Starting and resulting testosterone levels after androgen supplementation determine at all ages in vitro fertilization (IVF) pregnancy rates in women with diminished ovarian reserve (DOR)

Abstract

Purpose

To investigate whether androgen conversion rates after supplementation with dehydroepiandrosterone (DHEA) differ, and whether differences between patients with diminished ovarian reserve (DOR) are predictive of pregnancy chances in association with in vitro fertilization (IVF).

Methods

In a prospective cohort study we investigated 213 women with DOR, stratified for age (≤38 or >38 years) and ovarian FMR1 genotypes/sub-genotypes. All women were for at least 6 weeks supplemented with 75 mg of DHEA daily prior to IVF, between initial presentation and start of 1st IVF cycles. Levels of DHEA, DHEA-sulfate (DHEAS), total T (TT) and free T (FT) at baseline (_BL) and IVF cycle start (_CS) were then compared between conception and non-conception cycles.

Results

Mean age for the study population was 41.5 ± 4.4 years. Forty-seven IVF cycles (22.1 %) resulted in clinical pregnancy. Benefits of DHEA on pregnancy rates were statistically associated with efficiency of androgen conversion from DHEA to T and amplitude of T gain. Younger women converted significantly more efficiently than older females, and selected FMR1 genotypes/sub-genotypes converted better than others. FSH/androgen and AMH/androgen ratios represent promising new predictors of IVF pregnancy chances in women with DOR.

Conclusions

DOR at all ages appears to represent an androgen-deficient state, benefitting from androgen supplementation. Efficacy of androgen supplementation with DHEA, however, varies depending on female age and FMR1 genotype/sub-genotype. Further clarification of FMR1 effects should lead to better individualization of androgen supplementation, whether via DHEA or other androgenic compounds.

Introduction

While androgen effects on follicle maturation in humans have remained controversial [1], experiments in rodents strongly support an essential androgen function, mediated by the androgen receptor (AR) on granulosa cells, especially at early stages of follicle maturation, up to preantral stages [2].

Direct studies of androgen effects on human follicle maturation do not exist. Androgens, indeed, for the longest time, have been considered detrimental to normal follicle maturation [1]. Similarities of receptor expression in mouse and human follicles, however, suggest that small growing human follicles may be as much androgen-dependent as mouse follicles [1].

Such a conclusion is also supported by clinical evidence: Supplementation with dehydroepiandrosterone (DHEA), in women with objectively determined DOR improves IVF cycle outcomes and pregnancy chances in general (reviewed in reference 3). These investigations were initiated after a 43-year-old patient with severe DOR developed a typical polycystic ovary (PCO) phenotype over consecutive in vitro fertilization (IVF) cycles after self-medicating with DHEA [3]. DHEA has been utilized to induce PCO-like ovarian phenotypes in rats [4], and mice [5]. From subsequent observational studies we then concluded that DHEA significantly and beneficially affects functional ovarian reserve (FOR), egg and embryo quality and overall pregnancy chances in women with DOR [6]. How DHEA supplementation causes these effects has, however, remained undetermined.

In women DHEA in a majority converts to testosterone (T), and to a lesser degree to estradiol [7]. The hormone, therefore, can affect processes through AR or estrogen receptors [1]. That a minimum of 6 weeks of DHEA supplementation are required to see significant beneficial effects [6, 8], supports a DHEA effect on relatively early follicle maturation stages, coinciding with presence of ARs, likely only present at small growing follicle stages, up to early preantral follicles [1]. Beneficially affected follicles, therefore, still will require weeks to months to reach gonadotropin-sensitive stages, where they become clinically relevant, explaining the need for at least 6 weeks of supplementation prior to IVF cycle starts [6, 8]. During this time period DHEA conversion dynamics into other androgens offer a potential model to investigate androgen effects on follicle maturation.

We recently reported preliminary clinical data, suggesting that T levels after DHEA supplementation are statistically associated with IVF pregnancy chances [9]. Increasing androgen levels have previously been associated with increasing AMH levels, and, therefore, better FOR [10, 11]. Better androgen levels have also been associated with higher IVF pregnancy chances [12], though data have remained conflicted [13].

Ovarian genotypes and sub-genotypes of the fragile X mental retardation 1 gene (FMR1) have recently been demonstrated to affect ovarian aging [14] and IVF pregnancy chances [15]. They differ from classical expansive FMR1 genotypes of premutation and mutation ranges, traditionally used to predict neuro-psychiatric risks, including the fragile X syndrome. Because of their effects on female reproduction, we suggested that fertility studies should now be stratified for these FMR1 genotypes and sub-genotypes [16]. This was done in this study.

Here presented study offers a systemic investigation of androgen effects on IVF and, by doing so, offers insights into the pathophysiology of DOR, androgen effects on pregnancy chances, differences at younger and older ages and of FMR1 effects. Here presented data also offer evidence for the importance of androgens to normal follicle maturation and female fertility in humans.

Materials and methods

Patients

This study involves a cohort of 213 women with primary or secondary infertility and a diagnosis of DOR. A DOR diagnosis was reached based on abnormal FSH [17] and/or AMH levels [18], outside of the 95 % confidence interval (CI) for age and/or because of age above 40 years [19]. To prevent confounding biases from repeat cycles, only first IVF cycles, which offered a complete data set as described below, were included.

Androgen supplementation

DOR patients at our center, prior to IVF cycle start, are for at least 6 weeks routinely supplemented with 25 mg of micronized pharmaceutical grade DHEA, p. o., TID [6].

