Between 2009 and 2017, a restorative reproductive medicine (RRM) clinic (now known as NeoFertility) operated in two sites in Ireland, Galway, and Dublin. All patients are medically managed for underlying health issues and treated while attempting to conceive naturally. There were five physicians providing services during this time, all of whom were licensed for family medicine in Ireland. All had received training in RRM, specifically natural procreative technology, which uses the Creighton Model fertility chart to assess menstrual cycle and reproductive function (23, 24). They underwent mentoring with the clinic medical director (PB) to RRM approaches and clinic procedures.

The data used for this study were routinely collected from patients during their initial and subsequent visits as recorded in the medical records. Additional clinical information customarily collected through phone, mail, e-mail, or survey, and placed in the chart. Data required for this study were abstracted from the existing medical records and entered into a database, with secondary verification of the information occurring as needed. Extracted data were anonymized with unique identifiers.

Starting in 2009, the clinic routinely checked serum estradiol levels repeatedly in all pregnancies. Blood draw for serum estradiol analysis was performed in the morning. The analyses were performed by clinical laboratories throughout Ireland and reported in standard units of pmol/L. The precise day of gestation at each blood draw was determined, as described below. Levels were checked weekly in very early pregnancy, prior to an ultrasound to establish early pregnancy viability, which was usually performed around 7 weeks gestation. After the pregnancy was noted to be viable on ultrasound, estradiol levels continued to be checked at intervals throughout pregnancy.

To identify patients with low estradiol levels in early pregnancy, starting in 2009, a reference range for levels of estradiol in pregnancies that proceeded to term was used from prior work by Check and colleagues, as shown in (Table 1) (11). Additionally, for this current paper, we generated a reference range of estradiol in this clinic population by analyzing retrospectively estradiol levels from 104 pregnancies of women who had term live singleton births, who were not supplemented with estradiol, DHEA or pregnenolone, and who did not have gestational diabetes mellitus. For the first trimester of pregnancy, the estradiol level was analyzed in one week intervals: from 3 weeks 0 days to 3 weeks 6 days, from 4 weeks 0 days to 4 weeks 6 days, and so on. Also, for the entire pregnancy, the estradiol level was analyzed in 2-week intervals: From 3 weeks 0 days to 4 weeks 6 days, from 5 weeks 0 days to 6 weeks 6 days, and so on. Each estradiol level was assigned to its associated weekly or biweekly interval and all estradiol levels of all patients for each time interval were averaged as well as the precise day of gestation within each time interval. This resulted in a clinic-specific reference range for estradiol values in pregnancies, shown in Figure 1, which we used for all subsequent analyses in this paper. In this group, we also explored whether estradiol levels varied by maternal age.

Relative serum estradiol levels were obtained by dividing the absolute estradiol value for each patient and gestational age to the respective reference level of the associated gestational age interval in the clinic-specific reference data (reference levels shown in Figure 1), yielding a percentage of the gestational-age specific estradiol level. The relative serum estradiol levels before vs. after estradiol or DHEA supplementation were obtained by calculating the average of all relative estradiol levels obtained 1–10 days before vs. the average of all available relative estradiol levels obtained 1–6 weeks after supplementation, for each pregnancy.

The inclusion criteria for the main analysis were all pregnancies clinically identified with low estradiol levels in the first trimester of pregnancy, who either did not receive supplementation with estradiol or DHEA (2009–2011), or who were treated with estradiol (2013–2015), or DHEA (2015–2017). Exclusion criteria were having a basal relative serum estradiol >50% of the reference level (specific for gestational age), multiple gestation, receiving supplementation with both estradiol and DHEA, having received supplementation only in the second or third trimester, or not having serum estradiol levels available before or after supplementation was started, and presence of a severe genetic disorder of the fetus. The number of women excluded for each criterion is given in Figure 2.

Estradiol and DHEA are available by prescription only in Ireland. Supplementation was by prescription, with the medication obtained from a pharmacy.

Gestational age was determined clinically by estimated day of ovulation (peak mucus day) (25, 26), and confirmed by early pregnancy ultrasound with crown-rump length, usually at 7 weeks gestation. Miscarriage was defined as loss of pregnancy before 24 weeks of gestational age. The presence of small or large for gestational age at birth was determined by using the 10th percentile values from the Intergrowth-21st standards for newborn weight (27). We assigned the birth weight at a gestational age of × weeks and 0–3 days to week x, while days 4–6 were assigned to week x + 1.

