Obstetric and neonatal outcome of ART in patients with polycystic ovary syndrome: IVM of oocytes versus controlled ovarian stimulation

ABSTRACT

STUDY QUESTION

Does IVM of immature oocytes retrieved from small antral follicles in women with polycystic ovary syndrome (PCOS) have an impact on obstetric and neonatal outcomes compared to controlled ovarian stimulation (COS)?

SUMMARY ANSWER

Obstetric and neonatal outcomes after IVM appear to be similar to those after COS.

WHAT IS KNOW ALREADY

Women with PCOS have an increased risk of adverse pregnancy outcomes and congenital malformations in their offspring. For patients with PCOS who require IVF, IVM of germinal vesicle (GV)-stage oocytes retrieved from antral follicles has been adopted as a mild approach ART, with improved pregnancy rates over the last two decades. Although reports of obstetrical and neonatal outcomes after IVM have been reassuring, the limited sample sizes in previous studies preclude firm conclusions, and further study is warranted.

STUDY DESIGN, SIZE, DURATION

This is a retrospective observational study analysing obstetric and neonatal data from 1036 clinical pregnancies in unique patients with PCOS who conceived following a cycle of IVM or COS between January 2010 and December 2016 in a tertiary reproductive centre. In total, 393 singleton pregnancies with a gestational age beyond 20 weeks were included. A phenotypic approach was used for the diagnosis of PCOS. Pregnancies following oocyte donation, standard IVF (as opposed to ICSI) or preimplantation genetic testing and pregnancies requiring testicular biopsy in the male partners were excluded.

PARTICIPANTS/MATERIALS,SETTING, METHODS

Pregnancy outcomes were analysed in women with PCOS phenotype A, C or D, as defined by different combinations of the Rotterdam criteria. Data from 164 pregnancies beyond 20 weeks after IVM were compared with those from 229 pregnancies after COS. Pregnancies in the IVM group were obtained after minimal ovarian stimulation and IVF with ICSI of transvaginally collected GV oocytes that had reached the metaphase II stage in vitro after 28 to 40 h of culture. No hCG trigger was administered before oocyte retrieval. Outcome measures were analysed or reported in singleton pregnancies only and included adverse obstetric events and neonatal health parameters, in particular birthweight, prematurity, small-for-gestational age, large-for-gestational age, perinatal death and major/minor malformation rates. The incidence of hypertensive disorders of pregnancy (HDP) and birthweight was analysed by multiple linear and logistic regression, adjusted for relevant treatment variables and maternal characteristics.

MAIN RESULTS AND THE ROLE OF CHANCE

The IVM and the COS groups differed significantly (P < 0.001) for maternal circulating AMH levels and PCOS phenotype distribution, with more of the PCOS phenotype A in the IVM group. Pregnant women in the IVM group were younger than pregnant women in the COS group (P = 0.05). With regard to obstetric complications in singleton pregnancies, in the unadjusted analysis, mothers of infants in the IVM group more often had HDP (29/164 (17.9%) vs 22/229 (9.6%), P = 0.02) compared with mothers in the COS group. Singletons born after IVM and COS had a similar birthweight standard deviation score (SDS) (0.51 ± 0.94 after IVM vs 0.33 ± 1.05 after COS, P = 0.19). Preterm birth rate (32–36.9 weeks) and early preterm birth rate (<32 weeks) were also similar in both groups. The total malformation rate was 4.1% in singletons after IVM and 2.4% in singletons after COS. Multivariate linear regression analysis accounting for relevant confounders demonstrated that parity was the only independent predictive factor (P = 0.04) for birthweight SDS. Multivariate logistic regression analysis showed that BMI, parity and type of ART (IVM as opposed to COS) were significantly correlated with the incidence of HDP. Only patients with the PCOS phenotype A showed a tendency towards a higher risk of HDP in those who underwent IVM compared to those who had COS.

LIMITATIONS, REASONS FOR CAUTION

The study is limited by its retrospective nature and loss to follow-up of a subset of children with no information regarding congenital malformations. Furthermore, the paediatricians who assessed the children after birth were not blinded for the type of ART procedure.

WIDER IMPLICATIONS OF THE FINDINGS

This study provides further evidence that, compared to COS, IVM of oocytes derived from small antral follicles does not adversely affect the neonatal health of the offspring of patients with PCOS. The observed increased risk of HDP in patients with PCOS phenotype A following IVM treatment warrants further scrutiny.

STUDY FUNDING/COMPETING INTEREST(S)

Translational IVM research at Universitair Ziekenhuis Brussel (UZ Brussel) and Vrije Universiteit Brussel (VUB) has been supported by grants from the Institute for the Promotion of Innovation by Science and Technology in Flanders (Agentschap voor Innovatie door Wetenschap en Technologie—IWT, project 110680), the Fund for Research Flanders (Fonds Wetenschappelijk Onderzoek–Vlaanderen—FWO, project G.0343.13) and the Belgian Foundation Against Cancer (HOPE project, Dossier C69). Clinical IVM research was supported by research grants from Cook Medical and Besins Healthcare. M.D.V. reports honoraria for lectures from Cook Medical and Besins Healthcare outside the submitted work. S.S.R. reports honoraria for lectures by MSD and Besins and research grants by MSD, Ferring and Merck Serono outside of the submitted work. C.B. reports personal fees from Merck-Serono, Ferring, IBSA, Finox, MSD and Abbott outside the submitted work. H.T. reports grants from Merck, MSD, Goodlife, Cook, Roche, Besins, Ferring, Mithra (now Allergan) and the Research Fund of Flanders (FWO) and consultancy fees from Finox, Abbott, Obseva and Ovascience outside the submitted work. The other authors have nothing to disclose.