Assessments of androgen levels

Androgen profiles, including DHEA, DHEA-sulfate (DHEAS), total T (TT) and free T (FT) were serially evaluated monthly, at DHEA supplementation start (i.e., “baseline”), tagged with the subscript “_BL”, and at IVF cycle start, with subscript “_CS”. All androgens were measured by routine commercial assays.

Ovarian stimulation for IVF

With first menses, after at least 6 weeks of DHEA supplementation, women with DOR were started in a microdose agonist ovarian stimulation cycle (leuprolide acetate, 50 μg, s.c., BID, Lupron®, Abbot Laboratories, North Chicago, Illinois), with daily gonadotropin dosage of 450–600 IU, all FSH but 150 IU, given as human menopausal gonadotropin (hMG) (gonadotropin manufacturers varied, based on patient preference/insurance). In DOR patients our center performs oocyte retrieval whenever at least one mature follicle of ≥20 mm mean diameter is seen on ultrasound.

Definition of clinical pregnancy

A pregnancy was defined as clinical if at least one gestational sac and fetal heartbeat were confirmed on vaginal ultrasound.

FMR1 genotypes and sub-genotypes

FMR1 genotypes and sub-genotypes are based on a normal range of 26 to 34 (median 30) CGG triple nucleotides (CGG _{n = 26–34}). Women with both alleles in normal range are considered to have a normal genotype (norm), those with one outside range as heterozygous (het) and those with both alleles outside range as homozygous (hom). Het and hom genotypes, depending on whether abnormal alleles are above (high) or below (low) normal range, can then be further subdivided into sub-genotypes [14–16].

Statistical analyses

Where appropriate, statistical analyses were carried out by using Mann–Whitney U tests, t-tests or ANOVAs. All post hoc procedures utilized the Holm-Sidak test but were confirmed with independent t-tests for significance. Differences in distributions between categorical data were compared using Chi-Square tests. All tests were 2-tailed, with P < 0.05 considered statistically significant.

A series of logistic regressions was performed to determine whether various androgen levels affected clinical potential for pregnancy in association with IVF. In interactions between two contributing factors the effect of the response variable depends on the value of the other factor. For multi-factor ANOVAs interaction terms of α were set at 0.05 in regards to individual contributions (at α ≤0.05 the 95 % CI indicates that factors of interaction terms are significant) but at 0.10 for interaction terms.

The first series of logistic regression models were adjusted (in weeks) for duration of DHEA supplementation from DHEA start (_BL) to IVF cycle start (_CS). When examining changes in various androgen levels, models were adjusted (in weeks) for time intervals between blood draw _BL and blood draw _CS. To correctly assess potential effects of androgens and potential effects of the inter-conversions between androgens on pregnancy odds, models were, where indicated, further adjusted for length of DHEA supplementation, oocyte yields and patient age. (Pearson product moment correlation coefficients confirmed that there were significant negative correlations between age and DHEA_CS, r = 0.274, P ≤ 0.001; age and TT_CS, r = 0.291, P = 0.002; age and FT_CS, r = 0.381, P ≤ 0.001; and age and oocyte yields, r = 0.258, P ≤ 0.001. There was also a significant positive correlation between oocyte yields and FT_CS, r = 0.204, P = 0.035.)

To correctly assess the differences between low, normal and high DHEA levels at cycle start, DHEA_BL and DHEA_CS were binned, and re-coded into categorical variables. To determine whether there was change from androgen _BL to androgen _CS, the same grouping paradigm was used at both time points. To offer a fair comparison, optimal cut offs were calculated, based on percentiles to yield equal sample sizes. Low, normal and high groups were assessed independently to examine the effects of androgens _BL and androgens _CS on pregnancy odds. They were for DHEA_BL and DHEA_CS, respectively: for low DHEA levels: <461 ng/dL (n_BL = 113_, n_CS = 70); for normal 461–661 ng/dL (n_BL = 45, n_CS = 69); and for high >661 ng/dL (n_BL = 45, n_CS = 69). Five patients missed cycle DHEA data but did have DHEAS data.

To assess pregnancy odds, factor increases or decreases in pregnancy odds were calculated as e^β for increases and 1/e^β for decreases, using the β coefficients derived from the logistic regression. To directly assess an androgen conversion index from DHEA to T, we created the following formula: Ratio = % change T/absolute value of % change in DHEA. The actual formula is:

The same formula is potentially applicable to conversion calculations for any androgen. This is a unit-less formula because a unit of DHEA is not necessarily converted into a unit of T (i.e., the molar ratio of DHEA to T is not 1:1).

Institutional review board (IRB)

Here presented data only involved the retrospective review of an anonymized electronic research database. Such studies at our center only require expedited IRB reviews since patients at initial consultation are asked to sign a universal informed consent, which allows for medical record reviews for research purposes if the patient’s medical record remains confidential and her identity protected. These conditions were met, and the study, therefore, underwent only expedited review. Patients also signed a specific consent for FMR1 testing. The center’s clinical and research staffs are, in accordance with federal HIPAA rules, in writing committed to confidentiality of medical records and privacy of individual patients.

Results

Table 1 summarizes patient and IVF cycle characteristics as well as primary infertility diagnoses for the study cohort. The 213 women (IVF cycles) had a mean age of 41.5 ± 4.4 years and a mean BMI (kg/m²) of 24.4 ± 5.0. Out of 213 women, 140 (65.7 %) carried a primary infertility diagnosis of DOR, with remaining patients having DOR as secondary diagnosis or being above age 40. The cohort’s mean FSH (12.2 ± 9.4 mIU/mL) and AMH levels (1.0 ± 1.3 ng/mL) are confirmatory of the diagnosis of DOR.

Table 1.

