Is Ovarian Reserve Impacted in Anorexia Nervosa?

Abstract

Study Objectives:

Hypothalamic amenorrhea is common in adolescents and young adults (AYAs) with anorexia nervosa (AN), and ovarian reserve is not routinely assessed. AN increases rates of fertility problems, but how or when AN negatively influences future fertility is unclear. We sought to determine whether biomarkers of ovarian reserve were impacted in AYA with AN.

Design:

Cross-sectional study

Setting:

Tertiary care center

Participants:

Females with AN and amenorrhea (n=97) at the pre-intervention visit of a clinical trial, n=19 females without an eating disorder or menstrual dysfunction.

Main Outcome Measures:

Serum anti-Mullerian hormone (AMH) concentrations

Results:

AMH levels were higher in AYA with AN than unaffected adolescents (4.7 vs. 3.2 ng/mL; p=0.03). Neither FSH nor inhibin B differed between groups. In 19.6% of participants with AN, AMH levels were elevated above the normal range (>6.78ng/mL). These subjects had a longer disease duration than those with normal AMH levels (9 vs. 3 mos; p=0.03); age or degree of malnutrition did not differ between AN subjects with normal or elevated AMH.

Conclusions:

AMH levels appear to be normal or elevated in AYA with AN. Low AMH in a patient with AN should raise clinical concern regarding ovarian reserve, and should not be attributed to degree of malnutrition alone. Currently, AMH is not regularly assessed during routine AN clinical care. However, our findings suggest some clinical utility in identifying those patients with reduced ovarian reserve. Potential links between the hypothalamic amenorrhea suffered by patients with AN and PCOS should be explored.

Introduction:

Anorexia nervosa (AN), a disorder characterized by energy restriction, intense fear of weight gain, and disturbance of body weight perception, is becoming increasingly prevalent.¹ Disease onset often occurs during adolescence, during the window of time when pubertal development, reproductive health, and bone mineral accretion are all peaking. Multiple endocrine aberrations can occur in the setting of malnutrition and AN, including hypogonadotropic hypogonadism, acquired growth hormone resistance with low insulin-like growth factor 1 levels, elevated cortisol levels, and altered secretion of adipokines and appetite-regulating hormones such as adiponectin, leptin and ghrelin.² As a result, young women with AN are at risk for many long-term health consequences of their disease, including low bone mineral density (BMD) for age and cardiovascular disease. In addition, the multiple neuroendocrine alterations that occur with AN appear to impair reproductive health. Some studies have shown higher rates of fertility problems, later age at childbearing, and increased rates of obstetrical complications.^3,4 The etiology of these reproductive issues for women with a history of AN during adolescence is not well understood.

Typically, a woman’s fertility declines with age. Some disease processes interfere with fertility, and cause a premature drop in the number of follicles in the ovary, otherwise known as diminished ovarian reserve. Anti-Mullerian hormone (AMH) is produced by granulosa cells of follicles in females. AMH expression is highest in secondary, preantral, and small antral follicles. Circulating levels of this hormone peak during the third decade in healthy women, and decline to non-measurable levels by menopause. Clinically, AMH is used to measure indirectly the ovarian follicular reserve; lower AMH concentrations are associated with decreased ovarian reserve. Diminished ovarian reserve is associated with a lower likelihood of a successful live birth, and with reduced response potential of the ovary to stimulation such as would occur with in-vitro fertilization (IVF) protocols. AMH is a useful clinical marker as values are equally valid at any point in the menstrual cycle. Inhibin B is another biomarker that has been used to estimate ovarian reserve. It is produced by developing antral follicles of the ovaries, arising primarily from the dominant follicle. In healthy females, inhibin B concentrations increase as puberty progresses and play an important role in the regulation of FSH levels during the early and mid-follicular phase of the menstrual cycle. Unlike AMH, inhibin B concentrations vary with the menstrual cycle, peaking at ovulation and then falling to basal levels in the luteal phase.⁵

To our knowledge, no studies have evaluated the ovarian reserve in adolescents and young adult (AYA) females with AN. In the current study, we investigated AYA with AN to determine: 1) whether ovarian reserve is impaired in the setting of early AN; 2) whether AMH or inhibin B levels vary with degree of malnutrition; and 3) whether other hormonal values (estradiol, ghrelin, leptin) are predictive of ovarian reserve. We hypothesized that adolescents with AN would have lower AMH levels, indicating impairment of ovarian reserve in the setting of this disease.