Introduction

IVM of oocytes is receiving growing interest as a mild alternative to ovarian stimulation before IVF in suitable patients. Because of the reduced hormonal burden for the patient and the absence of monitoring of ovarian response, IVM has been described as a ‘patient-friendly’ treatment. Although the first pregnancy after IVM in a patient with polycystic ovary syndrome (PCOS) was reported almost 25 years ago (Trounson et al., 1994), IVM has not been widely applied in fertility clinics. First, oocyte retrieval rates and maturation rates are lower after IVM compared to controlled ovarian stimulation (COS). Second, oocyte competence following maturation in the IVM culture systems currently available is generally lower compared to the competence of oocytes that acquire their full developmental potential within the follicular environment (Walls et al., 2015a). Live birth rates are consequently lower after IVM, and reports of favourable outcomes are scarce and have mainly been published by expert IVM teams (Walls et al., 2015a; Tannus et al., 2018, Ho et al., 2018). Finally, IVM had initially been advocated as an ovarian hyperstimulation syndrome (OHSS)-free approach for patients with PCOS. Since GnRH agonist maturation triggering and freeze all protocols have redesigned the landscape of ART in high responders, resulting in a significant reduction of the incidence of OHSS, OHSS prevention is no longer the key incentive for IVM. However, urgent fertility preservation in cancer patients is now an emerging indication for IVM.

In order to endorse IVM as a safe fertility treatment, it is crucial to monitor the health status of children born after IVM. However, the reported series of IVM children are small and publications that have reported on the health of IVM offspring have been flawed by a lack of consensus of the definition of IVM and the lack of a standardized IVM protocol. More specifically, reports of hCG-primed variants of IVM have not consistently detailed the meiotic status of the oocytes at the time of oocyte collection, which makes it almost impossible to distinguish between oocytes that have resumed meiosis within the environment of the follicle and those that have undergone critical processes of acquisition of competence in the culture dish. Indeed, a substantial proportion of ‘IVM children’ have in fact been conceived after the fertilization of an in vivo matured oocyte. According to rough estimates, more than 5000 babies have been born worldwide after IVM so far (Sauerbrun-Cutler et al., 2015), although there is no registry of centres that perform IVM, and the reported IVM data are, at best, incomplete (De Geyter et al., 2018). In spite of the aforementioned limitations of published studies reporting obstetric and perinatal outcomes after IVM, the findings of these studies have been reassuring, and outcomes after IVM and COS have been comparable (Cha et al., 2005; Mikkelsen 2005; Söderström-Anttila et al., 2006; Shu-Chi et al., 2006; Roesner et al., 2017). The largest series of IVM children published so far reported on children conceived in non-PCOS patients who received an hCG trigger before oocyte retrieval. In that study, in which most births were derived from oocytes found to be mature at recovery, the authors observed a higher birthweight in IVM children (Fadini et al., 2012). Nevertheless, concerns that this observation may be linked to epigenetic alterations in IVM oocytes have not been confirmed (Kuhtz et al., 2014; Pliushch et al., 2015).

When reporting health of IVM offspring, the selection of appropriate controls is crucial. Patients with PCOS are a heterogeneous group, and the PCOS phenotype can have a substantial impact on obstetric and neonatal outcome (Naver et al., 2014). Moreover, parameters of neonatal health may be influenced by the endocrine and endometrial milieu during the peri-implantation period.

In this single-centre study, we report the obstetric complications and neonatal outcomes including birth defects of singletons born to infertile patients with PCOS, following the transfer of embryos generated after IVM of GV oocytes as compared to those after generated after COS.

Materials and Methods

Study design and study groups

The present analysis represents an obstetric and neonatal follow-up of singleton pregnancies resulting from transfer of embryos generated using either IVM of GV oocytes from small antral follicles or COS in women with PCOS. All singleton pregnancies beyond 20 weeks’ gestation that resulted from an oocyte retrieval between January 2010 and December 2016 were included in this retrospective analysis, irrespective of the ART treatment cycle rank. Subsequent pregnancies following a first clinical pregnancy of >20 weeks’ gestational age in the same patient as well as multiple pregnancies were excluded from the analysis. The study included women between 18 and 36 years with PCOS as defined by a follicle number per ovary ≥12 on ultrasound scan (Balen et al., 2003). Women were diagnosed with PCOS according to recommendations based on extended Rotterdam criteria (National Institutes of Health. Evidence-based methodology workshop on PCOS, 3–5 December 2012. Executive summary available at https://prevention.nih.gov/docs/programs/pcos/finalreport.pdf). Briefly, the following classification was used: phenotype A: clinical or biochemical hyperandrogenism + ovulatory dysfunction + polycystic ovarian morphology (HA/OD/PCOM); phenotype B: hyperandrogenism + ovulatory dysfunction (HA/OD); phenotype C: hyperandrogenism + polycystic ovarian morphology (HA/PCOM); and phenotype D: ovulatory dysfunction + polycystic ovarian morphology (OD/PCOM). Because in our centre, IVM is only offered to patients with polycystic ovaries, apart from cancer patients who seek fertility preservation (De Vos 2016a) and patients with resistance to follicle stimulating hormone (FSH) (Galvão et al., 2018), none of the patients in this study had PCOS phenotype B.

