categories. Numbers of oocytes retrieved were recorded as a binary variable, differentiating between women with and without high oocyte yields. High oocyte yields were defined as exceeding the 75th centile in oocyte numbers for each age. Implantation rates were calculated as number of gestational sacs per number of embryos transferred. Pregnancy outcomes were considered a cycle to have led to a clinical pregnancy only if at least 1 fetal heart beat was seen on vaginal ultrasound.

Statistical analyses We used Kruskal-Wallis, Mann-Whitney U, and ␹2 tests, where appropriate, to assess potential relationships between continuous and categorical variables, including all demographic and IVF cycle data. For the statistical evaluation of the FSH/ AMH categories, the Kruskal-Wallis test was used. Post hoc comparisons between FSH/AMH categories were calculated using the Mann-Whitney U test. Logistic regressions were used to examine the predictive value of age-specific FSH/AMH combinations (categories) for outcome variables. To assess whether FSH/AMH high/high is pre-

doi: 10.1210/jc.2013-1051


dictive of different outcomes from the other 8 categories, a bivariate variable was created, differentiating women with the FSH/AMH high/high combination and all other women without this combination. Probability equations are presented in the results for each logistic regression model. In the models presented, the ␤j parameter estimates the additive effect on the log of the odds for a unit change in the jth explanatory variable. Because these coefficients are not easily interpreted, we convert the parameter to an odds ratio by finding the exponential function (e to the power of the coefficient). All tests were 2-tailed, with a value of P ⬍ .05 considered significant. In logistic regression models, ␣ was set at .05 in regard to individual contributions of predictor variables but at .10 for interaction terms. SPSS (version 18; SPSS, Chicago, Illinois), was used for all statistical assessments.

Institutional review board Patients at our center, at the time of the initial consultation, sign an informed consent form, which, among other issues, allows for use of their medical records for research purposes, as long as they remain anonymous and the content of the medical record remains confidential. These conditions were met in this study, which, therefore, under the center’s institutional review board rules required only expedited institutional review board approval, which was obtained.

Results Of 544 first IVF cycles, 142 (26.1%) led to clinical pregnancies. Pregnancy outcome data were available for all cycles except for 12. Pregnancy outcome data, therefore, relate to 130 clinical pregnancies (Table 1). To assess potential associations between FSH/AMH and various clinical patient parameters, the Kruskal-Wallis test was used. Patient parameters included gonadotropin dosages (in international units) used for ovarian stimTable 2.

2139

ulations, oocyte yields, numbers of embryos transferred, numbers of embryos cryopreserved, and implantation rates. All of these patient characteristics were statistically different between the 9 FSH/AMH combinations at P ⬍ .001, except for the implantation rate, which almost reached significance at P ⫽ .05 (Table 2). Tables 1 and 2 offer additional interesting insights. The first relates to the distribution of categories. By far the most frequently encountered FSH/AMH category in infertile women is high/low (35.7%), followed by high/normal (14.7%), high/high (11.9%), and low/high (11.25%). Not unexpectedly, the least encountered category is normal/normal (2.4%). Notably, age did not differ among categories and probably because of the blunting effects of advanced ages of patients in all categories, BMI differences also failed to reach significance (P ⫽ .08). As one would expect, the highest FSH levels were seen in the FSH/AMH high/low category and the highest AMH levels in the low/high category. The lowest and practically identical AMH levels were, however, seen in the low/low and high/low categories. In addition, as one would expect, the highest oocyte yields were observed with FSH/AMH low/high (14.4 ⫾ 7.0), but, surprisingly, almost identical yields were obtained in the high/high category (13.0 ⫾ 7.2), followed by normal/high (12.0 ⫾ 5.7) and low/normal (10.3 ⫾ 5.5). The lowest oocyte yields were seen in the high/low category (5.0 ⫾ 4.2). The pairwise comparisons between FSH/AMH high/ high and the remaining 8 combinations were made using Mann-Whitney U tests (Figure 1, A–H). These comparisons demonstrated significant differences in FSH (P ⬍

Summary of the Demographic, Clinical, and IVF Cycle Data According to the AMH/FSH Categories

Cycles, n Age, y BMI, kg/m2 Estradiol, pg/mL FSH, mIU/mL AMH, ng/mL Gonadotropin dosage, IU Oocytes retrieved, n Embryos transferred, n Embryos cryopreserved, n Implantation rate Clinical pregnancy rate, n (%)

