Progesterone levels during ovarian stimulation in segmented ART cycles inversely correlate with normalized birthweight of neonates conceived through subsequent frozen-thawed embryo transfer

Abstract

Purpose

To evaluate the association between serum progesterone (P) at the day of ovulation trigger and neonatal birthweight in singletons born after frozen-thawed embryo transfer in segmented ART cycles.

Methods

A retrospective multicenter cohort study involving data from patients who achieved uncomplicated pregnancy and term delivery of ART-conceived singleton babies following a segmented GnRH antagonist cycle. The main outcome was birthweight’s z-score of the neonate. Univariate and multivariate linear logistic regression analyses were made to investigate the relation of z-score with variables inherent to the patient and to the ovarian stimulation. The variable P per oocyte was created by dividing the value of progesterone at ovulation trigger by the number of oocytes retrieved at oocyte retrieval.

Results

A total of 368 patients were included in the analysis. At univariate linear regression, the birthweight z-score of the neonate appeared to be inversely related to both P levels at the ovulation trigger (− 0.101, p = 0.015) and P levels per oocyte at trigger (− 1.417, p = 0.001), while it was directly related to the height of the mother (0.026, p = 0.002) and to the number of previous live births (0.291, p = 0.016). In multivariate analysis, both serum P (− 0.1; p = 0.015) and P per oocyte (− 1.347, p = 0.002) maintained the significant inverse association with birthweight z-score after adjusting for height and parity.

Conclusions

Serum progesterone level on the day of ovulation trigger inversely correlates with normalized birthweight of neonates in segmented GnRH antagonist ART cycles.

Introduction

Pregnancies achieved by assisted reproduction technologies (ART) are burdened by an increased risk of adverse perinatal outcomes, including low or very low birthweight (LBW), fetal growth restriction (FGR), and small for gestational age (SGA) after fresh embryo transfer compared to natural conception [1–6]. When focusing on frozen-thawed embryo transfer (FET), studies are increasingly demonstrating that singletons have a higher mean birthweight and lower risk of LBW and SGA, but an increased risk of large for gestational age (LGA) and macrosomia compared to fresh embryo transfer [7–12] as well as compared to natural conception in epidemiological studies [6, 13, 14]. While registry and population studies are inevitably exposed to underlying bias [6, 15], a thorough assessment of the safety of ART in terms of obstetrical and neonatal outcomes remains a major objective of clinical research. ART treatments use several pharmacological or laboratory interventions and are characterized by diverse hormonal milieus: it is therefore of utmost importance to determine what factors might correlate the most with the adverse obstetrical outcomes. To this purpose, clinical studies are needed to provide relevant insights into the associations involved, and an in-depth collection of patients’ characteristics is warranted to allow for a reliable adjustment [16].

Kalra et al. conducted a study on oocyte donor recipients, who were not undergoing ovarian stimulation (OS), showing no differences in the risk of LBW when comparing FET and fresh transfer cycles and suggested a potential negative effect of ovarian stimulation on neonatal birthweight [17]. Indeed, studies focusing on the hormonal milieu preceding fresh ET have shown that the supraphysiological estradiol (E₂) levels at the end of the OS cycle is an independent predictor for term LBW [18–20]. Moreover, high late-follicular progesterone (P) levels have been shown to be associated with lower birthweight after fresh embryo transfer (in case of P > 2.0 ng/mL) [21], and more recently with ischemic placental disease (defined as preeclampsia, placental abruption, and/or SGA) for values in the range between 0.64 and 1.05 ng/mL (but not for higher levels), compared with lower values [22]. Interestingly, FET implies the separation of OS from embryo transfer, thus avoiding the exposure of the endometrium to the supraphysiological hormonal milieu [23, 24]. Nonetheless, recent studies are emerging suggesting that a correlation remains between hormone levels reached during OS and birthweight of singletons born after subsequent FET. In fact, a negative association was observed between peak E₂ levels during OS and birthweight z-scores, risk of SGA, and LBW of singletons born after FET in a large cohort [25], and between peak E₂ levels and birthweight of full-term singletons born after FET in a smaller cohort [26]. However, none of the studies exploring the associations between the hormonal milieu during OS and neonatal outcomes after FET has accounted for late follicular P levels. We therefore conducted the first study assessing the relationship between P levels at the end of OS and birthweight of singletons born at term after FET by encompassing a secondary analysis of our previous study on the live-birth outcomes of a freeze-all (or segmented) strategy in presence of normal versus elevated P levels [27] and enlarging the study cohort with additional consecutive patients undergoing a segmented (or freeze-all) ART cycle. P levels at trigger was studied both in terms of actual circulating P levels and in terms of P levels per oocyte retrieved, as this might estimate follicular exposure of each oocyte to P, with implications on quality [28].