Clinical hyperandrogenism was defined as the presence of hirsutism (Ferriman–Gallwey score > 8) and/or severe acne and alopecia; biochemical hyperandrogenism was defined as total serum testosterone >52 ng/dl and calculated free testosterone >0.64 ng/dl, based on the distribution (mean ± 2 SD) of these parameters in our standard population. Analysis of serum testosterone was performed using validated automated immunoassay methods (Elecsys electrochemiluminescence immunoassays on Cobas 6000, Roche Diagnostics), whereas serum AMH was analysed using different test kits across the study period (predominantly AMH Gen II ELISA, modified version, Beckman Coulter Inc. or ECLIA, Elecsys® AMH assay, Roche Diagnostics). Published conversion formulas and formulas derived from the authors’ data sets were used to homogenize AMH levels across different time intervals. Oligo-ovulation or anovulation was defined as a cycle length > 35 days or variation between consecutive menstrual cycles of >10 days. Pregnancies from patients with congenital adrenal hyperplasia, Cushing’s syndrome and androgen-secreting tumours were excluded. All patients in this study underwent an ART cycle either because they had previously failed to become pregnant after ovulation induction using clomiphene, letrozole or gonadotropins (including failure to ovulate with this hormone treatment) or because of male factor infertility. Patients were allowed to enter the study only once. Pregnancies following oocyte donation, standard IVF (as opposed to ICSI) and preimplantation genetic testing were excluded. Pregnancies in women with partners suffering from non-obstructive azoospermia and requiring testicular sperm extraction were also excluded.

Patients consented to undergo either IVM or COS after discussion with the fertility doctor at the outpatient clinic regarding the risks and benefits of each procedure. The option of IVM was more frequently discussed with patients presenting with strongly elevated antral follicle count and serum AMH levels, compatible with a more severe PCOS phenotype (Guzman et al., 2013). The IVM treatment in our centre is part of a research project and subject to informed consent. More specifically, patients who underwent IVM consented to donate a proportion of their cumulus–oocyte complexes (COC) for IVM research if 30 or more antral follicles were visible on pelvic ultrasound scan before oocyte retrieval, although the majority of COC were allocated to the standard IVM protocol as part of patient’s fertility treatment (previously described by Sanchez et al., 2017).

Ethical approval

The study was approved by the institutional review board on 21 March 21 2018 (B.U.N. 143201835600). Due to the retrospective nature of the study, written informed consent was not obtained from the participants.

Setting

Baseline characteristics, cycle characteristics, and obstetric and neonatal data were extracted from the electronic medical record. Records were ascertained from all PCOS patients who had a pregnancy beyond 20 weeks following IVM or COS during the study period. All study data were collected by authorized staff and stored in a restricted directory on the hospital’s network system. Data analysis was restricted to patients who had a singleton pregnancy. A flowchart is presented in Supplementary Figure S1.

Ovarian stimulation protocol and oocyte retrieval

In the IVM group, subcutaneous injections of highly purified human menopausal gonadotropin (HP-hMG, Menopur, Ferring Pharmaceuticals SA, Aalst, Belgium) were started on Day 5 after 2 to 3 weeks of combined contraceptive pill pretreatment (which was prescribed for scheduling reasons) or on cycle Day 3 of the menstrual period in a subset of women with regular cycles. Minimal ovarian stimulation was performed for three consecutive days, with a daily dose of 150 IU or 225 IU HP-hMG. A pelvic ultrasound scan was performed in the morning of the third stimulation day to schedule the oocyte retrieval. If none of the follicles had reached a diameter of 6 mm, then one to three further days of HP-hMG administration could be added, but caution was taken for the diameter of the largest follicles not to exceed 12 mm. No ovulation trigger was administered in any of the included IVM cycles. Vaginal ultrasound-guided retrieval of COC from small antral follicles for IVM was scheduled 42 h after the last HP-hMG injection.

In the control group, COS was conducted using recombinant FSH (rFSH) or HP-hMG in a GnRH antagonist or agonist protocol. The gonadotropin starting dose and the GnRH analogue were selected at the physician’s discretion. Final oocyte maturation was induced by injection of 5000 to 10 000 IU hCG (Pregnyl; MSD, Oss, The Netherlands) or injection of 0.2-mg triptorelin (Decapeptyl, Ipsen Pharma, Merelbeke, Belgium), as soon as two to three leading follicles were 17–18 mm in size as observed on ultrasound scan. Oocyte retrieval following COS was carried out 36 h after ovulation trigger.

All oocyte retrieval procedures in this study were performed using a 17-gauge single lumen needle (Cook Medical, K-OPS-1230-VUB, Limerick, Ireland).

IVM laboratory procedure and fertilization

In the IVM protocol, follicular aspirates were collected in human tubal fluid (HTF) (IVF Basics® HTF HEPES, Gynotec B.V. Malden, the Netherlands) supplemented with heparin (5000 IU/ml, Heparin Leo, Leo Pharma, Belgium; final heparin concentration 20 IU/ml) and filtered through a cell strainer (Falcon®, 70 μm mesh size, BD Biosciences, CA, USA). After collection, COC were washed and transferred to four-well dishes (Nunc; Thermo Fisher Scientific; MA, USA) containing IVM medium (IVM System, Medicult, Origio) supplemented with 75-mIU/ml HP-hMG (Menopur), 100-mIU/ml hCG (Pregnyl) and 10-mg/ml human serum albumin (Vitrolife, Göteborg, Sweden). COC were cultured for 28–40 h in groups of 10 COC per well in 500-μl IVM medium with an oil overlay (Ovoil, Vitrolife) at 37°C under 6% CO₂ and 20% O₂.

Insemination of all mature oocytes in this study was carried out using ICSI as described by Van Landuyt et al. (2005). Although IVF has been shown to be a valid fertilization technique for the insemination of IVM oocytes from PCOS patients with normal semen parameters (Söderström-Anttila et al., 2005; Walls et al., 2012), we have used ICSI as the fertilization technique for the insemination of IVM oocytes in all IVM cycles as a standard approach from the start of our IVM program in 2010.