Patient, IVF cycle characteristics and primary infertility diagnoses

Characteristic	Results
	Total (n = 213)	Younger patients (n = 41)	Older patients (n = 172)
Age (years)	41.5 ± 4.4	34.6 ± 2.41	43.2 ± 3.01
BMI (kg/m2)	24.4 ± 5.0	23.4 ± 3.7	24.6 ± 5.2
Estradiol (pg/mL)	57.3 ± 38.0	60.3 ± 36.5	56.6 ± 38.4
FSHBL (mIU/mL)	12.2 ± 9.4	10.8 ± 6.0	12.5 ± 10.0
AMHBL(ng/mL)	1.0 ± 1.3	1.7 ± 1.62	0.8 ± 1.12
Gonadotropin dosage/IVF cycle (IU)	6011.8 ± 2330.8	4840.9 ± 1925.83	6304.5 ± 2336.43
Oocyte yield (n)	6.2 ± 5.6	9.0 ± 6.44	5.6 ± 5.24
Embryos Transferred (n)	2.0 ± 1.4	1.8 ± 1.2	2.0 ± 1.5
Embryos Cryopreserved (m_	0.9 ± 2.2	2.0 ± 3.35	0.6 ± 1.75
Race [n (%)]
Caucasian	159 (74.6)	31 (75.6)	128 (74.4)
Asian	31 (14.6)	8 (19.5)	23 (13.4)
African American	23 (10.8)	2 (4.9)	21 (12.2)
FMR1 [n (%)]
norm	106 (49.8)	22 (53.7)	84 (48.8)
het-norm/high	38 (17.8)	7 (17.1)	31 (18.0)
het-norm/low	58 (27.2)	8 (19.5)	50 (29.1)
hom	11 (5.2)	4 (9.8)	7 (4.1)
Primary infertility diagnoses [n (%)]a
Diminished ovarian reserve	140 (65.7)	19 (46.3)6	121 (70.3)6
Male factor	48 (22.5)	12 (29.3)	36 (20.9)
Endometriosis	9 (4.2)	3 (7.3)	6 (3.5)
Uterine factors	17 (8.0)	4 (9.8)	13 (7.6)
Polycystic Ovary Syndrome	11 (5.2)	4 (9.8)	7 (4.1)
Clinical Pregnancy [n (%)]
Per IVF Cycle	47 (22.1)	14 (34.1)7	33 (19.2)7

^aSome patients had more than one infertility diagnosis. Only 39 of 213 cycles ended up with cryopreserved embryos. Values are presented as means ± SD. Subscripts denote significant differences at P ≤ 0.05 (Chi-square and t-tests). ¹P ≤ 0.001; ²P ≤ 0.001; ³P ≤ 0.001; ⁴P = 0.003; ⁵P = 0.013; ⁶P = 0.004; ⁷P = 0.038

Younger (≤38 years) and older women (>38 years), of course, differed in mean age (P ≤ 0.001), but also AMH_BL (P ≤ 0.001), gonadotropin dosage (P ≤ 0.001), oocyte yield (P = 0.003) and number of embryos cryopreserved (P = 0.013). Younger women were less likely to have a primary diagnosis of DOR (P = 0.004)

A total of 47 clinical pregnancies were established (22.1 %). Younger women (≤38 years) conceived more frequently [14/41, 34.1 % vs. 33/172, 19.2 %; X² (1,N = 213) = 4.31, P = 0.038]. Pregnancy rates did not differ between FMR1 genotypes and sub-genotypes and between races.

Androgen conversion rates by age and FMR1

Two-way ANOVA for pregnancy outcome and age group demonstrated significant interactions [F(1, 58) = 5.32, P = 0.025]. Figure 1a demonstrates a significant difference between pregnant (M = 1.30, SD = 2.25) and non-pregnant groups (M = 0.33, SD = 0.86; P = 0.015), which, independently, was primarily observed amongst younger women (pregnant, M = 2.96, SD = 3.01; not pregnant, M = 1.05, SD = 1.44; P = 0.007), favoring better conversion in women who achieved pregnancy. Amongst pregnant patients, there was also a significant difference between younger women with better conversions (M = 2.96, SD = 3.01) and older women with lower conversion rates (M = 0.31, SD = 0.69; P < 0.001).

Fig. 1.

Androgen conversion rates by age and FMR1 genotypes and sub-genotypes. Two-way ANOVA for pregnancy outcome and age group demonstrated significant interaction (P = 0.025). a There was also a significant interaction between pregnant and non-pregnant women (P = 0.015), independently observed in younger women only (P = 0.007). Amongst pregnant women, older women also demonstrated a significantly lower androgen conversion rate from DHEA to T (P < 0.001). b demonstrates that androgen conversion rates differ significantly between FMR1 genotypes and sub-genotypes (P = 0.021), between younger and older patients (P = 0.003), and in the interaction term between FMR1 and age (P = 0.057). *Post-hoc comparison (Holm-Sidack) demonstrated that sub-genotype het-norm/high converts androgens significantly more efficiently than het-norm/low (P = 0.003), and younger convert androgens more efficiently than older women (P = 0.003)

Figure 1b demonstrates that the conversion rates from DHEA to FT significantly differed between FMR1 genotypes and sub-genotypes [F (3, 54) = 3.52, P = 0.021) and between younger and older patients [F(1, 54) = 9.55, P = 0.003]. The interaction term between FMR1 genotypes/sub-genotypes and age also almost reached significance (P = 0.057).

A post-hoc comparison (Holm-Sidack method) confirmed significant differences in conversion rates between het-norm/high (M = 1.38, SD = 2.60) and het-norm/low FMR1 sub-genotypes (M = 0.01, SD = 0.64; P = 0.003), with the former converting more efficiently than the latter; and between younger (M = 2.01, SD = 2.46) and older patients (M = 0.24, SD = 0.70; P = 0.003).