Materials and methods

Study design

This was a retrospective multicenter cohort study involving data from patients who achieved term delivery of ART-conceived singleton babies following a segmented ART cycle and FET.

Population

Data collection was carried out at three tertiary care hospitals (Centro Scienze Natalita, San Raffaele Scientific Institute, Milan, Italy; Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy; Brussels IVF, Centre for Reproductive Medicine, Universitair Ziekenhuis Brussel, Belgium). First, the cohort of women enrolled in our recent study on CLBR in freeze-all antagonist-suppressed ICSI cycles with or without P elevation was included [27]. Then, consecutive FET cycles following homologous Gn-RH antagonist protocol ICSI treatment from January 2019 to January 2020 and resulting in a singleton uncomplicated pregnancy that achieved delivery at early (37 to 38 + 6 weeks of pregnancy), full (39 to 40 + 6 weeks of pregnancy), or late term (41 to 41 + 4 weeks of pregnancy) were also included. Patients with pre-term birth or with pregnancy complications such as gestational diabetes, hypertensive disorders, or antiphospholipid syndrome were excluded from the analysis. Indications for the freeze-all strategy, Gn-RH antagonist ART protocols, laboratory procedures for insemination, embryo quality assessment, cryopreservation, and endometrial preparation for the FET were standardized among the three centers and described in a recent publication of our groups [27]. In more detail, all FETs were performed in a postponed cycle, not immediately following oocyte retrieval. The FET took place in either a natural, modified natural, or artificial cycle. In a natural cycle, ovulation either occurred spontaneously or was triggered artificially (with the use of hCG) in a modified natural cycle. In artificially supplemented cycles, the preparation of the endometrium consisted of sequential administration of E₂ valerate (Progynova, Bayer-Schering Pharma AG, Berlin, Germany) and micronized vaginal P (Utrogestan, Besins Healthcare, Belgium; Progeffik, Effik Italia S.p.A., Italy; Prometrium, Meda Pharma S.p.A., Italy). When the endometrial thickness was > 7 mm, vaginal P supplementation was initiated. Patients who had the pre-implantation genetic testing of embryos were excluded from the analysis. The serum P levels were assessed on the morning of trigger administration (between 9:00 am and 11:00 am), using the Tosoh AIA fluorimetric system with the ST-AIA-PACK immunoassay (Tosoh Corporation) in Centro Scienze Natalita, San Raffaele Scientific Institute, Milan, Italy, and in Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy, while electrochemilluminescence immunoassay (Cobas 6000VR, Roche) was used in the Centre for Reproductive Medicine, Universitair Ziekenhuis Brussel. A cross-validation of the two systems was performed on 52 fresh samples that revealed a high correlation (R² = 0.9859) between the two sets of values, with the Roche assay result being an average − 0.09 ng/mL compared to the Tosoh system (corresponding to a mean % bias of 10.1) [27]. Intra-assay and inter-assay coefficients of variation were 10.7 and 12.9% respectively between the two centers in Milan [27].

Variables

Variables of interest included patient’s characteristics: age, height, pre-pregnancy body mass index (BMI), main cause of infertility, number of previous live births; basal hormonal profile (FSH, LH, TSH, AMH); OS data: length of stimulation, starting gonadotropin dose, total gonadotropin dose, peak estradiol level, progesterone level at trigger, number of oocytes retrieved, indication to FA; endometrial preparation for the FET: natural, modified natural, artificial; number of embryos transferred; embryo quality: grade A blastocysts, grade B blastocysts, grade C blastocysts, day 3 embryos; neonate’s characteristics: birthweight, sex, gestational age; pregnancy complications and type of delivery. Data are presented as n (%) for categorical variables and mean ± standard deviation for continuous variables. Progesterone at trigger day was also divided by the total number of oocytes retrieved at oocyte retrieval (OR) to create the variable “P per oocyte.”