Embryo culture and embryo transfer policy

Embryos generated after IVM or COS were cultured in individual 25 μl droplets of sequential media formulations (Quinn’s Advantage™ Fertilisation, Cleavage and Blastocyst medium, SAGE or Fert™, Cleav™, Blast™ medium, Origio). Selection of embryos for transfer or vitrification was done in the morning of the day of transfer, according to the morphological criteria described in Belva et al. (2016). When embryos and blastocysts were eligible for vitrification instead of fresh transfer, they were vitrified then warmed using closed CBS-VIT high-security straws (CryoBioSystem, L’Aigle, France) in combination with DMSO–ethylene glycol–sucrose as cryoprotectants (Irvine Scientific® Freeze kit, Newtownmountkennedy, County Wicklow, Ireland), according to the method previously described by Van Landuyt et al. (2011).

Embryos were selected for fresh transfer on Day 3 or Day 5 after ICSI. Alternatively, all embryos of good morphological quality were vitrified electively on Day 3 or on Day 5/6 after ICSI. The decision regarding extended embryo culture to the blastocyst stage after COS was determined at the discretion of the physician. In the early years of our IVM programme (2010–2014), all embryos were used for fresh transfer on Day 3 after ICSI. From 2014 onwards, IVM embryos were cultured to the blastocyst stage if at least four cleavage-stage embryos of good morphological quality (i.e. with at least six blastomeres and ≤20% fragmentation) were observed on Day 3; if not, IVM embryos were vitrified electively on Day 3 after ICSI. In the COS group, a ‘freeze-only’ approach was used in case of excessive ovarian response, defined as 19 or more follicles ≥11 mm on the day of trigger (Griesinger et al., 2016) or when serum progesterone levels were ≥1.50 ng/ml on the day of ovulation triggering (Bosch et al., 2010).

Preparation of the endometrium

IVM

The protocol for endometrium preparation for fresh embryo transfer after IVM in our centre also evolved over time. Briefly, two unit doses of Oestrogel® (Besins Healthcare; one unit dose of the Oestrogel® metered-dosing pump corresponds to 1.5 mg of gel and contains 0.75 mg oestradiol) were administered three to six times daily until seven weeks’ gestation, after which the dose was gradually reduced and discontinued 1 week later. Administration of Oestrogel® was started on the day before oocyte retrieval or the day of oocyte retrieval onwards. Until March 2014, treatment with intravaginal micronized progesterone (P, 200 mg three times a day; Utrogestan®, Besins Healthcare) was started on the day of oocyte retrieval. From April 2014 onwards, intravaginal micronized progesterone was started 1 day later, i.e. on the evening of the day of the ICSI procedure.

Vitrified-warmed embryo transfer (frozen embryo transfer, FET) after IVM was performed in an artificial endometrium priming cycle initiated when baseline hormone levels were reached after the IVM cycle. Briefly, the endometrium was primed with transdermal Oestrogel® (2 units administered three times a day), or with oral oestradiol valerate (Progynova®, Bayer-Schering Pharma AG, Berlin, Germany) at a dose of 2 mg three times daily, based on the clinician’s decision. When an endometrial thickness of more than 6 mm was reached, luteal support was started using intravaginal micronized progesterone tablets (200 mg three times a day), and transfer of one or two embryos was scheduled between 3 and 6 days later, depending on the stage of the embryo. Transfer of a Day 3 vitrified embryo was performed 1 day after warming, whereas vitrified blastocysts were transferred on the day of warming. Administration of oestrogens and P was continued until a pregnancy test was performed and was continued until 7 weeks of gestation if the pregnancy test was positive, after which the dose was gradually reduced and discontinued 1 week later.

COS

In case of fresh embryo transfer in the COS group, vaginal micronized progesterone tablets (Utrogestan®) 200 mg three times daily were administered for luteal phase support from Day 1 after oocyte retrieval onwards, until 7 weeks of pregnancy, after which the dose was gradually reduced and discontinued 1 week later. In case of GnRH agonist triggering, enhanced luteal support was administered, or embryos were cryopreserved electively (Devroey et al., 2011; Humaidan et al., 2013).

Embryos that had been cryopreserved electively in ‘freeze-only’ cycles and supernumerary embryos, all of which were vitrified on Day 3, 5 or 6 as described by De Vos et al. (2016b), were warmed and transferred in artificially supplemented cycles (HRT cycles) or in natural cycles, depending on the menstrual cyclicity of the patient and at the physician’s discretion. Artificial endometrium priming for FET following COS was performed as described above for IVM.

Outcome measures

Gestational age was calculated from the day of oocyte retrieval which was defined as Day 14 of the cycle. Stillbirth was defined as intrauterine or intrapartum death of a child born at a gestational age ≥ 20 weeks (elective terminations not included). Preterm birth was a birth between 32 and 37 completed weeks of gestation. Early preterm birth was a birth before 32 completed weeks of gestation. The cut-off for low birthweight was 2500 g at birth, and the cut-off for very low birthweight was 1500 g at birth. Small-for-gestational age (SGA) neonates had a standard deviation score (SDS) ≤ −2, whereas large-for-gestational age (LGA) was defined as a SDS of ≥2. Hypertensive disorders of pregnancy (HDP) included pregnancy-induced hypertension, (pre)eclampsia and haemolysis elevated liver enzymes and low platelets (HELLP) syndrome. Abnormal placentation included placenta praevia and placental abruption.