Low/Low

Normal/ Low

Low/Normal

Normal/ Normal

High/Low

Low/High

High/ Normal

Normal/ High

High/High

P Value

41 38.8 ⫾ 4.9 24.8 ⫾ 4.8 62.1 ⫾ 5.4

27 36.7 ⫾ 4.4 25.9 ⫾ 5.6 50.6 ⫾ 23.6

43 37.2 ⫾ 5.5 25.1 ⫾ 6.7 48.5 ⫾ 29.1

13 36.1 ⫾ 4.9 24.5 ⫾ 3.4 49.3 ⫾ 12.5

194 39.0 ⫾ 4.6 24.0 ⫾ 4.8 49.9 ⫾ 27.4

61 38.8 ⫾ 5.1 25.9 ⫾ 5.2 53.9 ⫾ 24.8

80 38.3 ⫾ 5.7 23.7 ⫾ 4.5 55.3 ⫾ 24.3

20 38.9 ⫾ 6.7 25.4 ⫾ 6.7 48.2 ⫾ 19.4

65 38.3 ⫾ 4.7 23.9 ⫾ 5.0. 50.6 ⫾ 18.7

.11 .08 .32

5.8 ⫾ 1.3 0.5 ⫾ 0.4 6640 ⫾ 1840

7.4 ⫾ 0.6 0.8 ⫾ 0.5 5256 ⫾ 1884

5.7 ⫾ 1.2 1.8 ⫾ 1.2 4860 ⫾ 2367

7.4 ⫾ 0.6 1.9 ⫾ 1.2 3877 ⫾ 1906

14.8 ⫾ 8.1 0.5 ⫾ 0.4 6811 ⫾ 2420

5.8 ⫾ 1.2 3.8 ⫾ 2.4 4135 ⫾ 2249

12.2 ⫾ 6.4 1.7 ⫾ 1.3 5775 ⫾ 2362

7.5 ⫾ 0.8 3.0 ⫾ 2.1 4545 ⫾ 2233

10.9 ⫾ 5.5 3.4 ⫾ 2.2 4159 ⫾ 1991

⬍.001 ⬍.001 ⬍.001

6.1 ⫾ 4.7

7.6 ⫾ 4.1

10.3 ⫾ 5.5

8.5 ⫾ 5.0

5.0 ⫾ 4.2

14.4 ⫾ 7.0

8.4 ⫾ 5.6

12.0 ⫾ 5.7

12.9 ⫾ 7.2

⬍.001

2.4 ⫾ 1.1

2.2 ⫾ 0.8

2.7 ⫾ 0.9

2.5 ⫾ 0.9

2.1 ⫾ 0.9

3.0 ⫾ 0.9

2.5 ⫾ 1.2

3.2 ⫾ 1.1

2.7 ⫾ 1.0

⬍.001

0.6 ⫾ 1.5

1.4 ⫾ 1.9

2.5 ⫾ 3.3

1.6 ⫾ 3.0

0.7 ⫾ 1.9

4.3 ⫾ 4.3

1.9 ⫾ 3.6

3.3 ⫾ 4.3

4.1 ⫾ 5.0

⬍.001

0.4 ⫾ 0.2 10 (24.4)

0.6 ⫾ 0.3 5 (18.5)

0.6 ⫾ 0.3 13 (30.2)

0.8 ⫾ 0.2 6 (46.2)

0.6 ⫾ 0.2 34 (17.5)

0.5 ⫾ 0.3 22 (36.1)

0.5 ⫾ 0.2 23 (28.7)

0.5 ⫾ 0.3 4 (2.0)

0.5 ⫾ 0.2 25 (38.5)

.05 .009a

Unless otherwise noted, values shown are means ⫾ SD. A Kruskal-Wallis test was used for comparison of data between FSH/AMH categories. A ␹2 test indicated a significant association between pregnancy outcome (pregnant and not pregnant) and the 9 FSH/AMH categories: ␹2 (8, n ⫽ 544) ⫽ 20.32, P ⫽ .009.

a

2140

Gleicher et al



Figure 1. Significant differences in demographic and clinical data according to the FSH/AMH categories. A–H, Differences (P ⬍ .05) in patient demographics and IVF cycle outcomes data between the FSH/AMH high/high category and other combinations. Figures are shown as box and whisker plots; the mean and median are depicted with dotted and solid lines, respectively. The y-axis represents the following units: FSH, milliinternational units per milliliter, AMH, nanograms per milliliter; gonadotropin dosages (GD), international units; oocyte yield (OY), embryos cryopreserved (EC), BMI, embryos transferred (ET), and implantation rate (IR) reflect numbers.