Main outcome measure and sample size

The main outcome was z-scores of neonatal weight, calculated through the R software (version 3.6.0) and the R library “growthstandards,” using as reference the INTERGROWTH newborn standard (the likeness of fetal growth and newborn size across non-isolated populations in the INTERGROWTH-21st Project: the Fetal Growth Longitudinal Study and Newborn Cross-Sectional Study). We calculated sample size on the base of our largest time-series of neonates born by IVF/ICSI, with a mean birthweight of 3350 ± 417 g. For an expected difference of 200 g between the first and the last quartile and a statistical power of 90%, the required sample size was n = 360.

Statistical analysis

Exploratory univariate linear regression analysis was used to assess the relation between variables of interest and birthweight’s z-score of neonates. Categorical variables were analyzed in linear regression coded as dummy variables; the most numerous groups were used as reference. Variables showing a significant association with z-score (p value < 0.05) were included in the multivariate linear regression model. To avoid bias introduced by collinearity, serum P and P per oocyte were tested in two separate multivariate models (MVA 1 and MVA 2 respectively). Mixed models with the random-effects analysis of the three centers were made to account the possible differences of the sites. For descriptive purposes, quartiles of P at induction were calculated. Quartiles divide a set of observations into four groups, each of equal size and representing 25% of the observations. The SPSS software was used for the statistical analyses (IBM SPSS Statistics for Windows, Version 26.0. Armonk, NY, USA).

Results

Descriptive data

A total of n = 368 patients were included in the analysis. Characteristics of patients are listed in Table 1. The age was 34.8 ± 3.8 years (median ± SD) at OR and 35.5 ± 3.6 years at FET (median ± SD). The median ± SD length of stimulation was 10 ± 2 days; the starting dose and total dosage of gonadotropin during OS were 200 ± 75 IU and 2030 ± 870 IU respectively (mean ± SD). Levels of E₂ and P at ovulation trigger were 3221 ± 1725 pg/mL and 1.45 ± 1.14 ng/mL respectively (mean ± SD). The mean ± SD number of oocytes retrieved at OR was 14 ± 6.8. Thawing endometrial preparation was natural in 50 (13.6%), modified natural in 157 (42.7%), and artificial in 161 (43.7%). Gestational age at delivery was 279 ± 12 days (median ± IQR) and birthweight of neonate was 3374 ± 417 g (mean ± SD), with a birthweight’s z-score of 0.17 ± 0.91 (mean ± SD).

Table 1.

Patients’ characteristics, ART outcomes, and delivery outcomes

Age (years)
OR	34.8 ± 3.8
FET	35.5 ± 3.6
Height (cm)	164.7 ± 5.5
BMI (kg/m2)	21.6 ± 3.3
Main cause of infertility
Endometriosis	20 (5.4%)
Tubal factor	35 (9.5%)
Reduced ovarian reserve	1 (0.3%)
Oligoanovulation	43 (11.7%)
Recurrent miscarriage	9 (2.5%)
Genetic	10 (2.7%)
Male factor	142 (38.6%)
Idiopathic	108 (29.3%)
Previous live births (n)
0	311 (84.5%)
1	54 (14.7%)
2	3 (0.8%)
Hormonal parameters
FSH (IU/L)	6.5 ± 2.1
LH (IU/L)	6.1 ± 2.7
TSH (μIU/mL)	1.8 ± 0.7
AMH (ng/mL)	4.2 ± 3.1
ART cycle
Length of stimulation (days)	10 ± 2
Starting dose (IU)	200 ± 75
Total dose (IU)	2,030 ± 870
E2 levels at ovulation induction (pg/mL)	3,221 ± 1,725
P levels at ovulation induction (pg/mL)	1.45 ± 1.14
Number of oocytes retrieved (n)	14 ± 6.7
Indication to freeze-all
Increased risk of OHSS	147 (40%)
P elevation	125 (33.9%)
Clinician’s choice	58 (15.7%)
Irregular endometrium on ultrasound	21 (5.7%)
Patient’s illness around OPU	7 (1.9%)
Hydrosalpinx on ultrasound	5 (1.4%)
Impossible ET due to cervical stenosis	5 (1.4%)
Endometrial preparation (n)
Natural	50 (13.6%)
Modified natural	157 (42.7%)
Artificial	161 (43.7%)
Blastocyst transferred (n)
Grade A blastocyst	98 (25.9%)
Grade B blastocyst	181 (47.7%)
Grade C blastocyst	62 (16.4%)
Day 3 embryo	38 (10.0%)
Neonate
Gestational age (days)	278.5 ± 8.3
Birthweight (g)	3374 ± 417
Percentile	54.8 ± 27.3
z-score	0.17 ± 0.91
Delivery (n)
Vaginal	228 (62%)
Elective cesarean section	119 (32.3%)
Urgent cesarean section	21 (5.7%)