Obstetric and neonatal follow-up

The follow-up program of IVM pregnancies was similar to earlier studies and has been described previously (Belva et al., 2016). Briefly, early pregnancy hormonal and ultrasound monitoring was performed until 7–8 weeks’ gestation at the fertility clinic, after which antenatal care was provided by their hospital of choice. However, in a subset of pregnancies, this was not possible because patients came from abroad for fertility treatment and had pregnancy follow-up in their country of origin. Obstetric and neonatal data were therefore obtained from questionnaires sent to the parents after the due date. Questions related to demographics were filled out by the parents, and obstetric/neonatal data were provided by gynaecologists and/or paediatricians. After delivery, parents of Belgian children were invited for a detailed clinical/morphological assessment of their 4-month-old children at the institutional clinical genetics outpatient clinic which was run by certified paediatricians who were not blinded for the ART procedure (IVM or COS). Belgian parents who were unable to attend this clinic were asked to provide the results of a clinical exam of their baby at the age of four months. At the time of the outpatient clinic appointment, the written questionnaire information was verified with the parents and adjusted when necessary. Although the above strategy to ascertain obstetric and neonatal data resulted in available data for approximately 90% of all liveborns, children from patients living abroad were not invited for a clinical exam, and a substantial proportion of Belgian children did not attend the clinical genetics outpatient clinic, which led to lack of information about congenital malformations in 21/164 (12.8%) IVM singleton pregnancies and 63/229 (27.5%) singleton COS pregnancies. The definitions and classification of congenital malformations have previously been described (Bonduelle et al., 2002). Malformations were considered major when they generally cause functional impairment or require surgical correction. The remaining structural abnormalities were considered minor malformations. Cryptorchidism was considered a major malformation when a surgical intervention was planned. The total major malformation rate was calculated as follows: (affected liveborns + affected stillborns + elective terminations) divided by (total liveborns + total stillborns + elective terminations). In contrast with obstetric and neonatal outcome parameters, which are presented for pregnancies with a gestational age beyond 20 weeks (n = 164 IVM and 229 COS pregnancies), congenital malformations are presented for the entire duration of pregnancy, and the 20-week cut-off does not apply, as such a cut-off would potentially lead to an underestimation of the malformation rates in the groups under study. Hence, two pregnancies that were electively terminated before 20 weeks’ gestation because of a congenital malformation in the foetus are also included (n = 166 IVM and 229 COS pregnancies).

Statistical analysis

Descriptive statistical analysis was performed on main maternal and ART characteristics. Continuous data were presented as the mean value ± standard deviation (SD), and categorical data were presented by the number of cases and corresponding percentage. Categorical data and continuous data that did not show a normal distribution were analysed by Pearson’s χ² test/Fisher exact test or Kruskal–Wallis test as appropriate. Multivariate linear regression analysis was performed with a significance value (P value) set at 0.05 to identify variables related to birthweight. Birthweight was expressed as standard deviation score (SDS) in order to adjust for gestational age and gender (Niklasson and Albertsson-Wikland, 2008). Candidate predictive factors of birthweight SDS were type of ART (IVM or COS), parity, PCOS phenotype, patient’s age at embryo transfer, AMH, BMI, presence of HDP, embryo stage at transfer (Day 3/Day 4 or Day 5/Day 6), fresh or vitrified-warmed transfer and single embryo transfer or double embryo transfer. Multivariate logistic regression analysis using stepwise backward selection was performed with a significance value (P value) set at 0.05 to identify variables related to HDP. Candidate predictive factors of HDP were type of ART (IVM or COS), parity, PCOS phenotype, patient’s age at embryo transfer, AMH, BMI, embryo stage at transfer (Day 3/Day 4 or Day 5/Day 6), fresh or vitrified warmed transfer and single embryo transfer or double embryo transfer. All analyses were performed in IBM SPSS statistics version 24.

Results

Pregnancy outcomes

Among 393 singleton pregnancies in patients with PCOS that were ongoing after 20 weeks, 164 pregnancies resulted from IVM, and 229 pregnancies resulted from COS. After IVM, 160 pregnancies resulted in a liveborn. One IVM pregnancy was terminated because of agenesis of the corpus callosum; three were stillborn. After COS, 229 singleton pregnancies resulted in 225 singleton liveborns; four pregnancies resulted in stillbirth. To facilitate comparison of obstetric and neonatal outcome parameters between pregnancies after IVM and COS, data analysis was restricted to singleton pregnancies with confirmed outcome. All data are derived from unique patients. (Supplementary Fig. S1).

Patient characteristics

Circulating AMH levels and PCOS phenotype distribution of patients achieving a singleton pregnancy beyond 20 weeks differed significantly between the IVM and COS groups (P < 0.001), with a predominance of PCOS phenotype A in patients in the IVM group (Table I). Patients in the IVM group were younger than their COS counterparts (P = 0.05). Maternal BMI and parity were similar between groups. According to the patient records, dysovulation was mentioned as the indication for fertility treatment in 142 patients (85.5%) in the IVM group, with male factor being mentioned as an indication in only 3 (1.8%) IVM patients, which led us to suggest that male factor infertility was probably underreported in the IVM group. In contrast, dysovulation was recorded as the indication for fertility treatment in 85 patients (37.1%) in the COS group, and male factor was the indication in 101 patients (44.1%).

Cycle characteristics and embryo transfer policy

The number of mature metaphase II stage oocytes were similar between groups (12.0 ± 6.1 after IVM vs 9.6 ± 5.7 after COS, P = 0.29) (Table II). In the group of IVM pregnancies, patients more often underwent Day 3/Day 4 transfer (104/166 (62.6%) vs 75/229 (32.7%), P < 0.001) and vitrified-warmed embryo transfer (106/166 (63.9%) vs 69/229 (30.1%), P < 0.001) compared with COS pregnancies.