Materials and Methods:

Subjects

From 2009–2014, 837 female adolescents seen for eating disorder visits at Boston Children’s Hospital were screened for participation in a randomized clinical trial. Eligible patients were aged 11–26 years, had a body mass index (BMI) < 20 kg/m², and exhibited disordered eating, distorted body image, and fear of weight gain consistent with a diagnosis of AN (as defined by Diagnostic and Statistical Manual of Mental Disorders-V); n=317 met these study eligibility criteria (Figure 1a). Patients were excluded for other medical diagnoses (e.g., thyroid disease, celiac disease, cystic fibrosis, inflammatory bowel disease, diabetes mellitus) or medication use (e.g., hormonal contraceptives, glucocorticoids, anticonvulsants) affecting BMD. No subjects regularly consumed alcohol or used cigarettes. Study procedures were approved by the Boston Children’s Hospital Institutional Review Board. Written informed consent was obtained, with parental consent and subject assent for patients <18 years (clinicaltrials.gov NCT01100567).

Fig. 1.

Pathway to study entry for (A) 97 adolescents with anorexia nervosa and (B) 19 healthy comparison subjects.

The study was a single site, placebo-controlled trial in which 102 subjects underwent randomization; 99 completed baseline measurements (Figure 1a). Height (cm) was measured using standardized procedure with a wall-mounted stadiometer. Weight (kg) was measured each morning, post-voiding and in the fasting state, with subjects wearing a hospital gown. Percentage median body weight (% MBW) was determined using the year 2000 Centers for Disease Control and Prevention Growth Charts and the formula: % MBW = 100 x [Patient BMI (kg/m²) / Median BMI for age (kg/m²)].⁶ All participants responded to a semi-structured interview to obtain demographic information and health history, which included information about medication use and menstrual history.

In addition, we enrolled n=25 healthy similarly aged female adolescents without an eating disorder or known menstrual dysfunction to serve as a convenience sample. The convenience sample provided measurements against which to compare the results of the participants with AN in real time, rather than only utilizing historical control data. We excluded participants with BMI > 35 kg/m2 (n=5), leaving us with n=20 healthy subjects (Figure 1b). Of these subjects, 8 were using hormonal contraception: 2 were prescribed oral contraceptives, 2 received depot medroxyprogesterone acetate injections, and 4 had a levonorgestrel intrauterine device (IUD).

Hormone Assays

Baseline measurements used for this investigation from the participants with AN occurred prior to any intervention. We identified n=97 participants with AN and n=19 healthy subjects who had usable samples (Figure 1). We measured concentrations of AMH (immunoassay, Quest Diagnostics). Inhibin B was also measured via a commercially-available assay (chemiluminescence, Quest Diagnostics). Follicle stimulating hormone (FSH), a pituitary gonadotropin whose pulsatile secretion is essential for stimulating ovulation and ovarian function, was measured by electrochemiluminescence immunoassay. Given the very low concentrations of estradiol (E2) frequently observed in adolescents with AN, we used an ultrasensitive assay (liquid chromatography/tandem mass spectrometry) to allow for superior analytical sensitivity and detection. Samples were batched to minimize the chance of batch effects.

Fasting samples obtained from participants with AN were previously tested for serum concentrations of cortisol, insulin adiponectin, and ghrelin. These values were utilized to determine if serum hormonal assays correlated with predictors of ovarian reserve.

Statistical Analyses

Descriptive statistics were used to summarize all variables by AN status, and continuous variables were compared between groups using the Wilcoxon rank sum test. Categorical variables were compared between groups using Fisher’s exact test. Multivariable linear regression modeling was used to evaluate predictors of AMH; and backwards stepwise regression was used to select the final model, with variable entry into the full model at the p<0.10 level. Recursive partitioning was used to explore whether or not there were any clear cut-points in AMH associated with AN status. Statistical significance is defined at the two-sided 0.05 level for all tests, and no adjustments have been made for multiple comparisons.