doi: 10.1210/jc.2013-1051

.001), AMH (P ⬍ .001), gonadotropin dosages (P ⬍ .001), oocyte yields (P ⬍ .001), and number of embryos cryopreserved (P ⬍ .001), with women with FSH/AMH high/ high, of course, demonstrating not only higher FSH and AMH but also lower gonadotropin dosages for stimulation, higher oocyte yields, and larger numbers of cryopreserved embryos than women with FSH/AMH low/low (Figure 1A). Women with FSH/AMH high/high also demonstrated lower FSH (P ⫽ .02) but higher AMH (P ⬍ .001), lower gonadotropin use (P ⬍ .001), greater oocyte yields (P ⬍ .001), and more cryopreserved embryos (P ⫽ .001) than women with FSH/AMH high/normal (Figure 1B). Moreover, they also demonstrated higher FSH (P ⬍ .001), AMH (P ⬍ .001), lower BMI (P ⫽ .046), lower gonadotropin dosages (P ⫽ .02), greater oocyte yields (P ⬍ .001), more embryos transferred (P ⫽ .04), and more embryos cryopreserved (P ⫽ .02) than women with FSH/AMH normal/low (Figure 1C). They also demonstrated higher FSH (P ⬍ .001), AMH (P ⬍ .001), and oocyte yields (P ⫽ .04) than women with FSH/AMH low/normal (Figure 1D). Compared with women with FSH/AMH normal/normal, women with FSH/AMH high/high demonstrated higher FSH (P ⬍ .001), higher AMH (P ⫽ .008), greater oocyte yields (P ⫽ .02), and lower implantation rates (P ⫽ .004) (Figure 1E). Compared with women with FSH/ AMH high/low, they demonstrated lower FSH (P ⬍ .001), higher AMH (P ⬍ .001), lower gonadotropin usage (P ⬍ .001), higher embryo numbers transferred (P ⬍ .001) and cryopreserved (P ⬍ .001), higher oocyte yields (P ⬍ .001), and lower implantation rates (P ⫽ .02) (Figure 1F). Women with FSH/AMH high/high demonstrated higher FSH (P ⬍ .001) and lower BMI (P ⫽ .01) than women with FSH/AMH low/high (Figure 1G) and higher FSH (P ⬍ .001) than women with FSH/AMH normal/high (Figure 1H). In further investigation of high oocyte yields, ␹2 calculations demonstrated a significant relationship between oocyte yields (ie, high vs normal and low yields) and the various FSH/AMH categories (ie, FSH/AMH high/high vs all 8 other categories; ␹2 (1, n ⫽ 544) ⫽ 32.00, P ⬍ .001). Among women with FSH/AMH high/high, 61.5% demonstrated high oocyte yields, whereas only 26.9% of women with other FSH/AMH categories demonstrated high oocyte yields (Figure 2). By including all 9 FSH/AMH categories (high/high category and everybody else) in a logistic regression model, based on the probability equation logit P ⫽ ⫺0.998 ⫹ (1.468 䡠 FSH/AMH high/high), a significant effect was observed with FSH/AMH high/high (P ⬍ .001), suggesting that FSH/AMH high/high conveys a 4.34 times higher