FSH_BL and AMH_BL comparison with FSH_CS and AMH_CS, and FSH/androgen and AMH/androgen ratios

FSH_BL and AMH_BL differed significantly between pregnant and non-pregnant patients (both P < 0.001), but with cycle start only AMH_CS had improved (P < 0.001) (Fig. 2a). Comparing androgens at baseline with androgens at cycle start, only FT_CS was significantly increased, comparing pregnant and non-pregnant women (P = 0.006) (Fig. 2b).

Fig. 2.

Comparison of FSH at baseline and cycle start, and ratios between. FSH and AMH with androgens. a Pregnant and non-pregnant women differed significantly in FSH_BL (P = 0.006) and AMH_BL (P = 0.006) but only in AMH_CS (P < 0.001). b In comparing androgens at baseline and cycle start only FT_CS differed significantly between pregnant and non-pregnant women (P = 0.006). C1 and C2 demonstrate the large number of FSH_BL/androgen ratios that demonstrated significant or almost significant differences between pregnant and non-pregnant women (for detail, see text). None of the FSH_CS/androgen ratios proved different. D1 and D2 demonstrate a similarly large number of significant AMH_BL/androgen ratios, and only one AMH_CS/androgen ratio (for detail see text). Adjustment for age eliminated all significant ratios, except for AMH_CS/DHEAS_CS., pointing at the age-dependency of most ratios

Since, especially at small follicle sizes, FSH and androgens are believed to act synergistically [20], we investigated ratios of FSH over the various androgens. Significant differences were noted between pregnant and non-pregnant women in: FSH_BL/DHEA_BL (P = 0.039), FSH_BL/DHEAS_BL (P = 0.040), FSH_BL/DHEA_CS (P = 0.013), FSH_BL/DHEAS_CS (P = 0.009) and FSH_BL/FT_CS (P = 0.014). FSH_BL/TT_CS almost reached significance (P = 0.051) (Fig. 2C1 and C2). Analyses of FSH_CS over cycle start androgen levels, proved all not significant (data not shown).

Similarly, the following AMH/androgen ratios were found to differ between pregnant and non-pregnant women: AMH_BL/DHEA_BL (P < 0.001), AMH_BL/DHEAS_BL (P = 0.002), AMH_BL/DHEA_CS (P < 0.001) and AMH_BL/DHEAS_CS (P = 0.002). AMH_BL/TT_CS (P = 0.005). In contrast to FSH_CS, which had not shown significant ratios, AMH_CS did demonstrate one significant ratio, AMH_CS/DHEAS_CS (P = 0.038) (Fig. 2d).

A series of logistic regressions, adjusting for age eliminated significance in all previously significant ratios, except for AMH_CS/DHEAS_CS (β = −0.499, SE = 0.240, P = 0.048). For every unit increase in ratio, clinical pregnancy potential decreased by a factor of 1.7.

Evaluation of pregnancy chances per DHEA, binned low, normal or high

Assessments of pregnancy chances, based on binned DHEA_BL (or other androgens) did not affect pregnancy chances, either based on low, normal or high DHEA levels or based on FMR1 genotypes and sub-genotypes (data not shown).

DHEA_CS, however, did: In presence of low DHEA_CS, odds of achieving pregnancy were significantly influenced by FT_CS (β = 0.948, SE ± 0.361, P = 0.009), increasing pregnancy odds 2.6-fold. In presence of normal DHEA_CS levels, FT_CS missed significance (β = 0.456, SE ± 0.255, P = 0.074). With high DHEA_CS levels, odds of achieving clinical pregnancy were significantly influenced only by DHEAS_CS (β = 0.607, SE ± 0.193, P = 0.002), increasing odds 1.8-fold (Fig. 3a). With normal levels of DHEA_CS, change of FT_BL to FT_CS was almost significant in affecting pregnancy chances (β = 1.662, SE ± 0.858, P = 0.053). With high DHEA_CS, the ratio of DHEAS_CS/DHEA_CS also significantly affected pregnancy chances (β = 0.040, SE ± 0.014, P = 0.005) (Fig. 3b), increasing pregnancy odds 1.04-fold.

Fig. 3.

Pregnancy chances per low, normal and high DHEA values. Predicted probabilities of pregnancy odds are presented as means and 1 SE from the mean. Different levels of DHEA_BL or of other baseline androgen levels did not affect pregnancy chances. As a demonstrates, in presence of low DHEA_CS, FT_CS, however, significantly influenced odds of pregnancy (P = 0.009). With increasing DHEA_CS, the significance of FT_CS, declines (normal DHEA_CS, P = 0.074). With high DHEA_CS, odds of achieving pregnancy were only significantly affected by DHEAS_CS (P = 0.002; data not shown). b summarizes significant changes in hormones between baseline and cycle start: With normal DHEA_CS, ΔFT_BL to FT_CS almost reached significance in affecting pregnancy chances (P = 0.053). With high DHEA_CS, the ratio of DHEAS_CS/DHEA_CS also affected pregnancy chances significantly (P = 0.005), though pregnancy odds barely changed (data not shown). c demonstrates associations after adjustments for age, oocyte numbers and length of DHEA supplementation: With low DHEA_CS, FT_CS was still significantly associated with increasing odds of pregnancy (P = 0.028), and the same applied to rising DHEAS_CS (P = 0.004) and DHEAS_CS/DHEA_CS ratio (P = 0.006) in presence of high DHEA_CS, though pregnancy odds barely changed

After adjusting for age, oocyte yields and length of DHEA treatment (time between DHEA_BL and DHEA_CS), FT_CS was, in presence of low DHEA_CS, significantly associated with increased odds of pregnancy (β = 0.882, SE ± 0.401, P = 0.028) by a factor of 2.4. In presence of high DHEA_CS, increasing DHEAS_CS (β = 0.568, SE ± 0.196, P = 0.004) and the ratio of DHEAS_CS/DHEA_CS (β = 0.040, SE ± 0.015, P = 0.006) also increased odds of pregnancy by factors of 1.8 and 1.04, respectively (Fig. 3c).