Obstetric outcomes

HDP were more often reported in pregnancies achieved after IVM compared to pregnancies following COS (29/164 (17.9%) vs 22/229 (9.6%), P = 0.02) (Table III). After adjusting for potential confounders, multivariate logistic regression analysis using stepwise backward selection for HDP demonstrated that BMI and parity, as well as the ART method (IVM vs COS), were significantly associated with HDP (Table IV). Other parameters, including age at ET, serum AMH, fresh or vitrified-warmed ET, day of ET, SET or DET and PCOS phenotype were not associated with HDP. Patients with PCOS type A who had undergone IVM had a tendency towards a higher risk of developing HDP compared with those who had undergone COS (24.1% vs 9.6%, P = 0.06, Table V). The vanishing twin rates were similar in the IVM group (3/164 (1.8%)) and the COS group (5/229 (2.2%)). After IVM, there was a tendency towards a higher incidence of preterm labour (25/164 (15.2%) vs 21/229 (9.2%), P = 0.07). The incidence of other pregnancy complications, including haemorrhage, gestational diabetes, abnormal placentation, cholestasis and nausea and vomiting, was similar in both groups. The mode of delivery was also similar (Table III).

Neonatal outcomes

Neonatal outcomes of singleton pregnancies after IVM and COS are presented in Table VI. Gender distribution was similar between groups. Birthweight SDS was similar in neonates conceived after IVM and in those after COS (0.51 ± 0.94 vs 0.33 ± 1.05, P = 0.19). The proportion of SGA (0.0% vs 1.5%) and LGA singletons (7.4% vs 5.0%), as well as the rate of low birthweight (6.1% vs 7.8%), was comparable between the IVM and COS groups. Although gestational age at delivery of IVM liveborns was significantly lower (38.0 ± 3.7 versus 38.7 ± 2.7 weeks, P = 0.03), preterm birth rates were similar in both groups. After adjustment for potential confounders (type of ART i.e. IVM vs COS, age at embryo transfer, BMI, AMH, parity, day and mode of transfer (fresh ET vs FET), PCOS phenotype, number of embryos transferred and presence of HDP), birthweight SDS was positively correlated with maternal parity (adjusted odds ratio 0.119, 95% CI: 0.011 to 0.579, P = 0.04), but not with the type of ART (IVM vs COS) nor with other patient or cycle characteristics (overview in Table VII). Birthweight SDS was also not influenced by the IVM incubation time of immature oocytes (30 h IVM vs 40 h IVM vs COS; 0.44 ± 0.90, 0.67 ± 1.02; 0.33 ± 1.05, resp., P = 0.11).

Congenital malformations

Out of 145 IVM singleton liveborns with available information about malformations, six had a major malformation. After COS, four out of 166 liveborns with available information had a major malformation. The total malformation rate was 4.1% following IVM and 2.4% following COS. Minor malformations were observed in 8/145 (5.5%) IVM singletons and in 2/166 (1.2%) singletons after COS. An iterative list of malformations is presented in Table VIII.

Discussion

This is the largest study evaluating obstetric and neonatal outcome of pregnancies following IVM of GV oocytes in patients with PCOS. In line with previously published smaller series, our data demonstrate that children conceived after IVM have a similar adjusted birthweight compared to their counterparts conceived after COS. Our data show comparable rates of congenital malformations and preterm birth in IVM offspring compared to their counterparts conceived after COS, but pregnancies following IVM were more prone to hypertensive disorders compared to those after COS.

Initial concerns with the safety of human IVM had been raised from animal studies showing that in vitro production of embryos, including IVM, may affect embryonic gene expression and result in large offspring syndrome in cattle and sheep (Sinclair et al., 2005). Apart from the lower efficiency of IVM compared to COS, this safety concern has precluded a more widespread application in reproductive medicine. We have previously shown that IVM does not interfere with genomic imprinting establishment in human oocytes (Kuhtz et al., 2014). To investigate whether IVM may be associated with impairment or disruption of normal patterns of epigenetic imprinting, the epigenetic stability of functionally important DNA methylation patterns has previously been analysed in chorionic villus and cord blood samples from IVM newborns, compared with children born using conventional ART (Pliushch et al., 2015). Although these studies have shown reassuring findings, their small sample sizes preclude the generalizability of the results.

The first live birth after IVM in a woman with PCOS was reported in 1994; nevertheless, data related to neonatal health in children conceived using IVM are still scarce, for various reasons. First, in the early years of human IVM, the efficiency of IVM programs was deceivingly low compared to the success in ruminant animal IVM, and attempts in the late nineties to improve the pregnancy rates after IVM have focused on an array of hormone interventions combined with in vivo oocyte maturation. Indeed, clinical ‘IVM’ protocols developed at that time recommended the use of a single large bolus of hCG, 34–38 h prior to oocyte retrieval (Chian et al., 2000). This practice has been adopted by a number of other groups and typically results in a mixture of a few oocytes with expanded cumulus that have already completed meiosis and reached the metaphase II stage within the follicular compartment (these in vivo matured oocytes are associated with a higher probability of pregnancy) and a larger cohort of recovered oocytes with compact cumulus, most of which have not yet undergone germinal vesicle breakdown (GVBD). Consequently, for the large majority of IVM children conceived following this ‘hCG-primed’ approach, the oocyte of origin (in vitro matured versus in vivo matured) cannot unequivocally be assigned, and an unknown proportion of ‘IVM’ children have not been conceived using in vitro matured oocytes. The confusing semantic debate on the definition of IVM represents a significant limitation of any follow-up study focusing on the safety score of IVM (De Vos et al., 2016c). Second, in view of the reduced efficiency compared to conventional hormone-driven ART protocols, IVM has not been widely applied. Therefore, the low-level application of IVM has resulted in small series of IVM offspring scattered across fertility centres around the globe, and these children have not been registered in a large-scale database. Third, there is widespread belief that the greater potential application of IVM is to rescue the currently unusable immature oocytes retrieved in standard IVF, rather than IVM in unstimulated cycles (Escrich et al., 2018). Nevertheless, there is compelling evidence that immature oocytes collected in COS cycles have poor intrinsic potential (Jones et al., 2008), and IVM media are primarily designed to meet the nutritional needs of the somatic cells supporting the oocyte during IVM. Moreover, there are no published follow-up data of children conceived after rescue IVM.