2141

odds of having a high oocyte yield than all other 8 FSH/ AMH categories. By including again all 9 FSH/AMH categories along with age in a logistic regression model, the significant effect was preserved for FSH/AMH high/high (P ⬍ .001). A significant effect was not observed with age (P ⫽ .15) with the probability equation, logit P ⫽ ⫺0.0427 ⫹ (1.469 䡠 FSH/AMH high/high), suggesting that a woman with FSH/AMH high/high has approximately 4.34 times the odds of having a high oocyte yield compared with a woman of any of the 8 other FSH/AMH categories, assuming that patients are at equal age. Further. by including the gonadotropin dosage into the model, the significant effect was preserved for the high/high category (P ⬍ .001). This picture, however, changes if FMR1 genotypes and subgenotypes are added to the regression model: Once again a significant interaction is observed, this time with the FMR1 subgenotype het-norm/low and FSH/AMH high/high (P ⫽ .001) in addition to just FSH/AMH high/high (P ⬍ .001), based on the probability equation, logit P ⫽ ⫺0.972 ⫺ (2.391 䡠 FSH/AMH high/high 䡠 het-norm low) ⫹ (2.107 䡠 FSH/AMH high/high), suggesting that a woman with FSH/ AMH high/high who, however, also displays a het-norm/low FMR1 subgenotype, actually demonstrates a 0.75 times lower odds of having high oocyte yields than women with any 1 of the other 8 FSH/AMH categories and a norm FMR1 genotype. These results remained unchanged when race was added to the model (data not shown). When all of the other 8 categories are combined into 1 category, the FSH/AMH high/high category also affected pregnancy chances with IVF: comparing it with the other 8 FSH/AMH categories in a logistic regression model, a significant effect was observed (P ⫽ .02) with the probability equation logit P ⫽ ⫺1.129 ⫹ (0.659 䡠 FSH/AMH high/high), suggesting that a woman with FSH/AMH high/high has 1.93 times the odds of having a clinical pregnancy compared with women with the other 8 FSH/AMH categories. When age was added to the model, a significant effect was maintained for FSH/AMH high/high (P ⫽ .02), and in addition, a significant effect was observed with age (P ⬍ .001) with the probability equation logit P ⫽ 4.144 ⫹ (0.709 䡠 FSH/AMH high/high) ⫺ (0.140 䡠 age), which suggests that, assuming the same female age, a woman with FSH/AMH high/high has 2.03 times the odds of pregnancy compared with all other categories together. In an additional logistic regression model, including numbers of embryos transferred, this significant effect for women with FSH/AMH high/high was preserved (P ⫽ .03) and even marginally increased with addition of number of transferred embryos (P ⫽ .02). With addition of FMR1, age, and all FSH/AMH categories to the logistic regression model, a significant effect

2142

Gleicher et al



Using various statistical methods, we here report distinctively different IVF outcomes with different FSH/AMH combinations (categories). This observation suggests that different FSH/AMH categories reflect distinctively different FOR parameters. The FSH/AMH high/high combination not only proved most divergent but also most beneficial as an overall IVF outcome predictor. It denotes favorable IVF cycle outcomes compared with those for all other FSH/AMH categories, including higher pregnancy chances, and, in addition, also demonstrated again the importance of the FMR1 gene in achieving positive IVF outcomes. This importance was shown by our finding Figure 2. Percentage of women with high oocyte yields in FSH/AMH high/high and that carrying the het-norm/low subgenotype of other categories. the FMR1 gene completely eliminated the otherwise beneficial effects on IVF outcomes of for FSH/AMH high/high was again observed (P ⫽ .045) in FSH/AMH high/high. This observation does not come as addition to that for age (P ⬍ .001). FMR1, however, was a complete surprise and, indeed, indirectly confirms the not a significant predictor of pregnancy chances (P ⫽ .85). effects of this FMR1 subgenotype reported previously by When equal age was assumed, a woman with FSH/AMH our group. Het-norm/low, characterized by 1 allele in the high/high, therefore, has 2.14 times the odds of pregnancy norm (26 –34 CGG) range and the second allele in the low of a woman with any other FSH/AMH category. (⬍26 CGG) range, had in many different ways proven most interesting. One relevant prior noted association is a PCO-like nonobese ovarian phenotype at young ages, Discussion which then, because of excessive recruitment, prematurely results in DFOR. Another important association relates to The discussion of our results has to start with reemphasignificantly lower IVF pregnancy chances than with the sizing that in this study the patients do not necessarily represent a typical infertile patient population undergoing norm genotype, the most prevalent genotype of FMR1 in IVF. In this article, the principles described, involving the population, as reported previously (12), an observaFSH/AMH categories, should nevertheless apply to all in- tion that was independently confirmed again here. Because of abnormally high AMH levels, exceeding the fertile women. With more than two thirds of the patients investigated 95% CI for age, women with FSH/AMH high/high can be here having DFOR, this population has to be considered phenotypically viewed as either having PCOS or, at a minadversely selected compared with traditionally studied infer- imum, expressing an ovarian PCO-like phenotype in astile women. The clinical pregnancy rate of 26.1% in first IVF sociation with abnormally high FSH levels (⬎95% CI for cycles observed at our center (defined by fetal heart on ul- age). This study now suggests that this ovarian phenotype trasound examination) is, therefore, quite remarkable, and can be further subdivided into at least 2 subphenotypes. The first, in association with a norm FMR1 genotype, is more fitting to an average infertile patient population. The positive effects of elevated FSH levels we observed present in a little more than half of all women (12, 13) with in the presence of high AMH levels, to a degree, came as highly favorable FOR, with high oocyte yields and excelsurprise because neither elevated age-specific FSH (1, 8, 9) lent oocyte quality, reflected in approximately twice the nor high AMH (2, 7, 10), in isolation, is known to be odds of pregnancy compared with women with all other associated with favorable IVF outcomes. In clinical prac- FSH/AMH categories. The second is seen in the presence tice, however, we empirically had gotten the impression of the het-norm/low FMR1 subgenotype, with all advanthat patients with FSH/AMH high/high performed better tages of the FSH/AMH high/high category eliminated and in IVF than expected, especially in association with typical odds of pregnancy turned negative (0.75 times lower). This study, thus, defines well the effects of the norm PCOS or PCO-like ovarian phenotypes. This clinical study was, therefore, initiated to investigate the consequences of genotype and het-norm/low subgenotype of the FMR1 different FSH/AMH categories on IVF outcomes. gene in association with FSH/AMH categories, represent-