Pregnancy chances based on FMR1 genotypes and sub-genotypes

As Fig. 4a demonstrates, in women with het-norm/high FMR1 sub-genotype, TT_CS was associated with greater odds of achieving pregnancy (β = 0.106, SE ± 0.052, P = 0.043), increasing pregnancy odds, though only 1.1-fold. In contrast, norm and het-norm/low women demonstrated only a trend towards significance (data not shown). Even collapsed hom patients were too few in number to allow for valid statistical analyses. FT_CS almost reached significance in increasing odds of pregnancy in women with het-norm/low sub-genotype (β = 0.502, SE ± 0.267, P = 0.060), and a similar trend was observed in norm and het-norm/high women (data not shown).

Fig. 4.

Pregnancy chances based on FMR1 genotypes and sub-genotypes. Predicted probabilities of pregnancy odds are presented as means and 1 SE of the mean. a demonstrates that increasing TT_CS in women with het-norm/high FMR1 sub-genotype was associated with greater odds of pregnancy (P = 0.043). The collapsed hom genotype, overall, behaved differently from other FMR1 genotypes/sub-genotypes: Predicted probabilities for pregnant and non-pregnant groups in hom women were very similar, while other genotype and sub-genotypes demonstrated statistical trends. Rising FT_CS also almost reached significance in het-norm/low (P = 0.060) and so, though to lesser degree, did other genotypes and sub-genotypes. As b demonstrates, in het-norm/low sub-genotypes, interaction between DHEA_CS and FT_CS decreased odds of pregnancy (P = 0.056) and this conclusion remained valid after adjustments for oocyte numbers, age and length of DHEA supplementation (P = 0.064). Het-norm-high demonstrated a similar trend but not women with norm and hom genotype. Odds of pregnancy, however, did demonstrate a trend in norm women, based on ΔDHEAS_BL to DHEAS_CS (P = 0.062)

Moreover, in het-norm/low patients, interaction between DHEA_CS and FT_CS actually decreased odds of clinical pregnancy 1.3-fold (β = −0.295, SE ± 0.155, P = 0.056). Adjusted for oocyte yields, patient age, and time between DHEA_BL and DHEA_CS, the interaction between DHEA_CS and FT_CS significantly affected clinical pregnancy chances in women with het-norm/low FMR1 sub-genotypes (β = 0.333, SE ± 0.180, P = 0.064), actually decreasing odds of pregnancy 1.4-fold (Fig. 4b). A similar trend, not reaching significance was also observed for het-norm/high but not for women with norm genotypes (data not shown). Norm women, however, demonstrated almost statistically significant odds of increased pregnancy rates based on changes from DHEAS_BL to DHEAS_CS (β = 0.206, SE ± 0.111, P = 0.062) (Fig. 4b).

Androgens in FMR1 genotypes and sub-genotypes, stratified by age

In younger women (≤38 years), TT significantly changed between TT_BL and TT_CS assessments for the various FMR1 genotypes and sub-genotypes [F (2,10) = 5.53, P = 0.024]. Post-hoc comparisons (Holm-Sidak test) suggested the mean TT change to be significantly lower for het-norm/low (M = −116.69, SD = 152.29) than for het-norm/high FMR1 sub-genotypes (M = 28.00, SD = 57.98; P = 0.028) and norm genotypes (M = 26.38, SD = 25.59; P = 0.009) (Fig. 5a). As the figure suggests, het-norm/low women at younger ages demonstrate higher mean TT_BL than TT_CS.

Fig. 5.

FMR1 genotypes and sub-genotypes stratified by age. a demonstrates that in women ≤38 years Δ TT_BL to TT_CSFMR1 genotypes and sub-genotypes, overall significantly changed (P = 0.024). Post-hoc comparisons further demonstrated that this change in Δ of mean TT was significantly smaller for het-norm/low than het-norm/high (P = 0.028) or norm (P = 0.009) but mean ratio values between androgens did not differ. As b demonstrates, older women > age 38 behaved differently: DHEAS_CS differed significantly between FMR1 genotypes and sub-genotypes (P = 0.026), and post-hoc analysis confirmed that DHEAS_CS was significantly higher in hom than het-norm/high (P = 0.003) and het-norm/low women (P = 0.005). Older women, however, also demonstrated significant differences in mean androgen ratios: c demonstrates that DHEAS_CS/DHEA_CS differed significantly between FMR1 genotypes and sub-genotypes (P = 0.024), and post-hoc analysis suggested that hom women had a significantly higher ratio than het-norm/high (P = 0.003), het-norm/low (P = 0.005) and norm women (P = 0.013)

None of the mean ratio values between androgens showed significant differences for different FMR1 genotypes and sub-genotypes between baseline and cycle start.

Older women (>38 years) demonstrated a different pattern: They showed significant differences in mean DHEAS_CS between FMR1 genotypes and sub-genotypes [F (3,168) = 3.17, P = 0.026]. Holm-Sidak confirmed that with hom FMR1 genotypes (because of small numbers, hom sub-genotypes were not evaluated separately) DHEAS_CS was significantly higher (M = 578.29, SD = 475.85) than in het-norm/high (M = 334.96, SD = 182.53; P = 0.004) and in het-norm/low women (M = 349.06, SD = 188.51; P = 0.005) (Fig. 5b).