Our observation that IVM does not have an adverse effect on neonatal health in comparison with COS is in line with existing literature data. The largest study published so far and focusing on IVM births that had been derived with certainty from oocytes matured in vitro included 71 singleton liveborns (Fadini et al., 2012) following fresh embryo transfer on Day 2 or Day 3 after ICSI. The authors did not observe any statistically significant differences in terms of (unadjusted) birthweight, gestational age and minor abnormalities. None of the IVM children were diagnosed with major congenital malformations. Previous small series of neonatal follow-up of pregnancies after non-hCG-triggered IVM had shown normal obstetric and perinatal outcomes but had not included any control comparator group (Mikkelsen 2005, Cha et al., 2005, Söderström-Antilla et al., 2006). Outcomes of pregnancies obtained after hCG-triggered IVM have also been reported, and also in these small series, IVM was not associated with adverse obstetric outcomes or congenital malformations (Buckett et al., 2007; Son et al., 2007). A major limitation of these studies, apart from their small sample size, is the absence of adjustment for potential confounders. Indeed, there is accumulating evidence that birthweight and major obstetric complications such as HDP and preterm delivery may be modulated by maternal parameters and by hormonal treatment, embryo manipulation and embryo culture. IVM of oocytes has historically gained attention among fertility specialists for its favourable risk profile in patients with PCOS. Because oocytes are harvested from small antral follicles and there is no need for hCG as an ovulatory trigger, the risk of OHSS is theoretically absent and, so far, no events of OHSS have been reported in patients undergoing IVM. We have previously developed a prediction model (Guzman et al., 2013) showing a strong correlation between antral follicle number and circulating AMH levels, and eligibility for IVM, which led us to suggest that patients at the severest end of the phenotypic PCOS spectrum in terms of AMH and AFC are the best suitable candidates for IVM treatment. In the study presented here, 50.6% of patients in the IVM group had been diagnosed with PCOS phenotype A, compared to only 22.7% in the COS group. This phenotype is characterized by elevated levels of circulating AMH and free testosterone (Fraissinet et al., 2017) and an increased incidence of unfavourable metabolic markers, such as insulin resistance, resulting in a higher risk of long-term metabolic and cardiovascular complications (Moran and Teede, 2009, Lizneva et al., 2016). Moreover, these PCOS phenotype A patients exhibit a significantly increased risk of pregnancy complications, including pregnancy-induced hypertension and preeclampsia, gestational diabetes and premature delivery, compared to women with more favourable PCOS phenotypes (Palomba et al., 2015). In our sample, 14/164 (8.5%) pregnancies resulting from IVM cycles were complicated by preeclampsia, as compared to 8/229 (3.5%) pregnancies resulting from COS cycles. The incidence of HELLP syndrome was also higher after IVM (3.0% vs 0.9%). The incidence of pregnancy-induced hypertension was comparable between the two groups (6.1% after IVM vs 5.2% after COS). Our observation, in the unadjusted analysis, that pregnancies in the IVM group had an almost 2-fold higher incidence of HDP compared to pregnancies after COS led us to perform a multivariate logistic regression analysis to identify factors correlated with HDP. Based on this analysis, a significant correlation was identified between HDP and BMI, parity and IVM. No correlation was found between the PCOS phenotype and HDP. We hypothesize that the observed association between IVM and HDP in our study could potentially be linked to poor placentation due to unfavourable patient characteristics, suboptimal hormonal preparation of the endometrium or epigenetic dysregulation. It is plausible that patients with PCOS phenotype A in the IVM group had a more unfavourable hormonal (e.g. androgen levels) and/or metabolic profile compared to their COS counterparts. Patients with PCOS type A who underwent IVM (n = 83) had higher average serum AMH levels (12.49 +/− 7.88 ng/ml), compared to those who underwent COS (n = 52, 7.81 +/− 3.79 ng/ml). Nevertheless, serum AMH levels in women with PCOS have not been associated with an increased risk of HDP (Valdimarsdottir et al., 2019).

Pregnancies after IVM were predominantly realized after transfer of a Day 3/Day 4 embryos, in contrast with pregnancies in the COS group, which resulted mostly from Day 5/Day 6 transfer. Although favourable live birth rates have been described after transfer of blastocysts following IVM (Walls et al., 2015a), extending embryo culture after IVM to day 5 may result in a substantial risk of having no embryo available for transfer. Indeed, the same Australian group has demonstrated that embryos generated from IVM have an increased rate of early embryo arrest from day 3 to day 4, compared to their counterparts generated after COS (Walls et al., 2015b). This observation has led us to adhere to an adapted embryo transfer policy for IVM (as detailed in the material and methods section) resulting in a day 3 (fresh) or day 4 (after vitrification) transfer in the majority of IVM cycles. Some cohort studies reported a higher risk of being born LGA (Maheshwari et al., 2013; Mäkinen et al., 2013; Dar et al., 2014; Zhang et al., 2018) after extended culture to the blastocyst stage, although these observations were not confirmed by other studies (Chambers et al., 2015; De Vos et al., 2015; Maxwell et al., 2015; Oron et al., 2015; Ginström Ernstad et al., 2016).