doi: 10.1210/jc.2013-1051


ing approximately 75% to 80% of all women (12, 13). It, however, does not define adequately het-norm/high and the 3 subgenotypes of hom (high/high, low/low, and high/ low), leaving open the question of whether failure to observe specific phenotypes in association with these subgenotypes simply reflects an absence of specific associated gene functions (and, therefore, specific phenotypes) or is just a reflection of too small patient sample sizes in these subgroups in this study. The study raises the obvious question of why does presence of the het-norm/low FMR1 subgenotype so profoundly change the clinical phenotype of the FSH/AMH high/high category? We previously noted the reported association of the het-norm low FMR1 subgenotypes with a follicle-depleting nonobese PCO-like ovarian phenotype that occurs relatively quickly at a young age. This association is further strengthened in the presence of immune laboratory abnormalities, suggesting a possible additional association between the het-norm/low FMR1 subgenotype and autoimmune risks. Het-norm/high, in contrast, is highly protective against such risks (12). Autoimmunity, in turn, can affect reproductive success in a variety of ways (13). The ultimate conclusion of this study relates, however, to the so far unexplained paradox of some infertility patients presenting with abnormally high levels of FSH in presence of high AMH. High AMH levels are usually indicative of high antral follicle counts and, therefore, high FOR (2, 7, 10). In contrast, abnormally high FSH levels usually indicate low FOR (1, 8, 9). Reviewing the significance of AMH, generally considered the best reflection of small growing follicles and in

Table 3.

2143

turn best representing FOR (1, 2), may offer some preliminary answers: remarkable homogeneity in IVF outcomes of all 3 high AMH categories, independent of FSH levels, is noted (Tables 1 and 3), with oocyte yields ranging from 12.0 ⫾ 5.7 (normal/high) to 14.4 ⫾ 7.0 (low/high) and high/high women being in the middle with 12.9 ⫾ 7.2. Likewise, patient characteristics, including age and BMI, are practically identical, independent of whether FSH is low, normal, or high. On first impression, this finding reemphasizes the relative primacy of AMH over FSH in determining FOR. A look at the 19 women with primary diagnosis of PCOS, who did not fall into 1 of the 3 high AMH categories, shows that were younger (33.2 ⫾ 4.1 years) than the women in those 3 high AMH categories (mean age for all 3 groups, 38.6 ⫾ 5.1, P ⬍ .001); however, BMIs were identical (26.0 ⫾ 8.9 vs 24.9 ⫾ 5.3 kg/m2) and oocyte yields were the same (10.5 ⫾ 5.1 vs 13.4 ⫾ 7.0). These findings are of interest because they suggest distinctly different PCOS and or PCO-like ovarian phenotypes, characterized by younger and older age, yet otherwise indistinguishable in FOR and BMI. In FSH levels, they closely mimic women in the FSH/ AMH normal/high category but are distinct from the those in the high/high category (8.9 ⫾ 4.2 vs 10.9 ⫾ 5.5 mIU/mL, P ⫽ .007) and the low/high category (8.9 ⫾ 4.2 vs 5.8 ⫾ 1.2 mIU/mL, P ⬍ .001). Except for age, patients with PCOS, overall, closest resemble women in the FSH/AMH normal/high category, although women in that category, among all 3 high AMH categories, demonstrate the lowest mean AMH (Table 3).