Regression models

Table 2 reports results from univariate and multivariate linear regression models predicting birthweight’s z-score of neonates. At univariate linear regression, the birthweight z-score of the neonate appeared to be directly related to the height of the mother (0.026, p = 0.002) and to the number of previous live births (0.291, p = 0.016), while it was inversely related to P levels at ovulation trigger (− 0.101, p = 0.015) as well as to P levels per oocyte (− 1.417, p = 0.001). None of the other variables showed a significant relationship with birthweight z-score, including pre-pregnancy BMI, basal hormonal profile of the patient, peak estradiol level during OS, total dose of gonadotropin used, number of embryos transferred, embryo quality, and endometrial preparation for FET. In the multivariate analysis (MVA), both serum P (− 0.1; p = 0.015) and P per oocyte (− 1.347, p = 0.002) maintained the significant inverse association with birthweight z-score after adjusting for maternal height and parity. ICC of random effect of the centers in the mixed model was 0/(0 + 0.779) for MVA1 and 0/(0 + 0.7716) for MVA2. Figure 1 shows the distribution of birthweight across increasing quartiles of P levels in the study population.

Table 2.

Univariate and multivariate linear regression models predicting birthweight z-scores

	UVA [B; p (CI 95%)]	MVA 1 [B; p (CI 95%)]	MVA 2 [B; p (CI 95%)]
Age	− 0.022; 0.08 (− 0.047, 0.003)
Height	0.026; 0.002 (0.01, 0.043)	0.022; 0.04 (− 0.181, − 0.02)	0.021; 0.013 (0.004, 0.038)
BMI	0.023; 0.13 (− 0.007, 0.053)
Parity	0.291; 0.016 (0.054, 0.528)	0.278; 0.021 (0.042, 0.514)	0.273; 0.022 (0.039, 0.507)
FSH	− 0.025; 0.375 (− 0.079, 0.03)
LH	− 0.012; 0.566 (− 0.055, 0.03)
TSH	0.026; 0.781 (− 0.155, 0.207)
AMH	0.024; 0.231 (− 0.016, 0.064)
E2 at trigger	0.015; 0.592 (− 0.07, 0.040)
P at trigger	− 0.101; 0.015 (− 0.18, 0.02)	− 0.1; 0.015 (− 0.18, − 0.02)
Total gonadotropin dose	0.000; 0.752 (0.000, 0.000)
No. of oocytes retrieved	0.009; 0.197 (− 0.005, 0.023)
Embryo quality
Day 3 embryo	− 0.127; 0.505 (− 0.5, 0.25)
Blastocyst grade C	0.135; 0.282 (− 0.112, 0.38)
Blastocyst grade B	Reference
Blastocyst grade A	0.191; 0.195 (− 0.1, 0.48)
Number of transferred embryos	− 0.155; 0.286 (− 0.44, 0.13)
Thawing endometrial preparation
NC	− 0.159; 0.29 (− 0.452, 0.135)
mNC	0.139; 0.18 (− 0.063, 0.342)
HRT	Reference
P per oocyte	− 1.417. 0.001 (− 2.283, − 0.55)		− 1.347; 0.002 (− 2.209, − 0.485)

Bold values denote statistical significance at the p value level < 0.05

UVA, univariate linear regression analysis; MVA 1, multivariate linear regression analysis encompassing P levels at trigger; MVA 2, multivariate linear regression analysis encompassing P per oocyte (i.e., P at trigger divided by n of oocytes retrieved)

Fig. 1.