Almost two-thirds of pregnancies after IVM in our study resulted from FET, mostly elective FET (eFET), whereas almost two-thirds of COS pregnancies were obtained after fresh embryo transfer. The preponderance of elective FET in the IVM group should be considered in view of previous publications by our group (De Vos et al., 2011) and others (Walls et al., 2015a), showing a clear benefit in success rates of vitrified-warmed embryo transfer compared to fresh embryo transfer after IVM. The poor success rates after fresh embryo transfer in IVM cycles are most probably linked to suboptimal endometrial receptivity due to the short follicular phase. All FET cycles following IVM were performed in an artificial endometrium priming protocol (HRT protocol), whereas 68.1% of FET cycles in the COS group were HRT cycles and 31.9% were (modified) natural cycles. According to a recent meta-analysis investigating outcomes after elective frozen versus fresh embryo transfer in ART cycles, FET resulted in a significantly higher birthweight and a higher risk of preeclampsia, but not pregnancy-induced hypertension, compared to fresh embryo transfer (Roque et al., 2019). The authors suggested that the observed increased risk of preeclampsia after FET may be related to abnormal placentation following endometrial priming with oestrogens in HRT cycles and in the absence of a corpus luteum. Nevertheless, the mode of embryo transfer (FET versus fresh ET) was not correlated with HDP in our study, according to the results of the multivariate regression analysis. Exogenous hormonal priming of FET cycles, more specifically the use of vaginal progesterone for luteal support in FET cycles, may also have an impact on early pregnancy loss (Labarta et al., 2017) although the study of miscarriage risk after IVM compared to COS was not within the scope of this study. Finally, endometrial preparation for FET in the IVM group was consistently performed using transdermal estradiol gel, as opposed to a more frequent use of oral estradiol valerate in FET cycles in the COS group, although the route of oestrogen supplementation does not seem to have an impact on clinical outcomes, according to a retrospective study in oocyte recipients (Madero et al., 2016).

Our study illustrates the relevance of the phenotypical classification of PCOS in clinical studies not only when investigating the efficiency of reproductive treatments (Ramezanali et al., 2016; De Vos et al., 2018) but also when evaluating the impact of different fertility treatments on obstetric and neonatal outcomes. Pregnancies after IVM do not appear to be negatively impacted regarding neonatal health, and IVM per se does not affect birthweight, but the higher incidence of adverse obstetric outcomes in IVM pregnancies in our study warrants close surveillance and timely detection of hypertensive disorders in IVM pregnancies in patients with PCOS phenotype A in particular. In this respect, optimizing outcomes after IVM should not only focus on the improvement of laboratory culture conditions but should also relate to preconceptional efforts to mitigate the negative impact of metabolic problems, including hyperinsulinism and hyperandrogenism in women with PCOS.

Nevertheless, our study has a number of limitations. First, the results of our study need to be interpreted in the light of the retrospective design. Moreover, our study sample was subject to loss to follow-up of a subset of children, with no information about congenital malformations in 24/193 (12.4%) IVM singleton pregnancies and 133/568 (23.4%) singleton COS pregnancies. Furthermore, the paediatricians who assessed the children after birth were not blinded for the type of ART procedure. Finally, the results of this study cannot be generalized to all IVM offspring; IVM is not only performed in patients with PCOS but also in the setting of fertility preservation in non-PCOS women and in rare cases of resistance to FSH.

In conclusion, IVM of oocytes derived from small antral follicles in patients with PCOS results in in reassuringly normal neonatal outcome parameters including birthweight and congenital malformations. Patients with a severe PCOS phenotype constitute an important group of candidates eligible for this treatment, as their excess antral follicle count may, at least to some extent, compensate for the intrinsically lower potential of oocytes matured in vitro. Nevertheless, when evaluating obstetric and neonatal outcomes of IVM, compared to COS, the higher incidence of adverse obstetric outcomes in IVM pregnancies needs to be evaluated against the background of more unfavourable PCOS phenotypes. That said, follow-up of children conceived using this technology has to extend beyond the neonatal period. Our institution has a strong track record of follow-up of children after ART, and follow-up data of IVM children will be presented in a separate paper (Belva et al., in preparation). Given the clear incentive for IVM not only in patients with PCOS but also in patients with a diagnosis of cancer who request fertility preservation and have no time to undergo ovarian stimulation for oocyte or embryo cryopreservation, one should expect to see increasing numbers of live births after IVM, and safety studies in these children will therefore be a crucial focus of future research.

Authors’ roles

L.M. and M.D.V. are responsible for the concept and the study design. L.M. performed the data collection, and R.B. did the statistical analysis. M.D.V. drafted the manuscript. F.B. is responsible for the clinical follow-up program of children after ART. I.S., S.R.S., C.B., E.A., J.S. and H.T. contributed to the critical discussion, interpretation and editing of the manuscript.

Funding

Translational IVM research at Universitair Ziekenhuis Brussel (UZ Brussel) and Vrije Universiteit Brussel (VUB) has been supported by grants from the Institute for the Promotion of Innovation by Science and Technology in Flanders (Agentschap voor Innovatie door Wetenschap en Technologie—IWT, project 110680), the Fund for Research Flanders (Fonds Wetenschappelijk Onderzoek–Vlaanderen—FWO, project G.0343.13) and the Belgian Foundation Against Cancer (HOPE project, Dossier C69). Clinical IVM research was supported by research grants from Cook Medical and Besins Healthcare.

Conflict of interest

M.D.V. reports honoraria for lectures from Cook Medical and Besins Healthcare outside the submitted work. S.S.R. reports honoraria for lectures by MSD and Besins and research grants by MSD, Ferring and Merck Serono outside of the submitted work. C.B. reports personal fees from Merck-Serono, Ferring, IBSA, Finox, MSD and Abbott outside the submitted work. H.T. reports grants from Merck, MSD, Goodlife, Cook, Roche, Besins, Ferring, Mithra (now Allergan) and the Research Fund of Flanders (FWO) and consultancy fees from Finox, Abbott, Obseva, and Ovascience outside the submitted work. The other authors have nothing to disclose.

References