Moreover, older women also demonstrated significant differences in androgen ratios: Specifically, DHEAS_CS/DHEA_CS differed between FMR1 genotypes and sub-genotypes [F(3,164) = 3.22; P = 0.024]. Holm-Sidak post-hoc test suggests that mean DHEAS_CS/DHEA_CS for women with hom genotype (M = 126.4, SD = 126.2) was significantly higher than in women with het-norm/high (M = 63.5 %, SD = 39.6; P = 0.003) and het-norm/low FMR1 sub-genotypes (M = 69,4 %, SD = 42.7; P = 0.005) and with norm genotype (M = 76.8 %, SD = 47.5; P = 0.013) (Fig. 5c). As the figure demonstrates, this may suggest lower DHEA_CS in women with hom genotypes.

The change between DHEAS_BL and DHEAS_CS, in itself, indeed occurred significantly differently between FMR1 genotypes and sub-genotypes [F(3,165) = 2.98; P = 0.033). Holm-Sidak indicates that in hom women the increase in mean levels from DHEAS_BL to DHEAS_CS was significantly larger (M = 330.71, SD = 382.85) than in either het-norm/high (M = 57.49, SD = 210.99; P = 0.003) or norm women (M = 103.76, SD = 209.53; P = 0.009).

Discussion

Androgen conversion rates by age and FMR1

Women who conceived had better conversion rates than those who did not, and this was particularly apparent amongst younger patients (Fig. 1a). These observations are of significance because they suggest that: (a) better conversion from DHEA to T improves pregnancy chances in women with DOR; (b) younger women convert androgen more effectively than older women; (c) that older women may, therefore, potentially benefit from more direct T administration and (d) that the evaluation of androgen levels in women with DOR is beneficial.

Not shown in our result section is that we attempted by receiver operating curve (ROC) to determine whether, at least in younger women, below age 38 years, androgen conversion rates cold be predictive of pregnancy outcome. Likely due to small patient sample size, we were, however, unable to demonstrate significance (area under the curve 0.694, 95%CI 0.369–1.000; P = 0.262)

Conversion from DHEA to FT significantly differed between FMR1 genotypes and sub-genotypes (Fig. 1b), and the interaction term between FMR1 and age barely missed significance (P = 0.057). These observations support that, as previously reported [15, 16], variations in pregnancy rates between FMR1 genotypes and sub-genotypes in populations with better FOR would, be more apparent. This is also supported by the post-hoc analysis, demonstrating that, independent of age, het-norm/high sub-genotypes converted androgens significantly better than het-norm/low women, as did younger in comparison to older women.

Combined, these data point towards improved pregnancy chances for women who most efficiently convert DHEA to T. The data, however, also point towards differences in efficiency of conversion, based on FMR1 genotypes and sub-genotypes.

FSH_BL/CS, AMH_BL/CS, FSH/androgen and AMH/androgen ratios

This study also demonstrated that FSH_BL was lower and AMH_BS higher in patients who conceived (Table 1). Lower FSH and higher AMH, of course, denote better FOR and, therefore, better pregnancy chances are expected [17, 18]. By IVF cycle start, however, only AMH was left significantly associated with pregnancy chances (Fig. 2a), which correlates well with the earlier reported observations that AMH, but not FSH improves with DHEA supplementation, and that increases in AMH levels are associated with improved pregnancy chances with IVF [21].

Comparing androgen levels at baseline and cycle start between pregnant and non-pregnant women, only FT_CS differed significantly, demonstrating a significant increase from FT_BL in pregnant women (P = 0.006; Fig. 2b). This observation, of course, supports that efficient DHEA to FT conversion is beneficial for pregnancy chances.

Considering previously noted synergism between androgens and FSH [19, 22], and likely negative feedback of AMH on small growing follicles [23], we also attempted to investigate any possible interplay between androgens and FSH and androgens and AMH. Our hypothesis was that synergisms/antagonisms between FSH/AMH and androgens might be detectable by investigating FSH/androgen and AMH/androgen ratios. Neither had ever before been investigated, and since we presumed such interactions to occur at small growing follicle stages, we expected them to become more “visible” at baseline rather than at IVF cycle start.

Results, indeed, confirmed this by demonstrating significant differences in pregnancy chances based on the following FSH_BL/androgen ratios: FSH_BL over DHEA_BL, DHEAS_BL, DHEA_CS, DHEAS_CS and FT_CS. It barely missed significance with FSH_BL/TT_CS (P = 0.051) (Fig. 2C1 and C2). In contrast, FSH_CS ratios did not demonstrate any significant associations. This dichotomy of results is supportive of our hypothesis that androgen/FSH ratios at baseline may accurately reflect the small growing follicle pool, representing most of FOR [1]. AMH/androgen ratios further strengthen this concept. With only one exception, all pregnancy-associated AMH/androgen ratios related to AMH_BL, Those included ratios with DHEA_BL, DHEAS_BL, DHEA_CS, DHEAS_CS and TT_CS. Only AMH_CS/DHEAS_CS related to AMH_CS (Fig. 2D1 and D2).

FSH/androgen and AMH/androgen ratios, thus, appear to offer new, and quite direct assessments of the small growing follicle pool, representing true FOR. Currently, AMH is considered to best reflect this stage of follicle maturation [24]. Here presented data suggest that androgen/FSH and androgen/AMH ratios at baseline may have prognostic significance that warrants further investigation. These ratios, further, raise possible therapeutic benefits: For example, adjustments in androgen supplementation, based on FSH and/or AMH levels, may make it possible to bring ratios into, yet to be determined, therapeutic ranges.