Data are presented as mean with standard error of the mean (SEM), unless indicated otherwise. One-way ANOVA with Bonferroni correction was applied to evaluate serum estradiol levels between estradiol and DHEA supplemented patients before or after supplementation. One-way ANOVA with Tukey's Honestly Significant Difference (HSD) test was applied to evaluate pregnancy outcome parameters. Logistic regression was used to determine the odds of miscarriage for patients given estradiol, or DHEA, as compared to neither hormone, unadjusted, and adjusted for woman's age, history of prior miscarriage, and history of prior live birth, which are characteristics associated with the risk of miscarriage. We also adjusted for the prescribed use of other medications, besides estradiol or DHEA, which had substantial differences in prevalence between the treatment groups.

The study was approved by Beacon Hospital Research Ethics Committee (BHREC), reference number BEA0187. The analysis was conducted under additional ethics approval from the University of Utah, IRB #80947. Data were abstracted from usual clinical data sources and patient anonymity was maintained by recording data as de-identified prior to analysis. There was no requirement for written informed consent of participants.

For the main analysis, we analyzed pregnancies identified with baseline estradiol levels <=50% of the gestational age reference range, who received no estradiol or DHEA treatment (n = 22; 2009–2011), or who received estradiol (n = 52; 2013–2015), or who received DHEA (n = 40; 2015–2017). The clinical characteristics of these pregnancies (114 pregnancies from 110 women) are given in Table 2. The mean woman's age was in the range 36–37 years; over 80% had prior pregnancies, and the mean number of miscarriages among those with prior pregnancies was 2.1–2.4. Most women (86%–92%) received progesterone during pregnancy, as well as other treatments shown.

As described above, reference data for serum estradiol in pregnancy were generated from patients with full term pregnancies from 2012 to 2017, who did not receive estradiol or other medication that might increase estradiol levels, including DHEA and pregnenolone. With a goal of analyzing about 100 pregnancies, 126 pregnancies were selected. After excluding multiple pregnancies or those with preexisting or gestational diabetes mellitus, 104 out of 126 pregnancies were included. A complete overview of clinical characteristics and treatments for this reference population is presented in Table 3. Mean serum estradiol at one week after estimated ovulation (p7) was 745 pmol/L (n = 41) (Figure 1A). Throughout the first trimester these estradiol levels steadily increased, leading to a mean 7,643 pmol/L (n = 25) at the 12th week of pregnancy. In the second and third trimester, serum estradiol levels kept increasing until the 31st week to mean 50,986 pmol/L (n = 13), after which the number of data points was considered insufficient to calculate mean estimates (Figure 1B). In addition, to explore whether age affected serum estradiol levels in this patient population, the 104 patients were split in patients of ≤35 years (n = 51, average of 32,5 years) and >35 years (n = 53, average of 38,2 years) at the day of conception. Estradiol levels were not statistically different between these age groups, although they tended to be lower in the elder group after 19 weeks of pregnancy (Figure 3).

From 2011 onwards, pregnant woman with low blood estradiol levels were identified clinically based on thresholds which were obtained from Check et al. (11) (Table 1). Women with low estradiol were supplemented with oral estradiol in 2013–2015 and in 2015–2017 with oral DHEA. After having established an internal reference estradiol curve for full-term pregnancies as described above and shown in Figure 1, cohorts of estradiol- and DHEA-supplemented patients were established, based on inclusion and exclusion criteria noted above, with exclusion criteria presented in Figure 2. This led to the inclusion of 52 out of 91 estradiol-supplemented patients (last menstrual period between May 2013 and January 2015) and 40 out of 129 DHEA-supplemented patients (last menstrual period between April 2015 and September 2017).

On average, both estradiol and DHEA supplementation started at 6.9 weeks of gestation. Baseline relative estradiol levels were a relative mean of 32.3% and 33.2% of reference levels for the estradiol and DHEA group respectively, which was not significantly different. After supplementation for 1–6 weeks, this reached a mean of 41.6% (p = 0.14, compared to baseline level) and 90.2% (p < 0.0001, compared to baseline level) (median of 37.7% and 87.2%). DHEA supplementation led to significantly higher relative serum estradiol levels as compared to estradiol supplementation (p < 0.0001). Oral estradiol did not increase serum estradiol levels to 100% of reference level in any patient, while 35% of DHEA supplemented patients reached this level (Figure 4). This response to DHEA supplementation is illustrated by representative patients treated with estradiol or DHEA only and in patients who were first treated with estradiol, followed by DHEA (Figure 5). In 17 patients with estradiol supplementation there was a decrease in relative serum estradiol levels, while all patients supplemented with DHEA had an increase.