Patient Demographics and IVF Outcomes in High AMH Categories and Patients With PCOS

Cycles, n Age, y BMI, kg/m2 Estradiol, pg/mL FSH, mIU/mL AMH, ng/mL Gonadotropin dosage, IU Oocytes retrieved, n Embryos transferred, n Embryos cryopreserved, n Implantation rate Clinical pregnancy rate, n (%) Live birth rate, n (%)b Miscarriage rate, n (%)b

Low/High

Normal/High

High/High

Combination of 3 High AMH Categories

61 38.8 ⫾ 5.1 25.9 ⫾ 5.2 53.9 ⫾ 24.8 5.8 ⫾ 1.2 3.8 ⫾ 2.4 4135 ⫾ 2249 14.4 ⫾ 7.0 3.0 ⫾ 0.9 4.3 ⫾ 4.3 0.5 ⫾ 0.3 22 (36.1) 17 (81.0) 4 (19.0)

20 38.9 ⫾ 6.7 25.4 ⫾ 6.7 48.2 ⫾ 19.4 7.5 ⫾ 0.8 3.0 ⫾ 2.1 4545 ⫾ 2233 12.0 ⫾ 5.7 3.2 ⫾ 1.1 3.3 ⫾ 4.3 0.5 ⫾ 0.3 4 (20.0) 2 (50.0) 2 (50.0)

65 38.3 ⫾ 4.7 23.9 ⫾ 5.0 50.6 ⫾ 18.7 10.9 ⫾ 5.5 3.4 ⫾ 2.2 4159 ⫾ 1991 12.9 ⫾ 7.2 2.7 ⫾ 1.0 4.1 ⫾ 5.0 0.5 ⫾ 0.2 25 (38.5) 14 (70.0) 6 (30.0)

146 38.6 ⫾ 5.1 24.9 ⫾ 5.3 51.6 ⫾ 21.5 8.3 ⫾ 4.4 3.5 ⫾ 2.3 4202 ⫾ 2125 13.4 ⫾ 7.0 2.9 ⫾ 1.0 4.0 ⫾ 4.6 0.5 ⫾ 0.3 51 (34.9) 33 (73.3) 12 (26.7)

Patients With PCOS 19 33.2 ⫾ 4.1 26.0 ⫾ 8.9 50.6 ⫾ 21.8 8.9 ⫾ 4.2 2.0 ⫾ 1.2 4346 ⫾ 2511 10.5 ⫾ 5.1 2.3 ⫾ 0.7 3.4 ⫾ 3.6 0.6 ⫾ 0.3 8 (41.2) 6 (75.0) 2 (25.0)

P Value ⬍.001 .58 .73 .55 .002 .95 .10 .01 .76 .17 .54a

Unless otherwise noted, values are means ⫾ SD. Mann-Whitney U tests were used for comparison of data between categories (high AMH and patients with PCOS). A ␹2 test indicated a nonsignificant relationship between pregnancy outcome (pregnant and not pregnant) and group membership (high AMH and patients with PCOS): ␹2 (1, n ⫽ 165) ⫽ 0.38, P ⫽ 0.54. a

b

Outcomes for 6 pregnancies were unknown.