Remarkably, only AMH_CS/DHEAS_CS maintained significance after adjustment for patient age. This confirms the importance of age in predicting IVF success but also points to the possibility that the AMH_CS/DHEAS_CS ratio may, independent of age, predict IVF pregnancy chances. Like all FOR assessment tools, AMH levels, in principle, are age-specific [18]. At 1.05 ng/mL, AMH, however, independent of age, has been reported in DOR to differentiate between better and poorer IVF pregnancy chances [24], and, in comparison to FSH, may, therefore, be more reflective of age-independent characteristics of FOR, required for pregnancy success. These data, ultimately suggest that, aside from the AMH_CS/DHEAS_CS ratio, all other here reported desirable ratios, likely, will have to be established in age-specific ways.

DHEA and other androgens binned for low, normal and high levels

DHEA_BL, and other androgens at baseline did not affect overall pregnancy chances or in association with individual FMR1 genotypes or sub-genotypes. In contrast, DHEA_CS did: If DHEA_CS was low, increasing FT_CS improved odds of pregnancy (Fig. 3a). As DHEA_CS improved towards normal, FT_CS slowly lost significance (P = 0.074), with high DHEA_CS levels no longer showing any association with FT_FS, and odds of achieving pregnancy only affected by rising DHEAS_CS.

This again points towards the importance of effective DHEA conversion to T but, likely, only if androgens are initially low. As baseline androgens increase, importance of further androgen conversion declines, suggesting the possibility of a therapeutic androgen range for maximal pregnancy chances. That at normal DHEA_CS levels changes of FT_BL to FT_CS only almost reached significance (P = 0.053) and that at high levels of DHEA_CS, once again, only the DHEAS_CS/DHEA_CS ratio significantly affected pregnancy chances (Fig. 3b) also support such an interpretation. It is further supported by the observation that after adjustments for length of DHEA supplementation, female age and oocyte yields, in presence of low DHEA_CS, FT_CS, still, was significantly associated with increasing pregnancy odds, and in presence of high levels of DHEA_CS, only increasing DHEAS_CS and the DHEAS_CS/DHEA_CS ratio were positively affecting pregnancy chances (Fig. 3c).

The data, therefore, also suggest that in women who poorly convert DHEA to T, the bottleneck in sequential conversions appears to lie at the conversion of [DHEA ↔ DHEAS] to androstenedione. So identified patients may, therefore, benefit from direct administration of T rather than DHEA.

FMR1 genotypes and sub-genotypes

FMR1 genotypes and sub-genotypes appear to reflect different degrees of androgen effectiveness, as TT_CS in het-norm/high women was significantly associated with greater odds of pregnancy (P = 0.043), and in norm and het-norm/low patients only suggested a trend towards significance (Fig. 4a). Similarly, FT_CS almost reached significance for het-norm/low and demonstrated trends for norm and het-norm/high.

Interaction between DHEA_CS and FT_CS in het-norm/low patients decreased odds of pregnancy (P = 0.056). However, most importantly, after adjustment for length of DHEA supplementation, female age and oocyte yields, the interaction between DHEA_CS and FT_CS still significantly affected pregnancy chances for women with het-norm/low sub-genotype; but only trends were observed with het-norm/high and norm (Fig. 4b). Least affected by conversion from DHEA_CS to FT_CS appeared to be women with norm genotypes but they, interestingly, demonstrated almost statistically significant odds of increased pregnancy rates (P = 0.062) based on changes in the DHEAS_BL to DHEAS_CS ratio, suggesting that in these patients androgen conversion rates may differ from other FMR1 genotypes (Fig. 4b), and higher T levels may be less important.

FMR1 genotypes and sub-genotypes and age

Stratification by age is important because it allows insights into two distinct forms of DOR, so-called premature ovarian aging (POA) in younger women, and physiologic aging with advanced female ages [17, 19]. Such stratification, once more, demonstrates differences in androgen conversion: young women (≤38 years) demonstrated significant differences in progression from TT_BL to TT_CS. Specifically, women with het-norm/low sub-genotypes demonstrated significantly slower increases in TT between these two time points than het-norm/high and norm patients (Fig. 5a). These data, therefore, suggest distinct differences in how FMR1 genotypes and sub-genotypes process androgens at different ages.

Women with het-norm/low sub-genotypes have been reported to demonstrate significantly reduced pregnancy chances with IVF [15, 16]. Considering the poor androgen conversion rates observed in this study in het-norm/low women, lower pregnancy chances with this FMR1 sub-genotypes should, therefore, not surprise.

While young women demonstrated no differences in mean androgen levels between FMR1 genotypes and sub-genotypes, older women demonstrated distinct differences in DHEAS_CS. Specifically, collapsed hom patients demonstrated significantly elevated DHEAS_CS in comparison to both het sub-genotypes (Fig. 5b) and higher DHEAS_CS/DHEA_CS ratios than het sub-genotypes and norm patients (Fig. 5c). Moreover, hom patients increased mean levels of DHEAS_BL to DHEAS_CS to a significantly larger degree than het-norm/high and norm women (Fig. 5d).

Based on their FMR1 genotype and sub-genotype, younger and older women with DOR, thus, appear to convert androgens at different rates. FOR differs in young oocyte donors, depending on race. At somewhat older ages the differences have, however, already significantly diminished in women with infertility [25]. At least partially, FOR differences observed between races are due to varying FMR1 genotype distributions, with African women demonstrating disproportionally more het-norm/low sub-genotypes and Asian women more het-norm/high [16]. IVF outcome differences between races, partially attributed to such variations in FMR1 genotype distributions [25], may, therefore, ultimately, reflect differences in androgen processing.