Results

Demographic characteristics of the two groups are presented in Table 1. Consistent with their diagnosis, participants with AN had a percentage median BMI for age (median 80% [Interquartile range (IQR) 80%, 90%]) that was much lower than that of the healthy subjects (120% [110%, 140%], p<0.001). Subjects with AN had been diagnosed with AN for median 4 months (range 0 to 120 months) prior to study entry. In the cohort who had achieved menarche (n=77), the median duration of amenorrhea was 9 months (range 1 to 78 months). The n=9 participants who had no menses for <3 months were those who were receiving oral contraceptive pills at the time of recruitment who then discontinued them between 1 and 3 months prior to the baseline study visit without having subsequent spontaneous menses. Eleven girls had not yet experienced their first menses, and were not included in the calculation of duration of amenorrhea. These subjects had a bone age >13 years. In the healthy convenience subject sample, AMH levels did not differ between the girls who were using hormonal contraception compared to non-users (p>0.05; AMH range: 1.1–6.3 ng/mL). Based upon this result, we chose to utilize the convenience sample as a comparator group.

Table 1:

Demographic characteristics of adolescents with anorexia nervosa (AN) and healthy control subjects

	Control Subjects (n=19)	AN (n=97)	p-value
	Median (Q1, Q3)
Age, y	17.9 (16.4, 21.0)	16.6 (15.2, 18.2)	0.04
Height, cm	165.0 (159.0, 169.0)	162.7 (159.6, 168)	0.57
Weight, kg	65.7 (60.8, 75.7)	45.4 (40.6, 49.4)	<0.001
BMI, kg/m2	23.7 (22.4, 30.2)	17.0 (15.6, 18.4)	<0.001
Percentage of median BMI for age, %	120 (110, 140)	80 (80, 90)	<0.001
Duration of AN, mo		4 (1, 14)	---
Duration of amenorrhea, mo* (n=77)		6 (4, 10)	---
	Number (%)
Race
White	9 (50)	87 (90)	<0.001
Non-white	9 (50)	10 (10)
Missing	1	0
Ethnicity			0.28
Non-Hispanic	15 (88)	92 (95)
Missing	2	0

AN, anorexia nervosa; BMI, body mass index;

^*Excludes patients with primary amenorrhea

Contrary to our initial hypothesis, AMH levels were higher in participants with AN than in the healthy adolescents (Figure 2; p=0.03). Estradiol concentrations were significantly lower in subjects with AN (p<0.001). Neither FSH nor inhibin B measurements differed between subjects with AN and unaffected subjects (Figure 2).

Fig. 2.

Box and whisker plots comparing serum hormone concentrations between adolescents and young adults (AYA) with anorexia nervosa (AN) and comparison subjects. (A) Anti-Müllerian hormone; (B) estradiol; (C) inhibin B; and (D) follicle-stimulating hormone (FSH). Data presented include the median (dark black horizontal bar), values at the 25^th and 75^th percentiles (bottom and top edges of the colored box), and the minimum/maximum values (horizontal lines at the end of the whiskers).

We divided subjects with AN into two groups based upon the length of time they experienced no menses: shorter duration of amenorrhea (no menses for ≤ 3 months related to recent oral contraceptive pill use) and longer duration of amenorrhea (no menses for >3 months). Subjects with AN who had a longer duration of amenorrhea (median 7 months) demonstrated lower estradiol concentrations (medians: 16 pg/mL vs. 32 pg/mL; p=0.01) and a trend towards higher cortisol concentrations (medians: 11.7 ng/mL vs. 9.6 ng/mL; p=0.05) compared to subjects with a shorter duration of amenorrhea (median 3 months; Figure 3). However, AMH concentrations did not differ between these two groups (p=0.11). Only 3 AYA subjects with AN had an AMH < 1 ng/mL. Of these subjects, one had primary amenorrhea and two had secondary amenorrhea with duration 3 months.

Fig. 3.

Box and whisker plots comparing serum hormone concentrations between MA with a prolonged duration of amenorrhea (no menses for > 3 months) versus those with shorter duration of amenorrhea (no menses for ≤3 months). (A) Anti-Müllerian hormone; (B) estradiol; (C) inhibin B; and (D) cortisol. Data presented include the median (dark black horizontal bar), values at the 25^th and 75^th percentiles (bottom and top edges of the colored box), and the minimum/maximum values (horizontal lines at the end of the whiskers).