2144

Gleicher et al


All these observations confirm our assumption that all high AMH categories, as defined in the study, at younger ages, probably were associated with either PCOS or, at a minimum, a PCO-like ovarian phenotype. How patients with PCOS fare in association with IVF has remained in dispute, with published outcomes split between diminished and normal age-specific pregnancy chances (2). This study, therefore, offers a potential explanation for this divergence of outcomes reported in the literature: they simply may reflect different patient selections based on FMR1 genotypes and subgenotypes and/or FSH/AMH categories. Gallot et al (18) recently suggested that pregnancy success in IVF directly relates to the number of antral follicles responsive to FSH. Small growing follicles represent a patient’s FOR, a crucial stage of follicle maturation, during which androgens and FSH synergistically promote follicle growth and, thereby, predict IVF outcomes of a follicle cohort, which weeks to months later reaches gonadotropin sensitivity (1). These authors, indeed, reaffirmed an observation reported by our group, demonstrating that, quite surprisingly, a ratio of FSH per oocyte yield but not of AMH per oocyte yield was statistically strongly associated with pregnancy chance in IVF (19). These 2 studies, therefore, contradict the primacy of AMH in defining the small growing follicle pool, representative of the FOR, and suggest that FSH levels probably are more important for these early stages of follicle maturation than currently recognized. Although there is still a consensus that AMH is more predictive than FSH in assessing FOR (1, 2), increasing recognition of the importance of FSH at small growing follicle stages also comes from other sources: whereas follicles are, in principle, considered FSH-insensitive during small growing stages, it appears that androgens enhance the sensitivity of granulosa cells to FSH at these early developmental stages (1). Relatively high peripheral FSH levels may, therefore, have clinical advantages at these early stages of follicle maturation, and may at least partially explain the findings in high/high FSH/AMH category patients reported here. With FSH and androgens enhancing follicle growth at these early growing follicle stages, AMH is believed to oppose follicle recruitment and growth (1). How a high/ high FSH/AMH constellation, therefore, can produce the excellent oocyte yields reported in this study is, therefore, an intriguing question. As also demonstrated in this study, AMH levels in patients with PCOS decline with advancing age but, still, remain elevated in comparison to those in patients without PCOS (20). Based on the rather advanced age of our study group (Table 1) and with more than two thirds of patients


presenting with DFOR, the prevalence of true PCOS, according to the Rotterdam criteria and/or of high oocyte yields producing PCO-like ovarian phenotypes in the study population presented here can be expected to be low. Abnormally elevated AMH levels, exceeding age-specific 95%CIs, as noted before, nevertheless, still suggest an earlier presence of PCOS and/or at least of a PCO-like ovarian phenotype with comparatively high oocyte yields. Eilertsen et al (21), who recently reported that AMH represents an excellent substitute to polycystic ovary morphology in making a diagnosis of PCOS, support such a proposition. Taken together, these observations, therefore, suggest that the answer to the question of how FSH/AMH high/ high can produce such superior results in IVF may lie in the recognition that in reproduction absolute hormone levels are only rarely predictive of outcomes. Much better predictability may be reached in hormone ratios, reflective of agonist/antagonist interactions (22). Indeed, we recently demonstrated such predictability for FSH/androgen and AMH/androgen ratios in association with IVF (17). Under such a hypothesis, a maximally beneficial FSH/ androgen ratio may drive maximal follicle growth and maturation at small growing follicle stages. The inhibitory function of AMH may, therefore, result in a secondary rise in AMH levels, resulting in the FSH/AMH category of high/high. If correct, then the FSH/AMH high/high category of women should demonstrate different FSH/androgen and AMH/androgen ratios than women in other FSH/ AMH categories. As a testable hypothesis, we are currently exploring this question. In summary, the results presented here confirm our preliminary clinical impression that the combination of FSH/ AMH high/high, at least in the absence of a het-norm/low FMR1 subgenotype, results in surprisingly favorable IVF outcomes. The reason, however, ultimately still remains to be determined.

Acknowledgments Address all correspondence and requests for reprints to: Norbert Gleicher, MD, The Center for Human Reproduction, 21 East 69th Street, New York, New York 10021. E-mail: [email protected]. This research was supported by the Foundation for Reproductive Medicine, a not-for-profit medical research foundation, and by intramural funds from the Center for Human Reproduction. Disclosure Summary: N.G. and D.H.B are members of the Board of the Foundation for Reproductive Medicine. N.G., A.W., and D.H.B. received research support, lecture fees, and travel support from a variety of pharmaceutical and medical

doi: 10.1210/jc.2013-1051

device companies, none in any way related to the issues discussed in this manuscript. N.G. and D.H.B. are listed as coinventors on two already granted US user patents, which claim therapeutic benefits from DHEA supplementation in women with DFOR and DOR. Both authors are also listed on additional pending patents in regard to DHEA supplementation and on pending patents, claiming diagnostic and therapeutic benefits from the determination of CGG repeats on the FMR1 gene. N.G. is owner of the Center for Human Reproduction, where this research was performed. The other authors have nothing to disclose.