To determine the potential effect of increasing serum estradiol on the rate of pregnancy loss, pregnancy outcomes in patients supplemented with estradiol (2013–2015 cohort) or DHEA (2015–2017 cohort) were compared with untreated patients (2009–2011 cohort); all patients initially had low estradiol levels in early pregnancy. The patient characteristics and other treatments were presented in Table 2. As shown in Table 4, the latter group the miscarriage rate was 45.5%, while this was 17.5% for the patients supplemented with DHEA (p = 0.038) and 21.2% (p = 0.067) for those supplemented with estradiol. There was no significant difference in the rates of low birth weight or pre-term deliveries among these groups, although the numbers of these events were small. Comparing serum estradiol levels in patients with subsequent viable pregnancy or miscarriage we found that in both estradiol and DHEA supplemented patients, pregnancies resulting in miscarriage had a blunted increase of serum estradiol with supplementation (Figure 6). Finally, we also compared the start and dose of estradiol and DHEA supplementation in the viable pregnancies and the miscarriages. We found that the average start of estradiol and DHEA supplementation was respectively 7.3 and 7.0 weeks in the viable pregnancies and 5.6 and 6.4 weeks with miscarriage while the average maximal dose was respectively 4.1 and 24.1 mg in the viable pregnancies and 4.4 and 26.4 mg with miscarriage, indicating that the full-term pregnancies with low serum estradiol were not related to a higher dose or earlier administration.

In a logistic regression model, the odds ratio (unadjusted) for miscarriage was 0.32 for treatment with estradiol, and 0.25 for treatment with DHEA. In models adjusted for woman's age, prior miscarriage, prior live birth, and/or the use of naltrexone and prednisolone, the adjusted odds ratios were very similar, but not statistically significant when naltrexone and prednisolone use was added to the adjustment (Table 5).

A clinical follow-up was obtained 5–7 years after birth, using the same questionnaire for all patients who received DHEA during pregnancy. The questionnaire informed for birth parameters, mode of delivery, complications for fetus or mother during pregnancy or at time of birth, concerns for motor or cognitive development of the child and for any health-related issues of children. The questionnaire was completed for 29 out of 33 children. In around 20% of pregnancies complications were experienced during pregnancy and at birth (Table 6). Few concerns were reported regarding the child's cognitive or motor development, while in 34.5% of children health-related concerns were reported including asthma (n = 3), progressive hearing loss (n = 1), Blount's disease (n = 1), migraine (n = 1), hydronephrosis (n = 1) and enlarged tonsils (n = 1). There were no reports of any sexual abnormalities.

Miscarriage is estimated to occur in up to 30% of pregnancies, depending on the population, with about half of these occurring within the first weeks after conception (28, 29). Although genetic aberrations are thought to cause around 50% of miscarriages in the first trimester (30, 31), the underlying cause in the majority of miscarriages is mostly not identified (32). It is critical to further identify the pathophysiological mechanisms to ultimately provide treatment, especially for women with recurrent miscarriage. In 1991, Check et al. demonstrated a strong correlation between pregnancy outcome and serum estradiol levels (11). Low estradiol would strongly indicate a deficiency in steroid production due to a lack of LH stimulation, or cellular response to LH during early pregnancy. It could also indicate a deficiency of the fetal adrenal-placenta unit during and after the transition to placental production of estradiol at 8–10 weeks. Supplementation of progesterone, is a more common method of treating presumptive corpus luteum deficiency in early pregnancy to reduce the risk of miscarriage (33) and 86%–92% of the patients in this study were supplemented, with no differences between the three treatment groups (Table 2).

Several older studies demonstrated a rise in serum estradiol after conception followed by further increases until delivery (11, 34, 35). These studies used radioimmunoassay to determine estradiol, while currently serum estradiol is mostly measured by non-radioisotope labelled immunoassays or liquid chromatography/tandem mass spectrometry-based methods (36, 37, 38). The latter methods have also been used to determine serum estradiol in healthy pregnancies (39, 40). The estradiol assays in this study were conducted at different clinical laboratories across Ireland, introducing an unknown level of between-lab variation. Thus, the reference levels used for this analysis cannot be directly compared to other laboratories and settings.

Among women with low serum estradiol levels in early pregnancy, the mean relative serum estradiol level increased from 33.2% to 90.2% (p < 0.0001), after supplementation with DHEA.

Among women with low serum estradiol in early pregnancy and no supplementation, the incidence of miscarriage was 45.5%, as compared to 21.2%, and 17.5% for patients supplemented with estradiol and DHEA, respectively. In multivariable models, adjusting for the use of low dose naltrexone and prednisolone (other medications that differed between the treatment groups), and/or for women's age, prior miscarriage, and prior live birth, there was a consistent reduced odds (0.19–0.33) of miscarriage with the supplementation of DHEA or estradiol, although only some of the models had effect estimates that were statistically significant. There might be an improved outcome when supplementation is started very early in pregnancy, or prior to pregnancy. A previous study demonstrated that prolonged DHEA supplementation before pregnancy increased the number of spontaneous pregnancies in couples who were preparing for IVF (21). Another prior study across two centers found that women with diminished ovarian reserve who were supplemented with DHEA for up to two months prior to IVF treatment had an unexpectedly low rate of miscarriage (about 15%) (41).