We compared participants with AN who had normal/low AMH levels to those with elevated AMH levels. We used recursive partitioning to identify a primary cut point in AMH associated with AN status, and calculated a value of 3.17 ng/mL. Given that this value did not appear to be clinically meaningful, we utilized a previously published cut-off value to distinguish between normal/elevated AMH (>6.78 ng/mL).^7,8 In 19.6% of participants with AN (n=19), AMH levels were elevated above this range. Within this cohort, those with elevated AMH concentration had a longer duration of AN diagnosis (9 months [2.5, 55]) than those with normal AMH levels (3 months [1.0, 10.8]; p=0.03). Fasting insulin levels were lower in subjects with AN and high AMH levels (3.7 mIU/mL [2.6,9.9] vs. 8.8 mIU/mL [4.1, 23]), but this difference did not reach statistical significance (p=0.08). Neither age, degree of malnutrition, or duration of amenorrhea differed between AN subjects with normal or elevated AMH measurements (Table 2).

Table 2:

Comparison of adolescents with AN who had elevated AMH concentrations and non-elevated AMH concentrations

	AMH < 6.78 ng/mL (n=78)	AMH > 6.78 ng/mL (n=19)	p-value
	Median (Q1, Q3)
Age, y	16.4 (15.1, 17.8)	16.9 (16.2, 18.5)	0.16
Percentage of median BMI for age, %	80 (80, 90)	80 (70, 90)	0.37
Duration of AN, mo	3 (1, 10.8)	9 (2.5, 55)	0.03
Duration of amenorrhea, mo (n=77)	6 (3, 10)	4 (4, 9)	0.70
Age at menarche, y (n=77)	12 (11.2, 13.0)	12.0 (12.0, 14.0)	0.38
FSH, mIU/mL	5.5 (3.3, 6.3)	4.8 (3.2, 6.7)	0.82
Estradiol, pg/mL	16.0 (0, 29.5)	25 (0, 45.5)	0.47
Inhibin B, pg/mL	25.5 (0, 47.8)	29.0 (22.0, 86.5)	0.07
Insulin, uIU/mL	8.8 (4.1, 23.0)	3.7 (2.6, 9.9)	0.08
Adiponectin, ng/mL	8.3 (6.0, 12.4)	7.9 (6.8, 10.7)	0.70
Leptin, ng/mL	2.3 (1.7, 4.0)	2.6 (1.6, 5.9)	0.63
Cortisol, mcg/dL	11.2 (8.7, 14.7)	10.8 (6.5, 13.5)	0.44

AN, anorexia nervosa; AMH, anti-Mullerian hormone; BMI, body mass index; FSH, follicle stimulating hormone

Within the larger randomized trial of AN, n=51 subjects had the opportunity to participate in a 6-month, longitudinal portion of the study. Of those longitudinal participants, n=9 subjects had elevated AMH concentrations at baseline. At the 6-month time point, 3 reported irregular menses, 3 continued to have amenorrhea, and 3 had been lost to follow-up prior to the 6-month time point.

In our linear regression modeling, variables considered for inclusion in the initial full model included AN status (yes/no), age, percentage of median BMI, hormonal contraception (yes/no), race (white/other), FSH concentration, and estradiol concentration. The remaining variables considered were not included because they did not meet the pre-specified significance criteria, and thus did not demonstrate association with AMH. Backwards stepwise regression resulted in a final model that adjusted for both AN status and age. Even after adjustment for age, AN status remained a significant predictor of AMH in the final model.

Conclusions:

In this AYA cohort, we demonstrated that AMH levels were higher in subjects with AN compared to both healthy participants and to published normative data.^7,8 Almost 1/5 of participants with AN demonstrated an elevated AMH level, as can be seen in polycystic ovary syndrome. Our findings suggest that ovarian reserve in adolescents with AN may not be negatively impacted as a result of the malnutrition and hormonal aberrations associated with AN, as we had hypothesized.

AMH levels have assumed increasing clinical utility over the last several years. Clinical applications include its use as a marker of ovarian response in in vitro fertilization cycles, assessment of ovarian reserve following chemotherapy/radiotherapy, and diagnosis of PCOS in adults. AMH acts as an inhibitor of primordial follicle recruitment and is important in dominant follicle selection in the follicular phase of the menstrual cycle.⁹ Serum concentrations of AMH are proportional to the number of developing follicles in the ovaries, and reflect ovarian reserve.¹⁰ AMH is considered to be a marker for the process of ovarian aging.⁹

Although serum FSH and estradiol levels are among the standard tests used in the diagnosis of menstrual dysfunction, they vary with different points during the menstrual cycle. In contrast, serum AMH levels do not fluctuate with the menstrual cycle, making them an attractive alternative.¹⁰ Furthermore, FSH and estradiol levels are affected by the use of exogenous hormones. The impact of hormonal contraception on AMH measurements is controversial⁹, with some studies indicating a suppression of AMH in users of hormonal contraception¹¹ while others demonstrated no impact.^12,13 In this cohort, we did not see differences in AMH values between users and non-users of hormonal contraception in our healthy comparison group, and thus elected to include both groups in our analyses.