The increased importance of DHEAS_CS/DHEA_CS ratio in older women suggests that, lacking ability to convert [DHEA ↔ DHEAS] to T as well as younger women, they accumulate DHEAS. High DHEAS_CS/DHEA_CS ratios, independent of age, therefore, likely, suggest a poorer pregnancy prognosis. Younger women with poor androgen conversion, and older women in general may, therefore, benefit from direct T supplementation in place of or in addition to DHEA.

Here presented data offer also a first glimpse at the possible function of the hom genotype of FMR1. While functions of norm and het genotypes have previously been well defined [14–16], hom, because it is comparatively rare, is not well understood. Based on higher DHEAS_CS, this study suggests that hom women convert androgens poorer than either norm or het females. Hom patients, therefore, can be expected to demonstrate poorer responses to DHEA supplementation, and poorer clinical outcomes with IVF, though that remains to be confirmed by clinical studies.

Summary and conclusions

The evolving picture, therefore, suggests that younger women convert androgens better than older women, though how well androgens are converted appears to differ between FMR1 genotypes and sub-genotypes at younger as well as older ages, though with greater impact at younger ages. Better IVF pregnancy rates after DHEA supplementation in younger than older women support this observation [8].

Whether androgens are beneficial for normal follicle maturation and female fertility has in humans remained controversial [1]. Indeed, only recently, did two commentaries suggest that DHEA supplementation in women with DOR should not be considered [26, 27]. So far only one, underpowered clinical trial of DHEA supplementation has been reported [28], and most reported studies utilized lower power of evidence [6]. Lack of recruitment resulted in the abandonment of at least two prospectively randomized studies of DHEA in the U.S. and Europe, in which our center was involved (Government ID# NCT00419913). Two trials are still underway but recruitment is excruciatingly slow (Government ID# NCT00948857 and #NCY00650754), as, perceiving themselves “out of time,” DOR patients are difficult to recruit into randomizations to placebo.

It recently occurred to us that the investigation of androgen conversion rates after DHEA supplementation might improve our understanding of DHEA effects on ovaries. Like in the mouse, in humans granulosa cells of small growing follicles demonstrate AR, FSHR and AMHR [1]. Synergistic activity on granulosa cells between androgens and FSH at these stages appears likely [20, 22]. Effects of AMH are less well understood. Kristensen et al. recently reported that peripheral AMH correlates with peripheral androgens [11]. Androgens, FSH and AMH, through their interaction, likely, affect follicle recruitment and small follicle growth, androgens and FSH synergistically, and AMH inhibitory. In high recruitment situations, like polycystic ovarian syndrome (PCOS), AMH, therefore, is high, and in low recruitment environments, like hormonal contraception, pregnancy and DOR, AMH is low [25]. Genotypes and sub-genotypes of the FMR1 gene also appear to influence follicle recruitment, and are closely associated with ovarian aging patterns [14, 15]. That they vary in androgen conversion rates may offer a potential physiological link between these observations.

This study investigated a unique patient population of 213 DHEA-supplemented infertile women with DOR. All, because of a DOR diagnosis, had undergone FMR1 evaluations and had a complete set of baseline androgen data at DHEA supplementation start and, after at least 6 weeks of DHEA supplementation, at IVF cycle start. Finally, all patients completed a first IVF cycle, allowing for detailed outcomes assessments.

In such an unfavorable population the observed clinical pregnancy rate (positive fetal heart on ultrasound) of 22.1% is noteworthy, with women under age 38 actually conceiving in 34.1% of cycles. That pregnancy rates did not differ based on FMR1, may on first glance surprise since we previously reported that het-norm/low women experience significantly lower pregnancy rates than women with norm FMR1 genotype [15]. Those data were, however, generated in women with better FOR. Here presented data suggest that, once ovarian reserve is significantly compromised, ovarian FMR1 genotypes and sub-genotypes no longer matter to the same degree. This interpretation of results is also supported by the observation that FORs in young donors vary significantly based on FMR1; but in older infertility patients differences have largely disappeared [29].

DOR appears to represent an androgen deficient state, whether in younger or older women. To what degree, is currently under investigation at our center. Since androgens mostly derive from adrenal precursors, DOR, at least in part, has to reflect a clinical insufficiency state of the adrenal zona reticularis. DOR, therefore, mirrors in opposing features the hyperandrogenism of PCOS, for decades recognized as adrenal and ovarian in etiology.

FSH and AMH/androgen ratios are, to a rather surprising degree, predictive of pregnancy success. One can hypothesize that these ratios may better represent the small growing follicle pool than either FSH or AMH alone. They also offer potential therapeutic benefits by allowing for the individualization of androgen dosages, based on FSH and AMH levels, and, therefore, for diagnostic and therapeutic reasons deserve further investigations.

Abbreviations

AMH

Anti-Müllerian hormone

AMHR

Anti-Müllerian hormone receptor

AR

Androgen receptor

_BL

Baseline

_CS

Cycle start

DHEA

Dehydroepiandrosterone

DHEAS

Dehydroepiandrosterone sulfate

DOR

Diminished functional ovarian reserve

FMR1

Fragile X mental retardation 1

FOR

Functional ovarian reserve

FSH

Follicle stimulating hormone

FSHR

Follicle stimulating hormone receptor

het

Heterozygous

IVF

In vitro fertilization

hom

Homozygous

norm

Normal

PCO

Polycystic ovary

PCOS

Polycystic ovary syndrome

POA

Premature ovarian aging

T

Testosterone

FT

Free testosterone

TT

Total testosterone

Δ

Change