Together with progesterone, estradiol is essential for remodeling of the uterus to facilitate embryo implantation and placentation, also known as decidualization (42). Furthermore, estradiol stimulates the proliferative phase of endometrium, has vasodilation effects and is important for angiogenesis and therefore neovascularization (43–45), which is crucial to ensure transport of oxygen and nutrients to the growing fetus.

It is believed that DHEA primarily functions as a steroid precursor for the production of androgens and estrogens (20). During the luteal phase of the cycle and the first 10 weeks of pregnancy, most of the progesterone and estradiol are produced by the corpus luteum (46). Therefore, DHEA supplementation is most likely being utilized by the luteinized theca and/or granulosa cells for this steroid production. We cannot rule out peripheral and or adrenal utilization however, especially when ovarian function is deficient, as is for example the case in post-menopausal women (20). After 10 weeks of gestation, it is the placenta that utilizes DHEA to make estrogen, as discussed below.

A potential risk of estradiol supplementation is the potential development of thrombosis (47, 48), which could also negatively affect placental development (49, 50). This concern coupled with the fact that estradiol supplementation was less effective leads to a preference for the use of DHEA during pregnancy.

There is no literature available on prolonged DHEA supplementation in pregnancy. Nonetheless, there are various studies regarding long-term DHEA supplementation prior to IVF to improve pregnancy outcome (21, 51). In most of these studies 25 milligrams of DHEA was taken three times daily by mouth for up to six months (51–53). DHEA supplementation was associated with a decreased number of miscarriages (51, 53). Serum DHEA or estradiol levels were not measured in any of these studies (51). The side effects with these dosages were minimal; however, those reported, like hair loss, oily skin and acne vulgaris, are related to its androgenic effects (51). In postmenopausal women oral supplementation of 50 mg DHEA per day for 52 weeks was found to be safe (54). There are no published data regarding prolonged DHEA supplementation during pregnancy.

At the NeoFertility clinic serum estradiol was carefully monitored during pregnancy and DHEA was only supplemented with low serum estradiol levels. As DHEA supplementation in this study was restricted to woman with low serum estradiol levels and never exceeded 50 mg/day, we believe that abnormally high serum androgen levels and severe side effects are unlikely. However, androgen levels were not measured in this study. An increase of androgen blood levels could impact fetal development, particular for female fetuses.

The luteal to placenta shift in synthesis of estradiol is completed around 8–10 weeks of gestation (9, 10, 55). However, the placenta can only make estrogen if there are sufficient levels of DHEA for enzymatic conversion. The DHEA for estrogen production comes from a combination of maternal and fetal sources, with the fetal adrenal gland becoming predominant around 10 weeks of gestation (55). Furthermore, the production of DHEA by the fetal adrenal gland is tightly regulated (55, 56), therefore it is very likely that an increased maternal source of DHEA would reduce production by the fetal adrenal gland, preventing excess androgenization. In our long-term survey on children from DHEA-supplemented pregnancies, no sexual abnormalities were reported. In future studies, measurement of fetal androgens and the anogenital distance at birth would be valuable for addressing this potential concern.

Any drug or supplement used during pregnancy should be highly scrutinized to prevent harm to the developing fetus. An infamous example of this is the use of Diethylstilbestrol (DES) for the prevention of miscarriage and premature delivery from 1938 to 1971 (57). Diethylstilbestrol was also a form of estrogen supplementation, but it is a synthetic nonsteroidal estrogen which has a modified chemical structure and 3–4 times higher affinity for the estrogen receptor (57). This difference is likely the reason it was classified as carcinogenic to humans in 2000 (57) and is no longer in use. This is contrasted by DHEA which is bio-identical, itself considered a weak hormone and a physiologic requirement during pregnancy which is tightly regulated.

Our observations of the link between DHEA supplementation and estradiol levels were made in a clinical setting and were not found in previously published literature. A major limitation of this study, however, is its retrospective nature, and relatively small number of pregnancies treated with DHEA. The assignment of treatment was not randomized but was changed over time periods of the clinical operations based on clinical data and symptomatology. Although the treatment groups had similar clinical characteristics, the potential for other unmeasured confounding factors exist. Other temporal trends in treatment (as noted in Table 2) may have contributed to differences in outcome for the patients who were treated with DHEA. Randomized, controlled studies are needed to confirm our observations and to further assess for adverse effects. All patients in this study were being evaluated and treated for subfertility and/or history of pregnancy loss andthese results may not necessarily apply to other types of patients.