References 1. Gleicher N, Weghofer A, Barad DH. Defining ovarian reserve to better understand ovarian aging. Reprod Biol Endocrinol. 2011;9: 23. 2. Ledger WL. Clinical utility of measurement of anti-Müllerian hormone in reproductive endocrinology. J Clin Endocrinol Metab. 2010;95:5144 –5154. 3. Singer T, Barad DH, Weghofer A, Gleicher N. Correlation of antimullerian hormone and baseline follicle-stimulating hormone levels. Fertil Steril. 2009;91:2616 –2619. 4. Sabatini L, Zosmer A, Hennessy EM, Tozer A, Al-Shawaf T. Relevance of basal serum FSH to IVF outcome varies with patient age. Reprod Biomed Online. 2008;17:10 –19. 5. Gleicher N, Weghofer A, Barad DH. Discordance between follicle stimulating hormone (FSH) and anti-Müllerian hormone (AMH) in female infertility. Reprod Biol Endocrinol. 2010;8:64. 6. Barad DH, Weghofer A, Gleicher N. Comparing anti-Müllerian hormone (AMH) and follicle- stimulating hormone (FSH) as predictors of ovarian function. Fertil Steril. 2009;(4 suppl):1553–1555. 7. Lin YH, Chiu WC, Wu CH, Tzeng CR, Hsu CS, Hsu MI. Antimüllerian hormone and polycystic ovary syndrome. Fertil Steril. 2011; 96:230 –235. 8. Beydoun HA, Stadtmauer L, Beydon MA, Russel H, Zhao Y, Oehninger S. Polycystic ovary syndrome, body mass index and outcome of assisted reproductive technologies. Reprod Biomed Online. 2009; 18:856 – 863. 9. Barad DH, Weghofer A, Gleicher N. Age-specific levels of basal follicle-stimulating hormone assessment of ovarian function. Obstet Gynecol. 2007;109:1404 –1410.


2145

10. Barad DH, Weghofer A, Gleicher N. Utility of age-specific serum anti-Müllerian hormone concentrations. Reprod Biomed Online. 2011;22:284 –291. 11. Wittenberger MD, Hagerman RJ, Sherman SL, et al. The FMR1 premutation and reproduction. Fertil Steril. 2007;87:456 – 465. 12. Gleicher N, Weghofer A, Lee IH, Barad DH. FMR1 genotype with autoimmunity-associated polycystic ovary-like phenotype and decreased pregnancy chance. PLoS ONE. 2010;5:e15303. 13. Gleicher N, Weghofer A, Lee IH, Barad DH. Association of FMR1 genotypes with in vitro fertilization (IVF) outcomes based on ethnicity/race. PLoS ONE. 2011;6:e18781. 14. Hoffman GE, Le WW, Entezam A, et al. Ovarian abnormalities in a mouse model of fragile X primary ovarian insufficiency. J Histochem Cytochem. 2012;60:439 – 456. 15. Practice Committee of the American Society for Reproductive Medicine and Practice Committee of the American Society for Assisted Reproductive Technology. Criteria for number of embryos to transfer: a committee opinion. Fertil Steril 2013;99:44 – 46. 16. Gleicher N, Barad DH. Dehydroepiandrosterone (DHEA) supplementation in diminished ovarian reserve (DOR). Reprod Biol Endocrinol. 2011;9:67. 17. Gleicher N, Kim A, Weghofer A, et al. Diminished functional ovarian reserve (DOR) is at all ages characterized by androgen deficiency, responsive to androgen supplementation. J Assist Reprod Genet. 2012. 18. Gallot V, Berwanger da Silva AL, Genro V, Grynberg M, Frydman N, Fanchin R. Antral follicle responsiveness to follicle-stimulating hormone administration assessed by the Follicular Output RaTe (FORT) may predict in vitro fertilization-embryo transfer outcome. Hum Reprod. 2012;27:1066 –1072. 19. Gleicher N, Kim A, Weghofer A, Barad DH. Towards a better understanding of functional ovarian reserve: AMH (AMHo) and FSH (FSHo) hormone ratios per retrieved oocyte. J Clin Endocrinol Metab. 2012;97:995–1004. 20. Carmina E, Campagna AM, Mansuet P, Vitale G, Kort D, Lobo R. Does the level of serum antimüllerian hormone predict ovulatory function in women with polycystic ovary syndrome with aging? Fertil Steril. 2012;98:1043–1046. 21. Eilertsen TB, Vanky E, Carlsen SM. Anti-Mullerian hormone in the diagnosis of polycystic ovary syndrome: can morphologic description be replaced? Hum Reprod. 2012;27:2494 –2502. 22. Weghofer A, Gleicher N. Ovarian function: a theory of relativity. Hum Reprod. 2009;24:17–19.