Quartile 1 P < 0.72; quartile 2 P ≥ 0.72 and < 1.23; quartile 3 P ≥ 1.23 and < 1.8; quartile 4 P ≥ 1.8

Discussion

To the best of our knowledge, our study describes for the first time an inverse correlation between peak P levels at the end of the OS in segmented cycles and birthweight of term singletons born from uncomplicated pregnancies after subsequent FET. Our results show that late follicular phase peak P levels are related to neonatal weight in the subsequent FET cycle after controlling for confounders. Of note, previous studies suggesting supraphysiological E2 levels to be inversely correlated with neonatal birthweight after FET did not encompass P analysis [25, 26]: the extent to which concomitant elevated P might have contributed also to these previous findings cannot be determined. At the present stage, our results do not allow to infer any causal relationship. On one hand, the exposure of oocytes to elevated P during the follicular phase might reduce the growth potential of deriving blastocysts. The rationale for this hypothesis relies on studies suggesting that blastocyst trophectoderm quality positively correlates with neonatal birthweight [29], while elevated P at the end of OS is known to be associated with lower blastocyst morphological quality [27, 30]. On the other hand, common underlying conditions— – such as pPolycystic Oovarian sSyndrome (PCOS)— - might instead predispose a subpopulation of ART patients to both elevated follicular P during OS and lower neonatal birthweight. Serum P levels during the follicular phase are hypothesized to originate from different sources: from multiple follicular development and excessive ovarian steroidogenic activity in women with good ovarian response such as those with PCOS [31] and from disrupted inhibition of ovulation in women with poor ovarian response [32]. To explore a possible difference between these two mechanisms in terms of neonatal outcomes, we have separately analyzed both total P levels and P per oocyte and found that both variables were inversely associated with neonatal birthweight z-scores. Thus, women with higher P levels during OS might be at a higher risk of impaired placentation even after FET, based on underlying mechanisms that might occur in women with both good and poor ovarian response. One of such factors might be embryo quality, as an association has been described between poor blastocyst quality and placental findings such as villitis, distal villous hypoplasia, intervillous thrombosis, or multiple malperfusion lesions, compared to pregnancies deriving from transfer of good- quality blastocyst [33].

As a strength of this study, only freeze-all cycles were included: thus, our findings relate to the ET of the first choice embryo (rather than supernumerary) of each patient. Conversely, the main limitation of our study is its relatively small sample size. This hampers reliable dichotomous analyses on most interesting clinical outcomes such as the proportion of large or small for gestational age newborns. The clinical relevance of our findings needs to be investigated more in depth. Nonetheless, even relatively small clinical studies can be valuable in contributing to the current understanding the impact of patient-specific and cycle-specific ART parameters on obstetrical outcomes, if a thorough characterization of patients and ART cycles data is ensured. In fact, large cohort studies focused on obstetrical complications after FET have so far overlooked the hormonal milieu at the time of gamete maturation [34]. Another limitation of our study is the lack of correction for P at transfer day, as this was not routinely assessed during the study period. While previous data suggest that P levels on transfer day do not affect neonatal birth-weight [35] and thus reassure about the interpretability of our novel results, future studies on this subject should also encompass P levels at transfer day as a potential confounder. Furthermore, as the freeze-all policy is not routinely justified in ART, our study could not be conducted on an unselected population of patients but rather on patients with an indication to cycle segmentation. Whether specific indications for cycle segmentation could affect the relationship between P levels and neonatal weight will thus deserve further consideration. As an example, around 33% of patients in our study had P elevation as the indication to freeze-all and this represents a potential bias related to patients’ selection. As discussed above, common underlying conditions might in fact predispose a subpopulation of ART patients to both elevated follicular P during OS and lower neonatal birthweight. Nonetheless, the fact that P elevation indeed represents an indication to cycle segmentation strengthens the clinical significance of our results. As a further limitation of our work, we have excluded patients with pre-term delivery or with pregnancy complications. We deemed this approach as the most appropriate given the small sample size and the exploratory nature of our analysis, which is the first on the subject. However, additional larger studies also performing sub-analyses on whether elevated P might be associated with other obstetrical outcomes including pre-term delivery and pregnancy complications will be of interest. Further studies in this field are thus warranted, especially considering the emerging knowledge that neonatal weight is related to long- term health outcomes [36, 37]. In fact, P at trigger is to some extent a modifiable element in OS: previous studies that attempted to identify risk factors for P elevation during OS have found history of P elevation and basal P at the beginning of stimulation as variables that can significantly predict the occurrence of elevated P during OS, but also the total dose of FSH/hMG administered [38, 39]. If future studies confirm a correlation between elevated P levels and lower neonatal birthweight, preventive strategies to reduce the risk of P elevation should be evaluated.

Declarations

Conflict of Interest

The authors declare no competing interests. None