Considering another clinical model, serum AMH levels are two- to three-fold higher among women with PCOS than among ovulatory women, consistent with the increased number of small antral follicles seen in PCOS.⁹ The number of these small follicles positively correlated with the severity of menstrual disturbances in PCOS, especially in women with amenorrhea.¹⁴ Taken together, these findings imply that anovulatory patients with PCOS have an increased number of small antral follicles producing high levels of AMH, which interferes with the action of FSH on follicles, leading to anovulation and amenorrhea. It is possible that a similar cascade of events leads to the prolonged amenorrhea seen in some patients with AN despite weight recovery, and is an area of further study. Additionally, a positive correlation was noted between fasting insulin and AMH levels in women with PCOS.¹⁵ In contrast, an inverse relationship between insulin resistance and AMH was seen in women without PCOS, hypothesized due to an abnormal effect of insulin action on AMH secretion from granulosa cells.¹⁶ In this study, we observed a trend towards lower insulin concentrations in patients with AN who had high AMH levels (Table 2), suggesting that if insulin plays a role, it does so in a different manner than that which occurs in PCOS. Decreased insulin secretion is common in malnourished patients with AN. Future studies of adolescents with AN are needed to understand the roles of insulin and AMH in both AN and PCOS.

Patients with AN often have functional hypothalamic amenorrhea (FHA), characterized by low estrogen levels and low/normal gonadotropin levels. In most studies, serum AMH concentrations in women with FHA^17,18 or hypothalamic hypogonadism are elevated. In a recent evaluation of a large cohort of women with hypothalamic hypogonadism, AMH measurements were increased compared to those of healthy, ovulatory control subjects, but lower than those of women with PCOS.⁷

Some patients with FHA may have increases in serum androgens during gonadotropin administration, similar to the response seen in women with PCOS, and also have ovarian morphology similar to women with PCOS.¹⁹ During long-term follow-up, some of these women developed features of PCOS as their hypothalamic-pituitary-ovarian axis recovered.²⁰ In contrast, Alemyar et al. did not find increased follicle counts or ovarian volume in adults with hypothalamic hypogonadism.⁷ The authors hypothesized that their findings were explained by either 1) a large pool of small antral follicles that were not counted during ultrasonography or 2) increased AMH production per antral follicle.⁷

The predictive value of AMH in hypothalamic amenorrhea has also been demonstrated. In adults with hypothalamic hypogonadism, the probability of weight increase and recovery of menstrual cycles increased as baseline levels of AMH increased.²¹ In our cohort, 6-month longitudinal follow-up was possible for 9 subjects who had elevated AMH concentrations. At 6 months, 3 reported irregular menses, 3 continued to have amenorrhea, and 3 were lost to follow-up prior to the 6-month time point. As such, during this short follow-up interval, elevated AMH concentrations were not predictive of menstrual recovery for adolescents with AN. Longitudinal studies should be conducted to explore this further in adolescents with AN. However, this finding may explain the positive association we observed between duration of AN and AMH levels, but not with duration of amenorrhea and AMH. Additional study limitations include the inclusion of control subjects using hormonal contraception, and the lack of corresponding ultrasound data. Lastly, the small sample size, and in particular small number of healthy control subjects, should be acknowledged which may limit the generalizability of the findings.

In conclusion, this study supports the finding that AMH levels are not low in the majority of AYA with AN. Therefore, low AMH in a patient with AN should raise clinical concern regarding ovarian reserve as in other patient populations, and should not be attributed to degree of malnutrition alone. Currently, AMH is not regularly assessed during routine AN clinical care. However, our findings suggest some clinical utility in identifying those patients with reduced ovarian reserve. Additionally, the convenience subject sample using hormonal birth control agents did not have suppressed AMH levels. Whether or not AMH should be assessed with any regularity in the AYA years is beyond the scope of this paper, but a low AMH identified in a healthy individual on birth control would also be of concern. Finally, potential links between the hypothalamic amenorrhea suffered by patients with AN and PCOS needs further exploration.